Functional load: Descriptive limitations alternatives of assessment and extensions of application 9783111354163, 9783110998771

Citation preview

FUNCTIONAL LOAD

JANUA LINGUARUM STUDIA MEMORIAE NICOLAI VAN WIJK DEDICATA

edenda curai

C. H. V A N S C H O O N E V E L D INDIANA UNIVERSITY

SERIES MINOR 99

1970

MOUTON THE HAGUE • PARIS

FUNCTIONAL LOAD DESCRIPTIVE LIMITATIONS ALTERNATIVES OF ASSESSMENT AND EXTENSIONS OF APPLICATION by

R. S. M E Y E R S T E I N CALIFORNIA STATE COLLEGE SYSTEM NORTHRIDGE, CALIFORNIA

1970

MOUTON THE HAGUE • PARIS

© Copyright 1970 in The Netherlands. Mouton & Co. N.V., Publishers, The Hague. No part of this book may be translated or reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publishers.

LIBRARY OF CONGRESS CATALOG CARD NUMBER: 70-126053

Printed in The Netherlands by Mouton & Co., Printers, The Hague.

To Z

TABLE OF CONTENTS

1. Existing Work on Functional Load 1.1 Significance Stated 1.2 Definitions and Descriptions 1.3 Previous Criticism 1.31 Synchronic Validity 1.311 Lack of computation 1.312 Relevance of criticism 1.32 Predictive Validity 1.321 Diachronic claims 1.322 Uncertainty of claims and of counterclaims 1.4 Measurement of Functional Load 1.41 Formulas of Measurement Proposed . 1.411 Scope of previous measures . . . 1.412 Measures disproving diachronic predictability 1.42 Relevance of Measurement Proposed . . 1.421 Incompatibility of previous quantifications 1.422 Previous emphasis on irrelevance disregards relevance 1.5 Limitations of Established Theory and Approach 1.51 Principles of Limitation 1.511 Additional questions presented by previous scholarship 1.512 Seven principles of previous identification of functional load

13 13 15 17 17 17 18 18 18 19 21 21 21 23 26 26 27 28 28 28 29

8

TABLE OF CONTENTS

1.52 The Non-Unitary Principle 1.521 Non-unitary prevalence . . . . 1.522 Unitary precedent 1.53 The Principle of Non-Zero Counterparts . . 1.531 Non-zero prevalence 1.532 Zero-precedent 1.54 The Principle of Sub-Morphemic Level 1.541 Submorphemic restriction . . . . 1.542 Supraphonemic extensions proposed . 1.55 The Principle of Unique Function . . . 1.551 Formal-denotative prevalence . . . 1.552 Precedent beyond formal-denotative function 1.553 Function, functionalism, and functional load 1.56 The Principle of Frequency versus Load 1.561 Prevalence of load determined by frequency 1.562 Parallelism proposed: frequential and non-frequential load . . . . 1.563 Frequential and non-frequential determinants: ranking 1.57 The Principle of Complementarity Entailing Zero Load 1.571 Non-complementary prevalence 1.572 Non-emic precedent : etic-emic coequality. 1.573 Differentiated load of variants: frequency of functions 1.58 The Principle of Systemic Uniqueness . 1.581 Monosystemic dichotomy: competence versus performance 1.582 Lexical competence versus textual performance 1.583 Parallelism of monosystemic and bisystemic dichotomy: load points . . .

30 30 31 35 35 36 38 38 39 40 40 41 43 45 45 47 47 49 49 50 53 54 54 55 58

TABLE OF CONTENTS

2. Alternatives of Functional Load Assessment . 2.1 Criteria of Comparison 2.2 Form 2.3 Functional Load 2.301 Extended scope of load areas . . . 2.302 Typological insignificance of text-token quantification 2.303 Lexical token quantification . . . 2.304 Type quantification proposed: overload of load points 2.305 Formal and non-formal determinants of overload scores 2.4 Levels 2.5 Systems 3. Individual Correlations 3.1 Type 1 — Traditional Binary Oppositions {a :b — 2:1). 3.101 Traditional scope integrated and extended 3.102 Overload implications excluded 3.103 Overload considered; variable, pedagogic. 3.104 Overload in traditional 'zero-load' oppositions ; variable, formal (feature-spread) 3.105 Overload in non-traditional oppositions; variable, semantic (connotative) . . 3.106 Overload in zero versus non-zero oppositions; variable, bisystemic (interdialectal) 3.107 Supraphonemic overload; variable, significative contrastivity potential (word differentiation) 3.108 Typology of supraphonemic overload: unequal phonemic potential; lexical (non-textual) attribute ; level interaction 3.109 Supraphonemic overload: submorphemic versus morphemic function . . . 3.110 Supraphonemic overload: mono-set ver-

9 61 61 62 64 64 65 67 69 71 72 73 76 76 76 76 78 79 81

81

82

83 84

10

TABLE OF CONTENTS

sus bi-set function (word set differentiations) 3.111 Monosystemic versus bisystemic overload; variable, 'stylistic' factions . . . 3.112 Idiolectal variable of linguistic system (speech habits); implications, pedagogic 3.113 Idiolectal variable of exolinguistic situation (intent) 3.2 Type 2 — Other Binary and Non-Binary Oppositions (a:m — 2:1) 3.201 Phonemic overload in multiply distinctive oppositions; variable, pedagogic (exolinguistic) 3.202 Variable, linguistic (level-transcending: morpheme-directed) . . . . 3.203 Level-transcending: set-directed . . 3.204 Morphemic (significative) overload of phonemic units 3.205 Three cases of unit-overload(+morphemic —syntactic; —morphemic —syntactic; +morphemic + syntactic) . . 3.206 Morphosyntactic overload of morphemic units: functional overlap versus functional potential (set, sequence) . 3.207 Frequential overload; variables: formalsynchronic (syntagmatic, paradigmatic); formal-diachronic(duration of existence, paradigmatic); semantic; exolinguistic . 3.208 Overload, frequential and semantic; synchronic variable of cooccurrence in either system 3.209 Synchronic variable of cooccurrence in both systems 3.210 Variable, synchronic or diachronic; cooccurrence or non-cooccurrence in both systems

84 85 85 86 87

87 88 90 90

91

91

92

93 93

94

TABLE OF CONTENTS

11

3.211 Frequential overload; variables, dialectal (geographic, social) or idiolectal subsystems 94 3.212 Semantic overlap in bisystemic paradigmatic strings 95 3.213 Opposite scores of frequential and nonfrequential (semantic) load; additional variable, systemic 96 3.214 Frequential load as a variable of systemic typology 97 3.215 Typology of or by the system; variable, linguistic or exolinguistic . . . 97 3.216 Load shifts; variable, frequential-exolinguistic ('idiomaticity') . . . . 98 3.217 Formal similarity versus frequential-load dissimilarity; exolinguistic (pedagogic) response 99 3.218 Formal similarity and frequential or non-frequential load dissimilarity; unidirectional conditioning of load increase 99 3.219 Static versus dynamic potential of overload: productivity 100 3.220 Admissible versus inadmissible productivity; variables, frequential or nonfrequential ('correctness'), synchronic or diachronic 101 3.221 Load score increase versus maintenance: productivity versus survival . . 101 3.222 Load score fluctuation: survival, demise, and revival 103 3.223 Load score creation ('birth'); variable, non-frequential (level) . .104 3.224 Load score evaluation; variables, linguistic (system) and exolinguistic (admissibility) 104 3.3 Type 3 — Binary Sequences and Oppositions (ab'.a — 2:1) 104

12

TABLE OF CONTENTS

3.301 Extent inequality of load exponents; syntagmatic extension, simultaneous (marked form) or consecutive (long form) 3.302 Opposite scores of mark-addition and frequency-potential (survival); variable, systemic; implications, exolinguistic (medical) 3.303 Opposite scores of longitudinal addition and non-frequential (semantic) potential: abbreviations 3.304 Extent inequality of load exponents; variable, systemic; functions, identical (linguistically in denotation) or dissimilar (exolinguistically in connotation) 3.305 Scores, opposite; variable, systemic; functions, dissimilar 3.306 Extension, optional or obligatory; variable, systemic 3.307 Exolinguistic implications of various types

104

105

106

106 107 108 108

4. Conclusion

Ill

References Index

117 118

1. EXISTING WORK ON FUNCTIONAL LOAD

1.1

SIGNIFICANCE STATED

The concept of functional load is well established.1 The importance of work on this subject has been seen, primarily, as a matter of descriptive completeness.2 A considerable part of previous scholarship has been devoted to the presentation of functional load as an explanatory factor in language history and in developmental trends to be expected.3 Synchronically, a number of areas of significance have been pointed out. For example, functional load was said to be a criterion assisting us in deciding what we should try to discover in the spectrograph^ analysis of speech,4 in explaining statistical processes which underlie the decoding of speech sound waves,5 in promoting research on speech recognition,6 in accounting for 1

Functional load has also been referred to, in English, as functional yield, yield, functional burden, burden, communication load, weight; in French, as rendement fonctionnel, rendement; in Czech, as funköni zatizeni; in German, as Belastung, semantische Belastung, Belastungs- und Kombinationsfähigkeit, Tragfähigkeit. All of these terms will be represented in this study by the terms FUNCTIONAL LOAD or LOAD. Quotations on functional load may pertain to statements which do not specifically use any of the terms listed above yet refer to it by implication; cf., for instance, some of the references to frequency cited in Section 1.56, below. 2 Trubetzkoy (1932), 18; Hockett (1955), 218; King (1967 a), 832; Mackey (1966), 96. 3 KuCera (1963), 209; King (1967 a); King (1967 b); Wang (1967), 43, 50. * Hockett (1955), 218. 5 Wang (1967), 43. 6 Wang (1967), 51. Cf. Rigault (1962).

14

EXISTING WORK ON FUNCTIONAL LOAD

aberrations from phonological or morphosyntactic norms, 7 in fact, in understanding entire linguistic systems, regardless of levels of analysis, and beyond that, certain linguistic universals.8 In areas primarily classified as those of applied linguistics, functional load has attracted attention as a relevant criterion in language management, such as devising new alphabets,9 or for such other practical purposes as the identification of optimum arrangements on typewriter keyboards.10 In pursuit of objectives less remote from the linguist's preoccupations, functional load has been related to the study of poetry, 11 to deciding claims of disputed authorship, 12 and above all, to a number of aspects of linguistic pedagogy. Instructional benefits were seen in terms of occurrence frequency, range, coverage, availability in specific situations, relative learnability, and priorities of procedure, as well as the preparation of teaching materials based upon studies of this type. 13 It cannot, however, be said that the present record reflects consensus about the significance of functional load studies, and questions have been raised about the justification of measuring this load, 14 and the extent to which the concept of functional load is applicable.15 The present discussion will review the principal features of existing load work and attempt to suggest some of the reasons why a concept as long in existence as most of the other analytic notions in current use, and as potentially significant as any of these for descriptive theory and practical application, has not as yet been put to any effective use as an integral part of the contributions of linguistics. 7 8 9 10 11 12 13 14 15

Phonological: Hockett (1958), 64; morphosyntactic: Martinet (1962), 55. Kudera (1963), 217. Cf. Greenberg (1966 a); Greenberg (1966 b). Wang (1967), 51. Frumkina (1963), 89. Frumkina (1963), 109. a . also Chatman (1964); Doleiel (1967). Yule (1939). Hockett (1958), 47, 56, 63, 64; Gougenheim (1959); Frumkina (1963), 91; Mackey (1966), 189; Stockwell, Bowen, and Martin (1965), 5. Wang (1967), 50. Stockwell, Bowen, and Martin (1965), 5.

EXISTING W O R K ON FUNCTIONAL LOAD

15

1.2 DEFINITIONS AND DESCRIPTIONS Functional load, as currently understood in représentative publications o n the subject, is reflected in the following définitions a n d descriptions presented in chronological order. RENDEMENT FONCTIONNEL. ... Degré d'utilisation d'une opposition phonologique pour la différentiation des diverses significations des mots dans une langue donnée. 16 RENDEMENT FONCTIONNEL DES PHONÈMES — SON EXPRESSION STATIS-

TIQUE. ... Toutes ... particularités qui donnent à chaque langue sa physionomie particulière peuvent s'exprimer en chiffres. De même le degré d'utilisation distinctive (le 'rendement fonctionnel') des diverses oppositions phonologiques et le rendement moyen des phonèmes se laisse en général établir en chiffres pour chaque langue par cette méthode d'étude du vocabulaire. 17 Allgemeingefasst kann der Grad der Ausnützung von phonologischen Einheiten in der gegebenen Sprache an dreierlei Tatsachengruppen gemessen werden. Es kann sich um ihre Ausnützung (1) im System, (2) im Wort- und Wortgruppenbau, (3) in dem Strom der aktuellen Rede handeln. Was man unter AUSNÜTZUNG IM SYSTEM ZU verstehen hat, wird am besten ein Beispiel lehren. Der Stimmunterschied in der Konsonantenbildung, d.h. das Vorhandensein bezw. Fehlen der Stimme bei den Konsonanten, erscheint als phonologisches Merkmal in einer ganzen Reihe von Sprachen, aber der Grad der Verbreitung dieses Unterschiedes im phonologischen System ist oft sehr verschieden.... Der Stimmunterschied kann auch f ü r die STRUKTURALE AUSNÜTZUNG der phonologischen Einheiten als ein gutes Beispiel dienen. 18 On donne souvent à l'importance fonctionnelle d'une opposition phonologique le nom de 'rendement fonctionnel' (angl. functional yield ou bürden, all. funktionelle Belastung). Il n'y a pas accord complet sur la valeur exacte de ce terme. Dans son acception la plus simple et la plus naïve, il s'applique au nombre de paires du lexique qui seraient de parfaits homonymes, s'il ne se trouvait qu'un mot de la paire présente un membre A de l'opposition là où l'autre mot présente l'autre membre B : la paire blanc-blond est un élément du rendement fonctionnel de l'opposition /à/ - /£>/ en français, de même que dent-don, sang-son, banc-bon et une foule d'autres. Le nombre de telles paires /ä/ - /ô/ étant 16 17 18

"Projet" (1931), 313. Trubetzkoy (1939), 288. The heading appears in Vachek (1960). Mathesius (1931), 148-149.

16

EXISTING WORK ON FUNCTIONAL LOAD

considérable, on dit que le rendement fonctionnel de l'opposition /â/ - /ô/ est élevé. Si nous essayons de faire la même chose avec, disons, l'opposition /e/ - /œ/, nous ne trouverons qu'un petit nombre de paires comme brin-brun, empreint-emprunt. On dira que le rendement fonctionnel de cette opposition est faible. 19 The way in which some phonologie contrasts do more of the work of keeping utterances apart than others is the matter of FUNCTIONAL LOAD. ... Assuming that two phonemes, x and y, can contrast at all, then the functional load carried by the contrast will be greater if both x and y have relatively high text frequencies than if one has a high frequency and the other a low frequency, and greater under those second conditions than if both x and y have low frequencies. The large number of pairs of English words differing only in the presence of /p/ or of /b/ at a certain point obviously implies the potential existence of a larger number of long minimal pairs turning on /p/ and /b/ than on, say, Ipl versus the relatively low-frequency /s/; yet one will expect a larger number of the latter than of minimal pairs turning on /§/ versus the very low-frequency /z/. Therefore we can be pretty safe in concluding that the /p/ : /b/ contrast carries a higher functional load than the /p/ : /s/ contrast, and the latter a higher load than the /§/ : /z/ contrast — indeed, the functional load of the last contrast must be vanishingly small. ... 2 0 The terms functional yield or functional load are employed in current linguistic discussion to characterize in a general way the extent to which contrasts among members of a set of phonemes or a set of contrastive features contribute to the signalling of significant differences. 21 Another of the considerations that must enter into building a pedagogical sequence [is] "the more typically Spanish construction...". This is the criterion of FUNCTIONAL LOAD. Language is communication, and those patterns which carry a proportionately larger share of the burden of communication are more important and deserve to be emphasized. 22 [Functional load is] the extent to which a given sound ... is used to distinguish one word from another, the quantity of distinctive information it carries. 23 The notion of functional load is that a phonemic system ... has a 19 20 21 22 23

Martinet (1955), 54-55. Hockett (1955), 215. Greenberg (1959), 7. Stockwell, Bowen, and Martin (1965), 292-293. Stockwell and Bowen (1965), 16.

EXISTING WORK ON FUNCTIONAL LOAD

17

(quantifiable) job to do, and that the contrast between any two phonemes, say /a/ and /b/, carries its share.24 The term FUNCTIONAL LOAD is customarily used in linguistics to describe the extent and degree of contrast between linguistic units, usually phonemes. In its simplest expression, functional load is a measure of the number of minimal pairs which can be found for a given opposition. More generally, in phonology, it is a measure of the work which two phonemes (or a distinctive feature) do in keeping utterances apart — in other words, a gauge of the frequency with which two phonemes contrast in all possible environments.25

1.3

PREVIOUS CRITICISM

1.31 Synchronic Validity 1.311 Lack of computation. — As based upon the definitions and descriptions cited, the concept of functional load has to date prompted few evaluative comments. Whatever criticism is reflected in linguistic literature centers on the qualitative and impressionistic nature of earlier exposition, the lack of supporting evidence, and subsequent statistics which tended to disprove claims of explanatory or predictive applicability advanced in previous work on the subject. 26 We may assess complaints of this type for the shortcomings 24

Hockett (1966), 8. King (1967 a), 831. Other discussions, specific or passing, of functional load include Malmberg(1942); Herdan (1958); Francescato (1962); Newton (1963); Avram (1964); DaneS and Vachek (1964); Faure (1964); Dolezel and Prucha (1966); Haugen (1966); Shapiro (1967); Clivio (1968); Vendler (1968); Wang (1969); and others mentioned below in connection with particular problems. For general discussions of background questions, cf. Bloomfield (1933); Martinet (1945); Haugen (1951); Bally (1954); Whatmough (1956); Martinet (1960); Ray (1963); Ivic (1965); Jakobson et al. (1965); Bolinger (1968); Hockett (1968). For related technical terminology, cf. also Hamp (1956); Vachek (1960); Marouzeau (1961); Pei (1966). 26 King (1967 a); Wang (1967), 36, 37. 25

18

EXISTING WORK ON FUNCTIONAL LOAD

they note, as well as for those they leave out of account, and we may wonder whether remedies ofiFered provide answers which would give statements on functional load descriptive significance and practical applicability. Critical comment deploring the absence of supporting statistics bears upon previous statements unsupported but not disproved, as well as statements unsupported and subsequently found to be at variance with observed facts. 1.312 Relevance of criticism. — Labeling of functional load evaluations as 'impressionistic', for lack of precise computations, need not in and by itself unduly preoccupy us. Operations with theoretical concepts inaccessible to empirical verification at an early stage of research are consonant with accepted procedure in many fields. In the context of the present discussion we shall regard complaints centering on lack of quantitative support to be trivial. We shall, however, enter the caveat that what is seen as high frequential load may correspond to a low load value in a non-frequential sense. Either view reflects a purely descriptive formulation without pretensions of accounting for past or subsequent events, i.e. without promise of predictive validity. We shall hold that a number, possibly the great majority, of functional load identifications cannot at present be supported by measurement or similar numerical data, and that in the presence of equivalent identificational principles many of them need not be so supported.

1.32 Predictive Validity 1.321 Diachronic claims. — Criticism on record carries weight primarily against earlier assertions to the effect that functional load had been the cause, or one of the causes, of phonological developments which occurred in certain languages. This type of assertion transcends the confines of mere formfunction correlations in a synchronic context, or of diachronic extension of bilingual description. This time we are told, not merely

EXISTING WORK ON FUNCTIONAL LOAD

19

that functional load is to varying degrees carried by one or the other form, but that it causes or influences the development of the form. The assertion implies that by compiling statistics on functional load we shall discover the inevitability of past developments and ipso facto their predictive import. 1.322 Uncertainty of claims and of counter-claims. — It would ap-

pear that in the absence of data which validate the prediction, complaints about mere assumptions unsupported by numerical evidence are indeed justified. Still, we may hesitate to lend unquestioning endorsement to criticism along these lines. Whatever claims have been made have been so guarded or indefinite that little if anything may be verified. It is the very lack of precision that renders earlier cause-and-effect hypotheses impervious to counter-claims. Wherever, in turn, we do encounter assertions of a more specific nature, there arises a question of whether the data assembled to prove or disprove the claims are any more valid than the claims themselves. Existing statements about functional load as a cause of sound change (nowhere, it seems, ever as THE cause) to varying extents hedge their bets. A small functional load, for instance, is said merely to FAVOR the loss of an opposition.27 Other views, quoted below (with emphasis added) are similarly non-committal, uncertain, contradictory, or interrogative rather than assertive. Des extensions sémantiques, la composition, une réorganisation morphologique apportent SOUVENT des solutions faciles aux problèmes qui PEUVENT surgir lorsqu'une opposition importante fonctionnellement se trouve menacée. ...28 [It] has been proposed that the PROBABILITY that a given phonologic contrast will be extirpated through sound changes stands in inverse ratio to the functional load carried by that contrast.29

27

Jakobson (1931), 259. On multiple causation, cf. also Malkiel (1960); Mal kiel (1967); Wang (1969). 28 Martinet (1955), 58-69. 29 Hockett (1955), 216.

20

EXISTING WORK ON FUNCTIONAL LOAD

How HIGH trast]?30

MUST THE LOAD BE

[to prevent elimination of a con-

[TOUTES] CHOSES ÉGALÉS D'AILLEURS, une opposition phonologique qui sert à maintenir distincts des centaines de mots parmi les plus fréquents et les plus utiles N'OPPOSERA-T-ELLE PAS une résistance plus efficace à l'élimination que celle qui ne rend de service que dans un très petit nombre de cas?31 Dans le cas de CERTAINES confusions [/Ê/ - /œ/ dans le français de Paris], on constate un rendement fonctionnel extrêmement bas. ... Mais le rendement fonctionnel de l'opposition /ô/-/à/ est ... CONSIDÉRABLE, et la distinction se maintient bien... .32 On devra pourtant noter ... que ... deux phonèmes voisins ne tendent pas nécessairement à se confondre pour la seule raison que leur opposition est d'un rendement fonctionnel pratiquement nul : on ne voit pas que /s/ et /z/ en anglais tendent à se rapprocher en dépit du rendement exceptionnellement bas de leur opposition... , 33 [Le] rendement fonctionnel d'une opposition est L'UN des facteurs de sa conservation ou de son élimination... , 34 [If] a particular contrast is little used in the language, its elimination will do less harm than THE ELIMINATION OF A CONTRAST WITH A HIGH FUNCTIONAL LOAD. 35

In ... cases of complete or conditioned merger where, TYPICALLY BUT functional yield is low, it SEEMS to be the GENERAL rule that the merger is produced by the marked feature losing its mark.36 NOT ALWAYS,

Thus, functional load may be a factor in sound change, but there are other factors which may reduce or, in fact, undo the effect of functional load. This can scarcely be called an unequivocal claim, and as 'iffy' an effect as that suggested in the quotations preceding is hardly amenable to quantificational checks which recent scholarship has demanded and, indeed, attempted, Still, to the small extent that cause-and-eifect claims were disproved or, at any rate, not significantly upheld by diachronic load statistics, criticism 30 31 32 33 34 35 36

Hockett (1966), 1. Martinet (1955), 54. Martinet (1955), 57-58. Martinet (1955), 58. Martinet (1955), 78. Hoenigswald (1960), 79. Greenberg (1966 a), 96.

EXISTING WORK ON FUNCTIONAL LOAD

21

of the limitation of previous work on the subject seemed to be justified. The remedy of shortcomings noted, as the critics saw it, appeared to be the accumulation of further statistics of oppositions expressed in terms of occurrence frequency of opponents, and the presentation of resultant formulas by which functional load might be measured.

1.4 MEASUREMENT OF FUNCTIONAL LOAD

1.41 Formulas of Measurement Proposed 1.411 Scope of previous measures. — Formulas of measurement offered overtly or tacitly intend to fill the evidential credibility gap of earlier work on the nature and effect of load. Hockett's initial exposition does not claim any specific validity. Incidental to phonology in general, it limits itself to suggesting that "... a tabulation of all the phonological contrasts in a system, showing the relative functional load carried by each, would constitute a valuable addition to any description of a phonologic system, and might help us considerably in deciding what to try to find on spectrograms of natural rapid speech", but does not propose to account for diachronic development. As "the ratio of the entropy which would be lost if the contrast were abolished to the entropy of the unchanged system", the functional load which it measures appears to attach to phonological contrasts of any linguistic system. 37 Greenberg's outline of methodological problems raised by the measurement of functional load is likewise "concerned only with the synchronic aspect of functional yield"; yet, without being unduly exegetic, we may take the observation that "the diachronic thesis [of high functional load tending to preserve a contrast] makes little sense unless we can compare functional yields in respect to magnitude" to suggest that functional load thus measured may 37

Hockett (1955), 217-218.

22

EXISTING WORK ON FUNCTIONAL LOAD

support diachronic conclusions. Though the study concerns one American Indian and three African languages, its author presents functional load measurements evolved for these four speech types as being assignable to "any set of phonemes or features for any language". What is measured is the feature of tone, beginning with its relative yield, "the total number of valid tonal substitutions divided by the total number of phonemes which have significant tone for all the texts examined", which, "multiplied by the overall frequency of vowels", gives the measure of absolute yield.38 Rischel's presentation on functional load in phonemics, with a separate chapter on the load of phonemic contrasts, appears to endorse the objective of load measurement of a previous report on the subject: this load is simply "an important thing to measure" — as done, in order to label as carrying high functional load those contrasts which are distinguished by certain frequency and entropy values of their constituent members, i.e. without implications of explanatory power for language development. Illustrations offered are selected from Danish or English, or represent hypothetical sequences. Though no universal validity is claimed for the procedure, the article appears to support the conclusion that the measure is valid for other languages as well.39 Kucera's correlation of entropy, redundancy, and functional load in Russian and Czech proposes a computational procedure for those two languages only. Again we are not offered the resultant formula with the thought that it might explain or project trends of change nor, in fact, that it might label anything other than "the extent of uncertainty contributed to [a given] environment ... by [a given] distinctive quality ...", though apparently also with validity for languages other than those referred to. 40 In a subsequent quantification of the functional load of "phonemic, allophonic, and componential" contrasts, Hockett draws illustrations from English, and his computations result in "three 38 39

Greenberg (1959), 7-9. Rischel (1961), 17. Cf. Jensen (1960), and Jensen (1961), quoted by Rischel,

loc. cit. 40

Kuiera (1963), 191. On predictability and redundancy, cf. also Chatman (1964).

23

EXISTING WORK ON FUNCTIONAL LOAD

interrelated measures, one based on ... the entropy in binits, one based on ... the entropy in shannons, and one based on ... the relative entropy". The measure is evidently offered as being valid for any language and, as pointed out above, its author asks, but does not indicate, how high the functional load must be to prevent disappearance of the contrast which carries it; earlier hunches, he notes, "are incisive and suggestive, and perhaps in part wrong; but they cannot be confirmed or disproved merely by someone else's hunches". 41 1.412 Measures

disproving

diachronic predictability.

— I t is

not

until the publication of King's work, specifically devoted to functional load and sound change, that we appear to have been given the opportunity to assess the import of load value measurement in accounting for linguistic change. Except for one page on Old Saxon, King's earlier article is limited to a consideration of vowels and, though operating with functional load in a quantitative sense, does not outline the method or formula of quantification, for which King refers to previous unpublished material. We are merely told that in general terms ... the formula for functional load used here ... applies to oppositions between pairs of phonemes ..., though it can be used to derive an estimate of the functional load of a distinctive feature. The present formula is the product of two factors: the first measures the global text frequencies of the two phonemes involved; the second measures the degree to which they contrast in all possible environments, where environment means, roughly speaking, one phoneme to the left and right. This formula requires an input text of several thousand phonemes, the reliability obtained in the computation of functional load indices being proportional to the length of the input text provided.42 With reference to earlier and later stages of various Germanic languages, King establishes a three-way classification of previous assumptions on load effect: 41

Hockett (1966), v, 1, 2. Cf. fn. 67, below. King (1967 a), 836. For two earlier correlations of load and diachronic change, cf. Diver (1955); Abernathy (1963).

42

24

EXISTING WORK ON FUNCTIONAL LOAD

THE WEAK POINT HYPOTHESIS states that, if all else is equal, sound change is likely to start within oppositions bearing low functional load.... THE LEAST RESISTANCE HYPOTHESIS states that, if all else is equal, and if ... there is a tendency for [merger], then that merger will occur for which the functional load of the opposition is smaller... . THE FREQUENCY HYPOTHESIS states that, if an opposition ... is destroyed by merger, then that phoneme will disappear in the merger for which the relative frequency is smaller... . 43

King finds that to varying extents the three hypotheses are disproved by his data, and that "functional load, if it is a factor in sound change at all, is one of the least important [for] the cause and direction of phonological change". His data reflect the diverse effect of functional load in the developments of the respective languages of his corpus. If one wished to apply functional load data predictively, King adds, "this is bad predicting; a set of good hypotheses should explain things better than that". At this point King's evaluation, written as a reaction to previous "hypotheses [none of which] stands up very well in confrontation with empirical findings", appears to engage in some hypothesizing and generalizations of its own. While earlier work had advertised functional load as ONE of the factors in sound change of certain specific systems, and while this type of claim, cautious though it might be, was taken to task for lack of supporting evidence, we are now told that functional load, apparently in any linguistic system, "is best disregarded in discussions centering on the cause and direction of phonological change", that is, that it is NONE of the factors, or "one of the least important", a claim seemingly as sweeping and far less guarded than the earlier claims, on the basis of one relatively small set of data of relatively restricted scope.44 On one issue we are certainly not advanced: what is the relative impact of functional load on sound change ? According to King, It may well be that a more grandiose scheme, based on better and more complete knowledge of what communication in language REALLY is, 43

King (1967 a), 834-835, 844 (for the quotations of the following paragraph). 44 King (1967 a), 831, 847, 848.

EXISTING WORK ON FUNCTIONAL LOAD

25

would find that functional load is, after all, a useful tool in seeking answers to the 'why' of sound change. But it will not be easy to come by this more grandiose scheme. ... This brings up the difficult question of just what functional load ought to measure. Should we measure the functional loads of subsets of the classificatory distinctive features ? Or certain phonological rules ...? 45 In a subsequent proposal on the measurement of functional load, King offers what he terms the first "usable data" on the subject, rather than "purely theoretical discussion" or "speculation", and here he does frame a succession of formulas: [The] functional load of the opposition xi ^ xj is the fraction of phoneme occurrence of /xi/ among the N phonemes in the text, multiplied by the fraction of occurrence of the other phoneme /xj/; this is then multiplied by a sum of products, for each of which the multiplier is the relative frequency of the phoneme /xi/ in the k-th environment, and the multiplicand is the relative frequency of /xj/ in that same environment. ... [The] functional load of the opposition xi + xj is the reciprocal of the square of the number of phonemes in the entire text multiplied by a sum of products, for each of which the multiplier is the number of occurrences of phoneme /xi/ in the environment Ek, and the multiplicand is the number of occurrences of phoneme /xj / in the environment En.46 These equations, based on the same Germanic data as King's earlier study, but with additional data on consonants and, possibly, the same as for the latter, offered as applying to other languages as well, "allow only the computation of functional loads between pairs of phonemes, not of a distinctive feature like voice or gravity". The reliability of computations "is proportional to the length of the input text provided ... [A] text of 20,000 phonemes suffices for almost all purposes and a text of 5,000 for most". Interestingly enough, "no particular importance attaches to the absolute size of a functional load — for the most part the linguist will be interested only in the size of the functional load of one opposition relative to the functional load of another opposition". 47 45 46 47

King (1967 a), 849. King (1967 b), 4. King (1967 b), 8. Cf. Section 2.302, below.

26

EXISTING WORK ON FUNCTIONAL LOAD

Wang's exposition on the measurement of functional load also disclaims being "qualitative and impressionistic". This will expose us to problems; in the absence of solutions, can we expect reliable or indeed meaningful measures ? Since any statistic has to be extracted from a finite sample whereas language constitutes an infinite set of sentence tokens, we are immediately faced with a problem in sampling. ... In this paper we will by-pass the problem. ... The task of measuring [functional load] may be regarded as assigning numerical values to various [subsets] such that these values reflect the amount of 'work' that phonemes ... do in distinguishing phoneme strings in the text, and by inference in the language.48 Wang proceeds to state a series of formulas measuring functional load according to the amount of uncertainty contributed by a phoneme to its environment. Still, the difficulty in measuring [functional load] is enormous. While the measures examined here clarify certain issues and suggest avenues for further research, the problem has not been solved.49 1.42 Relevance of Measurement Proposed 1.421 Incompatibility of previous quantifications. — It was noted that recent work on functional load had traced the deficiency of earlier statements to the absence of quantitative data likely to yield generalizable formulas of measurement. At this point, eight proposals to remedy the deficiency have been reviewed, and we may ask: (1) Do the proposals measure comparable data? (2) Assuming that they do, are the respective measures compatible? (3) What is the explanatory adequacy of the measures? In other words, it was said that for lack of yardsticks the notion of functional load was of questionable descriptive relevance. Now that yardsticks are at hand, are we substantially advanced ? 48

Wang (1967), 37-38. Cf. Thatcher and Wang (1962), on which Wang (1967) is based. 49 Wang (1967), 50.

EXISTING WORK ON FUNCTIONAL LOAD

27

The answer to the first question is clearly affirmative. Apart from differences in terms, largely conditioned by different avenues of approach (e.g. expectable results in variations of entropy, rather than mere occurrence), there is little if anything to distinguish the respective objects of attention: phonological oppositions, or distinctive features as their exponents. The compatibility of material should entitle us to anticipate compatible measures. In reality, however, not one of the measuring formulas is even remotely comparable to any of the other quantifications. The very existence of eight dissimilar proposals said to measure similar data, in fact, permits the inference that even in the absence of mutual criticism none of the respective authors appears to lend much credence to the formulas of any of the others. By comparison, we may point to the measurement of phonological features in general; a great many views may be on record, yet scarcely with as much divergence on the subject of their quantitative evaluation as that noted for statements on load. Quite clearly, we are afforded no remarkable incentive to consider measures as proposed to date to be the key to functional load as a relevant concept of descriptive theory, let alone any practical applications. Thus, if after reviewing attempts to evolve load measurement formulas outlined by his seven predecessors, Wang concludes that the problem remains unsolved, we can readily agree. 1.422 Previous emphasis on irrelevance disregards relevance. — Uncertain validity, noted for existing measurements as such, is evident to an even greater degree for measures as the basis of accounting for linguistic change. Six of the eight proposals reviewed evaluate single-stage (synchronic) data, that is, label function in quantitative terms without mention of relevance to historical events. The remaining two proposals do relate to diachrony, but in a negative sense, denying that measures of functional load have much to contribute to explaining linguistic development. Inevitably, we are prompted to reexamine criticism of previous work in the light of remedies which the critics themselves and others have offered. We were told that lack of quantitative research alone

28

EXISTING WORK ON FUNCTIONAL LOAD

had called the value of previous work into question; without that research, it was said, functional load had no explanatory significance. Now we do have no less than eight detailed proposals of quantification, with the net result that the defects remain, and the explanations continue to be missing.

1.5 LIMITATIONS OF ESTABLISHED THEORY AND APPROACH

1.51 Principles of Limitation 1.511 Additional questions presented by previous scholarship. — One of the quantificational studies discussed above expressed the belief that previous (non-quantificational) work had "carried the matter as far as it can be carried without actual quantification". 50 It will be one of the principal objectives of the present discussion to take issue with that view. It is evident that recent quantificational work, both by its nature and by its divergent results, has failed to emphasize the value of functional load as a descriptive concept in the sense of widening the range of possible applications, but rather in at least two instances explicitly disavowed its value. It thus seems indicated to renew the inquiry into circumstances which until now have prevented functional load from acquiring the status of a significant and useful criterion. As the result of establishing these circumstances, we shall propose modifications of previous principles of load analysis, as well as additional avenues of approach. Our inquiry will assume the form of preliminary questions: Assuming that further quantificational work will not yield benefits in the areas listed in previous scholarship and summarized above, as well as in other fields to be mentioned, what other directions might appear promising? Will work along different lines be possible on foundations of previous theoretical work? To the extent that it will not, what conceptual reorientation will be required? To what degree should we be prepared to forgo the search for answers 50

Hockett (1966), 2.

EXISTING WORK ON FUNCTIONAL LOAD

29

not readily attainable, at least on the basis of present bibliographic support, in the context of criteria of functional load previously excluded? What is the number of areas of descriptive and practical application relevant but as yet left out of account which the injection of new criteria will make accessible? The discussion of these questions depends upon identifying the limitations of previous conceptual scope. Definitions, claims, and quantifications currently on record on the subject of functional load add up to the following common denominators, stated as principles. 1.512 Seven principles ofprevious identification of functional load. — (1) THE NON-UNITARY PRINCIPLE. What is defined and consti-

tutes the basis of operations pertains, not to units but to oppositions, i.e. binary relations between units. (2) THE PRINCIPLE OF NON-ZERO COUNTERPARTS. The units related to each other both represent plus-features. (3) THE PRINCIPLE OF SUB-MORPHEMIC LEVEL. The units related to each other are non-linguistic units in the Bloomfieldian sense, i.e. restricted to the phonemic level. (4) THE PRINCIPLE OF UNIQUE FUNCTION. The function referred to is based on the principle of occurrence alone, i.e. on the privilege of position in the sense of Bloomfield's definition. (5) THE PRINCIPLE OF FREQUENCY VERSUS LOAD. Functional load is distinguished from frequency of occurrence, in the sense of a correlation of frequency with a unit and functional load with the opposition of units. (6) THE PRINCIPLE OF COMPLEMENTARITY ENTAILING ZERO LOAD.

The absence of emic opposition conditions the absence, or zerovalue, of functional load. (7) THE PRINCIPLE OF SYSTEMIC UNIQUENESS. Statements of functional load are made in reference to forms of one and the same specific language only and, regarding the forms in question, the statements are undifferentiated within that language. These are some of the characteristics of previous load studies which bear upon the present discussion. None of the quantificational

30

EXISTING WORK ON FUNCTIONAL LOAD

proposals reviewed deviates in any noticeable way from any of the principles enumerated above, and the record of leading scholarship indicates that functional load appears inconceivable except in terms listed. We must, however, remain aware that established theoretical formulations on functional load, and the quantifications derived from these formulations, must be distinguished from coexistent theory and practice in related fields. The next step of this discussion will thus pertain to the comparison of analytic principles proclaimed, adhered to, or disregarded. 1.52 The Non-Unitary Principle 1.521 Non-unitary prevalence. — Most of the definitions cited clearly refer, not to any one unit by itself, but to a replacive (paradigmatic) concept of one unit related to another, a principle here termed non-unitary and reflected in the quantifications mentioned. They do not, then, acknowledge any such thing as the functional load of one unit. What we do find, instead, are references to functional load in terms of "utilisation d'une OPPOSITION", "[the work ofJ some phonologic CONTRASTS", "entropy ... lost if the CONTRAST were abolished", "the extent to which CONTRASTS ... contribute", among the non-quantificational definitions and descriptions; among the quantifications, this is followed up by operations based upon "[the degree of utilization] of the different OPPOSITIONS", "the CONTRAST between any two phonemes", and similar variations on the same theme, with emphasis added. Restated in terms of units, "[the functional load] of any [subset] containing only one phoneme is zero". 5 1 The non-unitary views examined above introduce functional load as a measure of the work performed by any pair of units relative to each other, i.e. upon the principle of BINARY opposition. Among publications previously cited (with emphasis added), we find references to "PAIRES du lexique" constituted by "membre A " versus "membre B " such as blanc-blond by /a/ versus /o/, 51

Wang (1967), 43.

EXISTING WORK ON FUNCTIONAL LOAD

31

"TWO phonemes, x and y, ... contrast [ing]", "the functional yield of b vs. p", "the differential power of each contrast on the basis of minimal PAIRS", "functional load of distinctive features [which] are BINARY oppositions [involving] a series of computations ... separately [for] each set of ... TWO qualities [i.e.] TWO elements [of any] functional load", "the contrast between any TWO phonemes [carrying] its share [of the job done by the phonemic system]", "the measure of minimal PAIRS ... for a given opposition"; and similar allusions to "the TWO phonemes" in other frames of reference, as well as statements implicitly equivalent based on the nature of the descriptive or quantificational procedure adopted, clearly reflect the binary orientation of virtually all of the bibliographic precedent.52 1.522 Unitary precedent. — So much for what appears to be a well-established principle consistently adhered to in a succession of treatments. It cannot, however, be overlooked that the notion of functional load of a unit, or the unitary sense of the term, also derives backing from the framers of the original definitions; there is, moreover, ample evidence that it has been applied to a variety of descriptive problems connected with our topic. The connection may be seen in extensions of the load concept to any relevant linguistic function — one of the chief objectives of this discussion — and the structural utilization and combinability of phonemes, the topic of the following two quotations, clearly qualifies as a linguistic function of relevance and measurable dimensions compatible with the phonological orientation of established avenues of research on functional load. Ein spezieller Fall der strukturellen Ausnützung von Phonemen liegt in ihrer Kombinationsfähigkeit vor. Dabei kann es sich 1. um die Stelle im Worte oder in der Wortgruppe, wo die Kombinationen auftreten, oder 2. um ihre Zusammensetzung handeln. Was die Erscheinungsstelle von Kombinationen betrifft, ist es eine längst bekannte Tatsache, dass einige Sprachen Anfangsgruppen und andere wieder Endgruppen von Konsonanten bevorzugen. So finden wir 52

Martinet (1955), 54; Hockett (1955), 215; Greenberg (1959), 7; Rischel (1961), 18; Hockett (1966), 8; King (1967 a), 831; Wang (1967), 45.

32

EXISTING WORK ON FUNCTIONAL LOAD

in Wörtern, die höchstens aus vier Lauten bestehen, im Deutschen 21 Anfangsgruppen und 47 Endgruppen von Konsonanten, im Cechischen dagegen 160 Anfangsgruppen und 16 Endgruppen.53 Aus der Vergleichung der Belastungs- und Kombinationsfähigkeit der Phoneme in einigen europäischen Sprachen, die Professor Mathesius in seiner Abhandlung dargeboten hat, geht hervor, dass der Vorrat der Phoneme, über den die einzelnen Sprachen verfügen, verschieden ausgenützt ist. Slawische Sprachen z.B. sind weniger ökonomisch als z.B. das Englische oder Französische. Das ist schon daraus ersichtlich, dass in jenen Sprachen die Wortbedeutung sehr oft erst durch die Folge der zwei oder sogar drei oder mehr Silben (welche meistens allein keinen Sinn haben) gegeben ist, während in den zuletzt genannten Sprachen die Worte oft nur aus einer Silbe bestehen. Da die meisten Worte im Englischen einsilbig sind, sind die Englischsprechenden gewöhnt, mit einzelnen Silben eine Bedeutung zu assoziieren. Die fremden mehrsilbigen Worte, welchen in breiteren Schichten des Volkes durchdringen, werden darum oft gekürzt, z.B. choc (chocolate), pram (perambulator), flue (influenza) usw. Diese Neigung des Englischen zur Einsilbigkeit, welche zu seinen charakterischen Merkmalen gehört, hängt von der relativ grossen Tragfähigkeit (und im Vergleich mit anderen germanischen Sprachen von der grösseren Zahl) der Phoneme ab.54 Statements on the order of those by Mathesius and Trnka, quoted above, will be of considerable significance as starting points for the development of such basic premises of this study as the applicability of the notion of functional load to additional frequential-semantic as well as non-frequential and non-semantic areas of analysis. For the moment, however, they are cited merely as instances of preoccupation with functionality based on relations which are not specifically or, indeed, necessarily binary, but instead are interpretable as 'one-to-all-others', i.e. unitary in the sense of this presentation. The unitary view, of course, does not depart from the notion of opposition as such. We should, in fact, remain aware that the earliest introductions of functional load, those of the "Projet" and Trubetzkoy, speak of "oppositions", but not necessarily in an exclusively binary limitation. Trubetzkoy, moreover, explicitly 53 54

Mathesius (1931), 149. Cf. Section 3.2, below. Trnka (1931), 152. Cf. Section 3.3, below.

EXISTING WORK ON FUNCTIONAL LOAD

33

recognizes "oppositions bilatérales et multilatérales", 55 an da section of Martinet's Economie is concerned with a "critique du binarisme". 56 Recent Prague School publications have returned to this issue and propose, along with binary oppositions, some multimember (especially ternary) correlations, recalling other models of trimodal contrasts as reflected, for instance, in the work of Pike. 57 There is, of course, no question about the convertibility of non-dual plurality into a succession of binary expressions, yet the fact remains that functional load descriptions in non-binary terms, i.e. in terms both of fewer and more than two opponents, is compatible with early and subsequent views on the subject, and that fact is of considerable interest for a number of cases to be discussed below. To some, the recognition of unitary load may be no more than a confusion of frequency and functional load. Later in this study we shall return to the distinction. For the moment it may suffice to let the reader judge for himself whether on that ground each and every statement of unitary implication is of necessity subject to challenge. Some statements along these lines have been presented above, and we recall, from the definitions and descriptions of Section 1.2, above, that Trubetzkoy related to "le degré d'utilisation distinctive ('le rendement fonctionnel') des diverses oppositions phonologiques ET le rendement moyen des phonèmes" [emphasis added], and that Stockwell and Bowen, in one of the most recent developments along similar operational avenues, look upon functional load as "the extent to which A GIVEN SOUND ... is used to distinguish one word from another, the quantity of distinctive information IT carries" [emphasis added], a view which, as we saw, the same two authors had expressed earlier in associating functional load with "the more typically Spanish construction".

55

Trubetzkoy (1939), 69. 56 Martinet (1955), 73-74. Cf. Mikus (1947); Halle (1957); Contreras (1969). Cf. also Hjelmslev (1939). 5' Horâlek (1964) ; Pike (1967).

34

EXISTING WORK ON FUNCTIONAL LOAD

For additional bibliographic precedent, the following passages are likewise instructive. Die strukturelle Ausniitzung eines bestimmten Phonems in einer bestimmten Sprache ... ist etwas anderes als DIE FREQUENZ seines Vorkommens. ... 58 It will be noted that [the "Projet"] definition speaks of the functional load not of phonemes but of phonological oppositions. In the Grundzuge, however, Trubetzkoy already speaks about the functional load of the phonemes... A second consideration of the functional load type is that some perfectly genuine ITEMS are so infrequent or so arcane, or both, that the beginning foreign learner will either not be exposed to them or will not hear them even if he is. [Emphasis added.]60 The syllable-initial j has a high functional load in the utterance Joel (vs. e.g. Gol), but a lower functional load in My name is Joe.61 Given [the] initially smaller frequency [of marked phonemes], their functional yield is necessarily small.62 The quotations preceding suggest that the non-unitary principle is anything but inherent in the notion of functional load. They allow for the applicability of the load concept to unitary comparisons as much as to sets of binary oppositions. This message is clearly conveyed, not only by some of the 'founding fathers' — the most authoritative type of witness to many — but by authors of subsequent work in the field of load analysis and, among the latter, by three who reported on quantificational research strictly along the lines of the opposite (non-unitary) philosophy. Against the background of criticism, cited above, which was leveled against unsupported and unwarranted assumptions, unquestioning acceptance of the non-unitary basis of operation with functional load is thus obviously another instance of assuming what is not borne out either by conflicting pronouncements on 58

Mathesius (1931), 151-152. Vachek (1966 a), 65. Hockett (1958), 64. 81 Rischel (1964), 86. 62 Greenberg (1966 a), 96. Cf. Carroll (1958); Herdan (1958); Rigault (1962); Roberts (1965); Card el al. (1966); Dolezel (1967); Bakker (1968). 59

EXISTING WORK ON FUNCTIONAL LOAD

35

record or, as this discussion will attempt to show, by considerations of pertinent data to be accounted for. To notice this is at once to remark upon the unilateral nature of quantificational proposals devoted to the measure of one type of load at the exclusion of any other relevant type. There is, to be sure, an apparent counter-argument: unitary statements likewise appear to be mere claims as presented. What, after all, is Mathesius' criterion for differentiating "die strukturelle Ausniitzung" from "die Frequenz" of the phonemic unit? How DOES Trubetzkoy distinguish the functional load of phonemes from that of oppositions ? How DO others visualize functional load of an initial j, of marked phonemes, of 'items' generally, in terms other than frequentiality ? Are we back to general pronouncements on a par with those which had been subjected to criticism? The record supplies no specific answers. In the present discussion we note that fact, and we observe that there are apparently two schools of thought, one of them non-unitary and at the basis of subsequent examination and development, the other unitary and thus far without any substantial influence on existing work on functional load yet, as this study will try to outline, of considerable significance for the development of the functional load concept itself. 1.53 The Principle of Non-Zero

Counterparts

1.531 Non-zero prevalence. — In order to measure functional load, Wang proposes five principles of conceptual delimitation. One of these is the non-unitary condition, noted above, that the load of any subset with one phoneme only is zero. To this he adds: Since [functional load] is used to characterize the extent of contrast, the [functional load] of a single phoneme would have to be zero since it has nothing with which to produce contrast.63 As Wang quite correctly points out, "[this principle is] implicit in most of the linguistic discussions on this subject", but the fact 63

Wang (1967), 43.

36

EXISTING WORK ON FUNCTIONAL LOAD

that it has considerable currency in existing work on functional load does not make it any less of a procedural assumption. The exactly opposite position may be, and in innumerable expositions has been, taken. 1.532 Zero-precedent. — Zero as a correctable entity is at the basis of operations with marked versus unmarked features, and these alone suffice to refute any axiomatic elimination of zero as an opponent. Functional load analysis of necessity cannot proceed outside of the mainstream of descriptive development, and if the body of load information currently at our disposal does not utilize certain established criteria such as zero correlation, this, it would seem, can only be accounted for as an accident of procedural tradition within the microcosm of functional load description, but hardly represents linguistic description as a whole. Thus, one scarcely depends on adducing the additional argument that there is nothing in the statements by the 'founding fathers' and their successors that would tend to rule out the recognition of functional load of a plus-zero contrast. Problems arising from the imposition of the non-zero rule have not, in fact, gone unnoticed by previous scholarship. These are problems inevitable whenever a correlate of binarity is lacking. With features such as voice upon which to establish presence versus absence, things are fairly simple and the identification of functional load may proceed along its traditional road. But what about entities without ascertainable correlates ? [Some] of the phonemes can occur only in limited numbers of positions, so that, for instance, some phonemes cannot be directly contrasted with one another (see, e.g., the ModE phonemes [h], found only in a prevocalic position at the beginning of morphemes, and [g], found only in postvocalic positions at the end of morphemes — or at least not at the beginning of morphemes). In Prague terminology one speaks of the relatively small functional load (or yield) of such phonemes.64 It is hard to see how, on the basis of functional load as "the degree 64

Vachek (1966 a), 64. Cf. also Vachek (1966 b). On zero, cf. Jakobson (1939 b); Frei (1950); Olmsted (1951); Haas (1957); Austin (1961).

EXISTING WORK ON FUNCTIONAL LOAD

37

of utilization ... for the differentiation of meanings", the load of /h/ versus /rj/ could be anything more than zero. Nor can we easily imagine /h/ and /q/ being correlated upon any common denominator of distinctive feature. And are there any respective opponents that would qualify as plus-correlates for /h/ and /q/, respectively? Does English present us, for example, with a voiced counterpart to /h/ or a voiceless one to /rj/? The answers being negative to both questions, how else may /h/ and /r)/ have even small functional loads except in contrast with zero ? The problem is closely connected with that of the range of functionality involved in functional load determination, an issue to be taken up specifically in the context of the principle of traditional analysis next to be examined in this discussion. At this point, recalling the semantic orientation of the "Projet" definition cited above, we may view as having contrastive effect severally each of the components of the non-distributional phoneme identification centering upon the substitution (here restated as a plus-plus relation), addition (minus-plus) or subtraction (plus-minus) affecting meaning. In that sense, English hill contrasts equally with bill and with ill, that is to say, /h/ contrasts with /b/ or with zero, and if it is contrasts that carry functional load, it is of some interest to evaluate in comparable terms the non-correlate plus-plus contrast /h/ : /b/ and the plus-zero opposition /h/ : / 0 / and the extent to which they differ or coincide in respective load values. It will be noticed that this type of inquiry is not identical with the determination of respective frequency values. We do not inquire into the frequency of any one unit (/h/, /b/, or /0/) by itself, but instead operate with the same order of functional correlation as that to which existent load evaluations have addressed themselves, though it will remain to be seen to what extent these evaluations cover the entire range of measurable relations as long as plus-plus relations alone have been assessed. If it is accepted that plus-zero contrasts are indeed on a par with plus-plus relations, the unitary point of view merely represents a terminological alternative to a plus-minus correlation. In other words, we may posit the functional load of any unit, on grounds

38

EXISTING WORK ON FUNCTIONAL LOAD

not only of compatibility with some earlier statements cited, but as a logical extension of correlative concepts of previous scholarship on the subject of load and on related topics which do not specifically acknowledge unit load but leave room for a conceptual development in that direction. The subsequent exposition of this discussion will outline the significance of that development in the context of widening previous scopes of analysis and applicability in the study of functional load. 1.54 The Principle of Sub-Morphemic Level 1.541 Submorphemic restriction. — From the earliest descriptions to the most recent quantificational studies, the functional load in question has been that of phonological, i.e. sub-morphemic, pairs. 65 The "Projet" definition is limited to phonological oppositions, in its application to functional load in general, and under the special heading concerned with the functional load of phonemes and its statistical expression based on Trubetzkoy's Grundziige. Mathesius and Trnka also limit their attention to such questions as the phonological characteristics of a language, or the extent of using phonological units, and wherever they consider non-minimal entities such as clusters or syllables, these are no more than extensions of analysis on the same functional level. Martinet's description is a development of a similarly restrictive principle: what may be compared are two words, e.g. blanc and blond, but merely as "un élément du rendement fonctionnel de l'opposition /a/ - /5/", so that the comparison centers, not on the functional load of the words blanc and blond themselves but merely on the respective non-word component which they do not share. 65

This observation is confirmed by DaneS and Vachek (1964), 21, and is voiced here notwithstanding some previous work on the subject with titles suggesting otherwise. Cf., however, previous suggestions on extension of submorphemic load in supraphonological frames of reference, as reported in the present section and in Section 1.55, below. Cf. also De Groot (1939); Bazell (1949 a); Pike and Warkentin (1961). For a submorphemic-graphic extension, cf. Togeby (1952); Dolezel and Prúcha (1966).

EXISTING WORK ON FUNCTIONAL LOAD

39

Quantificational work of a subsequent period consistently adheres to the phonological characteristics of the material examined. Kucera's report pertains to the "structure of the PHONOLOGICAL syllable as the linguistic segment which serves as the basis of the informational analysis", "statistical data on the distribution of segmental PHONEMES and SYLLABLES", and "functional load of distinctive [PHONOLOGICAL] features". 66 Hockett, we saw, studies "the way in which ... PHONOLOGIC contrasts do ... work", and restricts "the load carried by a contrast" to that of "PHONEMIC, ALLOPHONIC, and [phonologically] COMPONENTIAL" relations. In the work of King, "functional load [is] evaluated in its effect on a number of PHONOLOGICAL changes", 67 or is stated as "the functional load of PHONEMIC oppositions", 68 a characterization not unlike that by Wang. 69 (All emphasis in this paragraph is added.) King qualifies the customary use of the term functional load as pertaining to linguistic units which are "usually" phonemes. As an expression of descriptive caution, the word "usually" appears to be quite uncalled for, since none of the publications referred to by King or reviewed by the present writer yield any data other than of an EXCLUSIVELY phonemic nature. 1.542 Supraphonemic extensions proposed. — The question which inevitably comes to mind — although thus far it has apparently not been considered in any existing work on the subject — is whether units other than those which King terms "low-level phonological units" are either devoid of function with load comparable to load previously described and, in part, measured, or else whether the description and measurement of supra-phonemic entities presents problems unsolvable by currently available methods, and must on that account be omitted from consideration. 66

Kucera (1963), 191. Hockett (1955), 215; Hockett (1966), v, 3 (the inclusion of "allophonic" is interesting in the light of statements reported in Section 1.53); King (1967 a), 831. 98 King (1967 b), 1. 69 Wang (1967), 36. 67

40

EXISTING WORK ON FUNCTIONAL LOAD

The former is hardly a serious proposition; it takes less than boldness to proceed from the axiom that the load of "high-level" forms, i.e. "linguistic forms" in the Bloomfieldian sense,70 is identifiable and of significance on a par with, and in many respects exceeding, that of the "non-linguistic" forms examined in traditional load work. It is, then, once again evident that the matter of functional load has by no means been carried as far as Hockett felt it might be without quantification. In fact, it would appear that an obvious area of advance in directions other than ever more sophisticated and comprehensive mathematical expressions might be that of 'linguistic' forms with unquestionably differentiable (if not, in part, as yet computable) load values. To show this is one of the objectives of this study. We do not propose to suggest avenues of approach leading to specific measurements of 'linguistic' functional load comparable to those of the load of 'non-linguistic' (phonetic) forms. The problem of the feasibility of this type of measure is not of crucial import in the context of equivalent identificational features outlined below, or in the context of theoretical recognition of relevant relations. 1.55 The Principle of Unique Function

1.551 Formal-denotative prevalence. — The function involved in functional load work as developed to date is of a purely formal order. It is function as Bloomfield used the term: the privilege of a form to assume a given position. 71 It is of course understandable, indeed inevitable, that this should be so in the context of limitations noted above. Held to the submorphemic level, traditional analysis could not aim at anything but submorphemic, that is exclusively formal, function, although, as we shall try to show, non-formal extensions might be conceivable up to a point even within the confines of analytic tradition. 70 71

Bloomfield (1933), 138. Bloomfield (1939), 26.

EXISTING WORK ON FUNCTIONAL LOAD

41

With a view to transcending these confines we must keep apart the philosophical foundations of functional load from those of the underlying concepts of function. 1.552 Precedent beyond formal-denotative

function. — Most load

definitions in existence appear fairly unambiguous about the function in question, semantic differentiation, more or less disguised in such formal criteria as 'avoidance of homonymy'. Basically, however, the wording of the statements is non-committal. In referring to 'degrés d'utilisation', the "Projet" does not specify the way, numerous different ways in fact, in which this utilization may be imagined. The same is true for Trubetzkoy's 'utilisation distinctive'. Distinctive of what, of how many manifestations ? From the earliest days of preoccupation with functional load, the range of utilization has indeed been regarded as anything but an open and shut case. Mathesius presents this as a matter of system or structure, with similar formal applications such as voice as a distinctive quality, yet with functional implications beyond those of mere contrastivity: Lehnwörter und einheimische Wörter EMOTIONELLEN Charakters widerstreben den genannten Regeln, ja die emotionellen Wörter gewinnen dadurch, besonders bei der palatalen Reihe, geradezu eine eigentümliche 72 LAUTMALENDE oder VERGRÖBLICHENDE Färbung .... [Emphasis added.] Thus, the effects of phonological behavior may be 'expressive' beyond the functional range shared by phonemes, which is that of utilization to 'affect', rather than express, meaning. Assuming, for the moment, that phonemes invariably possess contrastive (and, as a result, semantically distinctive) potential, we observe that some, in addition, are potential carriers of emotional, onomatopoetic or vulgarizing messages. The number of phonological, meaning-oriented potentials of this type corresponds to the number of the different 'utilisations' or functions dischargeable. Needless to say, the four potentials just noted are far from exhausting the range of potential functions even on the phonological level. 72

Mathesius (1931), 151.

42

EXISTING WORK ON FUNCTIONAL LOAD

Depending on the linguistic system involved, there are genuinely f o r m a l (i.e. not semantically oriented) functions, c o m p a r a b l e to the potentials referred to, the same as those excluded f r o m traditional load analysis, yet the same as those deserving of evaluative recognition, such as indicators of lexical o r grammatical sets of special, e.g. diachronic, status: Im Cechischen ist in dieser Hinsicht merkwürdig, dass in ... Lehnworten (z.B. kólon, Gen. kóla ; lózé) die Länge von o, welche ihre phonologische Funktion schon im Altböhmischen verloren hat, zum Notmittel wird, die Lehnworte von den gleichlautenden Worten kolo Gen. kola, loze zu scheiden. 73 A s used by Martinet, the notion of functional load is strictly lexical (i.e., by implication, inspired by denotative semantics): two words, such as blanc a n d blond, b u t f o r the pair of distinctive sounds, /a./ a n d /ö/, would be h o m o n y m o u s . Yet, semantic denotation m a y itself assume various dimensions : II est intéressant de noter que, jusqu'à ce jour, les Français n'ont pas trouvé de solution 'officielle' à l'irritant problème résultant de l'homonymie de l'ami, l'amie, mon ami, mon amie. La composition ne permet pas de se tirer d'affaire comme le fait l'anglais avec boy friend et girl friend. Il semble que la solution la plus fréquente soit celle qui consiste à prononcer le e à'amie, d'où les formes 'hors-système' [ami's] ou [a'mi'œ]. 74 " H o r s - s y s t è m e " is one way of putting it; another way might be t o recognize the special m o r p h e m i c (indeed, grammatically subsetdistinctive) incentive or purpose, i.e. function, of the a d d e d units, every bit as 'useful' as M a r t i n e t ' s other 'oppositions utiles'. Still, manifestations of various ' f o r m a l ' a n d ' n o n - f o r m a l ' functionality beyond the single type based o n lexically contrastive occurrence frequency have n o t been translated into coequal criteria of load assessment, a n d we m a y a s k : what has prevented the incorp o r a t i o n of these additional criteria? The answer can scarcely be f o u n d in the use of the concept of 73

Trnka (1931), 154. Martinet (1955), 58. For further references to semantic or emotional load, cf. Shapiro (1967); Vendler (1968). 74

EXISTING WORK ON FUNCTIONAL LOAD

43

function in the literature of 'functionalist' orientation, nor can we look for it in the philosophical orientation of Bloomfieldian and post-Bloomfieldian authorship. 1.553 Function, functionalism, and functional load. — 'Functionalism', in fact, specifically disclaims a purely 'formal' basis and entails 'non-formal' purposes: [The] terms 'function' and 'functionalist', as used in the Prague School, are not employed in the mathematical sense, implying some dependence of the changes of x upon j>. In the Prague sense the terms simply point out the fact that any item of language (sentence, word, morpheme, phoneme, etc.) exists solely because it serves some purpose, because it has some function (mostly that of communication) to fulfil... , 75 There is another point on which the Prague theory has gone a different way from what had been, for a couple of decades, a basic maxim of the American descriptive group of Bloomfield and his followers. This point has been its steady concern for meaning, by which the Prague scholars indicate what is often termed content or, more exactly, the reference made by an utterance (and by the parts composing it) to what we call extralingual reality. ... This reference, indeed, is what the Prague linguists understand by the function of language.76 The referential function is merely one manifestation of interest to functionalist studies, side by side, for instance, with Biihler's expressive and conative functions.77 [The] three functions discussed above should not be regarded as mutually exclusive within one context or within some of its segments. As a matter of fact, in most instances two or even all of the three functions are found to coexist in the same utterance, but one of them is found to be predominant. As the main purpose of speech utterances is undoubtedly the conveying of objective information about extralingual reality, it is hardly a matter of surprise that in a majority of cases the predominant function will be that of reference. ... But the other two functions will very frequently coexist with it and in some instances assert themselves rather vigorously, sometimes even to the degree of obtaining predominance. As an example of predominance of the first, expressive function ... may 75 76 77

Vachek (1966 a), 6-7. Vachek (1966 a), 30. Bühler (1934).

44

EXISTING WORK ON FUNCTIONAL LOAD

be mentioned personal lyrical poetry, in which the author's main preoccupation is to express his own personal feelings and emotions. ... A similar situation may be found in commands in which there is a strong tendency of the conative function to predominate over the remaining two, although in this case the purely communicative, objective function must always uphold a relatively strong place in the hierarchy of functions. ... [It] has often been urged, and justly so, that the list of the functions established by Biihler can hardly be considered exhaustive. There certainly exist other functions of speech utterances which were left unnoticed by Biihler.78 Thus, the functionalist school identifies quite a number of functions deserving of descriptive recognition, and a hierarchy of these functions, evidently from two points of view: a hierarchy evolving from what we might term a functional (and actualized) predominance — one function is more in evidence than another — and a hierarchy evolving from formal domain — one function attaches to a wider range of forms discharging it than another. But as on various previous occasions in this discussion, we must once again remain aware that the considerable refinement of descriptive technique noticeable in the recognition of functional diversification by the Prague functionalists pertains to FUNCTION but not to the LOAD of any function. Martinet's development of functionalist views shows a variety of typologically different manifestations, some of them level-bound, others cutting across level divisions. Quite clearly, his application of the concept of function exceeds the confines of a 'formal' function of phonological units. So, of course, does the use made of the term in other European scholarship. 79 We must ask: why did the functionalists and scholars of similar orientation, to whom we owe pioneering work in the area of concern to this study, proceed from function as much 'non-formal' as 'formal', yet, when inspecting the amount (load) of that function, fail to transgress beyond self-imposed 'formal', phonological, limits ? Why was function, though identifiable on morpho-syntactic ('linguistic'), as much as on 'lower' levels, in relation to functional 78 79

Vachek (1966 a), 35-36. Martinet (1962). Cf. also Firth (1934); Helbig (1968).

EXISTING WORK ON FUNCTIONAL LOAD

45

load identified in a phonological sense only? What, apart from the limits imposed by current development of theory and 'hardware' of computation, could justify the discrepancy between philosophical potential and descriptive realization, between the 'steady concern for meaning' in its various aspects and, apart from 'lexical differentiation', the totally asemantic turn which functional load work had taken? Vachek's distinction between Prague theory and American descriptivismisindeed well taken, though, in the light of restrictions in load analysis, hardly in the sense suggested. Bloomfieldian and post-Bloomfieldian analysis, while professing to be founded on strictly 'formal' criteria, for the most part proceeds to 'higher levels' on the basis of definitions either openly semantic, 'situational' (i.e. founded on 'non-formal' notions ostensibly disavowed), or solely distributional yet of questioned feasibility.80 The work described by Vachek, in contrast, gives 'nonformal' foundations their due yet has done little if anything to extend them to load work. Here, then, we have another area to which non-quantificational discussion of functional load may be extended: function in ALL of the aspects of that term, rather than in the selective sense of existing studies on load. The indication of how this might be accomplished will be one of the objectives of this discussion. 1.56 The Principle of Frequency versus Load 1.561 Prevalence of load determined by frequency. — The prevailing view of frequency in relation to one unit, and of functional load as the index of more than one (traditionally, two) frequency values in correlation, is quite consistent with the principles of previous analysis reviewed. We have seen that the distinction is not, however, uniformly adhered to. At this point we may outline some of the ramifications of frequency evaluations with respect to non-frequential criteria. Bibliographic precedent speaks of frequency of units, contrasts, 80

Haugen (1951); M. Fowler (1952); R.S. Meyerstein (1964 a).

46

EXISTING WORK ON FUNCTIONAL LOAD

or entire classes.81 Unit (or non-binary) frequency, in turn, may be of a phonological or supraphonological (lexical or phrasal) nature. The underlying considerations may be associated with form (phonological structure, shape, length, or complexity), systemic restrictions (e.g. style), and semantic function (number of meanings or semantic extent of categories) or a number of similar variables more or less remote from the exclusively formaldistributional basis of existing load work (etymological class, age of unit, necessity, availability, range, coverage, learnability, popularity, and general usefulness).82 The connection between frequency and functional load is found to be stated in various ways. In its most general formulation, frequency HAS a 'functional value', i.e. besides the two functions of form and meaning there is a statistical function. 83 The predominant view on the relation between the two concepts, however, holds that frequency RESULTS in functional load. 84 The measure of frequency in conjunction with that of functional load has been discussed above. Additional references of recent date include the work of Frumkina (and a number of bibliographic items quoted there)85 and of Herdan; 86 guidance to earlier procedure is found in the citation of Gougenheim et alP

81

Units and contrasts: Greenberg (1959), 8; Rischel (1961), 19; Mackey (1966), 190; King (1967 a), 831; Wang (1967), 45. Classes: Mackey (1966), 169, 177-178. 82 Phonological, e.g. Dewey (1923); Hockett (1955); Saporta (1955); W. Fowler (1957); Carroll (1958); Manczak (1959); Guiraud (1960); Frumkina (1963); Kuöera (1963); Greenberg (1966 a); Hockett (1966); Mackey (1966); King (1967 a); King (1967 b); Wang (1967). Lexical: Gougenheim (1959); Guiraud (1960); Rischel (1961); Frumkina (1963); Gougenheim et. al. (1964); Beier et al. (1967). Phrasal and grammatical: Josselson (1953); Martinet (1955). Other variables: Martinet (1955), 54-56; Chao (1964); Heilman(1964); Mackey (1966), 173, 177-178,181-184, 189, 194. 83 Guiraud (1960), 16. 84 Hockett (1955); Greenberg (1959); Hockett (1966); King (1967 a); King (1967 b); Wang (1967). 85 Fiumkina (1963), 115-118, 176-179. 86 Herdan (1960); Herdan (1966). 87 Gougenheim et al. (1964). Cf. also fn. 62, above.

EXISTING WORK ON FUNCTIONAL LOAD

47

1.562 Parallelism proposed: frequential and non-frequential load. — In the context of the present discussion, the problems inherent in the frequency-load correlation are, once again, those of conflict between all-encompassing theoretic scope and narrow descriptive purview. What is described is numerical prevalence of occurrence; what, in addition, NEEDS to be the object of evaluative attention, is the large number of areas of a formal non-phonological nature as well as those of a non-formal type. The reasons for disregarding these other areas are familiar: how, indeed, do we compute frequency of meanings, of needs, of availability, range, learnability, popularity, or usefulness ? Should we do without coverage of those areas because we lack the computational know-how? Or should load description include them even though quantification of frequency may not as yet suggest itself? Our decision will be based on the latter alternative. Subsequent exposition will make it inadvisable to uphold the terminological distinction between frequency and load. Frequency, viewed as one of the functions of a form, will be seen to carry a load on a par with the loads of other functions. We shall speak of the lexical or textual frequency of forms, and of a frequency of functions (e.g. of certain meanings). This will represent a further widening of previous descriptive coverage, as illustrated below. 1.563 Frequential and non-frequential determinants: ranking. — Besides a distinction in terms, traditional views derive from a cause-and-effect relation between the two terms. As a result, functional load scores have been ranked according to certain indices of frequential attainment. As deduced from text frequencies, computations of recent date have injected the factor of degree of contrast ascertainable from the existence of minimal pairs among the most frequent words, with a high functional load depending upon pairs of words both frequent, and functional load in general " a gauge of the frequency with which two phonemes contrast in all possible environments". 88 88

King (1967 a), 831.

48

EXISTING WORK ON FUNCTIONAL LOAD

Above, we reviewed (and questioned) the import of frequential load as a diachronic force motrice. Synchronically, we may well be equally skeptical about frequency as a yardstick of the 'importance' of forms. Frequent words may be 'basic' words — but then again they may n o t : In determining the 'basicness' of words, a corrective measure of dispersion is used in addition to frequency. Thus, of two words having the same frequency, the more basic is the one that is more evenly distributed. ... By combining frequency and dispersion, [some] authors arrive at what they call 'word usage', i.e. a measure of 'basicness'. ... This illustrates how dispersion 'corrects' frequency and confirms the intuitive conclusion that alcoholic cannot be a more 'important' word than pretty in English, notwithstanding its twice as many occurrences. ... 89 If this is what we choose to label as basic, well and good, as long as it is understood that no statement of functional load per se is inferrable. Just about the only significance we may detect in this frequential measurement for our purposes is a degree of concurrence with the unmeasurable variable of intuition, Still, happiness even to that extent is not a foregone conclusion: The procedure seems reasonable enough, and yet there are some surprising results: words suchas pirenaico 'Pyrenean', mañanero 'matutinal' and sor 'Sister' are included in the basic vocabulary, while such words as manzana 'apple' and vaca 'cow' are excluded.90 What is basic is thus not always what we thought, and comparable data may lead to opposite conclusions. Among Spanish morphophonemes, for instance, "types of alternation are basic f r o m the standpoint of the frequency with which they are encountered in texts", 9 1 yet "is one justified ... in presuming a more frequent pattern like Se le perdió el libro to be a transformation of a 'basic', but less frequent, pattern like El libro se perdió ...?"92 89

Contreras (1966), 819. Contreras (1966), 819. 91 Frumkina (1963), 91. 92 Posner (1968), 142. Cf. also statements on typicality, e.g. Stockwell, Bowen, and Martin (1965), cited in the passage of fn. 22, above; and on hierarchical ordeiing, e.g. as suggested in Section 1.553, above. 90

EXISTING WORK ON FUNCTIONAL LOAD

49

Still, consensus is in the direction of ranking a form according to its frequency, and concluding to high functional load of oppositions containing opponents so ranked. This may be hard to reconcile with other considerations of equal significance to this discussion, for example those of information theory. The relevance of the factor of entropy as a gauge of functional load has found widespread acceptance; Rischel regarded entropy as "the keyword of the approach" to load determination, and instances of discounting its import are not numerous. 93 It goes without saying that the uncertainty factor, the information per se, is, if anything, reduced as the result of great occurrence potential, while items of relatively scant occurrence have been noted to carry a high uncertainty factor. With informativeness viewed as one of the functions of a form — and in view of the communicational objective we may in this instance indeed speak of a basic function — it is clear that high frequency 'basicness' is not bound to constitute an index of high functional load. In fact, as we shall attempt to show, high frequential scores may entail low non-frequential load in some respects, while on other occasions they may happen to be accompanied by a comparable degree of functional overload in non-frequential areas of investigation. 1.57 The Principle of Complementarity Entailing Zero Load 1.571 Non-complementary prevalence. — The doubtful value of frequency as a unique factor in determining functional load is further noticed in the instance of two forms substantially frequent yet in complementary distribution, hence, according to prevailing analysis, in a mutual relationship with the functional load of zero.94 93

Hockett (1955); Greenberg (1959); Rischel (1961), 17; Ku&ra (1963); Hockett (1966); Wang (1967). Discounting the import of entropy: King (1967 a), 849; King (1967 b), 7. Cf. also Paducheva (1963); Bar-Hillel (1964). 94 KuCera (1963), 194, 210; Hockett (1966), 17; Mackey (1966), 75; King (1967 a), 836, 849; King (1967 b), 5, 8; Wang (1967), 40, 44.

50

EXISTING WORK ON FUNCTIONAL LOAD

We may approach this problem in the context of non-emic, i.e. etic, features in general. Etic features, by definition, fail to function contrastively, thus under present ground rules would be doomed to zero functionality in all instances of occurrence, a contention in conflict with bibliographic precedent in related fields, as well as with demonstrable facts of functional load itself, inferred from previous work or ascertained in the context of analytic extension proposed in this study. 1.572 Non-emic precedent:

etic-emic coequality. — A descriptive

notion of long standing is that of demarcative function. Functionalist authorship has stressed the part played, in this respect, by such Grenzsignale as non-phonemic non-random Czech stress, or random or non-random French stress.95 For French, Martinet elaborates : La mise en valeur du début du mot avec l'intention démarcative est ... un fait observable dans le français contemporain. Il s'agit de ce qu'on a appelé l'accent d'insistance, phénomène bien distinct, par sa localisation, sa nature physique, et surtout par son caractère intentionnel nonautomatique, de ce qui reste dans cette langue de l'accent traditionnel sur la finale du mot. Cet accent a, on le sait, deux formes distinctes, une forme affective et une forme intellectuelle. Sous sa forme affective, il est caractérisé par une intensité et un allongement insolites de la première consonne du mot; sous sa forme intellectuelle, par un relief particulier donné à la première syllabe. ... Dans ce qu'on pourrait appeler le style didactique, la mise en relief de la première syllabe est assez fréquente et sa fonction est nettement démarcative. ...96 Since it would surely not do to speak of function without implying a degree (load) of that function, there is certainly no doubt about the functional load (taken as a unitary notion) of these various types of stress in French. Some of these are predictable and some are not; that is to say, we must recognize differences in status from the point of view of informative potential. All of the types are represented in variously high incidence of respective token frequencies, yet none of them is phonemically contrastive, in the 95 96

Vachek (1966 a), 64. Cf. also Anderson (1964). Martinet (1955), 167-168.

EXISTING WORK ON FUNCTIONAL LOAD

51

competential sense of that contrastivity, either relative to ano ther kind of stress or to any non-accentual feature. That non-phonemic functionality is by no means restricted to certain phonological manifestations or systems is shown by Martinet on the example of German demarcative glottal stop,97 less felicitously perhaps since besides a modicum of semantic signaling which glottal constriction provides, for instance, in verachten, it might also intrude non-functionally in Theater. The point is of interest to us as an illustration where 'basicness', a frequential conclusion, does indeed acquire a measure of functional status: with the considerable preponderance of demarcative over nondemarcative glottalization in the phonotactics of German, (nonphonemic) glottal constriction remains a marker, i.e. discharges a function and thus a load value, on practical statistic grounds. Statistical 'practicality' ties in with injection of such equally 'practical' variables as availability, learnability, and similar functional criteria encountered in the literature reviewed. Learnability offers a case in point. Principles of applied linguistics with regard to this descriptive objective (function)98 may be seen to concur with the recognition of functionality above a certain practical cutoff point. Focus on etic variance, possibly minimal or abandoned in the case of 'slight' distinctions, will tend to be emphasized in the event of relatively outstanding spread of respective articulatory areas, as in German [x] and [5], or structural parallelism of noticeably phonemic, reduced phonemic (i.e. noticeably neutralized), and virtually non-phonemic oppositions as, respectively, in the French series [e] - [e], [o] - [o], [0] - [ce]. Monolingual considerations of this sort have their counterpart in contrastive (bilingual) linguistics: For instance, it has been argued that in teaching Spanish to speakers of English it is well to indicate — at least for a while — the distinction between the stop and the spirant allophones of the phonemes /b d g/. For a stop-spirant difference to be nondistinctive and positionally 97

Martinet (1955), 167. Description conditioned by the objective-function: R.S. Meyerstein, (1964 b). 98

52

EXISTING WORK ON FUNCTIONAL LOAD

determined is un-English; the subphonemic alternation is one which the English-speaking learner must acquire and render habitual and unconscious. The argument says that the acquisition is difficult or impossible without overt mention and drill, and that transcription of the difference helps." The phonemic-phonetic progression does not, at that, present as much of a learning problem as the reverse, where the coequal status of etic and emic features is mandatory as, of course, it is on the morphosyntactic level of language acquisition. At this point, the distinction between the emic and the etic is a strictly taxonomic one, and of importance to respective status in the etic-emic classification as a whole, but not necessarily for the functional import (load) of certain specific relations. Within one and the same system such as that of English, we have noticed that the opposition /p :b/ has a load at variance with that of /s:z/. The units /h/ and /g/, which happen to lack correlates on the order of those for /p/ or /s/, nevertheless function (i.e. carry load) in opposition to zero. That is precisely what we may say about stress in French — in fact, as shown by Vachek and Martinet, we may say quite a bit more than that about other functions (load criteria) of any predictable or non-predictable manifestations of this 'non-contrastive' (non-emic) feature, based on its morphemic (Grenzsignal) and systemic (stylistic) relevance, in addition to its various connotatively semantic values, functions not unlike that of set-identification noted above for Czech ('contrastive') length. Then there is the factor of frequency — not the sole factor for arriving at load determinations, but one factor nevertheless. Immense for French 'word'-final stress, it is fairly low for English /h/ or /r)/ yet high for either of those units in comparison with the frequency of /z/. Again, emicness is not the decisive issue. Frequency as the traditional gauge of functional load, and functionality as the Bloomfieldian 'privilege of occurrence', indeed appears to render non-emic acknowledgments inescapable: it may be posited as a virtual axiom that any two etic elements of the same complementary set, the same as two of different sets, will exhibit 99

Hockett (1958), 65. Cf. also Shen (1959).

EXISTING WORK ON FUNCTIONAL LOAD

53

unequal occurrence frequencies, and that inequality alone would permit discriminations of respective functional values (loads). 1.573 Differentiated load of variants: frequency of functions. — Let us, moreover, recall that frequency is in the context of this discussion seen not merely as frequency of occurrence established paradigmatically as in the foregoing examples; it is, beyond that, frequency of functions (to a degree, Bloomfieldian cooccurrence, i.e. the result of additional syntagmatic criteria). For descriptive reorientation along those lines we may again refer to Vachek: [The term] 'allophone' seems to be equivalent rather to the Prague term 'combinatory phonemic variant' than to the term 'phonemic variant' at large. To put the matter more concretely, the ... two [l]-sounds of Modern English could both claim to be denoted by 'phonemic variant', while 'allophone', as often used in Anglo-American writings, could only be used for the 'dark' [lu]. In Prague terminology, the clear P] sound would be denoted as the 'principal phonemic variant' (in French, variante fondamentale), i.e. as that variant which depends least on the neighboring phonemes, and which, besides, is free from emotional coloring. The dark [lu], on the other hand, would come under the heading of 'combinatory phonemic variant' (in French, variante combinatoire), whose occurrence is conditioned by that of the neighboring phonemes — in the case of flu], by its word-final, post-vocalic position. Clearly, the Anglo-American term allophone would have to be adapted, if it were to serve equally well with the Prague rival terms — one would have to distinguish between the principal (main) allophone and the (combinatory) allophone, which would naturally result in its extension — at least occasional — into a two-word term. 100 To this it might be objected that all 'non-free' variants necessarily depend on one or the other cooccurrent feature and that, of the two variants of Vachek's illustration, neither one particularly carries 'emotional' connotations; but these objections would address themselves to imperfections in wording or coverage and thus be essentially trivial. The fact remains that one variant exceeds the other in range of occurrence — thus giving the impression of 100 Vachek (1966 a), 51-52. On similar terminological distinctions bearing upon the present study, cf. Prieto (1954) and Martinet (1960) distinguishing opposition and contrast. Cf. also Bloch (1953); Dykstra (1956).

54

EXISTING WORK ON FUNCTIONAL LOAD

relative environmental independence — and in so doing qualifies as the 'principal' variant, the one 'which most commonly implements the phoneme'. 'Implementation' as a further function clearly enhances the frequentially determined value (load) of one variant relative to the 'non-implementing' frequency of the other. With a view to the recognition of a range of functions, rather than the monofunctional basis of traditional load work, we thus observe that some of the additional functions are indeed pertinent for etic as well as emic conditions. To the extent of discovering the variables controlling respective functional manifestations, the widening of descriptive scope will lead to useful frameworks of integration and correlation. One of these variables was seen to be one of system: the identification of load, in need of elaboration within one language, will remain incomplete without giving intersystemic considerations their due. This was illustrated in Hockett's remarks on pedagogy in dealing with the presentation of allophonic features of Spanish to an audience of speakers of English where these features are phonemic. The impact of 'outside' determinants of this type will concern us in connection with the principle of traditional load analysis next to be examined. 1.58

The Principle of Systemic Uniqueness

1.581 Monosystemic dichotomy: competence versus performance. — Existing work on functional load agrees on relating formal functionality to one and the same undifferentiated system. The principle accords with traditional views, in Bloomfieldian and similarly oriented writings, of A language as A system of vocal symbols.101 Among positions to the contrary, we find those of de Saussure 102 and Chomsky: We... make a fundamental distinction between COMPETENCE (the speakerhearer's knowledge of the language) and PERFORMANCE (the actual use 101 102

Bloch and Trager (1942), 5. Saussure (1915). Cf, also Levin (1965).

EXISTING WORK ON FUNCTIONAL LOAD

55

of language in concrete situations). ... In actual fact, [performance] could not directly reflect competence. ... Observed use of language ... cannot constitute the actual subject matter of linguistics, if this is to be a serious discipline. The distinction I am noting here is related to the langue-parole distinction of Saussure; but it is necessary to reject his concept of langue as merely a systematic inventory of items and to return rather to the Humboldtian conception of underlying competence as a system of generative processes.103 The preceding passage indicates that although its author chooses to operate himself with no more than one single systemic corpus, it is not the one encountered, for instance, in computational load studies which precisely derive from 'observed use of language' ; he does, moreover, provide the philosophical foundation for a systemic dichotomy of significance to this study. Without dwelling upon generative power with which we are not concerned here, we may notice that Saussure, in effect, presents the problem as a difference between non-specific (system) and specific (phonological actualization of systemic potential), whereas Chomsky's diversification allows for the absence of specification in either sense. 1.582 Lexical competence versus textual performance. — With the latter correlation in mind, we may interpret previous analyses of load, especially computational ones, as reflections of phonological performance. In relating these to competence (likewise, in previous load studies, of a phonological order), we arrive at the correlation of textual and lexical foundations of description, i.e. text versus lexical frequency. Quantitative indices based on language fall generally into two types, text and lexicon counts. In the present instance a lexicon measure might be constructed as follows. If, for example, we wished to measure the functional yield of b vs. p in English, we would go through the entire dictionary, or sample it in some systematic way, counting the number of instances in which the replacement of a b by a p or a p by a b produced another word in the dictionary. The absolute number of such replacements or, more usefully for purposes of comparison, the number of replacements divided by the total number of phonemes in the sample, would give a quantitative estimate of the functional yield of the bip 103

Chomsky (1965), 4. For the distinction as adapted in this study, cf. R.S. Meyerstein (1968), 23, 30-34.

56

EXISTING WORK ON FUNCTIONAL LOAD

contrast. Two relevant factors, however, are not taken account of, factors which would figure in any method based on running text. In the lexicon method no account is taken of the relative frequency of occurrence of the particular forms examined, yet, other things being equal, a contrast which distinguishes two frequent forms is more significant than one which distinguishes two infrequent forms or a frequent from an infrequent one. The second defect is that contrasts between forms such as nouns and verbs which would rarely if ever actually occur in the same context are given the same weight as those between words of the same form class, some of which might frequently be in significant possible contrast. These considerations suggest that a method based on possibility of replacement in context will, on the whole, be more likely to measure what is significant for diachronic or other applications.104 There can be no question about the validity of these arguments whenever we deal with incidence in observed performance. But will this render interest in competence, in non-frequential potential in particular, obsolete? If not, may we assume that (textual) performance in each and every case adequately represents (lexical) competence? Questions have indeed been raised about the latter being obscured in text selections from one or the other style.105 Other critical comment has centered on the instability of frequency based on texts or samples of speech, as compared with the stability of 'situations' (or opinions about them), as well as on the influence upon numerical data and conclusions by such coincidental factors as diverse methods and procedures: word counts, for instance, might differ according to whether forms like German Bruder and Briider or English go and went were, respectively, counted as two different words because of their difference in shape, or as occurrences of one and the same word because of the meaning they shared.106 Other discrepancies might arise from diverging analytic judgments, or else might attest to the unreliability even of an extensive sample corpus. In the relative frequencies cited above for Modern 104

Greenberg (1959), 7-8. Martinet (1955), 56; Mackey (1966), 163. Cf. also Mathesius (1931) in the passage of fn. 18, above; Malmberg (1942); Pike and Warkentin (1961); Dolezel and Prucha (1966), 40; and fn. 62, above. 106 Mackey (1966), 180. 105

EXISTING WORK ON FUNCTIONAL LOAD

57

German, King records phonemic frequencies revealed in his text for '/?/' and '/x/', i.e. two phonemic units. Noting that neither frequency is near zero, the reader is surprised to learn that the functional load of the 'phonemic' contrast is nonetheless listed as zero. We are reminded of the relationship between English /h/ and /q/, except that, contrary to the latter pair, the two German units match up as articulatory correlates, and thus we wonder whether they are indeed phonemes rather than allophonic variants, i.e. whether, discarding junctural totems and intermorphemic taboos, we accept Tauchen and tauchen as a minimal pair, 107 or whether the text might not be adequate enough to reveal contrasts admittedly not of great lexical frequency but certainly above zero. Yet, a far more compelling question arises, once again, with respect to functions currently non-computable. How, for instance, do we quantify semantic features from a text performance with the expectation of arriving at correlations with lexical competence ? All things considered we may certainly allow for the integration of functional load analysis with the description of language as a problem of competence. We may further bear in mind that quite frequently emphasis on performance variation will inject diversity of analysis likely to becloud quite needlessly the invariance of the systemic potential. This is true especially about facts of interdialectal extent. Thus, with a subclass of French multiple-object verbs of the type dire which are capable of entering reflexive (or reciprocal) constructions such as ils se disent, the functional domain of the pronoun les in a referential sense, as a substitute for direct object phrases, is matched by that of the indirect object substitute leur (i.e. both represent masculine or feminine reference); lui, in contrast, comprises the domains of le and la by the same token; and se encompasses those of all of the preceding. Vagaries of text frequency may favor one or the other substitute and to that degree inject a factor of 'performance uncertainty', but that factor can hardly be claimed to be of overriding descriptive import against the back107

King(1967 b), 11. Cf. Bloomfield (1930); Moulton(1947); Leopold (1948); and fn. 64, above.

58

EXISTING WORK ON FUNCTIONAL LOAD

ground of relative structural potentials, as illustrated — a fact of competence. The tacit acceptance of this proposition in established analytic practice, including much of what pertains to functional load, is certainly beyond question. One example may suffice. Apart from studies centering on abnormal speech, the functional significance of pause manifestations is not determined according to the lexical frequency of pauses as expressions of individual hesitation, interruption, etc., nor does it, in fact, derive from the numerous instances of absence of pauses at structurally expectable points; instead it is, tacitly if not in so many words, a reflection of generalizable competence. 1.583 Parallelism of monosystemic and bisystemic dichotomy: load points. — At other times, to be sure, there is a very real competence-performance conflict. A t least, in allowing for nothing other than that dichotomy, this is how we would have to account for the data. In comparing, for instance, the two English expressions I haven't seen anybody (competence as performed) and I haven't seen nobody (a deviant, i.e. not 'well-formed', performance), we apply two functional statements to either expression: a count of constituent forms as exponents of comparable semantic functions ('negation'), and the resultant difference in respective marking. The former expression 'loads' the functional signaling upon one (grammatical) negative, -n't, and its status is describable as 'non-descript' (i.e. stylistically unmarked) in relation to the latter expression, where the 'load' is spread over two negatives (one grammatical and one lexical), with the result of marking 'sub-standard' style. One of the problems of opposing competence to performance, of course, is the comparison in binary terms of what in reality is a number of performances in excess of two, in relation to a more or less controversial unity of competence. Where does competential variation end and performance diversity begin? In the course of this discussion it will become evident that operations within traditional monolinguality have the effect of

EXISTING WORK ON FUNCTIONAL LOAD

59

excluding a large number of typologically diverse identifications apparent in a bisystemic frame of reference only, yet of acknowledged import in numerous areas of applications, none the least that of second language learning. If, for instance, we assign 'load points' to each identifiable 'feature of functionality', the opposition -n't: -n't ... nobody as English expressions of negation will rate two points, one point for the formal difference between numbers of 'load carriers' (one versus two), and one point for the systemic difference between stylistically unmarked or marked expressions. French, in contrast, will not score on either basis, for lack of similar formal diversity and attendant lack of systemic marking. It is the very comparability, in terms of similar 'surface structures' in a bisystemic sense, between the second English sequence and French je n'ai vupersonne that calls for bilingual description of this and related phenomena, for without the bilingual point of view the 'unopposed' French structure would scarcely attract notice. By standards other than inconclusive data of mutual intelligibility, one of the oppositions (between 'standard' and 'deviant' English) is essentially of the same type as the other (between English and French). There is thus not much to be gained from drawing a sharp distinction between 'monolingual' and 'bilingual' bisystemicness, and we do not propose to do so. We shall refrain from diversifying a monosystemic analysis wherever the introduction of additional performance dimensions would be insignificant or unfeasible. We shall deal with frequential features as manifestations of functions rather than (competence or performance) systems, and observe that on occasion certain functions will associate with certain systems in a frequential as in a non-frequential context. We shall take advantage of the competence-performance dichotomy, not in basing a discussion on this distinction, but in interpreting it as a challenge to the monolithic universe of discourse within which previous load evaluations have been offered, and in treating as bisystemic, besides data pertaining to more than one 'language' as usually classified, such not very 'monolingual' performance variety as social, geographic, or stylistic dialect

60

EXISTING WORK ON FUNCTIONAL LOAD

distinctions or diverse historical stages. Accordingly, the two English expressions of negation illustrate a bisystemic correlation contrasting English of Type A with English of Type B. The negative construction of French, by itself, would pose a monosystemic problem; as a component of a conventional inspection of bilingualism, it would enter into a bisystemic correlation of English and French. Further elaboration on this principle of operation will be presented in connection with the section on System in the second part of this study, Alternatives of Functional Load Assessment.

2. ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

2.1

CRITERIA OF COMPARISON

This study does not address itself to specific frequency counts or specifications of load statistics. We shall, instead, adopt an approach which may be characterized as DIFFERENTIAL and TYPOLOGICAL. 1 Difference, as an alternative to specificity, is a well-established criterion in such functional areas as that of semantic content. Semantically oriented morphemics, in this sense, does not depend on precise indication of all the meanings of English cat but merely on the fact that the semantic domain of cat is different from that of bat or cab. There may be objectives of study where it does matter to enumerate 'felis domestica', 'collective term for lions, tigers, etc.' and 'spiteful and gossipy person' as some of the semantic possibilities of cat — or as we might say, some components of the semantic load; for the limited objective of establishing the morphemic identity of this three-letter sequence within a semantic frame of reference, however, enumerations of this sort add nothing to the benefits derived from a differential statement alone. The present discussion similarly forgoes numerical precision, the objective of quantificational work which we do not attempt to duplicate, and expresses functional load differentials between any 1

On typology, cf. IsaCenko (1939); Whorf (1941); Bazell (1949 b); Garvin (1949); Voegelin (1955); Greenberg (1957); Malkiel (1957); PavloviC (1957); Jakobson (1958); Krdmsky (1959); Greenberg (1960); Uspenskij 0965); Home (1966V

62

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

two terms of correlation as the overload of one of the terms relative to the other. We shall attempt to extend the scope of previous load studies, not by quantitative diversification within one area of functional significance, but by diversifying the number of relevant areas. This extension will lead to three types of functional load expressions based on an equivalent number of formal relationships as exponents of load. In addition to FORM and FUNCTION we shall consider variables of LEVEL and SYSTEM. No attempt will be made to accomodate terminology to the descriptive philosophy of any particular 'school' of analysis past or present. The use of the concept of FORM will variously encompass, ignore or transcend Bloomfieldian operations noted above. FUNCTIONAL LOAD will be identified in a variety of manifestations including but exceeding previous criteria. Recent scholarship disagrees on the need or necessary extent of LEVEL and SYSTEM as used below. Correlations of these four areas of analysis, as well as illustrations ensuing, will be offered as samples and are not intended to imply limitation of applicability to the particular facts presented. As the next step we shall briefly introduce each of the four areas as significant components of functional load criteria in the extended sense of this study, with examples attempting, at this point, merely to clarify the approach adopted.

2.2

FORM

Under the heading of FORM we shall consider any type of function exponent, and its relation to a comparable exponent of dissimilar extent of function. This comparison will assume unitary or binary character, and will be concerned with paradigmatic (one-for-one) or syntagmatic (one-plus-one) relations. The first two of the three correlational types to be identified in this study will involve paradigmatic comparisons. Of these, the

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

63

first will compare traditional oppositions (pairs of phonemic units). Each opposition will be viewed as (single-feature) correlate-unit relatable to other pair-units of their phonological set, and the correlation of any two such pair-units, e.g. /s-z/ with /s-s/, will be designated as a :b. This will include binary comparisons with load as usually acknowledged, others with functional significance outside of the customary frequential domain, as well as some to which functional status has previously been denied. The notion of functional load will next be applied to a paradigmatic type of correlation based, binarily, on non-correlative (multiply distinctive) plus-opponents or plus-zero, or on the isolation of a unit matched non-binarily in the customary sense, i.e. with the entire inventory of a suprabinary set. As an illustration of the unit in this type of matching, here symbolized as a:m, we may take the suffixial morpheme -er. As a noun-derivator of the system of English, we may compare it with 'similar' derivational suffixes (-th, -ant, etc.) and note the dissimilarity of functional loads in respective text frequencies, lexical frequency potentials (productivity as lexemic formative) and other areas of function within the same nominal domain. Morphocentrically, i.e. viewed as one exponent with diverse syntactic potentials, -er as both a derivative noun-formant and an inflectional adverb or adjective ending carries a significantly higher load than non-homonemic formants within the respective sets, such as purely nominal -ant. We may similarly set offword-homonemes, exhibiting grammatical cleavage, against syntactically unambiguous, e.g. solely nominal or verbal, forms. 2 Type 3 represents binary comparisons combining paradigmatic and syntagmatic relations, in which 'long' expressions confront 'short' constituents or non-constituent replacives as carriers of comparable functions or, as we might say instead, in which similar loads are placed upon components versus components-plus-one. Monosystemically or in bisystemic frames of reference, morphological or syntactic strings on the order of Czech/Russian ja jedu, 2

Homonemes: Lord (1966), 35. Cleavage: Bloomfield (1933), 204.

64

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

examined in conjunction with their respective abbreviations or reductions (jedu), will lead to the formulation of functional load differentials. This typological relationship will be shown as ab:a.

2.3

FUNCTIONAL LOAD

2.301 Extended scope of load areas. — Earlier in this discussion, functional load was noted to raise problems in two respects: areas of functionality, and extent of function in the respective areas. The only area of functionality consistently inspected in traditional work on the subject, we found, was that of text frequencies of opponents leading to functional load assessments of oppositions. We did, however, encounter a number of additional areas to which previous scholarship had alluded without, for that matter, integrating them in correlative treatments of a topic which, after all, in and by itself implies functionality without discrimination in favor of any one particular function, or at least need not imply otherwise. Among references reviewed above, we cited descriptions based on privilege (but not actual text frequency) of occurrence; 'utilisation', implicitly in the sense of lexical frequency; morphophonemic or lexical marking, e.g. of loan words; semantic features of denotation per se and of extent of coverage, and of such connotational implications as those of 'emotion' or 'vulgarity'; educational values of 'learnability' or 'availability'; 'situational' factors of various kinds; and a number of additional, more or less exolinguistic considerations. The following exposition will recall these functional areas. In addition, it will address itself to such related or unconnected variables as synchronic and diachronic stability; unequal denotative or connotative contributions of phonemic constituents in lexical or grammatical respects; the speaker-hearer's response to his own or an outsider's communication, such as the paramountcy of one or the other distinctive feature; areas of disagreement between

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

65

speakers and hearers; monosystemic transformability; synchronic and diachronic bisystematicity; potentials on the order of replacement values or productivity; the hierarchy of the 'relative weight' of forms retained or disappearing; functionality 'according to Hoyle' as opposed to that of speakers 'not playing the game'; and other functions with different degrees of load. These degrees, as has been observed, are statable in various ways. The only way of stating them in previous work has been by computation of specific text frequencies of opponents. Even in allowing for the representativeness of the corpus, and granting the adequacy of the sampling, we have recognized two principal drawbacks of this procedure: the discrepancies of computations evolved, and the limitations of the computational criterion, part of one of the more than thirty areas of functionality even of the very superficial summary of the preceding two paragraphs. 2.302 Typological insignificance of text-token quantification. — In addition, we cannot overlook the very limited typological significance of the extensive load value specifications on record. King's tables of frequency loads may serve to illustrate this point. His table of the functional load of German phoneme oppositions3 consists of sixty entries, with a maximum load of 49.419 for /m:n/, and a minimum of 0 which he indicates for /?:x/. Sixty numerical indications on a similar order might support definable ranking criteria if we knew what the entry of 49.419 signifies in relation to the highest score of some other set — what is the typological conclusion for German as opposed to the top value for some other language? — and in relation to the other scores of the same system. There does not appear to be any one ascertainable factor of progression as we compare the load values of any two oppositions of any of King's tabulations. It is not possible, for instance, to say that the functional load of one opposition is precisely twice that of another within or outside of German, or that the line of ascend3

King (1967 b), 11. Cf. the passage of fn. 47, p. 25 above and Bolinger (1961).

66

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

ing load values is straight or in the shape of any definable configuration. Rearranging, by way of sampling, the first seven of King's German oppositions in descending order of load magnitude, and deducing value differentials from King's numerical indications, we observe: Oppositions /m:n/ /d:z/ /d:g/ /t:s/ /b:d/ /l:r/ /t:d/

King's Load Scores 49.419 31.148 27.154 16.984 16.652 13.263 10.984

Differentials 18.271 3.994 10.170 .332 3.389 2.279

Thus, the load of one opposition differs from that of another to a more or less significant degree. As we reach the tail end of King's score values, we notice: Oppositions /ë:ë/ 4 /y:0/ /p:pf/ /y:0/ /0:B/ /S:x/ /ç:x/

King's Load Scores 0.052 0.037 0.036 0.029 0.019 0.017 0.000

Differentials 0.015 0.001 0.007 0.010 0.002 0.017

As may be inferred especially from the latter set of load differentials, there appears to be, apart from interest in raw data, little descriptive justification (given the inevitable vagaries of the corpus and of sampling routine) and clearly no ascertainable practical significance in providing a precise ranking of, say, /y :&/, winning by the landslide of 0.001 over /p:pf/, especially when we bear in mind that, for the sake of improved readability, King multiplied the true load scores by 100,000; a differential of 4

King's symbol of length (:) is here replaced by a macron, to avoid confusion with the colon used in this study to separate the two terms of an opposition.

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

67

0.00000001 cannot seriously be invoked in defense of numerical precision of functional load from a typological point of view. In addition, we ask: what does 0.00000001 mean in the total task of language description, in the typological comparison of different languages, and in the application of data of this sort in a variety of fields, none the least those of language acquisition and pedagogy? The difficulty of answering questions along these lines is compounded as soon as we transcend the safe terrain of traditional load limitations and recognize the multiplicity of problems presented by the additional consideration of areas of functional relevance, or even just an extension of the traditional area. 2.303 Lexical token quantification. — For example, we may wish to establish numerical expressions of various syntactic functions of a German paradigm of ¿-words (der, die, etc.). How many of them possess homonemic potentials as representatives of one or the other set (relative pronouns or articles), or as exponents of one of the other gender or case ? The significance of this kind of investigation is readily assessed: high scores will reveal outstanding ambiguity, hence will serve as stimuli to search for cooccurrent (grammatically redundant yet semantically decisive) signaling. In one sense this will be a very simple computation if we restrict ourselves to the lexical type of frequency, a frequency of potentials, and if in matching the number of paradigmatic constituents (9) with the amount of genders (4, abbreviated f, m, n, and p, with plural taken as one gender subset on a par with the three singular subsets) times the amount of cases (4: A, D, G, and N), we compute the homonemic load as the percentage of the product of functional potentials ( 4 x 4 = 16). The resultant picture is as follows: Forms

Article Functions ( %)

Relative Pronoun Functions (%)

das dem den

nN, nA mD, nD mA, pD

nN, nA mD, nD mA

2 (12.5) 2 (12.5) 2 (12.5)

2 (12.5) 2 (12.5) 1 (6.25)

68

denen der deren des dessen die

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

mN,fG,fD,pG 4 (25) mG, nG

pD mN, f D fG, pG

1 (6.25) 2 (12.5) 2 (12.5)

mG, nG fN, fA, pN, pA

2 (12.5) 4(25)

2 (12.5)

fN,fA,pN,pA 4(25)

We see that die scores highest for pronouns and, tied with der, for articles; it also takes the prize in categorial and non-categorial functions combined, or, within one and the same homonemic ¿/-word set, ties with der ~ deren taken as syntactic variants. The homonemic potential of denen, in contrast, is zero (i.e., its loadpoint score is 1). But the limited value of this kind of computation is easily seen against the background of additional (text) frequency data, as well as of such exolinguistic criteria as stylistic preference, quantifiable in diverse incidence of relatively higher or lower occurrence of one or the other form, but not in invariant percentages as in the lexical comparisons presented. All else being equal, text frequency will be affected by the expectable preponderance of antecedent or modified nouns of one rather than another gender, and of certain cases rather than others. (The last-mentioned disparity is an even stronger counter-argument in a similar study of lexico-syntactic potential of Czech demonstratives;5 on the basis of 36 as the product of genders, numbers, and relevant cases, occurrence-potential percentages for the 16 forms of the set exhibit values of 8, 7, 6, 4, and 1, in round figures, except for the oblique t£ch which towers over all of the others with 17, an exalted status hardly likely to be confirmed by performance statistics.) 'Style' distinctions will further distort the picture; in the colloquial style of speech, occurrence of the German G case will be drastically reduced in favor, for instance, of von + D, and more or less 'sloppy' expression favoring das + infinitive over possible der/die nominals will likewise contribute to a modification of data. The 8

Carried out by the present author, based on data in Havränek and Jedliika (1963), 200. Cf. also Seiler (1967).

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

69

nature of the subject matter, in turn, will reinforce or weaken grammatically or systematically restricted potentials. We are thus likely to end up with statistics within limits individual to each corpus examined. This diversity of load data is inevitable; the question it presents is whether the specification of the diversity is of typological relevance in a descriptive framework, and whether anything is gained from it for practical objectives to which the description might be applied. How much are we advanced with a list of functional loads for 60 oppositions more or less arbitrarily selected (including affricate sequences, but excluding many non-sequential matchings), or with one of 9 forms with lexical frequency percentages of 6.25, 12.5, and 25? 2.304 Type quantification proposed: overload of load points. — The alternative to computational specificity has been noted earlier in this discussion. It may be termed a differential statement and is concerned solely with the fact that different form-units (or oppositions viewed as units) differ in respective functional significance. Differences recognized below will be based on the concept of 'load points', and will be expressed as statements that a given form scores one point in addition to the score of its 'peers', i.e. of replaceable or cooccurring forms of the same order. As expressions of addition versus subtraction of the load point, and without implications necessarily intended in the sense of mathematical doubling, or correlations of respective syntagmatic precedence or subsequence, or any chronological (diachronic) sequence, load point differentials will be symbolized as 2:1. In a sense, overload may be regarded as a 'distinctive feature of functional significance'. There are numerous precedents for the application of the principle, if not the term, of overload, including, for instance, suprasegmental diacritics. Conceived as relative muscular exertion, stress is effected by extra effort on the part of the speaker, a heavier work load distinguishing primary stress from non-primary, or secondary from tertiary, etc. In the physical sciences the unit of work or energy in the CGS system is the erg, and we could formulate the difference in two

70

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

manifestations of this type of energy expenditure as consisting of x ergs restatable as one work-point. We could then say that in a series of physical endeavors, one involves one work-point more than the point score of the other endeavors, or simply that it involves excess work to the extent of a differential (or ratio) which may be assigned the value of 1 point beyond the base rate of the other work scores. In acoustic perception this could be expressed as one additional decibel-point of sound-intensity. Whether the excess decibel count would reflect a corresponding excess of listening effort is, of course, anything but a foregone conclusion; very low intensity will, in fact, require a very high amount of perceptive exertion.6 Perceptional effort, contrary to amounts of other physical work, thus lacks established yardsticks of measurement; yet, there is certainly no question that in principle that effort, too, might be recordable and yield data at least to the extent of relative orders of magnitude. Even in the area where specific (decibel) degrees might be ascertained, however, suprasegmental description has been quite content with what amounts to a load-point scale similar to that proposed in the present study. The fact is implicit in terminology adopted: successions such as primary-secondary-tertiary reflect an arithmetic progression based upon a deduction of one point per step. Taken as binary opponents, primary and secondary score as 2:1, and so, respectively, do secondary and tertiary. Once an exact gradation of corresponding work loads is devised, it might well reveal that the point differentials involved are not comparable either from step to step within one system or between steps involving different systems; that is to say, 2:1 in one case will not correspond to a mathematically precise or even similar degree to 2:1 in the other case, a discovery without effect upon the differential character of the numerical expression. By the same token we may restate any pair of phonological oppositions previously tabulated in a descending order as 2:1 statements of functional load reductions from pair to pair. The disadvantage of this restatement is the absence of numerical 9

R.S. Meyerstein (1967), 75-86.

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

71

precision in each case, and the difference in precise values among various load points; it was seen that this precision is not essential either for typological or other descriptive purposes in this kind of study. The advantage of the reformulation is its adaptability to areas of load identification previously left out of account where at the present stage of computational development it would seem to represent a viable and practical alternative. 2.305 Formal and non-formal determinants of overload scores. — In supra-phonemic relations, 2:1 expressions may closely approximate or, indeed, exactly reflect the numerical ratio of one-plusone versus one. In comparing, for instance, Latin avunculus and patruus with French oncle we observe, first, the obvious fact that in point of vocabulary extent, Latin beats French 2:1. For the expression of semantic content, however, it is 2:1 for French over Latin since, assuming equality in respective denotational scopes, i.e. the scope consisting of the combined semantic spheres of the two Latin forms and that of the French form, oncle has to perform the work load shared by avunculus and patruus both with respect to frequencies and semantic loads. The paradigmatic comparison of the preceding illustration is matched by the correlation of syntagmatic and paradigmatic features inherent in German Handschuh versus English glove; the English form does the work spread over the two German constituents, hence in this respect achieves a work-load score of 2:1 over the latter. We may not assume that either avunculus or patruus singly equalled the frequency of oncle, though this assumption is legitimate in equating Handschuh and glove. In other words, load analysis in one area of functionality, such as frequency, may or may not coincide with that of another area, e.g. denotative force. The observation is virtually axiomatic and hardly in need of belaboring. It acquires interest in the context of this discussion to the extent of questioning the unilateral basis of previous load formulations, and in investigating the nature and degree of evaluational discrepancy traceable to dissimilar areas of function, and to the injection of additional variables.

72

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

Assuming frequencies of both Hand and Schuh to top that of Handschuh in German, we may suspect that with the introduction of the functional variable of leamability, the German child, learning the form and referential value of Handschuh, has a distinct advantage over the English-speaking child, whose effort in acquiring glove is not facilitated by similar componential hints; the same, of course, is true in respective second-language learning efforts on the part of adult students. Glove and Handschuh, though equal in respective frequencies, thus score 2:1 in acquisitional effort required, a paradigmatic relationship mirroring that of respective denotative force in a syntagmatic-paradigmatic sense, noted above. Frequential and denotational scores may coincide in such intercalations as you know and man!, viewed as having the additional load score point of 'expressive exclamation', though from the point of view of concrete information conveyed, frequency is the reverse of expressivity in this case.7 That being so, we shall, in fact, be prompted to inquire into the very presence of functionality as such. Is man! a functional constituent? 'Informationally', the answer undoubtedly is no; accordingly, we shall introduce the additional load-score comparison of presence or absence of functionality, (actually a case of '1: 0', yet typologically no more than a special instance of 2:1 relationship), and we shall compare zero values in functional significance with plus and zero features of form. These and other inspections of functional imbalance will be discussed in further detail in Section 3 of this study.

2.4 LEVELS

Forms will be further associated with specific LEVELS. Lower-case abbreviations will distinguish between submorphemic (sm) and morphosyntactic (ms) manifestations. Submorphemic identifications, usually matters of phonology, though in some instances pertaining to graphic constituents, will include, yet significantly 7

R. S. Meyerstein (1968), 24.

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

73

exceed, the scope of traditional load analysis. Morphosyntactic extensions, unknown in previous work on load, will encompass grammatical or lexical features.

2.5

SYSTEMS

The employment of the notion of SYSTEM has been outlined in some detail in Section 1.58, above. Under this heading we shall recognize any set of data commonly treated as linguistic or describable in similar terms of reference. Operations with one or the other type of set will proceed on a unitary (monosystemic) or binary (bisystemic) basis. Traditional statements on functional load, as has been noted, are of a monolingual nature, i.e. systemically unitary descriptions of one language. Transcending the limitations of established monosystemic analysis, we may either widen the scope of purview within one system, or we may carry the investigation to contrastive (bisystemic) load analysis, either extension representing new avenues of research suggested in this study. A broadened view of SYSTEM may, in actual fact, narrow the concept to that of a linguistic subset indicative of the behavior (performance or response) of subgroups of a linguistic community, such as children. For example, we shall recognize diverse degrees of functionality in such areas as acquisitional potentials manifested for Fernsprecher versus Telephon in a (sub-)system of 'German children' in opposition to the 'German' system at large, and one or the other of these opponent systems might be further matched with other German subsystems, or with the set of potentials of 'French children'.8 Existing work on bilingualism has explored the interrelation of linguistic sets and subsets in the areas of 'languages' as opposed to those of 'immigrant languages' and similar systemic designations. Work along these lines will be drawn upon from the point Martinet (1955), 89. On potentials in general, cf. Mathesius (1911); Pulgram (1967). 8

74

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

of view of functional load of the respective expressions, not normally stated in bilingual research as an objective in itself.9 In instances of systems not generally recognized as 'languages' in their own right, and in those generally regarded as dialects or stages of the 'same language', we shall operate with the indication of Language A : Language B, specified, whenever needed, e.g. as English A : English B, though left unspecified in observations independent of specific systematization, being in the nature of language universals or, what essentially may come down to the same thing, general theoretical statements. It has been noted that the systemic labeling adopted here does not imply any specific direction in historical sequence or in syntagmatic or transformational precedence or succession. Thus, 'English A: English B' said to be founded upon differences in the incidence of strong verbs, though in fact associated with different time periods, is here descriptively not kept apart from comparable synchronic relations, considerations of cause and effect being nonessential to this discussion. Systems will be referred to by two-letter sequences with first or both letters in upper case. MS and BS will, respectively, stand for monosystemic and bisystemic analysis, and will, in general, be followed by the indication of the specific systems in question in parenthetical upper-and-lower case abbreviations (e.g. En, for English; Lg, for any unspecified system; or Sn, a comparable systemic generalization of 'Situation'). These two-letter abbreviations may, in turn, be followed by capital A or B denoting diverse representations of the 'same' language (e.g. EnA : EnB, two systems of 'English'), as indicated by the ensuing comments. Identical abbreviations in different parts of this discussion need not refer to identical correlations. For example, EnA : EnB, 9

Mackey (1966), 190. The concepts of overload and underload (or zero load), as used in this study, have descriptive, though not necessarily conceptual, counterparts in the terms of over-differentiation and under-differentiation; cf. Weinreich (1953), 18; Haugen (1956); G. P. Meyerstein (1966). For earlier references to overload with denotations at variance with the present use of the term, cf. Deutsch (1966), 389-393,

ALTERNATIVES OF FUNCTIONAL LOAD ASSESSMENT

75

though pertaining to two systems of English whenever this label appears, need not refer to the same two systems in each instance, nor is En alone, repeatedly mentioned, necessarily one and the same monolithic, undiversified system. Systemic abbreviations used in this study, and their respective denotata, are as follows: Cz En Fr Ge It La Lg Po Rc Ru Sc SI Sn Sp

Czech English French German Italian Latin Language Portuguese Romance Russian Slavic Slovak Situation Spanish

3. INDIVIDUAL CORRELATIONS

3.1 TYPE 1 — TRADITIONAL BINARY OPPOSITIONS (a\b — 2:1)

3.101 Traditional scope integrated and extended. — In Chapter 3, a number of specific form-function correlations will be examined according to levels and in relation to their monosystemic or bisystemic scope. We should re-emphasize that this presentation is in the nature of selective sampling and lays no claim to exhaustive coverage in any area. In its exclusively phonological (basically, single-feature-distinctive) reference (sm), and restricted to a monosystemic corpus (MS), matching of Type 1, as was noted, is the only one consistently reflected in previous load analysis. As pointed out above, we shall symbolize this type of form-opposition as a :b, and the respective load-point ranks as 2:1. The bibliography and examples of traditional oppositions have been reviewed in Chapter 1, while Chapter 2 has dealt with suggestions of modification. At this point, we shall consider pros and cons of elaborating on matching of the traditional type, beginning with a few areas for which extensions will not be proposed here. 3.102 Overload implications excluded. — Any explanatory, cause-and-effect relations will be disregarded. An example of this would be the factor of ease of production or perception cited as the reason for the predominance of /t/ or /d/ in consonant usage. Besides adding a dimension unnecessary to this discussion, articulatory ease, while to a degree demonstrable in physical terms and

INDIVIDUAL CORRELATIONS

77

though a recognizable criterion of function, is not a reliable explanation in the light of diverse data of different systems which more often than not may exhibit parallelism of incidence in phonological complexity and rate of use. By the same tokens of irrelevance to this discussion and uncertainty of effects, we eliminate projections of communicational consequences in related areas. In his basic theoretical work on functional load, Martinet, comparing two French oppositions of similar articulatory instability caused by the feature of Up-rounding, /ë:œ/ and /â:ô/, noted that the former, with low functional load, may be developing toward eventual zero load, i.e. confusion, while the latter, with a high load score, does not exhibit any comparable trend toward zero.1 As was noted in reviewing work by Martinet and King, generalized inferences of 'low-to-zero', i.e. oppositions of low functional load being doomed to eventual extinction, are unsupported either by Martinet's own observations on the preservation of low-load /S :z/ in English or by King's diachronic data. In the present study, we record the functional status as it exists; we do not attempt to predict developments from that status. We are not, in fact, concerned with the essentially subjective concept of specific 'low' values in relation to a range of similarly subjective evaluations of 'higher' scores, beyond statable instances of 'onein-addition'. On that basis, Martinet's case of French /e:s/ as in faite : fête, which he labels hors système and, subject to increasing 'confusion', constitutes a further case without additional typological contribution. In the context of this discussion, length is simply another opponent within the system of French where it is heard to occur; in another system of French where it is not, we need not, from a monosystemic point of view, preoccupy ourselves with it. The point of this observation is that diachronic projections, apart from the uncertain foundations of historical and contemporary data, are descriptively incompatible with the monosystemic i

Martinet (1955), 57.

78

INDIVIDUAL CORRELATIONS

frame within which they are set up, a remark which, needless to say, does not preclude inclusion of diachronic data as merely additional bisystemic correlations. Ku&ra's focus on the Czech opposition /f:v/, characterized as carrying low functional load due to the relatively late status of /f/ as a separate Slavic phoneme, 2 differs from the French correlations by Martinet in the absence of conclusions either synchronically (the opposition being seen as part of the system the same as other oppositions) or in the sense of diachronic projection of increasing confusion or loss. Once again, however, no typological difference is added to this discussion, where we note functional load but make no attempt to explain how it came about. By the same token we shall exclude, on the morphosyntactic level, further reference to the case of loze : loze, mentioned by Trnka; from the present-day point of view, certain phonological or morphosyntactic oppositions may achieve low frequency scores because they are connected with certain historical events, but these scores will resemble those of other oppositions not so connectable. 3.103 Overload considered; variable, pedagogic. — sm, MS/BS (Fr, Ge). Wherever, in contrast, we do detect an additional load point earned by the discharge of a distinctive additional function, the case will be relevant to this study, and at that point explanations forgone above for lack of typological relevance, such as ease of production or perception per se, do acquire significance for their unequal load score ranks. An instance of this kind of discrimination would be the recognition of the pedagogic import of frequential predominance. Instructional and acquisitional effort obviously centers upon oppositions with high load scores based on traditional considerations of opponent frequencies, for the rewards of 'correctness' achieved over vast ranges of grammatical and lexical applications. Instances of 'incorrectness' may similarly be diagnosed and, 2 Kuiera (1963), 209.

INDIVIDUAL CORRELATIONS

79

hopefully, corrected as expressions of monosystemic discriminations in two systems and bisystemic conflict of the respective data. For example, given the two submorphemic (sm) monosystemic (MS) oppositions /b:p/ and /b:m/ in German, and notwithstanding a higher frequency load for the first (not listed by King) than for the second (.583, according to King), the second might yet score higher than the first on a bisystemic (BS) basis of conflicting phonetic impressions, and this non-frequential criterion of bisystemically determined load potential may aid in predicting correct or deviant performance among relatively untrained second-language learners. This is seen when, in describing the bundle of distinctive phonetic features of the English or French representative of /b/, we include mention of occlusion and voice. The question arising in the context of the present study is which of these two features is of greater functional import than the other. As related by Mackey, A purely phonological analysis of German ... would reveal the fact that German words do not end in voiced stops. ... Reasoning from this, one would predict that when a German learner imitated a word like herb or, better still, its French equivalent herbe, he would come out with the pronunciation [herp]. But he might just as well imitate it as [herm] as [Mackey's] experiments have in fact indicated. This is because his /m/ phoneme in certain positions has almost as much voice as the foreign /b/ which strikes his ears.3

There is an extensive bibliography on similar interlingual reassignments, involving identification of like phoneme pairs in two systems, yet articulation of the wrong equivalents, and we may attribute this deviant performance to distributionally enhanced load imbalance. 3.104

Overload

in traditional

'zero-load'

oppositions;

variable,

formal (feature-spread). sm, MS (Ge). With the injection of additional load variables, or of additional implications of variables traditionally recognized, we may review 'zero load' statements, made by previous scholarship, prompted by the exclusion of criteria which are suggested in this study. In a monosystemic (MS) frame, 3

Mackey (1966), 110.

80

INDIVIDUAL CORRELATIONS

a distinction suggests itself between pairs of submorphemic (sm) units which have been recognized as form entities though not necessarily as load-bearing functives, and pairs or units the existence of which has itself been in question. German [x : 5] is an instance of the first-mentioned set. King was cited as listing this as an opposition with zero load, and accordingly, we wondered whether the pair represented an allophonic (hence non-contrastive) relationship or whether it constituted a pair of phonemes. Depending upon assessable evidence of determinant factors (in this case, measurable and invariably present juncture clues) and the status to be accorded to data of this sort, we would, in previous terms of load, reach one or the other conclusion. A s a result, the point of articulation would, or would not, be the sole differentiator between tauchen and Tauchen, and the opposition would, or would not, carry functional load in the traditional (phonemic) sense. Our first objection to zero load conclusions even in the latter case is based on distinctions of respective lexical and textual frequencies of the opponents. In addition, a presentation of the problem solely on the basis of the analyst's phonetic data and his phonemic conclusions amounts to over-simplification in other descriptively relevant terms. The analyst may attach significance to factors irrelevant to the informant, and the latter may emphasize facts which are secondary to the analyst. 4 Concretely speaking, one may legitimately propose that to the speaker and listener, the difference between velarity and palatalization is far more noticeable than whatever junctural features may show up on the spectrogram, and it is unquestionably that difference which the naive informant will spontaneously cite when asked about the distinguishing feature between the two German words. Various sources of indirect evidence suggest that this has met with 4 R. S. Meyerstein, (1964 a); R. S. Meyerstein (1964 b). The distinction is also implied by Brtere (1968), 20, who notes that frequently it is the articulatory feature ignored for classification which may be paramount to the encoder or decoder. Cf. also fn. 100, p. 53 above.

INDIVIDUAL CORRELATIONS

81

tacit analytic acknowledgment. Pedagogic treatments of otherwise consistently phonemic orientation will include [x] as well as [ç] in German, or [0] and [œ] in French. What this comes down to is the description of the language accompanied or not by the description of the speaker's reaction to the language, or the informant's phonemics versus the analyst's phonemics, with corresponding decisions on respective functionality decisions (load points). 3.105 Overload in non-traditional oppositions', variable, semantic (connotative). — sm, MS (Fr). The existence, whether allophonic or phonemic, of German [x] and [ç] is generally acknowledged. This cannot be claimed for [e] versus [œ] in French, though Martinet's data clearly indicate the existence of this opposition, as well as the functional status of the latter opponent, e.g. in elle est très belle, expressed in such connotatively semantic terms as 'aping', 'ridiculing' or 'vulgar' impression conveyed.5 Since listener response in the sense of that impression, as an essential attribute of the communicational process, is a codeterminant of functionality, and since that response contrasts with the 'distinction' or even 'affectation' detected by many members of the speech community in corresponding [e] articulation, the problem once again is one of analytic differentiation, this time between denotative and connotative oppositions. There appears in this case to be a parallelism of denotation and competence in the system of French, where [ae] is not part of the phonemic inventory of predictable representations, and a comparable but contrasting parallelism of connotation and performance which includes that unit, a view implying the performance of non-competential variants. Alternatively, we are dealing with a bisystemic problem. Either way, a zero-load relationship is to be stated in one case though not in the other, based upon the consideration of non-frequential load variables. 3.106 Overload in zero versus non-zero oppositions', variable, bisystemic (interdialectal).—sm, BS (ItA :ItB). In describing the relation between [s] and [z] in Standard Italian, Hall observed : 5

Martinet (1955), 40.

82

INDIVIDUAL CORRELATIONS

This [currently allophonic] situation is growing increasingly precarious. Contrasts between [s] and [z] in intervocalic position are on the increase: more Central Italian speakers than formerly are beginning to use [z], under North Italian influence, and they are also beginning to sort out their use of [s] and [z] to reach at least a statistical consensus of usage. Should this regularization and consistency of usage become generalized, we shall have of course a classic instance of the 'phonologisation' of a contrast. We shall have been able to observe a borrowing of the sound [z] in a new position (intervocalically) into Central Italian; its persistence for a long time as a non-distinctive variant; and its eventual settling down into use under consistently contrastive conditions, so as to establish the two phonemes /s/ and /z/ in intervocalic position but not elsewhere.6 We may visualize this as a case of load (of [s:z]) conditioned bisystemically (BS) by Italian A versus Italian B or, as a distinction within the 'one' speech type of Central Italian, by Speakers A versus Speakers B. The 'phonemic faction' (System A) justifies the recognition of functional load which existing work on the subject would deny for the 'allophonic faction' (Group B). 3.107 Supraphonemic overload; variable, significative contrastivity potential (word differentiation). — sm-ms, MS (Fr). With a view to further extension of functional load distinctions beyond their present limits, we may recognize as cases of typological import all those relations which involve new dimensions in levels or systems. In regard to levels, this injects morphosyntactic (ms) criteria. Of necessity, our presentation will to some extent adopt positions at variance with earlier conclusions, due to the difference in descriptive objectives. Thus we find that Martinet, in discussing the French /o:a/ opposition, draws attention to a need for computational refinement which would distinguish relative frequencies in the case of blond: blanc versus emonder:¿mender.1 This need is uncontestable, but it does not necessarily accord with objectives of this study. For example, we may wish to emphasize the unequal functional relevance of the phonetic constituents of either set. Any one of the 6

Hall (1960), 195. Cf. Sections 3.113 and 3.223, ("morphologization"), below. 7 Martinet (1955), 56.

INDIVIDUAL CORRELATIONS

83

phonemes is endowed with more or less equal contrastive (or constitutive) potential, but not with equal grammatical or lexical (significative) possibilities.8 Thus, all phonemes have varying degrees of combinability as one function, while in the case considered /5:a/ has the additional function of lexical differentiator; that is to say, all opponent units, as representatives of potential oppositions, score one point on the list of functional loads, yet one among these scores two points, the additional point being earned by morphemic differentiation, a potential not shared, for instance, by /d/, in the absence of *émoncer or *émencer. The relative infrequency of émonder or émender, noted by Martinet as falsifying the functional load tabulation in which these two forms are given equal weight to that of blond and blanc, as well as the respective environmental potentials, cannot detract from the facts of overload noted. 3.108 Typology of supraphonemic overload: unequal phonemic potential; lexical (non-textual) attribute; level interaction. — Typologically, we observe several things. To begin with, differentiator potential, much like frequency potential, attaches to varying degrees to any opposition. Unlike traditional frequency load, it evolves from lexical rather than textual data. Also unlike frequential relations, it varies with each minimal pair of morphemes contrasted solely by the differentiating pair of phonemes, i.e. with conditions of variance unlike those previously inventorized. Further comparisons that come to mind are on the order of levels and directions of determination. Similarly to phoneme pairs with the additional load point of morphological Grenzsignal potential, 'phonology determines morphology' in certain non-morphological features with more than usual 'low-level' functions, i.e. morphological terms of reference which, in a reverse determination, enhance the functional load score of the features. Unlike Grenzsignale, syntagmatic in character (though, of course, compared paradig8

Constitutive and significative: Pottier (1948) as quoted by Krámsky (1965),

45. Cf. also Krámsky (1967).; and fn. 100, p. 53 above.

84

INDIVIDUAL CORRELATIONS

matically to other boundary signals), differentiator pairing may be paradigmatic only. Another two features of typological interest are those of 'morphology determining phonology' without reciprocity as noted above, and the notion of stability, previously questioned for its predictive (diachronically bisystemic) claims, yet defensible as a manifestation within one and the same system. It is in that sense that we interpret distinctions, proposed by Malmberg, between a stable /e:e/ contrast in French fée : fait and its unstable counterpart in -et as in billet or carnet.9 The distinctive factor is clearly morphemic: given the above-mentioned two types of expression alone, non-morphemic final /e:e/ opposition is mandatory, while originally or currently morphemic -e/ 10 need not be similarly contrasted. The functional potential of /e:s/, increased by differentiator status, is thus counterbalanced by reduced phonological contrastivity in morphemic differentiation. 3.109 Supraphonemic overload: submorphemic versus morphemic function. — sm-ms, MS (La). Although French /ô:â/ in ém-der achieves a double score because of its differentiating function, it does not acquire morphemic status. This, however, is precisely what /a:e/ does in the Latin environment voc-mus. There is, of course, nothing novel about one-phoneme morphemes, previously relegated to the list of insignificant coincidences, later recognized as significative phonemes, and in this discussion of interest as a further sub-type of functional overload expressions. Given a specific environment (in this instance, morphemically determined), one set of opponents discharges phonemic function only, while another set functions both phonemically and morphemically. 3.110 Supraphonemic overload: mono-set versus bi-set function (word set differentiations). — sm-ms, MS (En). The functional potential of /a:e/ in the preceding example is that of differentiating 9

Malmberg (1964), 78. Morphemic status of -et in billet : R. S. Meyerstein (1964 a); R. S. Meyerstein (1964 b). 10

INDIVIDUAL CORRELATIONS

85

categories within the same functional set (verbs). The English correlation /0 :ô/ in /mawB : mawô/ is of yet increased significance in performing similar grammatical differentiation between diverse sets (nouns and verbs). It is interesting to compare our emphasis on the functional importance of /9:ô/ with Martinet's low rating of the functional load of this opposition, based solely upon consideration of relative frequency. 11 The discrepancy in evaluation is not hard to trace: among numerical scores for one and the same type of phonemic frequentiality, /0:ô/ clearly ranks low, whereas on a range of scores reflecting potentials of grammatical differentiation, the pair is on a high order of functional significance. It goes without saying that this evaluation is contingent upon contrastivity in final position. Elsewhere, as in ether : either, the significance of the opposition is no more than that of lexical differentiator within a specific word-pair of a specific set (nouns), yet even as such superior to the significance of the other 'low-level' constituents, and on a par with that of /ô:â/ in French émonder : émender. 3.111 Monosystemic versus bisystemic overload; variable, 'stylistic factions. — sm-ms, BS (FrA : FrB). Another type of extension beyond traditional descriptive dimensions of functional load is represented by bisystemic (BS) analysis. Rather than simply state diverse (contrastive versus non-contrastive) options for fée : fait and billet, we might correlate the respective options with respective systems of 'French' (French A versus French B), identified by Malmberg as different styles. Again, there will be a difference in evaluation scores for /e:e/ as opposed to such pairs as /u:o/, based on the morphemic potential absent in the latter, and now seen as a consequence of systemic differentiation. 3.112 Idiolectal variable of linguistic system (speech habits)', implications, pedagogic. — sm-ms, BS (LgA : LgB). Differentiation »

Martinet (1955), 78.

86

INDIVIDUAL CORRELATIONS

between systems may derive, not from the inspection of specific 'languages', but of factions within 'one language', typologically nothing but the respective 'languages' of Speaker A and Speaker B as tokens of the corresponding factions. According to Hockett, Functional load is reduced when speakers of a language differ in their assignment of one or the other term of the contrast to specific lexical or grammatical situations. The speakers are trained to understand despite differences between what someone else says and what they would themselves say; this affords a usable latitude in deciding what habit to try to convey to the foreign learner.12 Hockett's illustration of this is the case of plus/minus distinctiveness of French /e:s/, previously noted but related by Hockett to morphemic tokens (épée : épais) and non-stylistic systematicity. We have left the terrain of 'languages' in the common-parlance sense of the term, though not that of relevant linguistic criteria, in this instance once again the association of options with corresponding systemic sets and load rates. 3.113 Idiolectal variable of exolinguistic situation (intent). — sm-ms, BS (SnA : SnB). Recognition of the correlational type under discussion may extend to non-linguistic (though by no means descriptively irrelevant) variables in terms of Situation A or B. Situational variables may be traced, for instance, to communicational motivations or objectives, and might be considered to govern one or the other performance within the competential range or indeed, as in the following illustration, display the competence normally obscured by performance. In a certain North German speech type, [e] and [s] appear to be in complementary distribution, the former unit alone being heard with length. Thus in mähen or nähen, the stressed vowel is phonetically identical with nee (local dialect for nein), and the same, of course, is true for any paradigmatic representative (er mäht, sie näht). Now, whenever der Bauer sät, there arises a morphological homonymy, [zët] = sät or seht, syntactically without consequence 12

Hockett (1958), 63-64. On pedagogical application, cf. also Dykstra (1956).

INDIVIDUAL CORRELATIONS

87

due to environmental incompatibility of singular subject plus seht, hence no stimulus for deviating from the [e] articulation noted. But how about säen and sehen, for instance, in the frame das wird man jetzt..., in an agricultural context? At that point the rendition of [zean] for säen, contrasting with [zean] for sehen, is a definite possibility. This, then, is the 'phonemicization' of the vocalic pair based upon a non-linguistic incentive in a unique lexical situation, the creation of an opposition with minimal functional load in the sense of occurrence frequency. The variable may be vaguely termed that of presence or absence (Situation A or B) of 'fear of confusion'. If in the speaker's judgment (performance), the general context, the 'universe of discourse', precludes misunderstanding, or else if the speaker is not particularly concerned about the ambiguity of a relatively unimportant statement, he is likely to maintain the [e] pattern in säen; preoccupation with adequate perception, on the other hand, will tend toward the exercise of the competential option of [e], to preclude optical interpretations in [z-an] or nautical ones in [z-man], relative text frequencies notwithstanding. In accordance with one or the other situation we shall observe an opposition traditionally regarded as being endowed with plus or minus functional load.

3.2 TYPE 2 — OTHER BINARY AND NON-BINARY OPPOSITIONS (a:m —2:1)

3.201 Phonemic overload in multiply distinctive oppositions', variable, pedagogic (exolinguistic). — sm, MS (En). The discussion of the preceding load type identifications has centered on traditional matching of correlative (single-feature) phonemic oppositions, extended to various non-traditional frames of reference. In this section we shall further depart from tradition in inspecting the functional load of non-correlate (multiply distinctive) or nonbinary oppositions. The first illustration of the former type of opposition is based on what might be termed polarity of occurrence of phonemes,

88

INDIVIDUAL CORRELATIONS

in which high-frequential /i/ in English is matched with the least frequent [z].13 Mackey presents [z] as one of the few sounds that may be restricted by the avoidance of words like rouge or measure. The nature and justification of this non-traditional correlation is seen in comparing it with the traditional pair /s:z/, which in a fairly high but non-extreme frequency is linked to an extremely low one, while /i:z/ links two (high versus low) extremes. To criteria of respective frequencies we may add pedagogic considerations, i.e. functionality as potential of language acquisition. Some phonemes, such as the five 'basic' vowels or the voiceless stops, may approach universal currency as high-frequency units, as compared to others infrequent intersystemically as well as within specific languages. 14 A cursory glance at the respective high or low scores suggests that pedagogic ranking is not necessarily related to frequential progression. Both high-score 'universal' and lowscore 'esoteric' sets contain phonetically disparate members, yet either set, though with arbitrary cut-off points, has its operational justification. The high-score type betokened by /i/ reflects an area of possible instructional slighting because of presumable ease of acquiring 'universal' sound-features. A quite similar degree of planned inattention, at least in the initial stages of language instruction, may pertain to /z/ as a token of the low-score type, for the opposite reason, its restrictability. In consequence, the 'correlation of extremes', /i:z/, represents the upper and lower boundaries of the large gray area on which acquisitional effort must focus, i.e. which has the additional load point. 3.202 Variable, linguistic (level-transcending: morpheme-directed). — sm-ms, MS (En). Another type of binary multiple-feature correlation of interest to this study is that of zero contrast, the counterpart to single-feature allophony and discussed above, English /h:i)/. Beyond recalling the functional overload of one of the opponents relative to the other as the result of frequential dissimilarity, 13

Frequencies of /i/ and /z/: Miller (1951), 86. Guiraud (1960), 30; Herdan (1966). Cf. also Aginsky and Aginsky (1948); Danes (1966); Greenberg (1966 a); Greenberg (1966 b). 14

INDIVIDUAL CORRELATIONS

89

the two units may serve as further illustrations of carriers of varying degrees of 'morpheme-directed' (boundary-marking) and morphemic (significative) potentials. Additional non-frequential load scoring, effected in increasingly supraphonological terms, is then noticed 'premorphemically', i.e. in the componential no-man's land of Grenzsignale, by morpheme-introductory /h/ or morpheme-terminative /q/, these two units, despite occasional exceptions (ahoy, dungaree), being intrinsically 'morpheme-directives'.15 In less generalizable fashion, Rischel has pointed to the load of syllable-initial j-, high in Joel (vs. Go!), relatively low in my name is Joe;u as an opening wedge to non-binary analysis of non-frequential load functions, Rischel's observation is of considerable interest to this discussion. In proceeding from phonologically minimal units to unit-strings, we may apply a very similar analysis to the English sequence -er, allegedly no more functional in hammer than ha-; that is to say, while in a word like singer the morphemic status of -er, in conjunction with that of sing, is not in doubt, there is, in contrast, no morphemic residue *hamm- hence no isolatable nor, in fact, semantically justifiable individuality of -er in hammer. Thus, -er taken as a unit in the latter word would appear to carry zero function. That indeed is true in the sense of inherent morphemicity (cf. /S-/, below), but not in that of morpheme-boundary marker similar to /h/. The separate status of -er in hammer has previously been, if not acknowledged, at least anticipated.17 Its functional significance may in the context of the present discussion be abstracted as an instance of plus-load (as a morphemic marker) versus zero load (as a morphemic unit); -er in both singer and hammer may be regarded as a plus feature in the former and as a zero feature in the latter from the standpoint of semantic morphemics alone.

15

Morpheme-introductive /h/: Martinet (1955), 166. Cf. also Chapter 1, fn. 95, p. 50 above. 16 Rischel (1964), 86. 17 Bloomfield (1933), 240, 243; R. S. Meyerstein (1964 a), 565, 569. Cf. also Bolinger (1968), 56-57; Malkiel (1968),7.

90

INDIVIDUAL CORRELATIONS

3.203 Level-transcending: set-directed. — sm-ms, MS (Fr). The case of French savant aveugle18 resembles the English examples cited in the increased significance of one of the phonemic components (/t/): the presence of the unit marks the word sequence as being representative of the adjective-noun structure, while its absence suggests the reverse arrangement. What establishes the typological individuality of this case is the potential of presence or absence, which English /h/ or -er do not share. In a discussion devoted to differences in functional LOADS, the recognition of a mere difference in function may appear out of place. The inclusion of the two French constructions hinging on /t/ does, however, acquire justification on the basis of frequency projections for one or the other sequence establishing the numerical primacy of expectable noun-adjective arrangements, hence a numerically distinct load conditioned by the presence of /t/, with the excess frequency factor for these arrangements being the additional load point of the 'marking' phonological unit. 3.204 Morphemic (significative) overload of phonemic units. — ms, MS (En). In proceeding to the type of phoneme previously described as significative, we have crossed over to the area of specifically morphemic overload measured against purely constitutive phonemes. Under Type 1 we presented the opposition /9:ô/ as having increased functional significance in most instances of final occurrence because of the grammatical contrastiveness absent from such other voice-distinguished pairs as /p:b/ or /k:g/. At this time we may single out certain phonological units in opposition not to their articulatory correlates but to all other members of the phoneme inventory. It will then be observed that English initial /ô/, in forms other than though, has a morphemic function in addition to the phonemic one shared with its peers. Similar functional distinctions, apart from that between significative and constitutive elements, have previously been established in 18

Vinay and Darbelnet (1958), 11; Léon (1966), 125. On related liaison phenomena, cf. Mok (1966).

INDIVIDUAL CORRELATIONS

91

connection with 'phonetic symbolism'19 and diversification among types of supraphonemic functives,20 though previous work on these subjects does not evaluate features with specific reference to score points of functional load. 3.205 Three cases of unit-overload (+morphemic —syntactic; —morphemic —syntactic; +morphemic -{-syntactic). — Within monosystemic analysis, we may suggest additional descriptive extension in recognizing three cases of overload correlated with distinctions of levels: Case 1, which is always morphemic; Case 2, which is never morphemic; Case 3, with two-level implications. Denotationally in Case 1, English /S-/, barring exceptions noted, always is a 'deictive' functive of one kind or another; connotationally, a certain intonation contour may always be associated with 'friendliness'. 21 In Case 2, overlap instances on the order of GOPolitician ('politician of the GOP', the Republican party in the United States) in a non-diachronic sense involve doubled nonmorphemic load on the graphic constitutent P (most likely of non-morphemic status in the abbreviation GOP, and certainly so in politician). The case of /S/- is systemically general, i.e. without performance options; the formation of Case 2, in contrast, is 'sporadic' in specific and optional lexical constructions. Similar systemic instances in Case 3 are encountered in constituents of enhanced transformational relevance, here seen as syntactic overload, of morphemic units like her in he told her to go, with actualized object function plus transformational subject potential; 'sporadically', interlevel overload is shown in fee-vee ('pay TV'), 22 reflecting the non-morphemic change from t(ee) to f(ee) on the nonmorphemic-morphemic correlation of tee with fee. 3.206 Morphosyntactic overload of morphemic units: functional overlap versus functional potential (set, sequence). — As lexical 19

Householder (1946). Bolinger (1948); Bolinger (1950). 21 Martinet (1962), 33, 37. 22 Fee-vee: Time Magazine 92, 26: 58-59 (1968). On cooccurrence and transformation specifically relevant to this part of the discussion, cf. also Harris (1964). 20

92

INDIVIDUAL CORRELATIONS

item, her in the sentence of the preceding paragraph compares to her as a class-cleavage homoneme (personal pronoun, possessive determiner), or to it with the distinctive score point of 'artificial subject' candidate, in diverse instances of excess function potential which other members of their respective paradigms do not carry. Another manifestation of supraphonemic function in excess of that of other cooccurring elements is the function of sentencereplacive potential. The examination, from that point of view, of Are you tired? readily illustrates the zero score for non-replacive Are, a relatively low plus-score for you (with non-specific replaceability concerning the question as a whole and antecedent dependence, also illustrated in Are youl), versus a high score for Tired'! as a free form both antecedentally and as a representative of any specific question.23 3.207 Frequential overload; variables: formal-synchronic (syntagmatic, paradigmatic) : formal-diachronic (duration of existence, paradigmatic)', semantic, exolinguistic. — Again, what has been evaluated here is not the frequency of the holophrastic replacement but the ability to function as such. This, of course, does not rule out the possibility of a certain parallelism of expressive potential of the sentence replacive; in some systems of 'English', perhaps governed by variables of certain kind of 'folksiness' or of psychophysical incentives reduced by laziness or fatigue, Tired? as a free interrogative utterance may indeed be more frequent than the sentence which it replaces. Comparable correlations have been suggested in previous scholarship. Guiraud's highest frequency scores attach to the shortest, oldest, morphologically least complex and semantically most inclusive forms. 24 Heilman and others also elaborated on the interdependence of frequency and phonic structure, phonic shape, etymological class, or style.25 In comparison with longer, recent, learned, or esoteric words, 23 24 25

Holophrastic Tired-. Gunter (1963). Guiraud (1960), 19, 30. Heilman (1964), 428,

INDIVIDUAL CORRELATIONS

93

we thus have four distinctive variables for a particular kind of functional load distinctions, i.e. that based on frequency. Additional determinants of frequency may be sought in such areas as degree of grammaticality, efficiency, exolinguistic situation, subjective impressions of 'naturalness', 'simplicity', folklore, or combinability.26 Some of these variables will be taken up below. 3.208 Overload, frequential and semantic; synchronic variable of cooccurrence in either system. — ms, BS (S1A:S1B). Bisystemic relations as understood in this study may or may not reflect traditional notions of bilinguality. In either case, comparable diversity of functionality will support the bisystemic view. In comparing, for instance, Standard Slovak (S1A) and many varieties of the Slovak spoken by immigrants to the United States (SIB), it is observed that Standard Slovak has two words, tu and sem, referring, respectively, to the locational and directional meanings of 'here', whereas the type of Immigrant Slovak to which this observation applies uses one word only, tu, in the meaning both of location and direction.27 This establishes the semantic overload of the Immigrant form against the background of Standard tu which shares the load with sem. 3.209 Synchronic variable of cooccurrence in both systems. — ms, BS (En:Cz/Sl) In the preceding illustration, tu COULD have a diverse number of functions, as determined by one or the other system (S1A or SIB). Czech or Slovak noha, in contrast, MUST have the functions of English leg and foot combined. In other respects, the two Slovak types and the Czech/Slovak-English matching represent the same kind of form-function correlation in which a single carrier of double load is opposed to two carriers with their respective single loads. 26 Grammaticality: Levin (1964), 310, 312. Efficiency: KuCera (1963), 194. Exolinguistic: Jakobson (1939 b), 151. Naturalness: Windfuhr (1967), 86. Simplicity: Mackey (1966), 71. Folklore: cf. Section 3.215, below. Combinability: Mackey (1966), 84, 175; Vachek (1966 a), 64. G. P. Meyerstein (1966), 6.

94

INDIVIDUAL CORRELATIONS

3.210 Variable, synchronic or diachronic; cooccurrence or noncooccurrence in both systems. — ms, BS (La:Fr). Much the same may be observed in another gray area of systemic differentiation, that of 'stages of the same language'. Alternatives in one or the other sense, i.e. as in S1A:S1B or Cz:En illustrated above, may or may not reflect different stages of historical split or coalescence, as shown in Latin avunculus and patruus in diverse functions now recovered by French oncle. It is then seen that tu:sem reflects a functional relationship historical (i.e. the result of shift due to interference from a third system, English) and paradigmatically cooccurrent (i.e. both monofunctionally with one load point and bifunctionally with two); noha : leglfoot, in turn, relate non-historically and cooccurrently, while in avunculus/patruus:oncle, as noted, we find a bond of history but not of cooccurrence. What all three of these cases have in common is the discharge of diverse functions by two forms in one system for one form in another system; as a result, the non-frequential load point assignment for one of the forms is in excess of the number of load points assigned for the other form. 3.211 Frequential overload; variables, dialectal (geographic, social) or idiolectal subsystems. — ms, BS (EnA:EnB). A number of frequential load evaluations acquire validity in a bisystemic context only. Again, we may first consider instances of bisystematicity 'within the same language'. Traditionally, they may be held to reflect one or the other style. We then recognize in child language the increased incidence of the feature of reduplication.28 Certain geographic or occupational 'dialects' may be characterized by the excess frequential load scores of specific vocabulary items, such as tonic in New England, invoice in commercial text, so or and in popular usage, and similar correlations listed by Mackey; 29 for the sake of simplification, we ignore Mackey's diversification of dialect, register, style, medium, and time. Other instances of increased incidence of occurrence are func28 29

Ferguson (1956). Mackey (1966), 177-179.

INDIVIDUAL CORRELATIONS

95

tional indicators of the speaker's age or group affiliation (you know or man! as interjections currently favored among certain segments of the population) or his personal or imitational idiosyncrasies {powerful as a recurrent vocabulary item in recent linguistic literature). Without gaining universal currency among speakers of a given language, some forms of lexical favoritism may to varying extents transcend the confines of individual 'styles' and encompass morphological sets; witness the predilection for possessives in many quarters of the English speaking community (she has HER nerve!, YOUR friendly undertaker, etc.). In all of these cases, the favorite item (or set) has the additional — in this instance, frequential — load point. 3.212 Semantic overlap in bisystemic paradigmatic strings. — ms, BS (En, Fr, Ge). With representatives of large ('open') sets, we may identify more or less extensive paradigmatic strings comparable in type if not in extent, in a non-frequential sense, to the frequential paradigms of traditional phoneme-pair loads. In the semantic paradigms of English danger-hazard-coincidenceaccident and French danger-hasard-coincidence-accident, bilingual informants may concur in locating the principal meanings of English danger and hazard in French danger; hence, by and large, danger (Fr) is to danger-hazard (En) as noha (Cz) is to leg-foot (En) from the point of view of load carried singly or spread over more than one form, though, of course, neither from the point of view of specific translatability nor from that of respective semantic relations between danger and hazard or leg and foot. Thus, danger (Fr), in relation to the two English words, acquires a load point rating which, in the sense of the general picture outlined above (i.e. disregarding subsidiary meanings, nonsynonymy postulates, etc.), is substantially the sum total of the respective ratings for those two words. (It is not the result of the addition of equal scores, since notably dissimilar semantic ranges, besides differences in text frequencies — both relatively high for English danger but low for English hazard — would not permit the recognition of functional equality of the two English forms.)

96

INDIVIDUAL CORRELATIONS

English coincidence, in turn, recovers areas served by French hasard (common) and coincidence (less common). French coincidence-accident and their English cognates, however, more or less agree in their respective combined areas of functionality; thus English coincidence substantially scores the load points of French hasard and coïncidence, i.e. En coincidence > (has a higher load score than) Fr coincidence, whereas the French coincidence score embraces values of English coincidence and accident combined, that is En coincidence > (has a smaller load score than) Fr coïncidence, a fact which represents the opposite of the preceding correlation of these two cognates. With the injection of German Gefahr-Zufall-Unfall, we obtain German Gefahr more or less matching French danger (Ge = Fr), though if we extend consideration to French péril, Gefahr = danger-péril (Ge > Fr); German Zufall — French hasard-coïncidence (Ge > Fr); German Unfall = French accident (Ge = Fr). With German compared to English, Gefahr — danger-hazard (Ge > En) ; Zufall = coincidence-accident (Zufall > coincidence) ; Unfall-Zufall = accident (Zufall < accident). Clearly, the load scores will vary with the increase of form inventory and systemic scope, an observation militating against the statement of numerical specifics but not that of simple overload recognized in specific cases. 3.213 Opposite scores of frequential and non-frequential (semantic) load; additional variable, systemic. — ms, BS (Fr:Sp). Morphosyntactic 'underload', redundancy in comparison with contrastive signaling, may be identified in the presence or absence of the preposition /a/. Wherever in French it is present, it may have or lack primary differentiating value (on discute à Paris, on discute Paris', versus on obéit à ... where, with an object, à or its representative, must be present, hence has no contrastive function). Bisystemically against the background of French, Spanish exhibits a shift in the direction of zero load for 'personal accusative' constructions (veo a Pablo), where a is similarly redundant. If, on the other hand, we refer to respective frequency

INDIVIDUAL CORRELATIONS

97

counts, and all other factors being substantially alike, we would have to acknowledge a shift in the direction of increased plus-load for Spanish where in this specific area of comparison the frequency of /a/ would exceed that of French. 3.214 Frequential load as a variable of systemic typology. — smms, BS (Sc, Rc, Fr, Ge, En). To the extent that it is possible to evolve language typology based upon recognition of relative functional predominance, lexical features such as those of the preceding section convey an impression in very specific semantic spheres only. Their grammatical counterparts, however, may play the part of systemic indicators which on a lower level attaches to such 'typical' units as nasal vowels or certain consonant sequences. As a matter of lexical as well as textual frequency, this may be illustrated, to a descending degree, by the incidence of reflexive verb constructions in the Slavic and Romance languages, typically 'reflexive languages', as opposed to German or English — in the last-mentioned system, reflexive verbs are few and far between and every so often indicative of non-popular usage; rather than accustoming ourselves we get used to something, we prefer to go away rather than 'betaking ourselves hence', etc. — and we may also point to the textual preponderance of possessive usage in English over alternate expressions in French (he cut his finger: il s'est coupé le doigt). 3.215 Typology of or by the system; variable, linguistic or exolinguistic. — ms, BS (SnA:SnB). Systemic typology may be understood, as in the preceding section, as typology OF the system or, conversely, as typology BY the system in which the linguistic community typifies certain forms of the system. In the latter sense, excess load may be recognized, not as a matter of actual linguistic fact, but as the assignment prompted by the speaker's (or the grammarian's) subjective beliefs. Exolinguistic views of this type are here referred to as respective 'situations' (Sn). Given two systems with comparable nominal and verbal forms (Latin and German), we notice that one, Latin,

98

INDIVIDUAL CORRELATIONS

presents Situation A in which the verb (verbum) is 'the word', whereas Situation B is encountered in the German tradition of regarding nouns as the 'chief words' (Hauptwörter) — evaluations undoubtedly prompted by respective frequential or stylistic preferences, though absent in other linguistic communities such as English or French. The conditioning factor in these instances might be termed folklore. 3.216 Load shifts', variable, frequential-exolinguistic (Hdiomaticity'J. — ms, BS (En, Fr, Ge). Bisystemic load score evaluations may be affected by two conditions of instability: synchronic attrition by interference, and the effects of diachronic change. 'Synchronic change', in the present context, refers to modifications of load assessment as the result of 'negative transfer' reflected in a modification of the relative frequencies, for instance, of English already and German schon. What we are witnessing is, not a contrast of presence or absence of a feature, in the sense of respective inventories or distributions — the usual interference problem — but features equally present, acceptably distributed as used, though unequally frequent hence unacceptably used if given equality of frequency not rightfully theirs. In other words, interference derives from an imbalance of respective functional loads in a unitary frequential sense. This is noted in the disturbing predilection or slighting of one or the other lexeme, disturbing in the English speaker's judgment of certain 'pleonasms' on the part of speakers with German language backgrounds, no less than in the German speaker's disapproval of 'incomplete' expressions heard from speakers whose native language is English. The correct constructions Did you eat ? and Haben Sie schon gegessen? related to their 'unidiomatic' (i.e., diagnostically speaking, overloaded or underloaded) counterparts Did you already eat ? and Haben Sie gegessen ?, may illustrate the point. Transferred to bilingual grammar, the problem may be noted, for example, in the use made in different languages of similar grammatical features. A purely enumerative presentation including

INDIVIDUAL CORRELATIONS

99

a listing of active-passive contrasts in English, French and German is then seen to fall just as short of an adequate accounting for the linguistic realities in a taxonomic sense as does simple transformational equation of the two voices. English is spoken, whereas on parle français and man spricht deutsch. ONE may speak English, but ONE does become a bore to ONE'S neighbor if ONE overdoes the one-constructions in English, while the passive transformations for the French or German equivalents cited are either impossible or preempted to carry a specific unintended (e.g. 'military') function. 3.217 Formal similarity versus frequential-load dissimilarity, exolinguistic (pedagogic) response. — ms, BS (En, Sp). The potential of negative transfer, deriving in the preceding examples from a matching of taxonomically or transformationally (but not statistically) correlatable functions, is reinforced whenever there is, in addition, a similarity of forms, as noted in my name is and mi nombre es (for me llamo) in Spanish. Stockwell discusses this type of interference from the point of view of the learner's psychological or absorptive reactions: should we teach the beginner the similar construction (mi nombre es) to build up his sense of accomplishment, or the dissimilar one (me llamo) to avoid resistance to this expression later on? 30 Stated in this frame of reference, the difference in constructions is evidently not one between incorrect and correct expressions; the factor of correctness will enter to the extent that respective statistics, i.e. frequential load criteria, are recognized on a par with criteria of inventories and distributions. 3.218 Formal similarity and frequential or non-frequential load dissimilarity, unidirectional conditioning of load increase. — ms, BS (PoA:PoB; GeA:GeB). In the preceding section, the two different load scores corresponded to respective differences in form. Load imbalance may be just as noticeable for one and the same form seen as a bisystemic homoneme in one of the systems. 30

Stockwell, Bowen, and Martin (1965), 292.

100

INDIVIDUAL CORRELATIONS

As non-frequential (semantic) overload, this may be illustrated once again in lexical load changes in immigrant languages. Evocative of the tu 'tu/sem' correlation in Immigrant Slovak noted above, though dissimilar in its formal-semantic aspects, American Portuguese grosseria takes on the added meaning of 'grocery' over and above usage in Portugal where the word does not include that meaning.31 Non-frequential increase, of course, in this case is bound to parallel increase in the frequential sense. The reverse, however, is not necessarily true. For example, a non-immigrant language, succumbing to external interference, may increase frequential load without adding non-frequentially. This may be inferred from such studies of frequency changes as that by Galinsky, who observed the increased incidence of the German suffix -er as the result of assimilated loans (e.g. Fernschreiber);32 -er spread, but not under the impact of any added meaning, since the ending with a meaning as in Fernschreiber has numerous established precedents (Anhänger, Fernsprecher, etc.). 3.219 Static versus dynamic potential of overload: productivity. — ms, BS (En, Ge). The example of grosseria suggests an increase in load stimulated by the accident of a similar form in English, but with 'one' functional significance equivalent to the added load point of the American Portuguese form. There is otherwise nothing suggesting that the form as used in Portugal might take on the added meaning and incidence, i.e. that it would be productive in a functional sense. Productivity — functional potential in the process of realization, and a further grey area between the synchrony of monosystemic facts and the diachrony of historically related bisystemicness — is, in contrast, evident in the case of Fernschreiber. The suffix -er (in German as, incidentally, in English) has a productivity potential to begin with. The 'pathology' (interference factor) may be seen in the excessive fulfillment of that potential prompted by external (bisystemic) stimuli. The problem of excessive productivity may be 31 32

Pap (1949), as quoted by Haugen (1956), 53. Galinsky (1964), 377. Cf. fn. 17, Chapter 3, above.

INDIVIDUAL CORRELATIONS

101

resolved in either of two directions: it may be carried to an 'inadmissible' degree, or it may settle itself as the result of historical development making the load change 'admissible'. 3.220 Admissible versus inadmissible productivity, variables, frequential or non-frequential ('correctness'), synchronic or diachronic. — Inadmissible productivity (as distinct from admissible 'excess') may then be seen to have a bisystemic origin, but monosystemic validity only; there may be one form with two load points in one system but only one point in the other, as in (Fernschreib)er, but without the same bisystemic acceptance as in the latter case. What this means in practical terms of language acquisition is, once again, the existence of different degrees of 'incorrectness' based on diversity of load scores. Frequentially, this is illustrated when we consider the relatively high incidence of German nouns of the masculine gender, known to prompt speakers of English to create pseudo-German constructions of similar type at variance with existing German syntax, as seen in American restaurant names like Der Wienerschnitzel. It is of little consequence that in German they prefer das Schnitzel; der Schnitzel, the owner will tell you, "sounds more German". 'Admission' as a result of historical shifts represents the other solution to productivity. Hoenigswald has noted, among various instances of this phenomenon, the combinatory superiority of such grammatical units as English plural/-s/ in relation to other units not so distinguished. 33 Recalling traditional views on the zero load of etic (alio) oppositions, we can scarcely overlook the nonbinary, non-phonemic, non-zero load of this allomorph, both frequentially (in actual occurrence) and semi-frequentially (in productivity potential), compared to other representatives of this morpheme. 3.221 Load score increase versus maintenance: productivity versus survival. — sm-ms, BS (LgA:LgB, La:Sp, FrA:FrB). Productivity, 33 Hoenigswald (I960), 62,

102

INDIVIDUAL CORRELATIONS

the increase in functionality of actual or potential occurrence, finds its counterpart in survival, the maintenance of functionality. For reasons of phonetic universals or the accident of specific development, some sound-units enjoy a higher survival chance than others.34 On the supraphonological level, we evaluate the load point of survival value in relation to any function recognized. Frequential survival is noted, for instance, in the preservation of Latin -ere and Ire conjugations in Spanish, as opposed to conjugations productive (-are) or lost (-ere). The relation between frequential and non-frequential (e.g. semantic) survival is seen against the background of an older stage of French when moudre (as now spelled) had the meanings of 'to milk' and 'to grind'. From the present-day point of view, we may say, frequentially, that moudre was lost wherever it had the first of the two meanings,35 or else, non-frequentially, that moudre was kept, but not as a homoneme, i.e. that only one of the two meanings endured. In the related area of grammatical meaning, we may similarly view the demise of homonemic (singular-plural) -s of mediaeval French nouns as a non-homonemic retention in one (plural) function. Homonemic development in the opposite direction of functional increase is illustrated in the additional load factor of grammatical ambiguity unresolved, i.e. 'intolerably' increased functionality (German sie hat das Kindgeschlagen, 'she beat the child' or 'she was the one the child beat'), or only partly resolved by injection of additional grammatical criteria (English deer, plus/minus e.g. many), or non-grammatical clues (English Spanish students / Spanish students). Paradoxically, increased functionality may be connected with non-survival, or at any rate not be apparent until the loss has occurred. This, in essence, recalls the notion of 'basic' forms or 34

Hoenigswald (1960); Greenberg (1966 a). Cf. Weinreich (1958). Hoenigswald (1960), 21. Cf. also Guiraud (1964), 74. For synchronic counterparts, cf. Malmberg (1963), 154; Malkiel (1967), 1241; and Section 3.212, above. 35

INDIVIDUAL CORRELATIONS

103

functions viewed above as instances of 'typology by the language' (verbum, Hauptwörter). That form-set evaluation in terms of respective significance may be on far more solid foundations than that of folkloristic vocabulary does not, of course, require emphasis. In a diachronic context, as noted by Hoenigswald, There is a hierarchy of relative weight of forms constituting environments. It is clear, for instance, that the disappearance of an English verb has a far stronger effect on the meaning of the nouns which function as its subject than it has on adverbs like always or certainly.36 Thus, zero occurrence is by no means necessarily tantamount to 'zero function' in a diachronic sense, a state of affairs reminiscent of non-zero identifications in a diachronic context noted above. 3.222 Load score fluctuation: survival, demise, and revival. — sm, BS (La, Sp, FrA, FrB). On the comparable level of phonology, the initially positioned sequence of/s/ followed by a consonant is a cluster admissible in Latin (spatha, spathula). In a subsequent developmental stage (FrA, an earlier type of French, or Sp, presentday Spanish), there is no similar occurrence, hence naturally no frequency scores are registered, which has left its trace in modern counterparts (épée; espada, espátula). In French B, present-day French, the cluster does again occur initially, and score readings thus become once again available for vocabulary introduced in French B (spatule). Taking the system of origin and terminal stages only, we observe that the Latin-Spanish progression is in the direction of plus to zero. From Latin to French, there is at first a similar form DEVELOPMENT of similar occurrence frequency; from the point of view of modern French, however, it is a formula of form ALTERNATIVES similar to others of this study. Since noninitially the sequence exists in Old French and Spanish as well, we simply have 2:1 correlations between Latin and French A or Spanish, as also between French B and Spanish, as a matter of environmental discrimination, and between Latin and French B as a matter of simple frequency. 36

Hoenigswald (I960), 20.

104

INDIVIDUAL CORRELATIONS

3.223 Load score creation ('birth')', variable, non-frequential (level). — sm-ms, BS (EnA:EnB). The historical development of non-frequential (e.g. significative) overload may be illustrated by the 'morphologization' of final /-z/ in English. When the ancestral form of French cerise was borrowed to give Modern English /tSeriz/, the morphemic function of the final phoneme of the original loan word was as much zero as in the generally accepted interpretation of /tSajniz/ it has remained to this day; /z/ subsequently acquired its current status as plural exponent in one, but not except aberrantly in the other, of these two English forms. 3.224 Load score evaluation; variables, linguistic (system) and exolinguistic (admissibility). — sm-ms, BS (FrA:FrB). Comparably to preceding correlations, we may view apparent zero-load scores as instances of instability produced by 'admission' through language evolution, or by 'inadmissible' deviation. Thus, from the vantage point of French, articles as normally used (plus-load) correlate with 'admissible' zero-load in an earlier stage of the language, or with a present-day 'telegraphic' system, as opposed to 'inadmissible' use in 'child language'. The latter, in a subsequent stage preceding complete adjustment to the speech behavior of the linguistic community to which it belongs,37 exhibits a reduced incidence of use of this part of speech, leading to a conversion to 'inadmissible' 2:1 load relationship, noted above.

3.3 TYPE 3 — BINARY SEQUENCES AND OPPOSITIONS {ab\a — 2:1)

3.301 Extent inequality of load exponents: syntagmatic extension, simultaneous {markedform) or consecutive (long form). — sm, MS. The preceding part of this study has been an examination of traditional oppositions and other correlations involving the replacement of units. Thus, a:b symbolized matched units such as /9:6/ Jakobson (1941); Gregoire (1948). Cf. also Leopold (1953).

INDIVIDUAL CORRELATIONS

105

in comparison with a pair like /p :b/. Operations with arrangements on the order of -er proceeded in perspective of a non-sequential status on a higher level. Without questioning the significance of determinant environments, we notice that load evaluations as suggested above did not pertain to the environments but rather to units as determined by environments. The prevailing viewpoint of these evaluations may thus be termed paradigmatic. At this time we add a few examples including syntagmatic relations, in addition to paradigmatic comparisons, for the specific load problem presented. Some oppositions, earlier seen as a:b (contrastive unit pairs), may in this context shape up as features marked (ab) or unmarked (a), with the mark (b) cooccurring simultaneously in a 'bundle' with other features or consecutively to them. Again, no specific order of precedence and succession is symbolized by the ab formula. There have been a number of general observations on the relevance of this analysis in the context of monosystemic load assignments. Greenberg, for instance, observed that in simultaneous cooccurrence the number of marked phonemes of a set of correlative pairs is usually less than or equal to the number of the unmarked. 38 Thus, as between the marked and the unmarked, the functional significance (load score) based upon lexical frequency is lower for the former than for the latter. 3.302 Opposite scores of mark-addition and frequency-potential (.survival); variable, systemic; implications, exolinguistic (medical). — sm-ms, BS (LgA:LgB). In a clinical context, Jakobson observed that as a result of aphasia there is a tendency to drop marked entities rather than unmarked ones, e.g. finite verb forms rather than infinitives.39 Again, then, the marked unit comes out on the short end of relative functional significance; the load of marking is, as it were, too heavy a burden to be carried by the disabled. This is of interest also in the context of diachronic durability 38

39

Greenberg (1966 a). Paradigmatic versus syntagmatic; Seiler (1967). Jakobson (1932), 80.

106

INDIVIDUAL CORRELATIONS

projections, discussed above. In terms of bisystemic load correlations the community's language (LgA) is seen to differ from the aphasic's (LgB) in text frequencies comparable to ratios involving community and child languages as non-simultaneous (consecutive) manifestations (cf. Section 3.2, above). Similar text-frequency imbalance within one system has been pointed out by Greenberg.40 3.303 Opposite scores of longitudinal addition and non-frequential (semantic) potential: abbreviations. — ms, MS (En). On the morphosyntactic level, long and short expressions in a consecutive (syntagmatic) sense may amount to identical functions. Many abbreviations bear witness to that fact: thus, in American (specifically, Western) usage, West Coast and Coast refer to the same geographical concept; similarly, in conventional references to publications of bibliographies or running heads, short titles are common representatives of full-scale citations. In terms of the present study, the short expressions carry the load that may be spread over these expressions plus the constituents of the longer structures of which they are a part; from a related point of view, we may say, synchronically, that the functional potential of the short expressions exceeds that of the remaining constituents of the longer structures since the short expressions, but not the remaining constituents, may represent the long expressions. 3.304 Extent inequality of load exponents', variable, systemic, functions, identical (linguistically in denotation) or dissimilar (exolinguistically in connotation). — ms, BS(EnA : EnB;. While the distinction between full forms and abbreviations frequently stays within monosystemic limits, on other occasions there is a fairly obvious distinction to be made among dissimilar systems, such as the written, epistolary or cartographic one of Calif, or the 'postal system' of CA recently put into effect, versus California, 'fussy' or, at any rate, 'formal' in either of those two systems though exclusively used in the system of 'spoken English'. Again, 40

Greenberg (1966 b), 513.

INDIVIDUAL CORRELATIONS

107

the hierarchy of respective representational potentials is clearly indicated by the progression CA-Calif.-California', again, the identification of the ascending or descending character of that progression derives from the specific area of functionality studied: if the area is that of ability to represent in specific systems, the progression of functionality of the three forms is descending, while from the point of view of passe-partout utility — California being a possible expression in any system of 'English', to the contrary of the abbreviations — the progression is of an ascending nature. Accordingly, there is, within any binary opposition of two of the three forms, a 2:1 load relationship in either direction depending on either functional area. 41 In a strictly synchronic comparison of doctor and doc, we may arrive at the conclusion that the longer (ab) and shorter (a) forms constitute monosystemic features with identical semantic functions. Additional criteria, however, would tend to question this analysis. To begin with, the systemic diversity needs to be recognized: the two forms are not representative of identical 'styles'. This, in turn, entails connotational differences between 'native' and 'borrowed' terms according to one or the other stylistic system, as well as corresponding frequential distinctions. At the very least we may detect differences in connotative content on which to base a distinction of respective load points; depending on one or the other system of style, the non-frequential load point distinction will be confirmed or contradicted by frequential scoring. 3.305 Scores, opposite; variable, systemic; functions, dissimilar. — ms, MS/BS (CzA:CzB/RuA:RuB). Comparable numerical implications may be encountered in the assignment of 'normal' versus 'expressive' functions to one or the other formal structure. Normalcy implies statistical preponderance; expressiveness, in contrast, may be associated with numerical minority status. Denotatively, of course, the relationship is the reverse: the infrequent structure acquires the additional load point of expressiveness. 41

Progressions ascending or descending: R. S. Meyerstein (1968), 42-44.

108

INDIVIDUAL CORRELATIONS

This may be illustrated in comparable monosystemic areas, the use or omission of personal subject pronouns in Czech and Russian, and the bisystemic correlation of the respective data. Both languages admit of pronominal or apronominal structures (Czech já jedu or jedu; Russian /ja jedu/ or /jedu/). However, as noted by Jakobson, en russe, c'est la construction à deux termes qui est le type 'normal', tandis que la variante à sujet-zéro est un procédé expressif. En tchèque, au contraire, le zéro d'expressivité se rapporte au sujet-zéro, et la valeur expressive s'attache au type já jedu. La première personne est mise en relief par la présence du pronom, qui est un pléonasme du point de vue grammatical. L'abus de ce pronom fait en tchèque l'impression d'un style vantard. Au contraire, en russe, c'est justement l'omission excessive du pronom de la première personne que Dostojevsky éprouve comme une morgue irritante.42 In terms of respective functional load scoring, it is then noticed that in one system the long form carries the increased signaling load, whereas in the other system the extent of the expression and that of the functional significance are inversely proportional. 3.306 Extension, optional or obligatory, variable, systemic. — ms, BS (Cz, En). The preceding contrast, based on presence or absence of a repetitious set (pronouns), finds its grammatical counterpart in the presence or absence of a non-repetitious (though similarly redundant) verb representative. Mathesius has pointed out the obligatory Czech verb-phrase preventing grammatical copying of such non-verbal English utterances as Much use, that\A3 In lexical-grammatical redundancy, addition of 'subjunctive' marking, omitted in English, at least as an obligatory feature, is a further case in point. 3.307 Exolinguistic implications of various types. — ms, BS (LgA:LgB). In the context of syntagmatic load studies, we may again inject criteria of varying degrees of 'admissibility'. Within 42 43

Jakobson (1939 b), 151. Vachek (1966 a), 89.

INDIVIDUAL CORRELATIONS

109

the range of the 'admissible', we may compare information concentrated in English glove yet spread over the two constituents of German Handschuh. Factors such as 'motivation' will tend to entail comparable enhancement of the significance of the bicomponential expression: [We] can all agree that glove and skate are completely unmotivated (in contrast with Handschuh and Schlittschuh), and that weekly or watery are quite transparent, but is thimble as utterly opaque as French dèi ... It looks rather as if motivation is not an absolute, but admits of degrees (including complete opacity as well as, perhaps, complete transparency). If so, the calculation of the required index is going to be rather difficult.44 Difficulties of numerical precision aside — a problem theoretically on the same order of that noted for stated specifications of the functional loads of binary phonemic oppositions, though based on less easily quantifiable psycholinguistic variables45 — motivation clearly constitutes a distinctive factor. So will learnability, discussed above, once a multi-constituent construction is identified as such, as pointed out by Martinet for a child's acquisition of French téléphone, compared to Fernsprecher learned by a German child: If it were shown that German children have less trouble with Fernsprecher than with Telefon, it would probably not be on account of the lack of motivation of the latter, but rather because Fern- and -sprechsound familiar, irrespective of what they mean. The French children of today, who constantly hear the phoneme combinations /tele/ and /fon/ in télévision and grammophone as well as téléphone, are probably just as well off as their German contemporaries even if they do not understand télé- and -phone. This, however, does not mean that motivation is to be rejected as a criterion for lexical typology, but rather that here is one more domain where one should not jump to conclusions.46 'Inadmissible' favoring of the long expression, in turn, may lead to such deviations as 'popular' or 'children's' French papa il dit for standard papa dit : 44

Householder (1966), 632. Saporta and Bastian (1961); Osgood and Sebeok (1965); Brière (1968). 46 Martinet (1962), 89. Cf. also Jakobson (1939 a); Jakobson (1941); Leopold (1953).

45

110

INDIVIDUAL CORRELATIONS

In popular French the equivalent of my father says is mon père il dit (pronounced [idi]) instead of the traditional mon père dit; since I have to use [idi] in reference to my father's saying something when I need not specify that the speaker is my father, why shouldn't I save myself the trouble of choosing between [di] and [idi] depending on whether I expressly mention my father or not? Since I can't help using [idi] at times, it is handier to stick to it, whatever the context, than to reckon every time with the context; pronouncing an extra phoneme is nothing in comparison with the output of mental energy required by the choice between [di] and [idi].47 In its stylistic system, [idi] thus carries a frequential load in excess of that of 'normal' (colloquially pronounced) French, with the first /i/ carrying, in addition, the potential, i.e. the further load point, of systemic marker, a potential actualized in environments such as that of the preceding quotation.

47

Martinet (1962), 55; cf. Grégoire (1948).

Ill

4.

CONCLUSION

In reviewing the preceding exposition we may ask three questions: How does extension of functional load analysis, as proposed in this study, lead to a redefinition of the underlying concept? To what extent is it of practical usefulness in the areas to which it was said to be relevant? What remains to be done? The description of functional load obviously derives from that of function. With more than just one relevant function to be recognized, there will, in principle, have to be as many descriptions as there are dissimilar and more or less correlatable functions. In previous work on the dissimilarity of concepts supporting supposedly all-encompassing terms, it was, for instance, pointed out that the incompatibility of definitional criteria was bound to lead to inconsistencies in defining the morpheme. We could speak of an informant's or an analyst's morpheme, and quite often the two sources of identification might agree on the candidate sponsored for morphemic status, yet every so often what would be a morphemic unit to one would not be so accepted by the other. As a result, we would have to view that unit as differing in nature from other units of identical designation, in accordance with their respective objectives of analysis. 'One and the same' unit might have to be regarded as being coextensive with, or including, or being contained in, non-units, or else it would have to accomodate 'equivalent structural function' as part of an amended morpheme definition based on semantic as well as non-semantic (or formal as well as non-formal) criteria.1 1

R. S. Meyerstein, (1964 a); R. S. Meyerstein, (1964 b).

112

CONCLUSION

What makes the determination of the morpheme comparatively easy is the fact that once we agree that in English the two analystmorphemes re- and -ceive are included in the informant morpheme receive, we thereby possess the respective physical dimensions, and barring extensions of descriptive purview the dimensions of the constituents are statable in terms of measurable criteria invariant within the limits of the description. Even the very limited correlation of text-frequency and semantic functions reflected in existing load analysis cannot lead to similarly clear-cut results. Sampling procedure, no matter how refined, is bound to alter the picture, and, be it ever so minutely, with each successive corpus. That is a matter of theoretical inevitability. That concrete data bear out the theory is shown, for instance, in King's zero load score for what is offered as the phonemic opposition of /x:q/. Phonemicity, by definition, must be demonstrable by contrast somewhere, and yet King's corpus, implicitly deemed sufficient to support his load value statements as the 'measure of minimal pairs', apparently is insufficient to unearth even one such pair to support the phonemicity in question. If, then, the 'same' functional load, as previously understood, is not the 'same' in different measurements, and if the discrepancy of supposedly homogeneous quantifications is accepted as the inevitable phenomenon it is, there can scarcely be philosophical objections to the inclusion of heterogeneous criteria entailing other discrepant data. It may, in fact, be claimed that many of these criteria, such as significative overload of certain phonemes, though not quantifiable in the same sense as textual occurrence contrasts, are not corpusdetermined as are the latter; on the contrary, each successive corpus, regardless of sampling procedure, will inevitably present identical results of overload. In other areas of functional extensions, to be sure, we are far less fortunate. Thus, as frequency increases, stylistic importance of certain forms will decline. But how large will have to be the increase? How pronounced will be the decline? How do we recognize, much less measure, the precise modalities of the interrelation ?

CONCLUSION

113

But then, how 'well-formed' does an utterance have to be to qualify for this descriptive accolade? Various degrees of 'wellformed' have been recognized,2 to which others might conceivably be added. Who could fail to agree with Bach that "the term 'intuition' is a loaded and ambiguous word", yet that "language as a cultural product cannot adequately be studied apart from the native speaker's judgment". 3 No one has yet presented us with quantificational procedures to measure variables of this kind; yet, for good and sufficient reasons, we operate with them. In practical terms we may ask: what measure of functional load is useful and attainable, and what measure is useful though not at present within the realm of feasible computations ? Measurements based on frequency and contrast of phoneme oppositions have obviously been attained. Their usefulness becomes obvious in proportion to certain more or less arbitrary degrees of Sequential discrepancy. An intuitive appraisal of lexical frequency data for French /s:a/ and /e:6e/ will reveal a drastic imbalance in favor of the former opposition, the latter, in fact, being demonstrable only by an exceedingly small number of minimal pairs. The instructional implications seem clear: attention should dwell more on the opposition of very high load scores than on oppositions scored low. The additional load criterion of 'learning relevance', inferred from these frequential load inspections, is then also measurable in amount of time budgeted or spent. But the inference itself cannot be generalized. The French opposition /e:ce/ is based on high-low frequency matching (high for /§/, low for /ce/). So is the English non-correlate (multiply distinctive) pair /i:z/ (high for /i/, low for /zf), and yet, for reasons suggested above, the conclusion of 'learning relevance' in the case of the English opposition may be just the opposite of that cited from French. On the supraphonological level, similar observations arise. Vocabulary is graded in accordance with word counts, and em2 3

Levin (1964), 309. Bach (1964), 4.

114

CONCLUSION

phasis focuses on high-frequency forms, to the detriment of esoteric items. Similar discrimination favors prevalent function: If one meaning of a word has a low frequency, then perhaps we do not have to teach that meaning, but another meaning of the word that has high frequency must be taught.4 Yet while unquestionably in a classroom situation involving students of English-speaking background, French, German or Spanish equivalents of the word class, fairly identical in forms and syntactic arrangements, are taught at the earliest opportunity, they are scarcely belabored; that is to say, the 'easy' term similar in form, function, and currency, ends up generating as little teaching and learning effort as the infrequent dissimilar 'hard' one, though for opposite reasons. Among worthwhile and attainable measures, we might include those of binary phonological oppositions outstanding AS A SET in their average load scores, compared to the remaining oppositions, e.g. the set of the seven pairs with highest scores in King's German data cited above. It seems pointless to dwell on individual scores within or outside of these sets: what, after all, would we gain, instructionally speaking, by distinguishing even two such outstanding scores as 49.419 and 31.148? In the evaluation of two sets arbitrarily limited as to floor or ceiling, respectively, one scores high and the other low and, all other things being equal, according time budgeting might be effected. As has, however, been shown, all other things are NOT equal: relative ease of 'phonetic universals', for instance, will prompt cuts in time allotments. Still, some correlations on the order suggested are at least conceivable and of demonstrable usefulness. This, then, is one area of applicability and needed development of the load studies proposed in this discussion. Other measures are statable as orders of magnitude corresponding to the number of certain additional functions: the added function of significativeness, for instance, provides the basis for the assignment of one additional load point. Measures apparently neither possible nor useful, at least on the 4

Hockett (1958), 56.

CONCLUSION

115

basis of data supplied by recent scholarschip, would include cause-and-effect assessments such as those offered in support of the predictability of diachronic events. The literature on the subject reveals no convincing case to justify these predictions, and has offered evidence militating against them. This of course does not preclude diachronic load-matching as an extension of bisystemic analysis, as matters of provable occurrence rather than prediction. Measures useful but not at present worked out would include such areas as 'stylistic relevance', and correlations between values previously measured (or measurable) and criteria of style, learnability, and other 'loaded and ambiguous' areas of unquestionable functional relevance. As we glance at the applications proposed, yet not effected, in previous literature on functional load, we notice that descriptive completeness, while obviously not attained in this study, depends on inclusion of load study areas previously left out of account. To some fields in which load data were said to play a part, such as spectrograph^ interpretation, statistical processes of sound wave decoding, or certain 'management' issues like alphabet construction, this discussion may not make a readily ascertainable contribution. For the discovery and understanding of aberrant phonological or morphosyntactic behavior, on the other hand, the present study has suggested numerous avenues of approach; quite clearly, limitation to 'correct' data of text-frequential phonology is not sufficient to cover submorphological pathologie and inevitably falls short of assisting us on similar morphosyntactic problems. This, of course, applies, as has been noted above, to the large area of pathologie prevention or cure known as linguistic pedagogy. What, in conclusion, might serve as a basis for defining functional load? Perhaps the question is not 'well formed': all functional features have load in the sense of significance. Copious listings of load statistics to a number of decimal points is one expression of load — not, perhaps, the most useful one as we have seen. Differential statements might suggest an alternative: in reference to a set of forms sharing a certain function, one sub-set (in this

116

CONCLUSION

study for convenience regarded as one unit-pair; one unit relative to a multi-member set; or a syntagmatic instance of long and short form matching, but not necessarily unique in the set) has an additional set of functions (again, for convenience, referred to as one additional function or load point), the overload. Taking this overload rather than respective load attainments — that is, actually conforming to the fundamental concept of opposition, commonly applied to forms and now carried to degrees of functionality — we may describe functional (overload as the statable quantitative difference (or at any rate, a difference possibly subject to future quantification) in function within a specific functional area, or, in a comprehensive sense, the composite picture of functional preferment or neglect for a selected range of functions, presented by one opponent relative to another. This statement, not unlike others that have preceded it, leaves many loose ends. It is not, in fact, offered as a definition, but rather as a suggestion of a possible avenue of approach to a definition. The problem is certainly more complex than previous description might have led us to believe. The complexity arises from the recognition of additional areas of relevance, a relevance scarcely in doubt. Any description of functional load cognizant of this relevance must either be highly complex, or else operate with evaluations sufficiently general to accomodate the range of relevant conceptual diversity.

REFERENCES

Abernathy, Robert, 1963 "Some Theories of Slavic Linguistic Evolution", in American Contributions to the Fifth International Congress of Slavists 1.7-26 (The Hague: Mouton & Co). Aginsky, Burt W. and Ethel G. Aginsky, 1948 "The Importance of Language Universals", Word 4.168-172. Akhmanova, O. S., I. A. Mel'chuk, R. M. Frumkina, and E. V. Paducheva, 1963 Exact Methods in Linguistic Research, translation from Russian by David G. Hays and Dolores V. Möhr (Berkeley and Los Angeles: University of California Press). Anderson, James M., 1964 "The Demarcative Function", Lingua 13.185-188. Austin, William M., 1961 "The Limits of Zero", General Linguistics 5.1-6. Avram, Andrei, 1964 "Some Thoughts on the Functional Yield of Phonemic Oppositions", Lingua 5.40-47. Bach, Emmon, 1964 An Introduction to Transformational Grammars (New York: Holt, Rinehart and Winston). Bakker, J. J. M., 1968 "Frequency in Usage and in the Lexicon", Lingua 21.13-22. Bally, Charles, 1954 Linguistique générale et linguistique française (Berne: A. Francke). Bar-Hillel, Yehoshua, 1964 Language and Information (Reading, Palo Alto, London: AddisonWesley Publishing Company). Bazell, C. E., 1949a "On the Neutralization of Syntactic Opposition", Recherihes structurales ( = Travaux du Cercle Linguistique de Copenhague), 77-86. 1949b "Syntactic Re'ations and Linguistic Typology", Cahiers Ferdinand de Saussure 8.5-20.

118

REFERENCES

Beier, Ernst G., John A. Starkweather, and Don E. Miller, 1967 "Analysis of Word Frequencies in Spoken Language of Children", Language and Speech 10.217-227. Bloch, Bernard, 1953 "Contrast", Language 29.59-62. Bloch, Bernard, and George Trager, 1942 Outline of Linguistic Analysis (Baltimore: Linguistic Society of America). Bloomfield, Leonard, 1930 "German ç and x", Le Maître Phonétique 29.27-28. 1933 Language (New York: Henry Holt & Company). 1939 "Linguistic Aspect of Science", in International Encyclopedia of Unified Science 1.26. Bolinger, Dwight L., 1948 "On Defining the Morpheme", Word 4.18-23. 1950 "Rhyme, Assonance, and Morpheme Analysis", Word 6.117-136. 1961 Generality, Gradience, and the All-or-None (The Hague: Mouton & Co). 1968 Aspects of Language (New York: Harcourt, Brace & World). Brière, Eugène J., 1968 A Psycholinguistic Study of Phonological Interference (The Hague: Mouton & Co.). Bühler, Karl, 1934 Sprachtheorie (Jena). Cantineau, J., 1955 "Le classement logique des oppositions", Word 11.1-9. Card, William, and Virginia McDavid, 1966 "English Words of Very High Frequency", College English 27.596-604. Carroll, John B., 1955 The Study of Language: A Survey of Linguistics and Related Disciplines in America (Cambridge, Mass.: Harvard University Press). 1958 "The Assessment of Phoneme Cluster Frequencies", Language 34.267-268. Catford, J. C., 1965 A Linguistic Theory of Translation : An Essay in Applied Linguistics (London: Oxford University Press). Chao, Yuen Ren, 1964 "Translation Without Machine", in Horace Lunt (ed.), Proceedings of the Ninth International Congress of Linguists (The Hague: Mouton & Co), 504-510. Chatman, Seymour, 1964 "Review of Poetics-Poetyka-Poètika", Linguistics 8.107-123. Chomsky, Noam 1965 Aspects of the Theory of Syntax (Cambridge: The M.I.T. Press). Chomsky, Noam and Morris Halle, 1965 "Some Controversial Questions in Phonological Theory", Journal of Linguistics 1.97-138.

REFERENCES

119

Clivio, Gianrenzo P., 1968 " A Note on Two Oppositions of Standard Italian with a Low Functional Yield", in Charles E. Gribble (ed.), Studies Presented to Professor Roman Jakobson by his Students (Cambridge, Mass.: Slavica Publishers), 70-75. Contreras, Heles, 1966 Review of Alphonse Juilland and E. Chang Rodriguez, Frequency Dictionary of Spanish Words, Language 42.817-821. 1969 "Simplicity, Descriptive Adequacy, and Binary Features", Language 45.1-8. DaneJ, Frantiäek, 1966 "The Relation of Centre and Periphery as a Language Universal", Travaux Linguistiques de Prague 2.9-21. Daneä, Frantiäek, and Josef Vachek, 1964 "Prague Studies in Structural Grammar Today", Travaux Linguistiques de Prague 1.21-32. De Groot, A. Willem, 1939 "Les oppositions dans les systèmes de la syntaxe et des cas", in Mélanges de linguistique offerts à Charles Bally (Genève: Georg & Cie), 107-127. Deutsch, Karl W., 1966 "On Social Communication and the Metropolis", in Alfred G. Smith (ed.), Communication and Culture (New York: Holt, Rinehart and Winston), 386-396. Dewey, Godfrey, 1923 Relative Frequency of English Speech Sounds (Cambridge, Mass. : Harvard University Press). Diver, William, 1955 "The Problem of Old Bulgarian st", Word 11.228-236. Dolezel, Lubomir, 1967 "The Prague School and the Statistical Theory of Poetic Language", • Prague Studies in Mathematical Linguistics 2.97-104. Dolezel, Lubomir, and J. Prucha, 1966 " A Statistical Law of Grapheme Combinations", Prague Studies in Mathematical Linguistics 1.33-43. Dykstra, Gerald 1956 "Perspective on the Teacher's Use of Contrast", Language Learning 6.1-6. Faure, G., 1964 "Le rôle du rendement fonctionnel dans la perception des oppositions vocaliques distinctives du français", in David Abercrombie, D. B. Fry, P. A. D. MacCarthy, N. S. Scott, and J. L. M. Trim (eds.), In Honour of Daniel Jones (London : Longmans), 320-328. Ferguson, Charles C , 1956 "Arabie Baby Talk", in Morris Halle, Horace Lunt, Hugh McLean, and Cornelis H. van Schooneveld (eds.), For Roman Jakobson (The Hague: Mouton & Co), 121-128.

120

REFERENCES

Firth, J. R., 1934 "Linguistics and the Functional Point of View", English Stuaies 16.18-24. Fowler, Murray, 1952, Review of Zellig S. Harris, Methods in Structural Linguistics, Language 28.505. Fowler, W., 1957 "Herdan's Statistical Parameter and the Frequency of English Phonemes", in E. Pulgram (ed.), Studies Presented to Joshua Whatnwugh on His Sixtieth Birthday (The Hague: Mouton & Co), 47-52. Fradis, A., L. Mihäilesco, and I. Voinesco, 1967 "L'entropie et l'energie informationnelle de la langue roumaine parlée", Revue roumaine de linguistique 12.331-339. Francescato, Giuseppe, 1962 "Notes on Relevant Features with Low Functional Yield", Lingua 11.118-127. Frei, Henri, 1950 "Zéro, vide et intermittent", Zeitschrift für Phonetik 4.161-191. Frumkina, R. M., 1963 "The Application of Statistical Methods in Linguistic Research", in O.S. Akhmanova, I.A. Mel'chuk, R.M. Frumkina, and E.V. Paducheva, Exact Methods in Linguistic Research, translation from Russian by David G. Hays and Dolores V. Möhr (Berkeley and Los Angeles: University of California Press), 80-118. Galinsky, Hans, 1964 "Stylistic Aspects of Borrowing", in Horace Lunt (ed.), Proceedings of the Ninth International Congress of Linguists (The Hague: Mouton & Co), 374-381. Garvin, Paul L., 1949 "Standard Average European and Czech", Studia Linguistica 3.65-85. Gougenheim, G., 1959 "La statistique du vocabulaire et son application dans l'enseignement des langues", Revue de renseignement supérieur 1.154-159. Gougenheim, G., R. Michéa, P. Rivenc, and A. Sauvageot, 1964 L'élaboration du français fondamental (Paris: Didier). Greenberg, Joseph H., 1957 "The Nature and Uses of Linguistic Typologies", International Journal of American Linguistics 23, 2. 1959 "A Method for Measuring Functional Yield as Applied to Tone in African Languages", Monograph Series on Languages and Linguistics 12.7-16. 1960 " A Quantitative Approach to the Morphological Typology of Language", International Journal of American Linguistics 26, 3. 1963 (ed.), Universals of Language (Cambridge: M.I.T. Press). 1966a "Language Universals", in Thomas Sebeok (ed.), Current Trends in Linguistics 3.61-112. 1966b "Synchronic and Diachronic Universals in Phonology", Language 42.508-517.

REFERENCES

121

Grégoire, André, 1948 "L'apprentissage du langage", Lingua 1.162-174. Guiraud, Pierre, 1960 Problèmes et méthodes de la statistique linguistique (Paris: Presses Universitaires de France). 1964 La sémantique (Paris: Presses Universitaires de France). Gunter, Richard, 1963 "Elliptical Sentences in American English", Lingua 12.137-150. Haas, W., 1957 "Zero in Linguistic Description", Studies in Linguistic Analysis 33.53. Hall, Robert A., Jr., 1960 "Italian [z] and the Converse of the Archiphoneme", Lingua 9.194-197. Halle, Monis, 1957 "In Defense of the Number Two", in Ernst Pulgram (ed.), Studies Presented to Joshua Whatmough on His Sixtieth Birthday (The Hague: Mouton & Co), 65-72. Hamp, Eric P., 1956 A Glossary of American Technical Linguistic Usage, 1925-1950 (Utrecht and Antwerp : Spectrum Publishers). Harris, Zellig S., 1964 "Cooccurrence and Transformation in Linguistic Structure", in Jerry A. Fodor and Jerrold J. Katz (eds.), The Structure of Language (Englewood Cliffs, N. J.: Prentice-Hall), 155-210. Haugen, Einar, 1951 "Directions in Modern Linguistics", Language 27.211-222. 1956 Bilingualism in the Americas: A Bibliography and Research Guide (Publication of the American Dialect Society 26). 1966 Discussion of John J. Gumperz, "On the Ethnology of Linguistic Change", in William Bright (ed.), Sociolinguistics—Proceedings of the UCLA Sociolinguistics Conference, 1964 (The Hague: Mouton & Co), 27-49. Havrânek, Bohuslav, and Alois JedliCka, 1963 Ceskâ Mluvnice (Praha: Stâtni Pedagogické Nakladatelstvi). Heilman, Luigi, 1964 "Statistical Considerations and Semantic Content", in Horace Lunt (ed.), Proceedings of the Ninth International Congress of Linguists (The Hague: Mouton & Co), 427-433. Helbig, Gerhard, 1968 "Zum Funktionsbegriff in der modernen Linguistik", Deutsch als Fremdsprache 5.274-287. Herdan, Gustav, 1958 "The Relation Between the Functional Burdening of Phonemes and the Frequency of Occurrence", Language and Speech 1.13. 1960 Type-Token Mathematics (The Hague : Mouton & Co). 1966 The Advanced Theory of Language as Choice and Chance (New York: Springer-Verlag).

122

REFERENCES

Hjeimslev, Louis, 1939 "Note sur les oppositions supprimables", Travaux du Cercle Linguistique de Prague 8.51-57. Hockett, Charles F., 1955 A Manual of Phonology (=Irtdiana University Publications in Anthropology and Linguistics: International Journal of American Linguistics 21, 4, Part 1). 1958 Discussion by Archibald A. Hill, Charles F. Hockett, and Harold V. King, "Some Linguistic Problems Involved in the Preparation of Teaching Materials", Language Learning, Special Issue: Linguistics and the Teaching of English as a Foreign Language, 47-77. 1966 The Quantification of Functional Load: A Linguistic Problem (Memorandum RM-5168-PR). (Santa Monica, California: The Rand Corporation). 1968 The State of the Art (The Hague: Mouton & Co). Hoenigswald, Henry M., 1960 Language Change and Linguistic Reconstruction (Chicago: The University of Chicago Press). Horâlek, K„ 1948 "La fonction de la 'structure des fonctions' de la langue", Recueil linguistique de Bratislava 1.39-48. 1964 " A propos de la théorie des oppositions binaires", in Horace Lunt (ed.), Proceedings of the Ninth International Congress of Linguists, (The Hague: Mouton & Co), 414-417. Hörne, Kibbey M., 1966 Language Typology: 19th and 20th Century Views (Washington, D. C.: Georgetown University, Institute of Languages and Linguistics). Householder, Fred W., 1946 "On the Problem of Sound and Meaning", Word 2.83. 1966 Review of Stephen Ullman, Language and Style, Language 42.632-639. IsaCenko, Alexander V., 1939 Versuch einer Typologie der slavischen Sprachen ( = Linguistica Slovaca 1). Ivic, Milka, 1965 Trends in Linguistics, translation by Muriel Heppell (The Hague: Mouton & Co). Jakobson, Roman, 1931 "Prinzipien der historischen Phonologie", Travaux du Cercle Linguistique de Prague 4.247-267. 1932 "Zur Struktur des russischen Verbums", in Charisteria V. Mathesio oblata (Prague: Cercle Linguistique de Prague), 74-83. 1939a "Le développement du langage enfantin et les cohérences correspondantes dans les langues du monde", Cinquième Congrès international des linguistes: Résumés des communications (Bruges: Imprimerie Sainte Catherine), 27-28.

REFERENCES

123

1939b "Signe zéro", in Mélanges de linguistique offerts à Charles Bally (Geneva: Georg), 141-152. 1941 Kindersprache, Aphasie und allgemeine Lautgesetze (Uppsala). 1958 "What Can Typological Studies Contribute to Historical Comparative Linguistics", Proceedings of the Eighth International Congress of Linguists (Oslo), 17-25. Jakobson, Roman, C. Gunnar M. Fant, and Morris HaUe, 1965 Preliminaries to Speech Analysis (Cambridge, Mass.: The M.I.T. Press). Jensen, Martin Kloster, 1960 Report to the Scandinavian Symposium on Statistical Linguistics. Cf. Rischel (1961), 17. 1961 Tonemicity. Cf. Rischel (1961), 17. Josselson, H. H., 1953 The Russian Word Count and Frequency Analysis of Grammatical Categories of Standard Literary Russian (Detroit: Wayne State University Press). King, Robert D., 1967a "Functional Load and Sound Change", Language 43.831-852. 1967b " A Measure of Functional Load", Studia Linguistica 21.1-14. Krâmsky, Jiri, 1959 "A Quantitative Typology of Languages", Language and Speech 2, 2. 1965 "On the Acoustic Identity of the Word", Linguistics 16.42-49. 1967 "Some Remarks on the Problem of the Phoneme", in To Honor Roman Jakobson 2.1084-1093 (The Hague: Mouton & Co). Kuôera, Henry, 1963 "Entropy, Redundancy and Functional Load in Russian and Czech", American Contributions to the Fifth International Congress of Slavists 1.191-218 (The Hague: Mouton & Co). Léon, Pierre R., 1966 Prononciation du français standard (Paris: Didier). Leopold, Werner F., 1948 "German ch", Language 24.179-180. 1953 "Patterning in Children's Language Learning", Language Learning 5.1-14. Levin, Samuel R., 1964 "Poetry and Grammaticalness", in Horace Lunt (ed.), Proceedings of the Ninth International Congress of Linguists (The Hague: Mouton & Co), 308-315. 1965 "Langue and Parole in American Linguistics", Foundations of Language 1.83-94. Lord, R., 1966 "Complex Homonymy in English Lexical and Semantic Structure", Studia Linguistica 20.35-56. Mackey, William Francis, 1966 Language Teaching Analysis (London: Longmans, Green).

124

REFERENCES

Malkiel, Yakov, 1957 A Tentative Typology of Etymological Studies (= International Journal of American Linguistics 23, 1). 1960 "Paradigmatic Resistance to Sound Change", Language 36.281-346. 1967 "Multiple Versus Simple Causation in Linguistic Change", in To Honor Roman Jakobson 2.1228-1246. 1968 Essays on Linguistic Themes (Berkeley and Los Angeles: University of California Press). Malmberg, Bertil, 1942 "A propos du système phonologique de l'italien", Acta Linguistica 3.34-43. 1963 Structural Linguistics and Human Communication (Berlin: SpringerVerlag). 1964 "Minimal Systems, Potential Distinctions, and Primitive Structures", in Horace Lunt (ed.), Proceedings of the Ninth International Congress of Linguists (The Hague: Mouton & Co), 78-84. Manczak, Witold, 1959 "Fréquence d'emploi des occlusives labiales, dentales et vélaires", Bulletin de la Société Linguistique de Paris 54.208-214. 1968 "Le développement phonétique irrégulier dû à la fréquence en russe", Lingua 21.287-293. Marouzeau, J., 1961 Lexique de la terminologie linguistique (Paris: Librairie orientaliste Paul Geuthner). Martinet, André, 1945 La prononciation du français contemporain (Paris: Librairie E. Droz). 1955 Economie des changements phonétiques (Berne: A. Francke). 1960 Eléments de linguistique générale (Paris: Librairie Armand Colin). 1962 A Functional View of Language (Oxford: Clarendon Press). Mathesius, Vilém, 1911 " O potenciâlnosti jevù jazykovych," translation, "On the Potentiality of the Phenomena of Language", in Josef Vachek (ed.), A Prague School Reader in Linguistics (Bloomington: Indiana University Press, 1964), 1-32. 1931 "Zum Problem der Belastungs- und Kombinationsfähigkeit der Phoneme", Travaux du Cercle Linguistique de Prague 4.148-152. Meyerstein, Goldie Piroch, 1966 "Bilingualism Among American Slovaks: Analysis of Loans", Publication of the American Dialect Society 46.1-19. Meyerstein, R. S., 1964a "Informant Morphemes Versus Analyst Morphemes", in Horace Lunt (ed.), Proceedings of the Ninth International Congress of Linguists (The Hague: Mouton & Co), 562-569. 1964b "Functional Parallelism in Descriptive and Applied Linguistics", General Linguistics 6.1-10. 1967 "Communicational Inequality of 'Like' Forms", Linguistics 37.73-88. 1968 "Peak Value Forms", Linguistics 46.21-46.

REFERENCES

125

Mikus, Francis, 1947 "Le syntagme est-il binaire?", Word 3.32-38. Miller, G. A., 1951 Language and Communication (New York: McGraw-Hill). Mok, Q. I. M., 1966 "Le rôle de la liaison en français moderne", Lingua 16.27-39. Moulton, William G., 1947 "Juncture in Modem Standard German", Language 23.212-226. Newton, B. E., 1963 "Patterns of Sound Change in Greek," Lingua 12.151-164. Olmsted, David L., 1951 "Covert (or Zero) Morphemes and Morphemic Juncture", International Journal of American Linguistics 17.163-166. Osgood, Charles E., and Thomas A. Sebeok (eds.), 1965 Psycholinguistics : A Survey of Theory and Research Problems (Bloomington). Paducheva, E. V., 1963 "Information Theory and the Study of Language", in O.S. Akhmanova, I.A. Mel'chuk, R.M. Frumkina, and E.V. Paducheva, Exact Methods in Linguistic Research, 119-179; translation from Russian by David G. Hays and Dolores V. Möhr (Berkeley and Los Angeles: University of California Press). Pap, Leo, 1949 Portuguese-American Speech (New York). Pavlovii, M., 1957 "Principe de corrélation et typologie linguistique", Zbornik za filologiju i lingvistiku 1. Pei, Mario, 1966 Glossary of Linguistic Terminology (Garden City, N.Y. : Doubleday & Company). Pike, Kenneth L., 1967 Language in Relation to a Unified Theory of the Structure of Human Behavior (The Hague: Mouton & Co). Pike, Kenneth L., and Milton Warkentin, 1961 "Huave: A Study in Syntactic Tone with Low Lexical Functional Yield", A William Cameron Townsend, Mexico, 627-642. Posner, Rebecca, 1968 Review of Robert P. Stockwell and J. Donald Bowen, The Sounds of English and Spanish, Journal of Linguistics 4.141-142. Pottier, Bernard, 1948 Communication in Actes du Sixième Congrès international des linguistes: Rapports sur les questions historiques et pratiques mises à Vordre du jour (Paris), 87. Prieto, Luis J., 1954 "Traits oppositionnels et traits contrastifs", Word 10.43-59. "Projet de terminologie phonologique standardisée", Travaux du Cercle Linguistique de Prague 4.309-323 (1931).

126

REFERENCES

Pulgram, Ernst, 1967 "Trends and Predictions", in To Honor Roman Jakobson 2.1634-1649. Ray, Punya Sloka, 1963 Language Standardization: Studies in Prescriptive Linguistics (The Hague: Mouton & Co). Rigault, André, 1962 "Rôle de la fréquence, de l'intensité et de la durée vocaliques dans la perception de l'accent en français", Proceedings of the Fourth International Congress of Phonetic Sciences, 735-748. Rischel, Jorgen, 1961 "On Functional Load in Phonemics", Statistical Methods in Linguistics 1.13-23. 1964 "Stress, Juncture and Syllabification in Phonemic Description", in Horace Lunt (ed.), Proceedings of the Ninth International Congress of Linguists (The Hague: Mouton & Co), 85-93. Roberts, A. Hood, 1965 A Statistical Linguistic Analysis of American English (The Hague: Mouton & Co). Saporta, S., 1955 "Frequency of Consonant Clusters", Language 31.25-30. Saporta, S., and Jarvis R. Bastian (eds.), 1961 Psycholinguistics (New York). Saussure, Ferdinand de, 1915 Cours de linguistique générale (Paris : Payot). Seiler, Hansjakob, 1967 "On Paradigmatic and Syntagmatic Similarity", Lingua 18.35-79. Shapiro, Michael, 1967 "Concatenators and Russian Derivational Morphology", General Linguistics 7.50-66. Shen, Yao, 1959 "Some Allophones Can Be Important", Language Learning 9.7-18. Stockwell, Robert P., and J. Donald Bowen, 1965 The Sounds of English and Spanish (Chicago: The University of Chicago Press). Stockwell, Robert P., J. Donald Bowen, and John W. Martin, 1965 The Grammatical Structures of English and Spanish (Chicago: The University of Chicago Piess). Thatcher, James W., and William S.-Y. Wang, 1962 The Measurement of Functional Load ( = Report 8) (Ann Arbor: University of Michigan Communication Sciences Laboratory). Togeby, Knud, 1952 "Le faible rendement des oppositions phonologiques et grammaticales en français", Bulletin de la Société de Linguistique de Paris 68.33-34. Trnka, B., 1931 "Bemerkungen zur Homonymie", Travaux du Cercle Linguistique de Prague 4.152-156. Trubetzkoy, N. S., 1932 "Charakter und Methode der systematischen phonologischen

REFERENCES

127

Darstellung einer gegebenen Sprache", in Proceedings of the [First] International Congress of Phonetic Sciences, 18-22. 1939 Grundzüge der Phonologie (= Travaux du Cercle Linguistique de Prague 7). Quotations refer to the translation by J. Cantineau, Principes de phonologie (Paris: Librairie C. Klincksieck, 1949). Uspenskij, B. A., 1965 Strukturnaja Tipologija Jazykov (Moscow: Izdatel'stvo Nauka). Vachek, Josef, 1960 Dictionnaire de linguistique de l'École de Prague (Utrecht: Spectrum). 1966a The Linguistic School of Prague (Bloomington: Indiana University Press). 1966b "Prague Phonological Studies Today", Travaux Linguistiques de Prague 1.7-20. Vendler, Zeno, 1968 Adjectives and Nominalizations (The Hague: Mouton & Co). Vinay, J.-P., and J. Darbelnet, 1958 Stylistique comparée du français et de Vanglais (Paris: Didier). Voegelin, C. F., 1955 "On Developing New Typologies and Revising Old Ones", Southwestern Journal of Anthropology 11.355-360. Wang, William S.-Y., 1967 "The Measurement of Functional Load", Phonetica 16.36-54. 1969 "Competing Changes as a Cause of Residue", Language 45.9-25. Weinreich, Uriel, 1953 Languages in Contact ( = Publications of the Linguistic Circle of New York 1) (New York). 1958 " A Retrograde Sound Shift in the Guise of a Survival", Estructuralismo e historia: Misceláneo homenaje a André Martinet 2.221-267 (La Laguna). Whatmough, J., 1956 Language: A Modern Synthesis (New York: St. Martin's). Whorf, Benjamin Lee, 1941 "The Relation of Habitual Thought and Behavior to Language", Collected Papers on Metalinguistics (Washington, D.C. : Department of State, Foreign Service Institute), 25-46. Windfuhr, Gemot, 1967 "Strukturelle Verschiebung: Konjunktiv Präsens und Imperativ im heutigen Deutsch", Linguistics 36.84-99. Yule, G. U., 1939 "On Sentence Length as a Statistical Characteristic of Style in Prose: With Application to Two Cases of Disputed Authorship", Biometrika 30.363.

INDEX

2:1, 59, 69-72, 76-110 1:0, 72 A:B, 74 a: b, 63, 76-87, 104-105 a : m , 63, 87-104 ab:a, 63-64, 104-110 Abbreviation, 32, 91, 106-107 Abernathy, R., 23 Aberration, 13-14, 58, 79, 104, 109, 115; see also Well Formed Actualization, 44, 54-55, 91, 101, 110 Addition (increase), 37, 42, 54, 69, 78, 88, 101-102, 107 African Languages, 21-22 Age, 46, 92, 95 Aginsky, B. and E., 88 Akhmanova, O., see Frumkina, R. Alternation (allomorph, allophone), 22, 29, 39, 49-54, 57, 79-82, 86-88, 101; see also Emic vs. Etic Ambiguity, 63, 67, 87, 102, 113, 115 American Indian Languages, 21-22 Anderson, J., 50 Application, 13-14, 17-18, 28-29, 32, 34, 38, 41, 44-46, 51-52, 59, 67, 69, 78-79, 98-99, 109-110, 114-115; see also Instruction Area, 62, 64-65, 115 Austin, W., 36 Availability, 14, 46-47, 51, 64 Avram, A., 17 Bach, E., 113

Bakker, J., 34 Bally, C , 17 Bar-Hillel, Y., 49 Basic, 48-49, 51, 53-54, 88, 102 Bastian, J., 109 Bazell, C„ 38, 61 Beier, E., 46 Binary (bilateral, 29-38, 46, 58-63, 70, 73, 76-110; see also Pair Unit Bisystemic (bilingual, intersystemic; vi. monosystemic), 18, 29, 51, 54-60, 63-65, 71-110, 115; see also System Bloch, B., 53, 54 Bloomfield, L„ 17, 29, 40, 43, 45, 52-54, 57, 62-63, 89 Bolinger, D., 17, 65, 89, 91 Bowen, J., 14, 16, 33, 48, 99; see also Posner, R. Briere, E., 80, 109 BS, see Bisystemic Bühler, K., 43-44 Card, W., 34 Carroll, J., 34, 46 Category (case, gender, mood, number), 46, 67-69, 84-85, 96-97, 99, 101-102, 104, 108 Cause and Effect (consequences, explanation), 13-14, 17-26, 28, 46-47, 74, 76-77, 114-115 Ceiling vi. Floor, 114 Chao, Y., 46 Chatman, S., 14, 22

INDEX Child Language, 72-73, 94, 104, 106, 109 Chomsky, N., 54-55 Cleavage, 63, 85, 92; see also Homonymy, Set Clivio, G., 17 Cognate, 96 Combinability, 31-32, 42, 53, 83, 93, 101 Communication, 16, 24, 43-44, 49, 64, 77, 79, 86 Compatibility, 26-27 Competence vs. Performance, 50-51, 54-60, 81, 86-87, 91, 112 Complexity vj. Simplicity, 46, 77, 92-93, 116 Component ( = part of a description), 37, 60-63 Componential ( = constituent), 22-23, 38-39, 43, 58, 63, 72, 85, 89-91, 106, 109-110 Cona,tive, 43-44 Connotation vs. Denotation, 41-43, 52, 60, 64, 71-72, 81, 91-92, 102, 106-108 Contreras, H., 33, 48 Correctness (admission), 78-79, 98-104, 108-110, 115; see also Aberration Coverage, 14, 46-47, 64 Cut-Off Point, 51, 88 Cz, see Czech Czech, 22, 31-32, 42, 50, 52, 63-64, 68, 75, 78, 93-95, 107-108 DaneS, F., 17, 38, 88 Danish, 22 Darbelnet, J., 90 De Groot, A., 38 Deictive, 91 Demarcative, 50-51 Dependence (bound), 53-54, 92 Descending (decline; vs. increase), 106-107, 112 Determinant (conditioning, variable), 35, 37, 45-47, 51-54, 65, 71, 76-110, 112-113 Deutsch, K., 74

129

Development (birth, creation, trend), 13, 19, 77, 86-87, 100-104 Dewey, G., 46 Diachronic vs. Synchronic, 13, 17-25, 27-28, 39, 42, 45-46, 48, 56, 60, 64-65, 69, 74, 77-78, 84, 91-94, 98-104, 106-107, 115 Diagnostic, 78-79 Differential, 61, 66, 69-70, 115-116 Differentiator, 31, 45, 83-85, 96 Direction, 24, 83-84, 89-90, 99-101, 107 Distinctive Features, 15-17, 22-25, 27, 36-37, 39, 51-52, 63, 69, 76-77, 79-81, 85-88, 113 Distribution (final, initial, position), 22-23, 25-26, 29, 31-32, 34-37, 40, 45-54, 56, 67-68, 79, 82-92, 98, 101, 103-105, 109-110 Diver, W., 23 Dolezel, L., 14, 17, 34, 38, 56 Dykstra, G., 53, 86 Economy (efficiency), 32, 93 Emic vs. Etic, 29, 50-52, 54, 101 En, see English English, 16, 20, 22, 31-32, 34, 36-37, 42, 48, 51-61, 63, 71-72, 74-75, 77, 79, 84-85, 87-100, 102-104, 106-109, 112-114 Entropy (uncertainty), 21-23, 26-27, 30,49 Exolinguistic (emotional, intellectual), 41-44, 50, 53, 64, 68, 81, 86-88, 91-93, 97-99, 103, 105-110 Expressive, 43-44, 72, 107-108 Faction, 82, 85-86 Faure, G., 17 Ferguson, C., 94 Finite, 26, 105 Firth, J., 44 Fluctuation (stability), 56, 64, 84, 103-104 Formative, 63 Fowler, M., 45 Fowler, W., 46 Fr, see French

130 Francescato, G., 17 Frei, H., 36 French, 15-16, 20, 30-32, 38, 42, 50-53, 57, 59-60, 71, 73, 75, 77-79, 81-86, 88, 90, 94-99, 101-104, 109-111, 113-114 Frequency (incidence), 14, 16-18, 21-25, 29, 31-37, 45-59, 61, 63-65, 67-69, 71-72, 78-80, 82-83, 85, 87-90, 92-108, 112-115 Frequency Hypothesis, 24 Frumkina, R„ 14, 46, 48 Function, 19, 29, 31, 37, 40-46, 49-54, 57-59, 61-65, 67, 69, 72-73, 76-85, 88-97, 99-108, 111-112, 114-116 Functional Load : definitions, 15-17, 29-31, 33, 38-39, 111, 115-116 equivalent terms, 13 extent, 15-116 importance previously stated, 13-14 Functionalist, 43-45, 50 Galinsky, H., 100 Garvin, P., 61 Ge, see German Geographic, 59-60, 94-95, 106 German, 31-32, 51, 56-57, 65-69, 71-73, 75, 78-81, 86-87, 95-103, 109, 112, 114, 116 Germanic, 23, 25, 32 Gougenheim, G., 14, 46 Graphic, 14, 38, 72, 91, 106-107, 115 Greenberg, J., 14, 16, 20-22, 31, 34, 46, 49, 55-56, 61, 88, 102, 105-106 Grégoire, A., 104, 110 Grenzsignal, 50, 52, 83, 89 Guiraud, P., 46, 88, 92, 102 Gumperz, J., see Haugen, E. Gunter, R., 92 Haas, W., 36 HaU, R., 81-82 Halle, M., 33 Hamp, E., 17 Harris, Z., 91 Haugen, E„ 17, 45, 74,100

INDEX

Havrânek, B., 68 Heilman, L., 46, 92 Helbig, G., 44 Herdan, G., 17, 34, 46, 88 Hierarchy (rank), 43-44, 47-51, 65-66, 78, 103, 107 Hill, A., see Hockett, C. Hjelmslev, L., 33 Hockett, C„ 13, 14, 16, 17, 19, 20, 21-23, 28, 31, 34, 39, 40, 46, 49, 51-52, 54, 86, 114 Hoenigswald, H., 20, 101, 102, 103 Holophrastic, 92 Homonymy (homoneme), 15, 41-42, 63, 67-68, 86-87, 92, 99, 102 Horälek, K., 33 Home, K., 61 Householder, F., 91, 109 Identification, 18, 29-63, 97, 103, 107, 112 Idiolectal (individual), 58, 85-87, 95 Idiomatic, 98-99 Immigrant Languages, 73-74, 93, 100 Impressionism, 17-18, 25-26 Inclusion, 111 Index (gauge), 26, 45, 47-49, 52, 97, 109 Informant vî. Analyst, 80-81, 111, 113 Information, 16, 33, 39, 43, 49-50, 72, 109 Inherent, 34, 89-90 Innovations (extensions, objectives of this study), 28-29, 35, 37-38, 40, 45, 47, 50, 52-54, 58-116 Instruction (acquisition, learnability), 14, 34, 46-47, 51-52, 54, 59, 64, 66-67, 72-73, 78-79, 85-88, 99, 101, 109, 113-115 Intelligibility, 59 Intent (motivation), 50, 86-87, 92,109 Interference (negative transfer), 94, 98-100; see also Under-Differentiation Intuition, 48, 113 Invariance, 57, 68, 112 Isacenko, A., 61 It, see Italian

131

INDEX

Italian, 75, 81-82 Ivic, M., 17 Jakobson, R., 17, 19, 36, 61, 93, 104, 105, 108, 109 Jedlicka, A., 68 Jensen, M., 22 JosseJson, H., 46 Judgment (belief, folklore, subjective), 56, 86-87, 93, 95, 97-98, 103, 113 Juilland, A., see Contreras, H. King, H., see Hockett, C. King, R., 13, 17, 23-25, 31, 39, 46, 47, 49, 57, 65-67, 77, 79-80, 112, 114 Krâmsky, J., 61, 83 Kucera, H., 13, 14, 22, 39, 46, 49, 78, 93 La, see Latin Latin, 71, 74, 84, 94, 97-98, 101-103 Latitude, 86 Least Resistance Hypothesis, 24 Length, 32, 42, 46, 50, 52, 63, 77-78, 86, 92, 104-110, 116 Léon, P., 90 Leopold, W., 57, 104, 109 Level, 29, 38-41, 44-45, 62, 72-110 Levin, S„ 54, 93, 113 Lexicon (vocabulary), 30, 42, 45-46, 58, 63-64, 68, 71-73, 78-79, 83, 85-86, 94-95, 98, 108-109, 113 Lg, 74-75, 101-103, 105-106, 108-110 Limitations, 28-61, 64, 67, 71-73, 115 Load Point, 51, 58-59, 61, 68-72, 76-110, 114-116; see also Overload Loadability, 32 Loanwords (borrowing), 41-42, 60, 64, 98-100, 104, 107 Lord, R., 63 Mackey, W., 13, 14, 46, 49, 56, 74, 79, 88, 93, 94 Malkiel, Y„ 19, 61, 89, 102 Malmberg, B., 17, 56, 84, 102 Manczak, W., 46

Mark, 20, 35-36, 51, 58-59, 64, 89-90, 97, 104-110 Marouzeau, J., 17 Martin, J., 14, 16, 48, 99 Martinet, A., 14, 16, 17, 19, 20, 31, 33, 38, 42, 44, 46, 50, 51, 52, 53, 56, 73, 77, 81, 82, 83, 85, 89, 91, 104, 109-110 Mathesius, V., 15, 32, 34, 35, 38, 41, 56, 73, 108 McDavid, V., see Card, W. Meaning (content, semantic), 15, 19, 32, 37, 40-47, 51-52, 56-58, 61, 64, 67, 71, 81, 83, 87, 89, 92-93, 95-97, 99-100, 102, 106-108, 111112, 114 Medical, 92, 99, 105 Mel'chuk, I., see Frumkina, R. Merger, 20, 24 Meyerstein, G., 74, 93 Meyerstein, R., 45, 51, 55, 70, 72, 80, 84, 89, 107, 111 Michia, R., see Gougenheim, G. Mikus, F., 33 Miller, D., see Beier, E. Miller, G„ 88 Mok, Q., 90 Monofunctional, 54, 94 Morphocentric, 63, 89 Morphologization, 82, 104 Morphosyntactic (grammatical, morphological, syntactic, supramorphemic, supraphonological), 14-15, 29, 31-32, 38-40, 42-44, 46, 52, 57-58, 61-64, 68-69, 71-73, 78-113, 115 Moulton, W., 57 MS = Monosystemic, see Bisystemic ms, see Morphosyntactic Multilateral, 32-33 Necessity, 46-47 Negation, 58-60 Neutralization, 51 Newton, B., 17 Norm, 14 Object vs. Subject, 91-92, 108

132 Objective (purpose), 14, 28, 43, 51, 61, 82, 86, 111

Obligatory vs. Optional, 51-52, 84, 87, 91, 93, 108 Olmsted, D., 36 Opponent, 33, 36-37, 49, 63-65, 70, 73, 76-110,116; see also Opposition Opposition (contrast), 15-17, 19-25, 27, 29-35, 37-39, 41-42, 45, 47, 50-53, 55-59, 62-66, 69-73, 76-110, 112-114, 116 Order (precedence), 69, 74, 105 Osgood, C., 109 Overlap, 91-92, 95-96 Overload, 49, 58-59, 62-72, 74,76-110, 112-113, 116

INDEX Prediction, 18-21, 23-26, 50, 52, 77, 79, 81, 115 Predominance, 31-32, 43-44, 46-47, 60, 64, 68, 76, 78, 90, 97, 101, 107, 109, 113, 116 Premorphemic, 89; see also Morphosyntactic, Submorphemic Preservation (duration, resistance, survival; vì. instability, loss), 20, 23-24, 42, 56, 64-65, 77-78, 82, 84, 92-93, 101-106; see also Diachronie Prieto, L., 53 Principal Variant, 53-54 Procedure (method), 14, 18, 21-23, 28, 33-34, 36, 40, 56, 60, 62, 73, 88, 112

Paducheva, E., 49; see also Frumkina, R. Pair Unit, 63, 69, 95, 104-105, 116 Pap, L., 100 Paradigmatic (replacive), 30, 53, 55-56, 62-63, 67-69, 71-72, 83-84, 86-87, 92, 94-96, 105-106; see also Syntagmatic Parallelism, 51, 77, 81, 92 Pathology, 100-101, 115 Pavlovic, M., 61 Pei, M., 17 Perception v.r. Production, 13, 34, 70, 76, 78, 87 Performance Uncertainty, 57 Phonologization, 82, 87 Pike, K., 33, 38, 56 Po, see Portuguese Poetry, 14, 43-44 Polarity, 87-88 Popularity, 32, 46-47, 92, 94, 97, 109-110 Portuguese, 75, 99-100 Posner, R., 48 Potential (candidate), 19, 24, 41-42, 45, 49-50, 54-56, 58, 61, 63, 65, 67-69, 73, 77, 79, 82-85, 87-92, 99-101, 105-107, 110, 115 Pottier, B., 83 Practical, 42, 51, 58, 66, 69, 71, 111 Prague School, 33, 36, 43-45, 53

Productivity, 63, 65, 100-103 Progression (degree), 47, 51-52, 65-66, 69-71, 88, 103, 107-109, 112-113, 116 Projet de terminologie phonologique standardisée, 15, 32, 34, 37, 38, 41 Prùcha, J., 17, 38, 56 Pseudo-Language, 101 Psycholinguistic, 109; see also Exolinguistic Pulgram, E., 73 Quantification (computation, measure), 13-32, 34-35, 37-40, 45-47, 49, 52, 55-58, 61, 65-71, 80, 82-83, 90,95-96,99,105, 107, 109, 112-116 Range (domain, scope), 14, 28, 37, 41, 44, 46-47, 53-54, 57, 61-63, 71-73, 76, 85-86, 95-96, 109, 116 Ray, P., 17 Rc, see Romance Languages Redundancy, 22, 67, 96, 108 Reduplication, 94 Referential, 43-44 Reflexive, 57, 97 Reinforcement, 69, 99 Related Fields, 30, 36, 50, 69-70, 77 Reliability (validity), 23, 27-28, 56, 76-77 Repetition, 108

INDEX Representative (implementation), 54, 65, 83, 86, 90, 92, 106-107 Restrictability, 87-88 Revival, 103 Rigault, A., 13, 34 Rischel, J., 22, 31, 34, 46, 49, 89 Rivenc, P., see Gougenheim, G. Roberts, A., 34 Rodriguez, E., see Contreras, H. Romance Languages, 75, 97 Ru, see Russian Russian, 22, 63-64, 75, 107-108 Sampling, 26,55-57,62,65-66,76,112 Saporta, S., 46, 109 Saussure, F. de, 54-55 Sauvageot, A., see Gougenheim, G. Sc, see Slavic Languages Sebeok, T„ 109 Seiler, H., 68, 105 Set (including Form Class), 25-26, 42, 46, 52, 56-57, 63, 65-68, 73-74, 82, 84-86, 90, 92, 95, 101, 103, 105, 108, 114-116; see also System Shape, 46, 56, 65-66, 92 Shapiro, M., 17, 42 Share, 16, 31, 56, 93, 115 Shen, Y„ 52 Shift, 94, 96-99, 101 Significative vi. Constitutive, 64-65, 82-84, 89-91, 104, 112, 114 Similarity, 99 Simultaneous, 105-106 Situation, 14, 45, 56, 64, 74-75, 86-87, 93, 97-98 SI, see Slovak Slavic Languages, 32, 75, 78, 97; see also Czech, Russian, Slovak Slovak, 75, 93-94, 100 sm, see Submorphemic Sn, see Situation Social (occupational), 59-60, 94-95 Sp, see Spanish Spanish, 16, 33, 48, 51-52, 54, 75, 96-97, 99, 101-103, 114 Speaker-Hearer, 34, 54-55, 63-64, 69-70, 80-81 Specificity (precision), 55, 61, 65-67,

133 69-71, 74, 109 Sporadic, 91 Spread, 51, 58, 71, 79-81, 95,106, 109 Stage, 23, 27, 74, 94, 102-104 Standard, 58-59, 64-65, 105-106, 109-110 Starkweather, J., see Beier, E. Stimulus (incentive), 86-87, 100 Stockwell, R., 14, 16, 33, 48, 99; see also Posner, R. Structure, 15, 31-32, 34, 39, 41, 46, 51, 58-59, 92, 111 Style, 46, 50, 52, 56, 58-59, 68, 85-86, 92, 94-95, 98, 106-110, 112, 115 Submorphemic (phonological), 13-18, 20-27, 29-47, 50-57, 63, 65-66, 69-70, 72, 76-91, 97, 102-106, 110, 112-115 Subphonemic, 52 Substitution (replacement), 37, 55, 57, 65, 69, 92 Symbolism, 90-91 Syntagmatic (cooccurrence, strings), 26, 31-32, 38-39, 46, 53, 62-63, 69, 71-72, 74, 83-84, 89-90, 92-93, 95, 97, 103-110, 114, 116 System (language, langue; vs. parole), 14-15, 21, 24, 26, 29, 41-43, 46, 51-60, 62-63, 65, 69, 73-110 Ternary (trimodal), 33 Thatcher, J., 26 Togeby, K., 38 Trager, G. 54 Transformation (generative), 48, 5455, 59, 64-65, 74, 91, 99 Transparence, 109 Trnka, B„ 32, 38, 42, 78 Trubetzkoy, N., 13, 15, 33-35, 38, 41, 78 Typology (token, type), 16, 32-33, 38-39, 44, 47, 50, 59-63, 65-73, 76-110 Ullman, S., see Householder, F. Under-Differentiation vs. Over-Differentiation, 51-52, 74 Unique (unilateral), 29-35, 40-45,

134

INDEX

54-60, 86-87, 116 Unitary (unit), 15-17, 29-35, 37-38, 45-46, 50, 52, 62, 69, 73, 86-87, 89-92, 98, 105, 111, 116 Universal, 14, 74, 88, 102 Usage, 48, 82, 97 Useful, 18, 20, 26, 28, 35, 42, 46-47, 107, 111, 113-115; see also Application Uspenskij, B., 61 Utilization, 15, 30-37, 41, 64, 111, 113-115 Vachek, J., 15, 17, 34, 36, 38, 43, 44, 45, 50, 52, 53, 64, 93, 108 Values, 37, 46, 54, 65, 71, 96, 112 Vendler, Z., 17, 42 Vinay, J., 90 Voegelin, C., 61

Wang, W., 13, 14, 17, 19, 26, 27, 30, 31, 35, 39, 46, 49 Warkentin, M., 38, 56 Weak Point Hypothesis, 24 Weinreich, U., 74,102 Well Formed (grammatical), 58-59, 93, 113, 115 Whatmough, J., 17 Whorf, B., 61 Windfuhr, G. 93 Work Load (effort), 17, 26, 30, 39, 69-71, 76-78, 88, 105, 110, 114; see also Load Point, Overload Yule, G., 14 Zero vs. Plus Feature, 29-30, 35-38, 49-54, 57, 59, 63, 68, 72, 77, 79-81, 86, 88-90, 92, 96-98, 101, 103-104, 108-109, 112