A Phonology of Italian in a Generative Grammar 9783110810417, 9789027907370


253 4 3MB

English Pages 96 Year 1970

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Abbreviations
1. Studies of Italian Pronunciation
2. Problems in Italian Phonology
2.1. Italian [z], [ε], [ɔ], and Partial Complementation
2.2. Length or Stress?
2.3. Assimilation and Ellipsis
2.4. Constituent Stress
3. A Phonology of Italian
3.1. The Phonological System
3.2. Phonological Redundancy
3.3. Precyclic Rules
3.4. Phonological Cycle
3.5. Postcyclic Rules
Bibliography
Recommend Papers

A Phonology of Italian in a Generative Grammar
 9783110810417, 9789027907370

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

A PHONOLOGY OF ITALIAN IN A GENERATIVE GRAMMAR

JANUA LINGUARUM STUDIA M E M O R I A E N I C O L A I VAN W I J K DEDICATA edenda curat

C.H. VAN S C H O O N E V E L D INDIANA

UNIVERSITY

S E R I E S PRACTICA 93

1970

MOUTON THE H A G U E · PARIS

A PHONOLOGY OF ITALIAN IN A GENERATIVE GRAMMAR by

MARIO SALTARELLI

1970

MOUTON THE H A G U E · 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: 77-95010

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

PREFACE

This book is a slightly altered version of my doctoral dissertation. The general aim is a description of the sound system of a variety of modern Italian within the framework of generative-transformational grammar. The writer's major thesis is the following: vowel length rather than stress is the 'linguistically significant' feature in Italian. Such a hypothesis stands in controversy with former structural descriptions, which in general uphold the converse. The content of the book is specifically directed to show that the primacy of length over stress is well supported on empirical as well as theoretical grounds. Phonetic and perceptual data is documented in classic works by Panconcelli-Calzia, Josselyn, Parmenter, and Porena. Generalizations analogous to those formulated here can be found more or less overtly stated in traditional descriptions from Bembo to Merlo. Of particular theoretical relevance are the specific generalizations that the proposed length hypothesis leads us to as regards major phonological phenomena like word and phrase stress, vowel and consonant length, rafforzamento, diphthongs, assimilation and ellipsis. Although the description is strictly synchronic, one can appreciate its diachronic implications by considering, for example, that / öiv+täd+e / is the underlying representation for cittd 'city' and / d i k + t + o / for detto 'said'. It is important to note that the claims made in this analysis of Italian are independent of the particular notation used. They are however crucially positioned on the following guidelines. There is no explicit procedure at present for obtaining THE grammar of a language; any such grammar is a hypothesis about the language, subject to paradigmatic alternatives. The excellence of a metatheory is measured best in terms of the 'linguistically significant' generalizations it leads us to. Montecatini August, 1969

M A R I O SALTARELLI

TABLE OF CONTENTS

Abbreviations

9

1. Studies of Italian Pronunciation

11

2. Problems in Italian Phonology

21

2.1. Italian [ζ], [ε], [o], and Partial Complementation

21

2.2. Length or Stress?

26

2.3. Assimilation and Ellipsis

30

2.4. Constituent Stress

31

3. A Phonology of Italian

36

3.1. The Phonological System

36

3.2. Phonological Redundancy

45

3.2.1. Morpheme Structure Rules 3.3. Precyclic Rules 3.3.1. Inflection of Nouns 3.3.1.1. Inflection of Adjectives

47 63 63 66

3.3.2. Inflection of the Verb

67

3.3.3. Word Formation

74

3.3.4. Alternations

77

3.3.4.1. Palatalization 3.4. Phonological Cycle

77 78

8

TABLE OF CONTENTS

3.4.1. Glides

80

3.4.2. Constituent Stress

81

3.4.2.1. Primary Stress

81

3.4.3. 'Dittonghi Mobili'

83

3.4.4. Affrication

83

3.4.5. Ellipsis

84

3.4.5.1. Vowel Ellipsis

84

3.4.5.2. Consonant Ellipsis

85

3.4.6. Consonant Length and 'Rafforzamento'

86

3.4.7. Apocope

86

3.4.8. Glide Residue

87

3.4.9. Secondary Stress

87

3.4.10. Phrase Stress

88

3.4.11. Cyclic Convention Rules

89

3.5. Postcyclic Rules Bibliography

89 93

ABBREVIATIONS

Adj: AFF: C: C.: comp: ens: cnt: dif: E: env.: f.: Fem: grv: Imperf.: impf.: Ital.: L: Lat.: Lex.: Igt: m.: Masc.: MS: n.: N: nas: Neutral.: NP:

adjective affix consonant cycle compactness consonantality continuancy diffuseness ending environment feminine feminine gravity imperfect imperfect Italian liquid Latin lexical (entry) length masculine masculine morpheme structure noun noun, nasal nasality neutralizing noun phrase

nsl: O.L.: P: Pf: ph: pi.: Plur.: p. pi.: PREF: Prep.: Pres.: Pst: PS: S: SF: shp: Sing.: SS: strid: Subj.: SUF: tns: TNS: TV: V: voc: voi: W, Χ, Υ, Z:

nasality Old Latin phonological prefix phonetic plural plural past participle prefix preposition present past primary stress sentence suffix sharpness singular secondary stress stridency subjunctive suffix tenseness tense thematic vowel vowel vocality voice string or segment

1. STUDIES OF ITALIAN PRONUNCIATION

This is intended as an introduction to the study of Italian pronunciation. It covers briefly the XVI century descriptivists, the XIX century historical phonologists, the phoneticians, phonemics and normative grammar. Although the material is intended to be comprehensive, critical attention is paid to works of insightful support to the proposal made in this dissertation. Pietro Bembo's (1549) description of the volgar lingua represents a precise account of the language of his time. His linguistic insights about certain phonological problems have only partly been acknowledged in modern studies of Italian pronunciation. Two topics are of greatest interest to us: stress and raddoppiamento (or rafforzamento). The discussion of the latter transcends most accounts of the phenomenon, in the writer's opinion. Bembo's description of the seven stressed vowels identifies their articulatory relationships as any modern phonetician would. He recognizes that syllables may be long or short according to either the number of segments which form them, stress, or both. For each word in this language the syllable on which the stress falls is long, the other syllables are short. The importance of stress and length is illustrated with reference to poetry. Normally stress can fall on the antepenult, penult and ultimate. If the stress is on the antepenult a verse has one extra syllable, but the two syllables which follow the stress are so light that one could say they count as one. On the other hand if stress falls on the ultimate a verse cannot have more than ten syllables, which means that the syllable on which the stress falls is so heavy to count as two syllables. The penultimate stressed syllable is the medium heavy. Heavy are also stressed monosyllables; in fact they count as two syllables at the end of a verse. Italian stress cannot fall on short syllables. If it is put on a short syllable this counts as a long one. Bembo notices first that certain prepositions cause doubling of the initial consonant in the following word. He deduces this by observing verbs like accogliere which he analyzes as a+cogliere. This explanation of the rafforzamento is a purely phonetic one rather than a syntactic one, which is the modern explanation. Bembo argues that when the stress falls on the last syllable of a verse it is so strong that this counts as two

12

STUDIES OF ITALIAN PRONUNCIATION

syllables. Thus, when another word is added, the strength of the final syllable is reflected in the doubling of the initial consonant of the added word, so that the syllable may become fuller. This seems to imply that syllables for Bembo must meet certain phonetic requirements of length, substantially influenced by stress. Bembo explains certain apparent exceptions, in a manner similar to that followed by Bianchi (1949) and hailed as a discovery by Merlo (1949). Bembo explains that in Dante's quetdmi, levdmi, direlo the rafforzamento does not take place in the initial vowel of the pronouns # m i # and # l o # because the expressions are reduced forms of quetdimi, levdimi, direilo in which the stress does not fall on the last vowel. Fiorelli (1957) recognizes striking similarities between Castellani's (1956) 'phonemic' analysis of Italian (one of the few attempts by Italian linguists) and the XVI century analyses of Tuscan speech done by Trissino, Tolomei, Giambullari, Rhoesus, and especially by Giorgio Bartoli in his Degli Elementi del Parlar Toscano (Firenze, 1584). Fiorelli says that Bartoli's elementi are in substance Castellani's fonotipi or allophones. The phonemes coincide with the alphabetic spelling reforms proposed especially by Trissino, such as ε and ω for open midvowels, ν for consonantic u, ζ for ti,f for intervocalic voiced [z], etc. Of these reforms, tending toward a rational orthography, ν and ζ have come down in today's Italian alphabet. Bartoli gives a full and coherent inventory of the allophones of Tuscan. Bartoli enumerates 35 elementi, while Castellani gives 32. However, while the latter considers ρ/φ and t/θ as possible phoneme pairs, the former definitely decides that they are each variants of a single phoneme. While Castellani considers [y, w] as phonemes, Bartoli identifies them as variants of the vowels /i/ and /u/. Bartoli groups all preconsonantal nasals as variations of the same elemento. Vanzon's Grammatica Ragionata della Lingua Italiana (Livorno, 1834), is one of the few important grammatical works since those of the XVI century, yet hardly their match. In the chapter entitled ortologia (revealing its normative intention) the author demonstrates a clear intuition of the notion underlying alphabetic writing. Vanzon says that the human voice is capable of more articulations than are needed for the communication of ideas; in order to keep his mind from confusion man had to limit the variations of human voice to those that were strictly necessary for language, and to prescribe the elements, which are called letters or characters. In keeping with the normative intention serious attempt is made to give rules of pronunciation. The greatest effort is always made, in the attempt, to instruct people in how to read the letters e and o. When Trissino (1528) proposed additional symbols for the alternate pronunciation of these letters, he had probably entertained the idea that their pronunciation cannot be fully predicted by general rule. Although in his own writings he used ε and ω, the reform did not take hold. This seems to have led orthologists in the XVII, XVIII, and XIX centuries, and even today, to formulate rules which are neither fully, nor even approximately, general in certain pronunciations of Italian. These rules are always lengthy, so much so that they can be considered simply lists. The bases of these rules vary but they rely mostly on morphological knowledge.

STUDIES OF ITALIAN PRONUNCIATION

13

The phonetic knowledge of the XVI century descriptivists has no echo in Vanzon's description of letters/segments. They are described as follows: 'mute' B, C, D, G, P, Τ, Z; 'semivowels' F, L, Μ, N, R, S; 'dentals' C(i), D, G(i), S, Τ, Z; etc. They are said to be 'mute' because their name begins in a consonant, 'semivowels' because their name begins in a vowel, and 'dentals' because they are pronounced with the help of the teeth. Scientific precision was achieved again in the study of sounds during the second half of the XIX century. It is particularly with the rise of phonetics in these studies that phonological work on Italian developed. Once again in keeping with the works of XIX century linguistic science the trend is away from a normative attitude. Now Demattio (1875) defines phonology as "that science which STUDIES and ANALYZES the various combinations of the prime elements, that is, basic sounds of words". Following the trend the treatment is carried out diacronically, but straight forward and without any attempt to explain the 'why'. Demattio observes that the change of accent from one syllable to another brings about important phonetic changes. Stressed vowels are constrained to fixed rules or 'phonetic laws', in order to establish them however we must distinguish them according to quantity. Liquids are treated according to the traditional usage. They include 1, r, m, n. The same phonetic analysis, in a historical light, is given by Stoppato (1897) whose phonetic research is meticulous. He recognizes 19 vowel qualities. His sounds inventory presents a major dichotomy between voiced/voiceless sounds. Consonants are subdivided into continuant/explosive. Buscaino-Campo (1873) in his Regole makes a serious but unsuccessful attempt to discover the 'laws' for the pronunciation of [ε]/[ε], [s]/[z], [dz]/[ts], etc. His confidence in the regularity of phonetic laws and the failure to discover them for segments which we now call phonemes are settled by means of a compromise in which he explains that "the laws by which pronunciation is governed are very deeply set, as the laws of all things in nature; indeed their discovery is not an easy undertaking. Sometimes we must be satisfied only with the knowledge of the facts, although they may appear at times imcomprehensible". Josselyn's Etude sur la phonetique italienne (Paris, 1900), is still today an important undertaking in the study of Italian sounds. This work was carried out with the Rousselot equipment for phonetic research. The principal points of this study are the following. Sounds thought to be identical present marked differences even for speakers of the same region and dialect. The same sound is often pronounced in variation by the same person. In the two series of Italian sounds, beside the open and closed vowels, there is an intermediate series not recognized before. The difference in articulation between voiced and voiceless sounds is not only in the point of articulation, but it extends to the entire articulatory tract. The distinction between voiceless and voiced sounds is in the pressure as well as in the position of articulation. Among the stops one can distinguish four classes, taking as basis the time relation between the beginning of the vibrations in the larynx and their appearance in the breath. One finds a weakening of occlusion in p, b, t, d, analogous to that of k, g which is more marked.

14

STUDIES OF ITALIAN PRONUNCIATION

δ and g are sounds consisting of two parts: a prepalatal stop which is never t or d and a fricative which is never sh. The quality of m and η is often indicated only by the nasality of the following vowel. Μ and η after a vowel can take a form intermediate between the vowel and the consonant. This is the first step of an evolution which has been completed in French. The palatal consonant is a simple consonant and is completely distinct from consonant plus j. Except for a few cases, there is only one diphthong in Italian when the 'accent' of the word falls on either one of two contiguous vowels. There are no triphthongs in Italian. There are three degrees of nasality which are quite distinct, and which occur not only under the influence of nasal consonants, but also in cases in which such influence cannot be felt. The double consonant is a single consonant, more intense and longer in all of its elements. Under the influence of a double consonant, the preceding vowel loses from one fourth to one third of its length. Panconcelli-Calzia contributed to the physiological and physical study of Italian sounds. Nasalization (1904) and variations in interrogative sentences (1921) are some of the topics experimentally examined. A summary presentation of his work to date appears in his Italiano (1911). There the phonetics and morphology of the language are accompanied by phonetic transcriptions of texts: from Dante to Manzoni. There are four characteristics of sounds which he proposes to dusty: quality, quantity, pitch and intensity. They are examined first physiologically and then physically. In the introduction he regrets that for typographical reasons he had to adopt the sign for stress before the syllable, although he preferred to put it on the stressed vowel. In fact, he says, that the concept of 'syllable', does not exist physiologically. In his treatment he studies sounds not only in isolation but also in groups (words), phrases, and sentences. With every regrouping of sounds he studies the changes which occur. Commenting on length he points out that the duration of isolated sounds is always relative; that is, a sound is more or less short only with respect to another sound. Length undergoes relevant variations as two or more sounds are grouped together. The process goes from the vowels to the consonants and viceversa. If a vowel is pronounced long, the following consonant will be relatively short. The converse occurs if a vowel occurs before a consonant characteristically long or before two or more consonants. In the latter case the second consonant must not be r or 1. Then the vowel loses about one third of its length. The rafforzamento is explained as a purely phonetic phenomenon, yet modern phonologists define it as a 'syntactic' phenomenon. The rafforzamento for Panconcelli-Calzia consists in the influence of a word-final vowel on the length of the initial consonant of the following word when they are in the same phonetic group. It is a condition for this phenomenon that the final vowel be stressed. Porena (1942) adds more phonetic distinctions to the articulatory description of consonantal sounds. He observes that 1, m, and η have a different phonetic composition in the following two positions V V (ala) and V C {alba). In the first position the sound can be analyzed into two components: an occlusive and a continuous

STUDIES OF ITALIAN PRONUNCIATION

15

one. In the second position only the continuous component is present. In voiced stops b, d, g, g, gy the sound consists also of two elements, a 'preparatory' element and an occlusion. The preparatory element is voicing in this group. In the voiceless stops the preparatory element is just a pause. He has an interesting explanation for double consonants and their origin. Phonetically they are lengthening of simple consonants. The origin of lengthening in Latin and in Italian is the compensation for the loss of a consonant by strengthening the remaining one in the group. Old Latin conligo > Latin colligo, O.L. adnecto > L. annecto, L. factus > It.fato. The change is explained as the loss of n, d, c respectively and the lengthening of the remaining consonant to preserve the length of the syllable. Porena refers to an earlier (1908) observation which apparently went unnoticed. He makes a distinction between what he calls 'grammatical syllable' and 'metric syllable'. The first is the syllable in the ordinary sense. The second one comprises the interval between the beginning of a vowel and that of the next vowel. This is the only interval which the ear can perceive as a well-defined measure. The word amore is in syllables -am-or-e. -e does not constitute a metric syllable, but in running speech it will form a syllable with the initial consonant of a following word: #-am-or-e#f-ed-el e # . It is evident that Porena's 'metric' syllable does not coincide with the 'grammatical' syllable. The phenomenon called rafforzamento or strengthening of word initial consonants is discussed in treatises of pronunciation as early as the XVI century. Both Bembo (1594) and Tolomei (1547) observed and discussed extensively this problem. Norman (1936) offers a comprehensive summary in English: Josselyn (1900), Meyer-Llibke (1890), Panconcelli-Calzia (1911), and Metz (1914). These authors all seem to agree on the fact that it is a phenomenon in which stress and length play the major role. Camilli (1941), who had examined the subject earlier, doubts or disregards the generality of the phonetic explanation and favors the syntactic alternative. He says that when two words are adjacent, one ending in a vowel and the other beginning in a consonant, very often the initial consonant is reinforced. However these two words must be 'united' by the meaning and the pronunciation. The first of the two words is said to be reinforcing. Accordingly he gives a list of monosyllables which govern the rafforzamento ('strong' monosyllables) and a list of monosyllables which do not cause rafforzamento ('weak' monosyllables). Apocopated plurisyllables are treated in a similar manner. He points out that in spite of the continuity of pronunciation this phenomenon is blocked by the lengthening of a stressed vowel or difference in intonation. Merlo (1949) praises Bianchi's article (1947) on the explanation of irregularities in Florentine rafforzamento, when he shows that the lack of the phenomenon in certain cases where one would expect it may be an indication of ellipsis. (We mentioned that Bembo (1549) proposed a similar explanation for apparent exceptions.) Merlo took this paper to be "a lesson to the many who still doubt the regularity of phonetic facts", echoing Ascoli's teaching on the regularity of sound laws.

16

STUDIES OF ITALIAN PRONUNCIATION

In a more recent article Fiorelli (1958) recognizes Tolomei's (1547) observation of the phenomenon but favors a syntactically-based definition. For Fiorelli it is a phenomenon of syntactic-phonetic characteristic of Italian: given word-initial consonants are reinforced when they follow words ending in a vowel, in the context of a "phrase". Fiorelli points out a further kind of rafforzamento: that which occurs in the position C # V , the converse of the more common type. For example: we have tram but [trameletriko], gas but gassoso [gasoso]. Italian intonation has been discussed in articles by Parmenter and Trevino (1930) and by Camilli (1950 and 1956). Stimulated by the work of Panconcelli-Calzia, Parmenter and Trevino carried out an experiment at their University of Chicago laboratory which led to the following conclusions about Italian intonation. Posttonic -a, which in isolated words has the lowest pitch, has its fall shortened by the presence of a following initial unstressed vowel. The chief modification in intonation in passing from isolated words to phonic groups occurs in syllables that are unstressed in the group. In relation to stressed syllables, monosyllables do not rise so high, and unstressed syllables do not fall so low. The pitch intervals between successive vowels have a sufficient range to make them easily perceived. From the graphs drawn from the experiment it is concluded that there are as many peaks in the group as there are stressed syllables. The pitch peaks coincide with the stress peaks. The difference between primary stress and neighboring syllables is large. The distinction in pitch between a pretonic and a tonic vowel is greater than that between a tonic and posttonic one, except when the posttonic is low and final. Stressed syllables tend to maintain a high pitch. Nowhere does the informant place two successive stressed syllables on the same pitch. Camilli (1950) presents three types of sentences fundamentally differentiated by intonation: (1) indicative (Sei stato a scuola.), (2) interrogative (Sei stato a scuolcft), and (3) exclamative (Sei stato a scuolal). (1) indicative sentences has two subtypes: (a) enunciativa and (b) emotiva. It seems that type (1) (b) has completely replaced (1) (a) in certain dialects. Camilli (1956) distinguished syllables into strong (F), medium (S), and weak (D). If one counts from right to left the syllables which precede the F syllable, then the even numbered will be S and the odd numbered will be D, or ... S, D, S, D, S, D, F .... However DSD may become SDD either to interrupt the monotony, or by influence of F or S syllables in derivatives: precipitoso. The posttonic syllables have the following pattern: F: F D: F D S: F D D S:

faro fa-ro a-ni-ma fab-bri-ca-no

Notice the absence of the following sequences: FF, SS, FS, SF. They are always

STUDIES OF ITALIAN PRONUNCIATION

17

separated by one or two D. This permits an analysis of Italian intonation by meter (FD, FDD, SD, SDD): notice also that the initial measure can be DFD, DFDD, DSDD, and the final can be F, S. The phonetic phrase has the same rhythm as the single word. It must be pointed out however that the F of the final word is always higher than those in the preceding words. In cases in which the sequence F ' # F " , occurs, F' becomes D ( D D # F " in closed syllable) or S(DS#F"). Examples: fuggi svelto fuggi svelto fuggir via

FD#FD DS#FD D S # F D (or DD # FD)

Quantity has been examined extensively by Josselyn (1900), again by Parmenter and Carman (1933), and by Camilli (1954). On Josselyn's and Parmenter and Carman's contributions we shall comment in Chapter 2. Along with Panconcelli-Calzia (1911) and Metz (1914) we find an insistence on the sensitivity of the Italian ear to mistakes in this aspect of pronunciation. This very sensitivity seems to indicate that a distinction in quantity constitutes an outstanding characteristic of the Italian language. Parmenter and Carman conclude that the combination long vowel plus consonant is shorter in duration than the combination short vowel plus long consonant in the proportion of 4 to 5. A long vowel is about twice as long as a following consonant. A short vowel is about one fourth shorter than a long vowel. Double consonants are longer than a preceding vowel by one third to one half. Double consonants are about twice as long as single consonants. Camilli (1954) views the relation of stressed vowel to the following consonantal segment in a different perspective. He says that a stressed vowel is longer in open than in closed syllables. In Italian quantity is always predictable from the type of syllable. As a consequence two words never differ only by length. Quantity is not distinctive in Italian. It was distinctive in Latin. The opposite is true of stress. When we read the words capita 'it happens' and capita 'understood' the important thing for us is only where we put the stress: antepenultimate and penultimate respectively. Camilli seems however to overlook the fact that Italian orthography does not mark the relative length of segments. If it did the two words would be respectively written as ca:pita and capi:ta, or the like, and they would differ also in stress and length. It is plausible to say that length is predictable from the type of syllable and stress. It seems equally plausible to say that stress is predictable from length. In fact, for these two words, given only length, we can assign stress by general rule: i.e. in a word a long vowel receives the stress. The preference of stress or length as to phonological primacy must be motivated. Given the pairs [fäto] and [fäto] we see first of all that the stress is not necessarily the distinctive feature. One could say that the distinction is t/t, on which the distinction ä/a depends; but it is just as reasonable (assuming this is all we can say about the two items) to state the opposite: they differ by ä/a on which the distinction t/t is based. In other words on

18

STUDIES OF ITALIAN PRONUNCIATION

purely superficial grounds one can subordinate vowel length to consonant length or viceversa; likewise one can subordinate vowel length to stress or viceversa. Only deeper considerations or new observations can motivate a choice. By far the most complete treatment of accentuation, both phonological and graphic, is given by Malagoli (1946). 'Expiratory', or 'dynamic', stress is the topic, as opposed to musical, or intonational, stress. On orthographic grounds six stress patterns are identified. Chapter 2 contains normative rules for stress, roughly on a morphological basis. Particular attention is paid to words of Greek provenience and their patterns in Italian. In conclusion lists of accentually ambiguous words are given, for which the author marks the 'correct' stress. The lack of theoretical sophistication is balanced by the thoroughness of treatment. The topic of elisione e troncamento, or ellipsis and apocopation, has stimulated less response. Malagoli (1913), Peruzzi (1960), and Leone (1963) have paid particular attention to this phenomenon. According to Leone the "phenomenon of syntactic phonetics in which the ending of a word is modified before another word, in a situation of semantic relationship" is one of the most confused chapters of normative grammars. It is well understood that ellipsis occurs before a word beginning in a vowel (quelVamico) and that apocopation occurs before a word beginning in a consonant (quel giovane). What is not understood is why apocopation occurs also before a vowel (buon amico, un amico). Leone brings up an observation of a common spelling 'mistake' among young Italian students. He asks: when they write un'amico (instead of the required un amico) are they really doing wrong? Or are they following a regularity about our language improperly represented in the writing conventions? The author observes that both apocopation and ellipsis occur only when the two words are part of the same syntactic phrase (in fact: nessun uomo but nessuno e ...). Furthermore ellipsis seems to be the most advanced stage of sinalefe. In conclusion the final vowel of a word is never apocopated but elided before the initial vowel of the following word. Thus forms like un uomo, buon uomo, qual eta, Suor Agnese, are instances of ellipsis and not of apocopation, as the lack of apostrophy would suggest. We can no longer say that both ellipsis and apocopation occur before a vowel. It seems that the orthographic apostrophy is not used in cases of ellipsis when they coincide with apocopated forms. Italian e's and o's have been always recognized as having an open and closed quality under stress. The fact that the orthography does not account for this distinction prompted spelling reforms, such as Trissino's (1528), or attempts to find general rules for their pronunciation. However, the problem of the quality of unstressed e's and o's has been vaguely approached. Most grammars fail to recognize this problem. Others imply that they are always closed: Grandgent (1924), Panconcelli-Calzia (1911), D'Ovidio and Meyer-Liibke (1890) etc. Others notice differences in quality: Grandgent (1927), Camilli (no date) and Josselyn (1904). Davis (1937) wrote an article on the study of the phonetic character of unstressed e's and o's. The author detects five

STUDIES OF ITALIAN PRONUNCIATION

19

phonetic degrees of quality for e's and o's (see Chapter 2 for details and theoretical discussion of midvowels also by Trager [1939] and Hall [I960]). Phonemics, as it is generally understood in America, is relatively emphasized. A phonemic analysis of Italian, was done by Porru (1939) of the Prague school and Hall (1944), (1948) of the American view. Camilli (1947) published perhaps the most complete study of Italian pronunciation to this date. As a phonetician Camilli brings together most of what is known of Italian phonetics. His study includes a table of possible vocalic and consonantal combinations, or phonotactics, as well as intonation, quantity, and dialect variations. A brief presentation of the phonemes is given. Castellani (1956) published one of the most complete systematic analyses of Italian phonemes done by an Italian. The linguist gives first a sketch of his techniques and then applies them to yield the sistema fonotipico e fonematico italiano. Castellani adheres basically to the proceduralistic view of American phonemics. However, he proliferates the terminology according to the kind of phonetic features primarily involved in the distinction: fonema, dinema, cronema, and tonema. According to Fiorelli (1957) Castellani's phonology is "the science of the sounds of human languages, in the larger sense", whereas 'phonematics' is structural phonetics. In Italian [ε], [ο], [ζ], [j], [w] are problematic, [ε] and [o] are generally accepted as phonemes on the basis of the Florentine system, [z] and [s] are considered by Porru (1939) as variants of /s/; [z] is normal intervocalically. Malmberg (1942-43) insists on two phonemes /s/ and /z/ on the basis of minimal pairs such as fuso (n.) [fiiso] and fuso (p. pi.) [fiizo] in Florentine. Migliorini (1945) includes /z/, Hall (1944) (1948) has no /z/, /j/, /w/. Castellani (1956) includes /z/, /j/, /w/. Arce (1962) assumes that Tuscan has seven vocalic phonemes in stressed syllables, and five in unstressed syllables. In literary Italian [s] and [z] are separate phonemes. Double consonants are to be interpreted as separate phonemes but semi-consonants are only phonetic variants of the vowels /i/ and /u/. American contributions to Italian phonology are represented by Hall, Trager, and Castiglione. Trager's (1939) comment on Italian vowels will be discussed below (Chapter 2). Castiglione's (1957) Italian Phonetics, Diction and Intonation presents the phonetics of Italian articulatorily and acoustically. The author presents what is generally known about Italian phonetics to serve as basis for pronunciation drills. Hall (1948) gives a complete descriptive study of modern Italian. The book has chapters on phonemics, inflection, derivation, phrase structure, and clause structure. The approach is structuralistic. The phonology follows closely the post-Bloomfieldian orientation. Revised versions of this work are in Hall (1959), (1960), (1961). Of particular theoretical interest is Hall's (1960) "Italian [z] and the Converse of the Archiphoneme" (see Chapter 2.1.). Normative responsibility is strongly felt among linguists in Italy. Polemics on the

20

STUDIES OF ITALIAN PRONUNCIATION

subject are common in scholarly publications. Fiorelli (1957) searches for the true basis of a linguistic norm. He argues that a scientific approach to the problem still remains to be discovered. However more advanced normative linguistics should direct the speaker to a conscious decision of what is the best choice. Ever since the sixteenth century, normative tradition has been based excessively on historical factors which have a purely external effect on the development of the language, such as the authority of writers and grammarians, or regional and social influence. Today, according to Fiorelli, we prefer objective structural, historical, etymological and analogical criteria. These criteria should inspire and guide the choice of pronunciation in each region. Fiorelli suggests Migliorini's Pronunzia Fiorentina ο Pronunzia Romana (1945) as the herald of innovating ideas on norms of Italian pronunciation. Fiorelli (1951), engaging in polemics with Camilli about the pronunciation on nationwide radio broadcasts, defends the trend toward Florentine features such as intervocalic [z] instead of [s]. Pointing out the lack of uniformity for intervocalic sibilants in the various regions of Italy, Fiorelli states that radio must impose among rival norms an intermediate one which is in accordance with the historical phonetics of literary Italian (and which happens to be his Florentine variety). Camilli retorts that the choice among rival norms should be based on taste rather than logic, esthetics, historical phonetics or the like. After World War II there has been copious production of orthophonic guides. This eems to indicate a feeling for a unitary pronunciation of Italian. However, while in 939 "a solution meeting Mussolini's wish and consistent with the 'foundations of the mpire' could be adopted without hesitation," Migliorini (1949) asks what would now be the basis for a choice. A survey of grammars for secondary schools reveals a more unorthodox approach to grammatical analysis. This is particularly evident in the lack of systematicity in the description and in the view about language held by current philosophical doctrines. Manna (1951) in his II Puro e Dolce Idioma states that phonology is subdivided into ortoepia and ortografia which 'teach' how to pronounce and how to write correctly. For Manna vowels are those letters which have a sound of their own, which can be pronounced by themselves. Consonants on the other hand acquire their sound in connection with another vowel. Battaglia (1951) writes that interest in grammar has diminished in the last fifty years as a just reaction to a schematic and mechanical mentality that deals with the facts of language as if they were outside the never-ceasing creative source which determines and sustains them. He explains that in the pronunciation of vowels people seem not to make as great a physical effort as when they pronounce consonants. For the pronunciation of the latter we make a physical effort and one feels a certain opposition to be overcome or maintained. Panzini (1953) writes that ellipsis and apocopation occur in order to avoid cacophony or to ensure euphony. Similar statements are found in practically all school grammars. It seems to be the aim of Italian school grammars to explain a philosophical 'why' rather than a descriptive 'how'. But their answers to 'why' are too shallow, in a linguist's opinion.

2. PROBLEMS IN ITALIAN PHONOLOGY

Whereas Chapter 1 served as an introduction to studies on Italian pronunciation, Chapter 2 is designed to pave the way for the grammatical formulations proposed in Chapter 3. Four topics are discussed here. In section one a distinctive feature analysis is proposed in the light of partial complementation and the solutions suggested by the 'Prague', 'American', and 'semi-componentiaF view. Section two discusses the factual bases of length and stress and the case for the phonemic primacy of length over stress. Section three reconsiders assimilation as ellipsis on the basis of internal motivations. And section four is a detailed statement on primary and secondary stress and the rules which predict it. 2.1. ITALIAN [ζ], [ε], [ο], AND PARTIAL COMPLEMENTATION

In this section we discuss various phonological analyses with particular attention to the handling of partially complementary segments. Specifically we shall present the solutions given by the Prague school, the American view and R. A. Hall's 'semi- componential' analysis; finally the problem is considered in terms of a 'distinctive feature' representation, and the results are evaluated. The phonetic bases of partial complementation in this language involve the following segments: (1) [ε],[ο], [e], [o], [s], [z] In modern 'standard' Italian [ε] and [o] are in contrast with [e] and [o] only in stressed position, [s] and [z] are in complementary distribution everywhere; however intervocalically these two sounds seem to be acquiring the status of contrast for some speakers of central Italy. Thus we have here a double example of partial complementation. The distinction [ε] [ο] is phonemic if the vowel is stressed, but dependent on the following segment if unstressed. Likewise [z] and [s] seem to be in contrast intervocalically, but clearly in complementary distribution elsewhere, [z] occurs only before voiced consonant, e.g. sdentato [zdentdto] 'toothless', and [s] elsewhere, e.g. stentato [stentato] 'forced', salato [saläto] 'salted', lapis [lapis] 'pencil'.

22

PROBLEMS IN ITALIAN PHONOLOGY

Italian linguists in general believed that mid-vowels were always close in unstressed position. The problem of partial complementation arose when Ε. B. Davis (1937) reported that unstressed mid-vowels are not always phonetically close, but that certain phonetic laws determine their degree of openness. In his investigation degrees of openness are established on an increasing scale from 1 to 5, where the lowest degree is identified with [e] and [o] and the highest with [ε] and [o]. The results can be abbreviated as follows: (2) (i) J[e]4 {[ο]3 I (ii) f[e]2 j[o] 2 /

LC

i.e., unstressed [e] and [o] have degree 4 and 3 of openness respectively if they are in a syllable closed by a liquid. They have degree 2 elsewhere. G. L. Trager (1939) commented on this fact and proposed a phonemic analysis which reflects the American view. Italian has seven vowel phonemes; in the case of the correlative pairs e, ο and ε, a ... the presence of phonemic distinction of openness in stressed syllable makes it possible to assign the unstressed vowel also to one or the other phoneme. Accordingly a phonemic orthography should distinguish the open e and open ο also in unstressed syllable. This analysis imposed a requirement of biuniqueness on the phonemic transcription; that is, there must be a one-to-one correspondence between the phonemic transcription and the physical sounds, and in addition phonetic [ε] and [o] must be identified with their respective phonemes even in unstressed position where such difference is not distinctive, on the basis of the principle that "once a phoneme, always a phoneme". This 'strong' biuniqueness requirement was thought to be indispensable for linguistic analysis. But even on its own grounds Trager's analysis fails in that there is no criterion for assigning [o] 3 either to the close phoneme /o/ or to the open phoneme /o/. In the variety with intervocalic contrast the application of this analysis to the phonemicization of [s] and [z] is likewise inadequate. In fact, our phonemic transcription would have to represent the words stentato and sdentato as /stentäto/ and /zdentäto/ respectively in accordance with the 'strong' requirement of biuniqueness. The result is that we have two phonemes in complementary distribution in certain environments for having to represent a difference where it is not distinctive. Therefore Trager's analysis would violate the principle of complementary distribution, which is at the very basis of phonemic analysis. Robert A. Hall takes up again the issue of partial complementation in Italian in a further effort to reach a feasible solution. I refer to his article in Lingua (June 1960), entitled "Italian [z] and the Converse of the Archiphoneme". Interpreting C. F. Hockett's (1955) discussion of the problem he rejects both the American solution (which incidently was adopted in his Descriptive Italian Grammar) and that of the

PROBLEMS IN ITALIAN PHONOLOGY

23

Prague school (which would involve setting up an archiphoneme, in neutralized position, in addition to the phonemes /s/ and /z/ intervocalically). Hall (1960) decides that: The 'Prague-type' archiphoneme has the disadvantage of adding extra units to one's analysis ... The 'American' procedure, on the other hand, is wasteful in that it insists on symbolizing differences where they are not significant ... and proposes: A more fruitful procedure is to establish the phonemic contrasts which are valid in all positions in the language under analysis, and then regard cases of further partial differentiation under any given set of circumstances as involving the addition of a further phonological component, which is significant only in those special circumstances and not elsewhere. Thus, if Italian [s] and [z] come to be in contrast only intervocalically, we can solve the problem best by recognizing only one phoneme ... and then, between vowels, we recognize an added contrast between voice and voicelessness ... by adding a dot underneath our symbol /s/ to indicate the additional component of voicing. Hall proposes the same solution for e and ο by adding an additional mark of openness under stress. For this procedure to apply, however, the 'strong' requirement of biuniqueness must be abandoned and the result is a 'semi-componential transcription'. This procedure is identified by the author himself with 'rephonemicization' (see Hall, Leave Your Language Alone, pp. 87-88 and "The Reconstruction of Proto-Romance", p. 306) as discussed by Z. S. Harris in Structural Linguistics (1951), p. 90. Harris writes: Rephonemicization breaks up some segments into two elements, each of which is assigned to a different phoneme. The effect is to regularize the distribution of phonemes. Purpose: eliminating exceptional distributional limitations, or more exactly to increase the freedom of occurrence of exceptionally restricted phonemes. We would like to eliminate some of these exceptional restrictions not by modifying our operational definition of a phoneme (7.5), nor by changing the criteria which we seek to satisfy, but by performing a further operation, if possible, on the restricted segments in order to make them amenable to those phonemic groupings which would satisfy our preference. An operational definition of phoneme requires only a 'weak' principle of biuniqueness on phonemic writing, in Harris' (1951, p. 72) views. The criteria which are to be satisfied and maintained and for which this further operation of 'rephonemicization' is motivated are (1) number and freedom of occurrence of phonemes, (2) symmetry in the representation of sounds, (3) relative and complete phoneme stock, and (4) symmetry of environment (Harris, 1951, p. 62). Let us now observe more closely Hall's analysis in comparison with the American view and see if the former's motivation for 'rephonemicization' is justified in view of the scope and place in a linguistic analysis of this remedial operation, as discussed by Harris. The American view would set up two phonemes: the consonants /s/ and /z/ which occur in the following environments;

24

PROBLEMS IN ITALIAN PHONOLOGY

# (3) (i) Ν

/

V C (voiceless) V

V

# C (voiced)

(ii) /ζ/ / I

Hall sets up also two contrasts as follows: /s/ and / . / (the latter is just the feature voice), which occur in these environments: # (4) (i) Μ I

(ii) / . / /

V C (voiceless and voiced) ν ν # V _ s _ V (only intervocalically; never as a single segment, but always in connection with /s/).

In this case Hall's analysis results in no reduction in the number of contrasts. In addition symmetry of environment is reduced, which is the very opposite of the purpose for applying 'rephonemicization'. Finally we have a phonemic unit /. / whose environment is shared by no other phoneme in the language. In the case of the vowels the American analysis would be as follows: (5) (i) /ε,ο/ / ί _ 1 _ (under stress) [ +liquid -f-C (unstressed, closed by liquid) (ii) /e, of I j [

'

(under stress) (unstressed, not closed by liquid)

Hall analyzes as follows: (6) (i) /e, ο/ I L l _ (stressed) [ (unstressed) (ii) 1^1 / e^jS (only under stress; never as a single segment, but always in connection with e,o). In this case we have one phoneme fewer but again we pay in symmetry of environment, which is contrary to the purpose of 'rephonemicization'. Furthermore we have a phonemic unit /"/ which is maximally limited in occurrence. In view of the above we maintain that with regard to the phonemic system of Italian as a whole Hall's 'semi-componentiaF analysis of [s], [ζ]; [ε], [ο], [e], [ο] is not motivated. Further it should be avoided since it defeats its own purpose by introducing highly restricted phonemes and consequently reduces symmetry of environment in the system. However, one might argue that even though Hall's analysis is not a desirable one when we consider the phonemic system as a whole, still it is more suited than former

PROBLEMS IN ITALIAN PHONOLOGY

25

analyses to the particular purpose of solving the problem of partial complementation in Italian. The argument seems to hold on its own, but in comparison with the American solution does not offer any advantage, in that to achieve its purpose the price which is to be paid in freedom of occurrence of phonemes and in environmental symmetry is very high. A more fruitful approach would be one which meets the problem without compromising on the validity of the analysis in some other way. Accordingly we propose a distinctive feature analysis in accordance with the tenets presented by M. Halle for a phonological component in a generative grammar, according to which a set of ordered rules predicts morphologically and phonologically governed features. Thus, in the case of Italian /s/ /z/, we regard these two segments as bundles of distinctive classificatory features, distinct because the feature voice is — (minus) for one and + (plus) for the other. Accordingly we would have in our analysis the following morpheme-structure rules:

—cons] —comp (ii) (B)

—cons -dif —comp —stress

where a variable coefficient has been used for voice, tenseness and vocality. Rule (A) is to be read: /s/ is voiced or voiceless if followed by a voiced or voiceless consonantal segment respectively. It is voiceless finally and initially before vowels. Rule (B) reads: unstressed /e/ and joj are closed or open depending on whether they are followed by a cluster the first member of which is a liquid or a non-liquid. These rules assign binary values of voicing, and closeness. Later indices of closeness, etc. are assigned by rule to the abstract distinctive features to reproduce the exact physical qualities of the sounds. Our transcription will contain no coefficient for the features voice and closeness where they are environmentally predicted by the rules above. Clearly this phonemic writing coincides with the level of analysis traditionally known as the morphophonemic level, which is here followed by an ordered set of rules mapping it directly into a phonetic representation. Linguistic theories agree on the necessity of both levels of analysis. None imposes a biuniqueness requirement (neither 'weak' nor 'strong') on the former but everyone does for the latter, as we do. The results show that our analysis of Italian [s], [ζ], [ε], [ο] accounts for exactly the same phonetic facts with which other theories dealt. In addition we do not add extra phonological units in our phonemic writing as the Prague solution did; we do not have two phonemes in complementary distribution which Trager and the American

26

PROBLEMS IN ITALIAN PHONOLOGY

school would need in Italian; nor do we reduce environmental symmetry among our phonological elements, maximally restricted in the case of voice /. / and openness as in Hall's 'semi-componentiaP analysis. On the contrary, our phonological elements (the classificatory features) on which the set of rules operates acquire extended freedom of occurrence and maximal possible symmetry (balanced with the rules) in our phonemic system. We conclude that partial complementation is automatically handled in our analysis. It can only become a problem if we insist on a biunique phonemic-type representation, rather than on a feature representation.

2.2. LENGTH OR STRESS?

Phonological studies generally describe stress and consonant tenseness as functional features in Italian. Vowel length is considered as a feature concomitant with stress and predictable from it. Some linguists analyze contrasts which distinguish pairs like fata, fatta and casa, cassa as single vs. geminate consonant, positing a syllable boundary thus: /fa-ta/ /fat-ta/ and /ka-sa/ /kas-sa/ (Hall, 1948). From the point of view of a distinctive-feature representation, we would discard the geminate analysis in favor of tense/lax (or long/short) on the strength of feature economy in the transcription. In fact, this contrast is apparently functional only in intervocalic position, predictable initially and never occurring finally. In the case of jij the intervocalic representation would be (on a given phonological tree, e.g., p. 40) for the 'geminate' analysis: —voc —strid —grave —comp —voice "—voc —strid — grave —voice —comp

-voc -strid -grave -comp -voice

(2)

/ W

—voc —strid —grave —comp —voice +tense

whereas for the 'tense/lax' analysis: 3

=

—voc —strid — grave —comp —voice —tense

While the 'geminate' representation uses 15 features, the 'tense-lax' representation uses only 12. And the occurrence of the alternate segment(s) is about the same. This means a saving of 3 features multiplied by the number of occurrences.

PROBLEMS IN ITALIAN PHONOLOGY

27

Thus a distinctive-feature analysis would distinguish the following segment types, as things stand now: ttppklcd3bbnnmmssffvv6SlIrfggn§c^gIeeiio0uÜ60aä with a tense/lax contrast for the first 28 consonantal segments and a stress contrast for the vowels. Such a description would accordingly contain rules for predicting vocalic length: (3) (i) (A) [—cons] -* (ϋ) (B)

[—length]

elsewhere

The rules read: a vowel is long if it occurs within a word and it is in a free syllable and stressed. Elsewhere a vowel is short. Length, however, is a more systematic phenomenon in Italian. Kymograph tracings done by Josselyn (1900) and by Parmenter and Carmen (1932) show a 'duration rhythm', a proportional difference in length between a stressed vowel and the following consonantal segment(s). Josselyn's results are as follows: V (3) V + C {pane) 26 V + C 2 (panni) 17

C 14 24

(where the numbers represent hundredths of a second); i.e., stressed vowels are long before single consonants, short before long consonants. Parmenter and Carman gave the following results (for words in isolation): (4) V: + C(pane) = 1 . 8 + 1 = 2.8 V + C: (panni) 1.4 + 2.2 Ϊ6 (where the length of a single consonant is the unit). The formula for words in sentences was about the same as above. The individual duration was as follows: (5) long single short double

V: C V C2

20 11 15 25

In the case of carta, casa, cassa ([kärta], [käsa], [kässa]) we can say, according to the conclusions given by the investigators, that:

28

PROBLEMS IN ITALIAN PHONOLOGY

(1) a long vowel is twice as long as a following consonant, and (2) double consonants are about twice as long as single consonants. On the basis of these facts, we formulate a rule in our description to account economically for the phenomenon generally called 'syntactic rafforzamento', or strengthening of word-initial consonants when preceded by certain words. Grammars usually give lists of words which govern this occurrence. Such lists contain words stressed on the final vowel or stressed monosyllables. For example: [pärlo kyäro] but [parl5 kkyäro] and [ä kkäsa] but [la käsa]

Ί speak clearly' 'he spoke clearly' 'to the house' 'the house'

The rule given earlier for predicting vowel length states that word-final vowels are always short. Consequently when this word-final short vowel comes to be in the continuous physical stream of speech sounds it will require that the immediately following non-vocalic segment(s) be twice its duration, in accordance with the 'duration rhythm'. I emphasize 'non-vocalic' in order to exclude vowels and liquids. In fact, the rafforzamento does occur in clusters consisting of consonant plus liquid: /a+presto/ [appresto]

a presto '(see you) soon'

It seems reasonable to suppose that this same 'duration rhythm' law accounts for rafforzamento. In fact the strengthening does not take place when the reinforced word begins with a cluster of consonants: parlo svelto [parlo zvelto]

'he spoke rapidly'

where the non-vocalic segments following the short, stressed vowel are twice its length, thus complying with the phonetic 'duration rhythm'. On this view we restate the rafforzamento not as a 'syntactic' phenomenon, but as a phonetic rule. The importance of this rule will emerge in the remaining portion of this section. It is generally agreed that the features of stress and vowel length are concomitant. We documented the phonetic relevance of length in Italian. Now we want to question the criterion on the basis of which stress is given as the functionally significant feature and length as the predictable one. It seems to me that for reasons of economy the order should be the opposite in a descriptive analysis of Italian. That is, length should be the phonemic feature, while stress should be the predictable one.

29

PROBLEMS IN ITALIAN PHONOLOGY

Accordingly, our analysis would contain a rule for predicting stress: (QVOVCC)

(6) (i) (A)' [—ens] -» [+stress]

/ (C)V(C)

(ii) (B)'

#

i.e. if the penult is 'light', stress antepenult; otherwise stress penult; where 'light' means short vowel in open syllable. After all phonological rules have been given we introduce the 'duration rhythm' rule to predict the rafforzamento: (7)

(C) [+cons]

[+length]

/

-cons —length +stress

[+voc]

The rule reads as follows: single consonantal segments are long if the preceding vowel is short and stressed, and if they are also followed by a vocalic segment. But now we see immediately that this rule is sufficient to predict all tense/lax consonantal contrasts in the language. Accordingly casa 'house' would be derived as follows: /käs+a/ (Β)' and cassa 'box': /kas+a/ (B)'

[käs+a] käs+a

(C) -* [käs+a]

A simple reordering in the features of stress and length permits us to capture an important regularity in the language without adding any rules, but merely recognizing the more general importance of a phonetic phenomenon which is a physical and, according to Porena (1908, 1942), a perceptual reality in Italian. The phonemic system formulated on this basis would consist of 14 instead of 12 vowels: (8) e e i i o ö u ü e e o ö a ä but the advantages are far greater if we consider that the number of consonantal segment types is now reduced by 14: (9) t p k d b n m s f v C l r g n ä c ^ g l In addition, since tenseness is concomitant with length in consonantal segments, we are motivated to choose the second feature over the first. Thus one feature, length, would be used in dealing with both vocalic and consonantal segments, resulting in a more parsimonious analysis. This analysis, besides being far more economical than the previous one, leads us to a deeper understanding of Italian phonology. For example let us consider some minimal (or near minimal) pairs often given as bases for considering consonantal tenseness as a phonemic feature in Italian: fato 'destiny' and fatto 'made', or dote 'dowry' and dotte 'led' (which appears in condotte, ridotte, etc.). The phonemic

30

PROBLEMS IN ITALIAN PHONOLOGY

representation for the second pair is /dote/ and /dotte/, where double vs. single consonant stands either for tense/lax or geminate/simple in different analyses. These pairs turn out to be not 'minimal' in our analysis. In fact, morphophonemically dote is a noun consisting of the formatives / d ö t + e / , whereas dotte is a past participle consisting of / d o k + t + e / . This must be the representation on the basis of other forms: condotta [kondotta] 'conduct', conduce [konduko] Ί lead', conducente [konduöente] 'conductor', conduttura [conduttüra] 'water pipe', etc. Then the derivation would be this: /dok+t+e/ dok+t+e dok+t+e d0+t+e [dote]

(both vowels are specified short in their respective environments), (the stress rule applies) (k is elided) (the juncture is erased and the general 'duration rhythm' rule applies.)

It is simpler to say that k goes to zero and let the 'duration' rule apply (which is independently motivated) than to resort to a rule of assimilation; the reduction-to-zero rule will be used for all other consonants in the language whose disappearance is usually explained as assimilation. The rules of assimilation would be different for every consonant on both sides of the operational sign, whereas our rule will always have zero as output. For details on assimilation see 2.3. The given minimal pairs are, thus, not 'minimal'. The truly distinguishing feature is not consonant tenseness (or length) but rather the length of the immediately preceding stressed vowel, for we have shown that this analysis is to be preferred on the basis of economy. The phonemic system of Italian proposed here is given on p. 40.

2.3. ASSIMILATION AND ELLIPSIS

This section is an extension of the analytical solution proposed above. A reformulation of the so-called consonantal 'assimilation' is here discussed on formal bases. An 'assimilation' rule would be of this form: (A) C,

C,

/

+ Cy,

i.e. certain consonants will become like the following consonants at derivational morpheme boundary. Theoretically all combinations of consonantal segments can occur at derivational morpheme boundary. However there are severe constraints on which combinations are actually allowed. Thus for every consonantal segment type occurring in Cy. only certain others are allowed in C x . For example if Cy< = /1/, then C x can only be filled by /s, r, 1, n, t/, which means every other segment type will have to be converted into /t/.

PROBLEMS IN ITALIAN PHONOLOGY

31

An 'ellipsis' rule on the other hand would be as follows: (A)' Cx -> 0

+Cy

I

i.e., certain consonantal segments will be elided before certain consonantal segments when a derivational morpheme boundary intervenes. Here the output of the rule will always be the same. Thus the rule is notationally more parsimonious and more economic in the use of features. This calculation includes cost of a later lengthening or doubling rule, which however is independently motivated. Our choice of the 'ellipsis' rule is further motivated by the greater generality which the whole theory will acquire. In fact, a theory which posits an 'assimilation' rule would also need to give a rule for vowel elision as follows: (B)

Γ-cons] [-stressJx

-

0

/

J+VJ |#V J

i.e., an unstressed vowel is elided if it precedes a derivational boundary and an identical vowel. An unstressed vowel is also elided if it is followed by a word boundary and any vowel. If we accept the 'ellipsis' rule, however, we can efficiently collapse consonant and vowel 'ellipsis' into one general rule: (A)"

[+cons] x

I

cons —stress J x

0

+ [+cons] y + [—cons],# [-cons]

Thus we have now formal reasons to consider the phenomenon traditionally called 'assimilation' as 'ellipsis', again on the basis of greater simplicity and consequent generality in the phonological component of the grammar of Italian as a whole. In other words, a better view of the excellence of this analysis would be: if we formulate a rule of length (or gemination) after short V in order to capture the generalization VC/VC, then we automatically gain two further simplifications, namely, assimilation may be replaced by ellipsis, and vowel ellipsis can be collapsed with consonant ellipsis, with substantial saving in features.

2.4. CONSTITUENT STRESS

The motivation for reversing vocalic stress and length as to phonological primacy is partly based on the feasibility of assigning stress by rules. In this section we examine in detail the stress problem and a scheme for the stress-assigning rules. First, we wish to speculate on what the notion of stress in previous works might be formally based. Second, we shall make explicit our theory of stress, how it will be incorporated in the analysis, and what results it leads to.

32

PROBLEMS IN ITALIAN PHONOLOGY

It is difficult to reconstruct a coherent notion of stress from any one published work. Our conclusion will be an interpretation of the various sources. Every description refers to primary stress as phonemic. Rules for predicting primary stress are often given, but only for certain traditionally recognized morphological environments, such as the third person singular of the preterit, etc. It is always claimed that primary stress (or accento tonico) is biuniquely identifiable with certain physical characteristics in the sounds of the language. It is said to be mainly a stress of intensity, usually recognized as the most intensely pronounced vowel in a 'stress group', roughly coinciding with constituent phrases. A second type of stress is mainly a pitch stress and is recognized as the highest vowel in a 'phonetic phrase', roughly coinciding with higher constituent nodes. The pitch stress always coincides with one of the intensity stresses. From what precedes we may conclude that stress is not 'locally determined' in the physical sounds by a strictly phonetic criterion. One cannot listen to a certain vowel and determine whether it is stressed or unstressed independently of a context and of a definition (A) according to which the most intense (or highest) of two or more vowels is assigned the primary stress. Thus to recognize stress one needs in addition to the sounds a definition and the correct word boundaries. We shall see that additional knowledge seems to be required in Italian phonemics. In fact the context in which the definition could be applied is minimally two vowels. Yet we find that certain bisyllables are termed 'unstressed' and some monosyllables are termed 'stressed'. This implies an additional definition (B) of pro- and en-'clitics': any word is said to be stressed or unstressed according to whether (in combination with other words in a 'stress group') it carries or does not carry the primary stress. Definitions (A)-(B) are both used in phonological descriptions, yet they intersect contrastively. Thus it seems that stress is characterized by higher grammatical definitions and not by a strictly phonetic criterion. My understanding of the notion of stress which my predecessors appear to have entertained is summarized in the following statements, to which no scholar, to my knowledge, subscribes explicitly. Some might agree in part. (1) Stress is identified in speech sounds as intensity or pitch of vowels. (2) Primary stress is recognized by definition, (a) It is that of the most intense or highest of two or more vowels in the same uninterrupted sequence of speech sounds, and (b) this sequence must be a 'stress group'. (3) Monosyllables or longer words are defined as stressed or unstressed according to whether they ever carry the primary stress in 'stress groups'. We said that strictly speaking the notion stress is purely definitional and not 'locally determined' in the sounds at all. In fact a word like della 'of the' [dela] where the first vowel is more intensely pronounced than the second one is termed unstressed, according to stement (3), yet words like da 'he gives' [da] and da 'from' [da] which are phonetically similar in isolation are termed stressed and unstressed respectively on the

PROBLEMS IN ITALIAN PHONOLOGY

33

basis that one carries the main stress in 'stress groups' and the latter never does. In another view one is a verb, the other a preposition. This is to imply that actually the rules of procedure for abstracting the phoneme stress ARE based on the morphologysyntax of the language and not at all on a purely phonetic (audible) criterion. The relevance of certain features of pronunciation such as stress, length, tones, etc., the necessity of having to describe them, and the impossibility of accomodation in a given descriptive theory, is another severe shortcoming of data-centered phonemic analysis. In fact features like those just mentioned are usually recorded in inventories separately from other phonemes. They are termed suprasegmental phonemes and introduced by statement. A distinction between these phonemes and the segmental ones is admitted in the terminology and schemes of presentation. The explanation given is a shallow one of geometrical segmentation. The true motivation I interpret to be the recognition of the fact that there are certain obvious features of pronunciation, which one needs to term phonemic, which cannot be recognized directly and only from the actual speech sounds. If this is the correct interpretation, then the taxonomic phonemic transcription is not biunique with the phonetic transcription. It contains symbols (the suprasegmental phonemes) which are not 'locally determined' in the speech sounds. One inevitable conclusion is then the following: there is probably a step by step procedure of segmentation and abstraction of certain speech sounds which leads to a biunique phonemic transcription of those sounds (barring complications of partial overlapping, etc.). However the claim that such a procedure leads to a full (correct, and simple) analysis of the features of Italian pronunciation is not legitimate. I interpret this to be implied in the segmental/suprasegmental dichotomy. We identify phonetic stress as the intensity (pitch in longer units) of vowels. Primary stress (PS) is defined to be theoretically a feature of the most intense (or highest) vowel within constituent boundary. Monosyllabic constituents automatically are assigned PS. For polysyllabic constituents PS is assigned by formalizing observed or inferred regularities. This is carried out by general rules based on vowel length for the greatest part and on morphological evidence. We recognize other levels of stress in addition to PS, namely secondary (SS) and tertiary stress (TS). These subtypes reflect regularities in the intensity (or pitch) component of the vowel in the constituent. It is observed that all vowels preceding (in time or space) the PS vowel within the same constituent boundaries are lower in intensity with respect to that vowel: thus they are termed SS vowels. Moreover all vowels following a PS vowel are lower in comparison with SS vowels: thus they are termed T S vowels. The formulation of the rules for assigning SS and TS is solely based on the PS vowel. Further, independently motivated rules characterize internal regularities of SS and TS vowels. This is a result of the cyclic application of rules using and erasing the syntactic tree structure. Cyclic phonological rules in general are given to apply repeatedly from the smallest to increasingly larger constituents, in other words from the lowest to the highest node in the syntactic tree or partial tree of the sentence or

34

PROBLEMS IN ITALIAN PHONOLOGY

partial sentence. This is a condition peculiar to the phonological transformational rules. Syntactic information is thus used in describing the regularity of a certain feature of pronunciation. No theoretical distinction will be made between stress of intensity (in stress groups) and stress of pitch (in phonetic phrases). In fact they reflect the same pattern in speech sequences of different lengths. The physical distinction is formulated by a phonetic rule. G. Malagoli discusses six general types of 'word' stress. In our analysis only three stress patterns are recognized as having independent standing. Type I: Ultimate A. virtu B. canto veritä udi cittä teme

C. canterä udirä udiro

D. caffe ragii palto

E. astracan bazar nadir

Type II: Penultimate A. amore B. carretto pensiero civile minestra amico Type III: Antepenultimate A. comodo B. bambola Lepanto edile fulmine civico 'Words' in modern Italian follow stress patterns basically the same as those of Classical Latin with two major deviations: (1) Italian has acquired an almost completely new pattern in that of type I, and (2) a good many words such as Lepanto, Taranto, etc., have PS stress on the antepenult although the penult is in closed syllable. For this reason we could not give a rule predicting penultimate stress on words like carretto on the basis that a penultimate short V is stressed in a closed syllable. It seems that type II and III can only be handled in a general way by assigning stress partly on morphological information and partly on vowel length. Rules (A) (B) (C) should take care of these two more common stress types. Rule (A): Assign PS to penultimate in the case of affixes + e t + , + ε 1 + , . . . . Rule (B): Assign PS to antepenultimate if penultimate is short. Rule (C): Assign PS to penultimate of remaining words. Type I words are seriously problematic. Type I, D are borrowings and all end either in an open midvowel or in /u/. In my vocabulary every word ending in /ε/ or /a/ or /u/ is type I. Rule (D) assigns primary stress to type I, B, C and D in environment (a) and (b).

PROBLEMS IN ITALIAN PHONOLOGY

35

Rule (D): Assign PS to the ultimate vowel if the word (a) ends in /ε/, /ο/, /u/, (b) is 3rd person singular Preterit, (c) belongs to the class of exceptions N z : astracan,.. Type I, A words are of special interest. Words with suffix + t u could be taken care of under Rule (D), env. (a). For words in + t a this suffix would have to be introduced in an additional environment in Rule D. One can easily motivate an analysis of I, A words thus: vir • veri + Affix, where Affix = /tu/, /ta/ öiv but it is the case that we also have words such as 'cittadino, solitudine' etc., where we find evidence that the affix has two alternants: tad/ta, tud/tu, the second one occurring only when no other affix follows. We find it then not implausible to introduce a phonological rule deleting what follows the primarily stressed segment. This rule would reduce all type I, A words to type II (as far as stress is concerned) and in addition would predict the alternation. Finally type I, Ε are a restricted peculiarity of my idiolect. I recognize them as borrowings and pronounced them as type I. Some speakers regularize them as type II. They will be marked in the lexicon as a special class and this class is introduced as a further environment (c) (=N Z ) in Rule D. To treat type I, Ε as exceptions seems to be a plausible formulation also on other grounds. In fact in the last ten or fifteen years the type I, Ε list has become shorter because of a tendency to regularize them with the more common type II and III (Hall, 1960). It is thus possible that type I, Ε will be eventually eliminated. Rules (E) and (F) assign SS and TS stress. Rule (E): SS to vowel(s) preceding the PS vowel. Rule (F): Assign TS to remaining vowels. The stress rules are said to be cyclic in that they apply from the smallest to increasingly larger constituents, for theoretical purposes as far as the # S # node: they erase the tree structure. Physical and other limitations usually permit reapplication up to Noun or Verb Phrase constituents, which appear to be the structural content of breath groups. The cyclic application of stress rules seems to formalize some basis for Italian intonation. In breath groups different intensities coincide with the peaks of basic intonational patterns. They can be analyzed into three recurrent degrees which can be obtained by reapplying the PS rule and adding some adjustment rules operating on secondary stresses. A more precise formulation of the stress assigning rules is given in Chapter 3.

3. A PHONOLOGY OF ITALIAN

This chapter provides the phonological component of a grammar of Italian. It includes most of the morphology, morphophonemics, and general phonological processes. The framework for the phonological description adopted is a distinctive feature theory as proposed by M. Halle in The Sound Pattern of Russian (The Hague, 1959), Chapter 1, and "Phonology in Generative Grammar", Word (1962). Since a substantial number of the rules of phonology are formulated on phrasestructure, a sketch of the syntactic bases has been developed as a separate work by the author and is assumed in this dissertation. The syntactic analysis accepts, in general, the generative-transformational view of description in the form proposed in N. Chomsky's Aspects of the Theory of Syntax (Cambridge, 1965).

3.1. THE PHONOLOGICAL SYSTEM

In the course of this dissertation we use two types of alphabets: a syntactic and a phonological one. In syntax one deals with the data of Italian by identifying utterances in terms of a vocabulary consisting of symbols like S, NP, VP, V, AD J, [ + N ] , [ + _ N P ] , .... In this section we introduce a phonological alphabet to represent the same data but with a view to describing the sound patterns of the language. The main point of contact is the lexicon. This consist of a list of entries. Each entry has the form (D, C). D is a phonological matrix compatible with the syntactic (etc.) marker C. The matrix D is a rational transcription of the sounds of the language in question. This transcription is in terms of a phonological alphabet of symbols. Notice that it is not just a mapping of physical noises into any finite set of symbols. The relation between the physical sounds and the 'spelling' chosen is not arbitrary but rather of 'linguistic' relevance. Each matrix D consists of a concatenated sequence of segmenttypes. A segment is a bundle of unordered acoustic/articulatory distinctive features. A matrix in our sense consists of a set of + and — distributed in a cross arrangement of rows and columns. The columns represent the segments and the rows the features. A segment i is positively specified with respect to the feature j if the intersection of i

A PHONOLOGY OF ITALIAN

37

with j is marked with + · A segment i is negatively specified with respect to the feature j if the intersection of i with j is marked with —. A segment i is unspecified (i.e. the matrix does not reveal whether the segment is positive or negative; it gives no phonological information) with respect to the feature j if the intersection of i with j is unmarked (or marked with zero). The general motivation or task for the phonological component of Italian is to assign PHONETIC REPRESENTATIONS to utterances generated by the syntactic component. A phonetic representation is taken to be a linguistically satisfying characterization of the regularities and generalities which identify the sounds and the processes (rules) of pronunciation. It is assumed that the following are reasonable conditions to be required of a phonological description. They are not entities of the particular phonological theory of Italian, but rather universals of the general theory of phonology assumed. (1) (i) Speech events are represented as sequences of segments and boundaries. (ii) There is a restricted and fixed set of properties or binary distinctive features that are used to characterize segments. (iii) The representation of an utterance must be independently and unambiguously readable. (iv) The phonological description is an integral part of the grammar. (v) The number of specified features must be minimized compatibly with (iii) and (iv). (vi) Syntactic boundary markers are replaced by phonological ones, when significant, or deleted. In accordance with (1) (i) the description contains symbols (i.e. segments) and spaces (i.e. boundaries). A segment is a fragment of speech identified in terms of articulatory/acoustic properties. It is a segment-type representation of an infinity of segmenttokens. Segment-types are to be understood as part of the representation of speech rather than actual slices of physical sounds. They are theoretical constructs in an abstract representation of sounds. The boundaries can be regarded as spaces where the sequences of segment-types break up. These are assigned no acoustic properties. They are not bundles of distinctive features. They are important in that the phonetic shapes of adjacent segments might depend on them. On the basis of condition (1) (ii) the distinctive features are only a few. They are a fixed set of properties chosen as hypothetical assumptions. They are selected a priori on the basis of observed phonetic parameters. A hypothetical set of distinctive features will be considered comparatively feasible if a set of rules can be formulated which maps the phonological transcription into a phonetic one in a comparatively simple way. Accordingly the distinctive features of Italian are chosen as a result of many abstract linguistic considerations rather than, say, of only a spectrographs investigation. Furthermore the distinctive features adopted are binary. They have a two-value coefficient: + , —. Rules of phonology operate on a domain of specified features.

38

A PHONOLOGY OF ITALIAN

Since a binary system has been chosen they recognize (can operate on) only features marked + or —. Condition (1) (iii) requires that the representation of an utterance must be readable independently of anything that is not in the representation exclusively of the knowledge of reading the symbols, of course. Having assumed condition (1) (iv) we consider the phonology to be just an addition to the sequence of rules and structures developed in the syntax. Phonological rules will be constructed which avail themselves of contexts developed earlier and still represented in terms of that syntactic alphabet, if this leads to simple phonological explanations. For example, the formulation of rules assigning the feature stress to Italian verbs can only be achieved by recourse to syntactic information. For example the items canto jcantd differ phonetically in stress (and concomitant effects) and syntactically in PRESENT/PRETERIT. Assuming (1) (iv) we assign stress on the basis of PRESENT/PRETERIT. If (1) (iv) is rejected we cannot assign stress by rule, must take it as a phonemic feature of vowels and assume the primacy of stress over length. But only the reverse leads to satisfactory explanations for basic phonological phenomena, as discussed in 2.2. Condition (1) (v) specifies that the phonological transcription must be minimally redundant, that is to say, that it must be a 'rational orthography', one containing only information not statable by rule. This goal is consistent with the choice of a binary rather than a n-ary system. Furthermore the most symmetric hierarchical arrangement of features will be chosen for lexical representation, consistent with the phonological rules. Finally the phonological redundancy (morpheme structure) rules will further allow reduction of lexical feature content (as well as characterizing the notion 'pronounceable in Italian'). Finally condition (1) (vi) insures that the output of the syntactic component has been acted upon by general rules turning boundary symbols into phonological ones where they are relevant. It seems reasonable that this requirement is to refer not only to the boundary symbol $ but also to the concatenation symbols. We then would accordingly require a complete syntactic description of Italian to include a specific set of rules converting & and the concatenation symbols (or spaces). It seems that boundaries which have phonological relevance in the present analysis of Italian are phrase boundary (%), word boundary ( # ) and derivational boundary (+). The following is an approximation of the boundary rules which a complete grammar of Italian would contain. (2) (i) S % I (ii) + - # (iii) -*· + (iv) $ 0

X

, where X = 0

D

, where D = phonological matrix

(2) (i) transforms the sentence boundary S into a phrase boundary before or after pause. (2) (ii) transforms the concatenation symbol + into a word boundary. (2) (iii)

A PHONOLOGY OF ITALIAN

39

inserts a derivational boundary next to phonological matrices. (2) (iv) deletes the remaining sentence symbols The phonological alphabet in terms of which the data under study is spelled consists of segment-types, the 'letters' of the alphabet, and boundaries, the 'punctuation'. The boundaries are provided by a late rule of syntax approximately in the form of (2) (i) of this chapter. The segments are introduced in sequences of unique combinations, or matrices, as discussed earlier in this section. They are introduced in accordance with the following criterion: (3) The set of all phonological matrices D in the lexicon has the same cardinality as the set of all unique matrices D. On the other hand the cardinality of all syntactic (etc.) markers C may be the same or greater than that of all unique syntactic (etc.) markers. This means that D matrices are stated only once whereas C markers may be repeated in cases of homophony. The phonological representation of D matrices, or morphemes, will appear in accordance with the following phonemic system of Italian (4). (4) consists of (i) a decision tree-diagram, (ii) a set of abbreviations and (iii) a feature matrix. The treediagram (i) is a binary decision-tree with just two branches converging on each labeled node. The numeric node labels correspond to the set of features marking the rows of the matrix (iii). The convention is that the branch departing to the right of a given node represents a positive specification ( + ) for the feature labeling that node, and that the branch departing to the left of a given mode represents a negative specification (—) for the feature labeling that node. It can be seen that a set of unique (unambiguous) paths through the tree is obtained if one connects each terminal (non-diverging) branch to the highest node or viceversa. We have marked each path with a letter abbreviation (ii). The same set of paths can be given in the form of a matrix (iii), where the complete set of features (labeling the nodes) is listed to the left, and under each terminal branch a corresponding column of feature values is constructed identifying each path. Since each path is unique, likewise each column will be unique. The paths of the tree-diagram reflected as columns of the systemic matrix (iii) are the underlying segment-types of Italian or the bundles of distinctive features. We shall represent the lexical matrices D in terms of the notation used in (iii), that is in terms of columns of distinctive features. Thus (4) (iii) will be the official notation, (4) (ii) the unofficial one, used for ease of reading. The tree-diagram (4) (i) is also unofficial material. The phonemic system of Italian which we have developed and are assuming in this analysis has been presented in (4). It consists of 34 decodable, pairwise distinct, maximal segment-types (roughly corresponding to the Prague School morphophonemes, and in effect characterizing the notion 'phoneme'). A segment-type is a unique configuration of + and —. A given segment-type is MAXIMAL in a given system if it cannot be converted into a different one of the system by specifying one or more

A PHONOLOGY OF ITALIAN

, + ++ ~

+

+

κ

+

+



+

1

α

+

I

+

+

1

+ 1

+

+ i

+

+

•ο

+

I

+

1 1

+

Ό

+

I

+

1 1

+

•Ή

+

ι

«ο

+

|

1

1

+

•ω +

I

I

1

+

I

β»

+

I

+

*

+

I

+

+

+

+ +

1

ο

+

1

+

1 1

1

+

I

+

1 1

1

+

1

o

+

|

1

1

I

ω

+

|

I

1

ι

0

1

>

1

CM

I

X>

[

1

+

+

I

+ + +

++ ++ ++

+

+

1

+

1

A

|

+

>60

I

+

+

1

>0

[

+

+

1

ao

| 1

+

1 1

ο -

+

+

1

I

+

1

1

+ + + I + I + I + I

+



|

1

+

+

ΧΩ

[

1

+

+

N

1

1

+

1

+

Ο

I

1

+

1

]

c

1

1

1

en

I

+

1

1

•tf

[

1

1

1

**

1

1

1

1

ο ο >

> ^ t o

ft »fi CO

Η η

η

+

+

+

** Ο nj a3 Sa Ό S3 fell > c o rSr rt vi ίο

C-

00

• k/ we can simply state [+diffuse] [—diffuse] in environment [..., + grave, ]. We have chosen diffuse instead of high for vowels because the former is used for vowels as well as consonants. (vii) CONTINUANT (cnt). Continuant sounds are accompanied by narrowing not in excess of occlusion. The onset of continuant is gradual. Non-continuant (interrupted) sounds are accompanied by CONTACT narrowing and abrupt onset. The segments specified for this feature include both consonants and liquids, but not vowels. As for liquids this feature separates /r/ from the rest of the liquids. There is in fact only a [+continuant] [—continuant] order of liquids, which this sub-classification handles in a simple way. In addition we identify [+continuant] segments as including both fricative and nasal segments. (viii) LENGTH (Igt). Long vowels are held through a longer sound interval (and display larger energy) than short vowels. Length is specified only for vowels. There are in this analysis seven short and seven long vowels. The choice of the feature length (instead of stress) for vowels is the major motivation for developing this dissertation. Specific discussion is found in 2.2., and in section 4 of this chapter. As for consonants and liquids length will be assigned by general rule. (ix) VOICE (voi). Voiced sounds are accompanied by periodic vibration of the vocal cords. With voiceless sounds such vibration does not occur. The distinctive feature voice is marked only for consonants. All other segments, i.e., vowels and liquids do not use voice distinctively. Of the consonants nasals are also

44

A PHONOLOGY OF ITALIAN

normally voiced, and therefore this feature is not marked distinctively. Voice is specified to distinguish the traditional contrasts of stops and fricatives in the case of /s/, however, we have a conveniently negatively specified segment for the rule giving its voiced allophones. (x) NASALITY (nas). The articulation of nasal sounds involves lowering the soft palate or velum, thus releasing the air stream through both mouth and nose. Consequently both oral and nasal cavity act as resonators. In non-nasal sounds the velum is not lowered and resonance is mainly through the oral cavity. Nasality specifies positively the three nasal segments /m, η, ft/. Because of the nature of binary oppositions some non-nasal segments are negatively specified. All other segments will be assigned negative nasality by rules filling in the matrix non-distinctive features. (xi) TENSE (tns). Tense vowels display a larger energy (and are held for a longer time) in comparison with lax vowels. Tense/lax distinguishes minimally close/open midvowels. Thus the rule closing open midvowels in unstressed position involves operating on only one feature [—tense] [+tense]/W. Whereas in (5) phonetic and linguistic comments about the individual distinctive features of the system (4) of Italian were given, (6) is a list of the classificatory distinctive features and the opposing segment-types of the phonological system (4) in abbreviated notation. (6)

(i) vocalic/—vocalic: ε, e, i, o, o, u, a, έ, e, i, 5, ö, ü, ä, r, 1,1/1, d, s, n, c, z^, §, ή, ρ, b, f, v, m, k, g, δ, g. (ii) conson./—conson.: r, 1,1/ ε, e, i, o, o, u, a, έ, e, i, 5, ö, ü, ä. (iii) grave/—grave: ρ, b, f, ν, m, k, g, δ, g, o, o, u, a, 5, ö, ü, ä/1, d, S, n, c, s, η, ε, e, i, έ, e, i. (iv) sharp/—sharp: c, s, ft, δ, g, 1/ t, d, s, n, k, g, 1. (v) compact/—compact: a, ä/ a, o, u, 5, ö, ü. (vi) diffuse/—diffuse: ρ, b, f, v, m, i u, i, ü/ k, g, δ, g, ε, e, a, ο, έ, e, 5, ö. (vii) contin./—contin.: s, η, ή, f, v, m, 1,1/1, d, c, p, b, r. (viii) length/—length: έ, e, i, ö, ö, ü, ä/ ε, e, i, ο, ο, u, a. (ix) voice/—voice: d, b, v, g, g/ t, c, p, f, k, δ. (χ) nasal/—nasal: n, n. m/ s, s, f, v. (xi) tense/—tense: e, o, e, 5/ ε, ο, έ, 5.

Finally, (7) complements (6) with lexical items exemplifying phonological oppositions of near minimal contrasts. The feature oppositions do not include the most basic ones: vocalic/—vocalic, consonantal/—consonantal.

45

A PHONOLOGY OF ITALIAN

(7)

(i) grave/—grave: p'mo, bart,/ino, vino, mina, coppa, gemo, c/'occo, G/anni/ iino, ί/are, sino, Nino, Gina, zoppa, scemo, gnocco, Janni; rosa, sono, sano, oro, moro, mwro/ resa, seno, s/no, ero, mero, mz'ro; ... (ii) sharp/—sharp: mazzo, zio, coscia, sogno, taccio, regg/o/ matto, dio, cosa, sono, tacco, reggo; ... (iii) compact/—compact: chicco, gotta, Cina, Gino/ Pippo, 6otta, /ina, vino, Aiino; ... (iv) diffuse/—diffuse: lfzza/ Lecce, letto; p/ni/ seni, beni; vmi/ reni beni; fwmo/ Roma, Como;... (v) contin./—contin.: jino, Nino/ iino, Dino; moic/a, ragnoj mozza, razzo ;/ino, vasta, Mino//>ino, basta; pa/a, /ana/ para, /ana; ... (vi) length/—length: pene, anelo, Pina, caro, roso, nono, fumo/ penne, anello, pinna, carro, rosso, nonno, fummo; ... (vii) voice/—voice: dopo/ topo, razzo/razza, bere/ pere, vano/ /ano, gallo/callo, iaggioj facc/o; (viii) nasal/—nasal: mano/ iano, gnocco/ sciocco, mare//are, mano/ vano; ... (ix) tense/—tense: e/ e, o/ ho, pero/ zero, coda/ moda,...

3.2. PHONOLOGICAL REDUNDANCY

We face the problem of reducing to a minimum the amount of phonological information to be given in the lexicon. This is consistent with condition (1) (v) adopted in section 1. In phonology the task of minimizing redundancy is carried out by the Morpheme Structure (MS) rules for the most part, and by the Phonological rules to a minor degree. The Morpheme Structure rules are simple processes which specify phonological features in a general manner. They do not delete, permute or transform. All MS rules are of the rewrite type. They extend the specification content of segment types. The generality of the rule is based on distributional configurations. Given a phonological matrix D, features can be predicted on the basis of (1)

(i) phonological boundaries (ii) features of adjacent segments (iii) concomitant features within the same segment.

For example given the matrix /stank/ (2)

#

s —voc +cns —grv —shp +cnt —nas

t —voc +cns —grv — shp —cnt — voi

a +voc —ens +grv +cpt

η —voc +cns —shp +cnt +nas

k —voc +cns +grv —shp -dif —voi

+

46

A PHONOLOGY OF ITALIAN

we predict:

—grv

(3) (i) Γ-voc] [_+cnsJ

I

#-

—shp

+cnt

—voc +cns —cnt

—nas i.e. if a consonant (/s/) is word initial (#_ consonant can only be /s/. (ϋ)

—voc +cns +nas

—y

—shp -dif +grv

/

_) and followed by a stop (

jtj), that

—voc +cns -dif + grv

i.e. IN/) is homorganic to /k/ and cannot be /ή/ (the sharp nasal) in the environment before /k/. (iii) (a) [—voc] -»• [+cns] i.e. all segments marked [—voc] are also [+cns]. Rules (3) characterize phonological redundancy in the matrix /stank/. Now we introduce the matrix (2) in the lexicon simply as

(4)

-voc

—voc —grv —shp —cnt —vol

+voc —ens +grv

—voc +nas

— voc" +grv —shp —dif —voi

and the rules (3) into the grammar. Rules (3) fill in the general information pertinent to (4). However since rules (3) have been stated generally and their statements represent bona fide characteristics of Italian, all such configurations can be entered in abbreviated form (4) for every lexical entry. Thus we first formulate the MS rules on the basis of genuine Italian morphemes and then enter all phonological entries of the lexicon in a form which considers the MS rules. Morpheme Structure rules serve two purposes. They eliminate redundancy consistently with condition (1) (v) and characterize concomitantly the distributional restriction of Italian. They are parallel to phonotactic statements but the approach is different. Whereas phonotactic statements seek to establish all possible combinations, MS rules determine obligatory constructions. Comparatively Syntactic Redundancy Rules and Phonological Redundancy Rules serve an identical purpose in reducing the amount of primary information that the lexicon must provide. However whereas MS rules characterize the notion 'pronounce-

A PHONOLOGY OF ITALIAN

47

able in Italian' there seems to be at the moment no parallel function which the Syntactic Redundancy Rules serve.

3.2.1. Morpheme

Structure

Rules

Consider words like (1) rosa lama luna

ridere rendere leva etc.

They illustrate a generalization about liquids. An initial liquid can be followed only by a vowel, or MSI

[X]

Γ-cnsl |_+vocJ

inenv. # Γ+cns" [_+vocJ 2

Thus the feature consonantality is omitted from the sequential representation of morphemes in the environment stated by MSI. One feature is saved from lexical statements for every word of that shape. The total cost or complication which this rule brings into the grammar is one rule involving three features. A rough estimation of the feasibility of MSI is obtained by considering the economy achieved in the lexical representation of the partial list (1). That is enough motivation for introducing the rule. The membership of the set exemplified in list (1) is certainly more than tenfold, and as a result the predictability range of the rule would be extended accordingly, thus consolidating the motivation. Notice incidentally that the extension of the lexical prediction range does not import consequent increase in grammatical machinery. Let us further consider liquids in list (2) parla Carlo Orlando

tuorlo merlo ciarlatano, etc.

They exhibit non-initial clusters of liquids (intervocalically, V V, and before derivational boundary, V +V). The generalization is that the segments adjacent to the cluster may be only vowels. The rule is as follows: MS2

[X]

Γ-cnsl |_+vocJ

inenv.

Γ+cnsl (+) [_+vocj

Every segment in such environment will be represented phonologically with the feature consonantality unspecified. That is, the value of those phonological categories which must be marked or known in order to insure pairwise distinction in the segmental representation are now omitted in the environment of MS2 in compliance with the

48

A PHONOLOGY OF ITALIAN

principle of economy. Pairwise distinction is insured by the knowledge which the independently motivated MS rules synthesize. Because of the last rule two features are saved for every such environment. The partial list (2) is alone a justification for the rule. The claim which is made about the sequential structure of Italian is that two adjacent liquids do not cluster with consonants. In addition such combination of liquids occurs medially. And further two liquids do not occur initially. Further consideration of items in list (2) shows that two-term clusters of liquids are always of one fixed structure, namely -rl- and never *-lr-. Therefore in clusters of liquids the feature distinguishing /r/ and /l/ is omitted from the lexicon and introducted by the following general rules: MS3'

r+cnsl |_+vocJ

[—cnt]

in env.

[~+cns~| L+vocJ

i.e., a liquid is non-continuant (/r/) when it precedes another liquid. MS3"

Γ+cnsl |_+vocJ

[+cnt]

in env. r+cns~| L+vocJ

i.e., a liquid is continuant (/l/) if it follows another liquid. Since the input as well as a portion of the environment in MS3' and MS3" are identical, the two rules are collapsed into one as follows, with some economy gain: MS3

[

+cns~| +vocJ

'[-cnt] in env.

+cns +voc

[+cnt] i.e., assign [—cnt] to the first of two adjacent liquids and [+cnt] to the second one. By virtue of this rule we can now represent liquids in this particular sequential environment simply by referring to the features consonantal and vocalic. Let us now consider list (3) treno truppa tranello trono tromba

draga dramma dritto droga dromedario, etc.

This list is presented to illustrate that word initially one finds in Italian the clusters -dr- and -tr- but not *-dl- nor *-tl-. In considering this distribution we shall discuss briefly the various possibilities available to the analyst in the formulation of a rule for that constraint, and what criteria govern the choice in this analysis. One might see that the constraint is on the liquid. That is to say that only a certain kind of liquid can

A PHONOLOGY OF ITALIAN

49

occur after alveolar stops (d_, t_). In that case one would formulate morpheme structure rule MS4.1 Γ + c n s l -»· [—cnt] |_+vocJ

in env. #

—voc —grv —shp —cnt

i.e., a liquid following the subset the elements of which are the segments /d/ and /1/ and none other is necessarily the interrupted variety, namely /r/. Alternatively one might say that the constraint is on the kind of consonant that can occur before /l/ in that particular initial environment and state the following morpheme structure rule instead: MS4.2 [ - v o c ]

-grv -shp -cnt

in env. #_

+ VOC

-fens +cnt

i.e., a non-vocalic segment (a consonant, as opposed to a liquid which is 4-vocalic and + consonantal, and as opposed to a vowel which is -{-vocalic) is the complement of the subset the elements of which are /d/ and /t/ in the environment preceding /I/ in initial position. In other words MS4.1 asserts that in Italian only /r/ may occur in position following alveolar stops, whereas MS4.2 alternatively states that neither /d/ nor /t/ occurs before /I/. The duality of interpretation leads to the search for motivation of choice, a formal reason for adopting one rule over the other. The first criterion one might consider is that of simplicity, that is, which rule allows the simplest grammar. A crude interpretation of this criterion is a computation of the number of rules and of the plus and minus which rival analyses employ. In our particular case both MS4.1 and MS4.2 use an equal number of feature specifications (plus and minus). Notational parsimony is disregarded since its bearing on simplicity (a direct correlate of generality) is probably nonexistent, in the present state of understanding. We adopt MS4.1 because this allows us to limit further the distribution of liquids. The alternative specifies no natural set and does not really specify what may occur before r. On the basis of MS4.1 we omit from the lexical representation specification of the feature continuant in that specific environment. The partial list (3) warrants the rule. We notice however that the same constraint seems to occur when alveolar stops, namely /t/ and /d/, are not initial after a consonant or a liquid, as is illustrated by the following list (3)' astro coltre

altro feltro

50

A PHONOLOGY OF ITALIAN

contro androne

dentro Andrea, etc.

In consideration of this additional environment for the same contraint on the clusters of alveolar stops plus liquid we can conveniently alter MS4.1 to cover this new environment. Consequently MS4.1 is reformulated as follows: MS4.3

r+cns |_+voc

[-cnt]

#

in env.

+voc~| +cnsj [—voc]

—voc —grv —shp —cnt

i.e., a liquid is /r/ after /d/ or /t/ in the following environment: initially, and medially following a consonant or a liquid. In view of the features saved in the lexical representation of morphemes even if one considers only the partial lists (3) and (3)' the above rule is justified. The increase of environmental types as we progress in alternate formulations from MS4.1 to MS4.2 in which only -dr-, -tr- occur implies a decrease of types of environments in which j\J (as well as /r/) may follow /d/ or jij. Namely, adopting MS4.2 the only environment not forbidden to the occurrence of alveolar stop plus /l/ is intervocalically. In effect, even this environment seems dubious. For a list of items exemplifying such clusters one would have difficulty finding more members than those in the following list (3)" atleta atlante It would be even more difficult to find an item exhibiting a voiced alveolar stop plus /I/. I am quite certain that the cluster -dl- does not occur in any item in my personal vocabulary. I could have said the same for -tl- were it not for the list of very frequent words (3)". In this analysis we want to characterize and point out the generalities as well as the exceptions. The majority of Italian lexical items do not have the clusters *-dl-, *-tl-, except for a few borrowings which have introduced -tl- intervocalically. Should one choose not to recognize this fact in a statement of morpheme structures the grammar would then contain the rules MS4.2 given above and in addition the following one: MS4.4

+cns L+vocJ

[-cnt]

in env.

—voc —grv —shp —cnt +voi

i.e., a liquid is of the interrupted variety /r/ whenever it follows /d/.

51

A PHONOLOGY OF ITALIAN

In review, MS4.3 specifies the environments for the archiphoneme liquid in -dl- and -tl-, and MS4.4 states the general exclusion of -dl-. Notice that the order of these two rules is not relevant. That one precedes the other will not affect feature economy. In this analysis we choose to treat -tl- as a peculiarity of a few Greek loan words. Therefore we state a rule: MS4

r + c n s ! -»· [—cnt] L+vocJ

in env.

—voc —grv —shp —cnt

i.e., a liquid which follows /d/ or /tl is /r/. This choice seems advantageous in that it recognizes a regularity about the majority of lexical entries by a simple rule. Items in list (3)" will be lexically subcategorized as [+Greek z ]. MS4 will then apply only to items which are the complement of those marked [+Greek z ]. The appeal of this analysis is that the grammar will characterize the distinction between peculiarly Italian and acquired sound patterns. In addition, by treating, say, Greek loan words as a group, general rules can be isolated specifying which and how Greek distributional features are adopted in Italian. This seems to be a more realistic approach to lexical study. The next rule asserts the regularity that clusters of the general shape Liquid-Consonant-Liquid are only of the more specific type /l/-Consonant-/r/, as illustrated by the following partial list (4): altro sepolcro

veltro feltro, etc.

This observation allows us to omit the feature distinguishing /r/ and /I/ in that context. Only the features which are common to the two liquids are necessary (i.e., their archiphoneme). The distinguishing features are then assigned unambiguously by a grammatical process as follows: MS5

r+cns~j |_+vocJ

[+cnt]

[—voc] in env.

[-cnt]

+cns +voc —voc]_

i.e., assign the feature [+cnt] to a liquid preceding a consonant and a liquid; assign [—cnt] to a liquid following a liquid and a consonant. Items like perplesso are analyzed as composites, consisting of two formatives, partly on the independent fact that there are items like complesso, etc. A consideration of the prepositions per, con, in, indicates further how one might judge the feasibility of certain morpheme structure rules. The three prepositions given seem to be the only items in the basic Italian vocabulary (if we choose to enter them in

52

A PHONOLOGY OF ITALIAN

it) which end in a consonant. An MS rule based on the prepositional class PREP, would predict most features in the final segments. Another rule would then apply to all remaining items and predict the vocality value of every native word-final segment. The underlying phonological representation for per, con, in, is the following: Lex. χ

Lex. y

Lex. ζ

([+Prep., ...],

([+Prep., ...],

([+Prep., ...],

—voc +grv +dif —voi +dif

+VOC —ens -grv

)

per

+lgt -dif +tns

—voc +grv —shp -dif —voi

+VOC —ens +grv —cpt

+voc —ens —grv

—voc —grv —shp +cnt +nsl

+lgt +dif

+ VOC +cns —cnt

—voc —grv —shp +cnt +nsl

+lgt -dif +tns

con

in

)

Suppose we introduce the following MS rules predicting features of the final segment in Prep, x, y, ζ in this form: MS6.1* r + c n s " l |_+vocJ

[—cnt]

in env.

#

Prep.

MS6.2* [—voc] ->·

—grv —shp + cnt + nas

in env.

#

Prep.

then the phonological representation for per, con, in, would be as follows: Lex. x' ( [ + P r e p

]X

Lex. y' ([+Prep., ...]X Lex. z' ([+Prep., ...]X

Y

"+voc Ί) +consJ Y —voc]) [-voc])

per con in

Comparing Lex. x, y, ζ and x', y', z' we see that the latter set uses nine features less than the former one by virtue of MS6.1* and MS6.2*. But since Lex. x, y, z, constitute

A PHONOLOGY OF ITALIAN

53

the exhaustive lexical range or domain for MS6.1* and MS6.2* seven features is the total economy that the two morpheme structure rules can contribute to the analysis. A more realistic appraisal of the value of such economy is better understood against the grammatical machinery that it requires. In our case this involves two rules, MS6.1 * and MS6.2*, using a total of eight features. It would appear that by introducing those two rules we are still saving features from the lexicon, namely one. However if we consider the number of features employed in the analysis as a whole to specify per con, in using the alternate solutions Lex. x, y. ζ and Lex. x', y', z' plus MS6.1, 2* we compute that the total is exactly the same. In conclusion, if we choose the latter we are just taking features from the lexicon and specifying them in the grammar. In addition since MS6.1*, 2* do not predict anything, they fail to qualify as grammatical rules, in that there is no motivation for them in the grammar. The obvious result is the abandonment of the proposed alternative. In order to incorporate in the analysis the fact that all Italian words end in a vowel except for per, con, in (and of course loan words), and at the same time do away with MS6.1*, 2*, this analysis will state the regularity restricted to the complement of the exceptions on a lexical basis. This results at least in a fair understanding of the facts. In the morpheme rules given so far we seem to have exhausted the sequential constraints that one could impose on liquids. In the event further restrictions are recognized on the occurrence of this subset of segment-types they are to be incorporated before the next rule on the basis of simplicity, unless a greater overall gain is obtained by doing otherwise. For the most part the rules which follow apply to vowels, consonants, or consonants-and-liquids. In order to characterize these three subsets with the smallest possible number of features we need to introduce at this point a phonological rule specifying the non-distinctive feature [+cns] for consonants, defined as non-vocalic segments. Such a rule needs to be given in the description sooner or later. We are assigning to it a fixed order with resulting simplicity in the formulation of the rules. (Notice that this is not an MS rule; in fact, it does not contribute to lexical economy). At this point in the analysis the major classes of segments are characterized as follows: vowels are [—ens, +voc], consonants are [—voc], liquids are [+voc, +cns], liquidsand-vowels are [+voc], and liquids-and-consonants are [+cns, +voc], [—voc]. This arrangement has been descriptively convenient since we have dealt with all of the above subsets, except the last one, namely, the subset encompassing both liquids and consonants. The next rule introduces information vital to a simpler formulation of the next group of rules. ph6

[—voc] ->· [+cns]

i.e., specify the feature consonantality as plus in segments which are non-vocalic. It seems that, although ph6 'fills in zeros' as MS rules do, a precise simplicity mechan-

54

A PHONOLOGY OF ITALIAN

ism must optimally interpret them in their proper function. Whereas MS rules introduce by general grammatical process features which are indispensable to the rational orthography of the language (i.e., distinctive features), phonological rules of the ph6 type introduce features dispensable to the phonemic representation (i.e., non-distinctive features), but vital for rule formation. Redundant information is used sometimes for distinctive purposes. After ph6 has applied the information is available to the analysis that non-vocalic segments are [-(-consonantal]. Accordingly the following is now the characterization of the five major subsets: vowels are [—ens]; consonants are [—voc]; liquids are [+cns, +voc]; liquids-and-vowels are [+voc]; and liquids-and-consonants are [+cns). The simpler characterization of the classes vowel and liquids-and-consonants as a result of having introduced rule ph6 is quite convenient since the majority of rules referring to these classes of segments are given after the application of ph6. Therefore the analysis exhibits an ordering of the rules which allows a simpler formulation. One might charge that the analyst has purposely arranged this desirable situation. That is precisely the intention of the analyst. In fact the goal of the theory is the simplest description of the phenomena of Italian pronunciation, and the presentation, BUT NOT THE CONSTRUCTION, of the description is claimed to be algorithmic. The fact that to discover the simpler order is difficult is so only for the linguist. There is no implication of difficulty for the language user. These rules and their ordering are part of an abstract model attempting to capture a formalization not of the language user himself but rather of the competence he exhibits in language behavior. We first consider consonants ([—voc] segments). The following list (5): strada scrigno sdraia

spronare Strega stretto, etc.

shows three-term consonantal clusters, the initial element of which is /s/, a noninterrupted segment. At first glance this seems to be a generality about three-term initial consonantal clusters. Thus one could introduce the value of the features continuant, nasal, compact and grave by a simple general rule with consequent substantial reduction in the feature content of the phonological representation of the lexicon. A more general view of the sequential constraints on the non-interrupted consonants is illustrated by the following list (6): gonfio penso guancia ronza

asta pasta Antonio stanco

ambizione aspide toscano spina

astro lampreda Andrea mandra, etc.

in which one can observe two-term clusters of consonants in which the first segment is invariably non-interrupted. We also notice in the same environment only /N/ or /s/. Incidentally we need not specify the members of the set /N/. It is an incompletely

A PHONOLOGY OF ITALIAN

55

specified segment representing the class of nasals. Featurewise it contains information common to all nasals and no other segment. A later MS rule characterizing the homorganic quality of nasals will specify the feature gravity for these segments in accordance with the gravity value of the following consonant. Such a rule assigning either plus or minus to nasals before plus or minus consonants distinguishes /m/ from /n/. By virtue of the following rule the feature specifications non-interrupted, non-grave, non-compact are omitted from the lexical representation of non-vocalic segments (consonants) occurring before a segment of similar quality and assigned by rule. The features omitted are those common to both /s/ and /n/, which in the arrangement of the phonological tree share all features except nasality. MS7

[-voc]

+cnt —grv —shp

in env.

[—voc]

i.e., a consonant is of the superclass consisting of the elements /s/ and /n/ when occurring before another consonant. Word list (5) was given to point out that /s/ is the only segment which can occur as first member of an initial three-term consonantal sequence. However, if we consider only the non-vocalic segments, the class which has consonants BUT NOT LIQUIDS as elements, and if we keep in mind also list (6), then we can formulate a more general observation, namely, only the archiphoneme of /n/ and /s/ may precede a consonant, regardless of its distributional environment. MS7 states exactly this. Consequently in initial position the class membership of the first segment in non-vocalic clusters is narrowed down to just /s/, which is expressed by the following rule: MS8

[—voc] -»• [—nas]

in env. #

[—voc]

i.e., assign [-nas] to the first member of a word initial non-vocalic cluster. By virtue of MS8 a consonant can only be /s/ in first position in an initial non-vocalic cluster (consonants, but not liquids). We want to insist further that the observation in list (5) can be interpreted in two ways. One might say that the first segment of an initial three-term consonantal cluster (consonants and liquids) is /s/, and accordingly propose the rule which follows: MS8.1* [+cns]

—voc —grv +cnt —shp —nas

in env. #

[+cns] 2

which represents the usual and shallowest statement. Or, as it is here proposed, one might exclude liquids from the statement and consider only clusters of non-vocalic

56

A PHONOLOGY OF ITALIAN

segments (consonants, but not liquids), in which case one recognizes the greater generality of the independently motivated MS7 and its distributional exception in MS8. The facts observed in list (5) have been analyzed in this theory in a simpler and apparently more integrated manner. As a result of MS7 and MS8 /s/ is represented in the lexicon only by the feature [—•voc], when it appears as the first member of an initial consonantal cluster. It is the motivated proposal in this analysis that length is the phonologically significant feature in vowels on the basis of which among other phenomena one can account for consonant length by general rule. Thus our underlying representation does not contain identical segment-types immediately contiguous. There does not seem to be in the realm of MS rules a statement of that fact since it would in no way increase lexical economy. (MS rules state what MUST necessarily occur in a given context rather than what MIGHT alternatively occur. In this they differ diametrically from the statements of phonotactics, although the notion of 'pronounceability' is characterized in both). The fact that this is part of the intuition of the native speaker, that he is somehow aware of this, is formalized in the Phonological (P) rule accounting for the phenomenon of (identical) segment ellipsis. The same observation is valid for nasals in Italian: i.e., nasals are not found in adjacent position. On the basis of nasality in a segment we can predict the value of the nasality feature in adjacent segments: MS9

[—voc] -> [—nas]

[+nas]

Although this rule theoretically applies to all segments in the given environment, in practice it will contribute to the lexical economy by allowing omission of the nasality coefficient for those segments which, due to the particular configuration of our phonological tree, have that feature specified, namely, /s/, /n/, /§/, /ή/, /f/, /v/, /m/. Actually, due to the particular distributional arrangement of the formative only /s/, /f/, /v/, (and /s/ if we consider conscio) will be affected by the rule. The rest of the segments mentioned will be assigned the classificatory value of [+nas]. This last effect of MS9 does not contribute to lexical economy, in that there exist no formatives with such a consonantal arrangement. Nevertheless it performs a function which is independently motivated in the description, that of specifying non-distinctive features in the matrix, a function carried out mostly by phonetic and, to a lesser extent, by phonological rules. In sum MS9 does in part the job of a morpheme structure rule and in part that of a phonetic rule. It is a well known distributional fact in this language that elements of the palatal and affricate subset do not form clusters with each other. A rule formalizing this would predict the non-palatal and/or non-affricate feature of segments adjacent to palatals and affricates. The rule would state in this theory that consonantal segments adjacent to palatals and affricates must be elements of the complement class. The complement set of palatals and affricates includes every other segment in the language. In other words anything may precede or follow palatals and affricates,

A PHONOLOGY OF ITALIAN

57

which means that there is no restriction on its adjacent segments, or that nothing can be predicted about them. This immediately disqualifies such a statement from the MS rules, by definition. Furthermore such a statement would be false, for there are restrictions on what can follow or precede palatals and affricates. In conclusion, a statement saying what cannot occur as a combination is not translatable into MS rules directly. The fact is, however, represented by MS rules in that they specify what classificatory features the segments must necessarily have in a given position. The next rule involves the assignment of the gravity feature to nasals, characterizing their homorganic nature in Italian before consonantal segments. (However, notice that this phenomenon is also in part taken care of by Ρ rules; in fact, whereas the following MS rule accounts for this fact within word boundaries, and therefore for items in the lexicon, the Ρ rule 41 accounts for the same phenomenon across word boundary in terminal strings. This separation of the rules for homorganic segments seems to be a necessary one.) Since the gravity feature is not distinctively marked for liquids in the phonological matrices the following phonetic (ph) rule is made to precede, and it specifies gravity for all segments relevantly involved in the rule for homorganic segments, with resulting simplicity in the statement of that rule. phlO

r+cns~| -» [—grv] |_+vocJ

i.e., liquids are non-grave. The homorganic nature of Italian nasals is stated in the following rule which introduces the gravity feature of nasals when they precede other consonantal segments. MSll

i.e., the coefficient of gravity for nasals is the same as that of the following consonants. In this view homorganicity of nasals is taken as a distribution problem (in part), that is treated in the MS rules, rather than (entirely) as an 'assimilation', transformationally. The motivation is that in this theory ellipsis is here feasibly considered and adopted (cf. 2.3). Incidentally, phlO must be given in the grammar sooner or later on independent grounds. We are fixing its order before MSI 1. The beneficial result of the ordering is the full applicability of MSI 1. Failure to give this sequence in the rules reduces the generality of MSI 1; in fact another similar rule would be required for liquids only if phlO were to follow MSI 1. In addition there seems to be no phonological convention of analytical import warning against introducing phlO at this state of the analysis. It is a peculiarity of the Italian which I speak, that only the voiced members of the affricates appear in word-initial position, thus the list (7)

58

A PHONOLOGY OF ITALIAN

zio zuppa zaino

zeppo zoccolo zappa, etc.

has always voiced and never /c/. This regularity allows omission of the feature voice from the lexical representation of initial affricates. Post-word-boundary affricates are thus partially specified as #[—voc, —grv, +shp, —cnt]. In other words and /c/ are neutralized, (represented by the features common to both: their archiphoneme). One of the advantages of the classificatory distinctive feature analysis is exactly this. It allows the theory to avail itself of the important notion of the archiphoneme without having to add extra units to the inventory as the 'Prague-type' did (cf. Chapter 2. 1). The following morpheme structure rule based on the particular environment specifies the value of the voice coefficient. MS12

—voc —grv +shp —cnt

[+voi]

in env.

i.e., an affricate is [+voice] word-initially. Suppose speaker A pronounced all initial affricates voiced, speaker Β pronounced them all non-voiced, and for speaker C they were voiced in some words and voiceless in others. This dialect difference can be formally characterized if we compared the respective grammars of A, B, and C. Specifically we would observe that A includes MS 12 or a similar one, Β differs from A in that the output of MS 12 is [—voi]. C does not contain MS 12 at all, and consequently its lexicon contains affricates specified as to the feature voice also in environment # , initial in words. The following is based on an observation of consonantal clusters, specifically those involving palatals and affricates. Palatals and affricates may be preceded only by the consonant segment /n/ and only if this is non initial. The clusters as illustrated in list (8) pancia calcio mancia gancio

danza calza manzo terzo, etc.

must necessarily be followed by a vowel. For simplicity of statement we shall give now a lower, phonetic rule specifying the nondistinctive feature of sharpness for all unspecified segment-types: namely, vowels, /r/, and /p, b, f, v, m/ formulated in the following rule:

59

A PHONOLOGY OF ITALIAN

phl3

[—ens] —voc +grv _ —cmp] +voc +cns —cnt

l_

[—shp]

The fact that only vowels can follow affricates and palatals can now be stated simply as follows: MS14

[X]

[

—cns~| +vocJ

in env. [+shp]_

i.e., any [X] segment following a [+shp] segment (the category which after ph!3 distinguishes pairwise palatals and affricates from every other segment in the phonological matrices) is a vowel. Three-term consonantal clusters are of the following shape: either CCL both in initial and medial position or LCL only in medial position. Thus the third consonantal segment is always a liquid regardless of its distribution. This we can state as follows: MS15

[+cns]

[+voc]

in env. [+cns] 2

i.e., a consonantal segment, the archiphoneme of C and L, is invariably [+voc], a liquid, when it follows at least two segments which are elements of the class [+consonantal]. Items in list (9) schiena scoppio prato frate

aspro altro contro compro, etc.

exhibit the following. The first of two consonantal segments in word initial position is non-vocalic, as well as the middle member of a three-term consonantal cluster. The following rule formulates the observation. MS 16

[+cns] -* [—voc]

in env. [+cns] _[+cns]

Notice that the three-term consonantal cluster type CCL as illustrated in the second column in list (9) also occurs initially. The value of the features in the first segment is almost fully predicted by MS7 and MS8 and the last segment in part by MS 15. Thus the environment in MS 16 is limited to consonantal clusters types LCL and LL. Since MSI implies that a liquid cannot be first segment of an initial consonantal cluster,

60

A PHONOLOGY OF ITALIAN

then the consonantal segment in the first environment in MS16 is a consonant (and not a liquid), as stated by the rule. Word initial two-term consonantal clusters are of two general types: CC and CL, as list (10) illustrates scope spina stupido

trono cratere plurale, etc.

MS7 and MS8 predict the first segment of the CC type as well as the first segment of the CCL type. The following rule applies only to # C L (initial cluster) illustrated in ist (10), second column. MS17

[+cns] -* [+voc]

in env. # Γ—vocl L-cntJ

i.e., a [+consonantal] segment (unspecified as to consonant or liquid) is to be assigned the feature [+vocalic] (which identifies it as a liquid) when it follows a word initial interrupted consonant. Notice that the environment is limited to /t, d, k, g, b, ρ/. Other interrupted consonants do not appear in this combination in the phonological inventory of the lexicon. Such restriction is characterized by MS 15. Further the type of liquid which may occur is also restricted in the environment of /d/, or /t/ by MS4. Three-term consonantal clusters exhibit the regularity that their middle member is a stop, as in list (11) entrare Australia Stromboli Strega, etc. which we state in the following rule: MS18

[—voc] -* [—cnt]

in env. [+cns]

[+cns]

It seems that non-initial three-segment clusters with [+continuant] middle segment such as in list (12) conflitto infrangere, etc. are best analyzed as con+flitto, in+frangere, etc., thus not qualifying as three-term clusters in these morpheme structure rules. Further sequential constraints are observed leading to increased lexical economy. It seems that /f/ may be followed by either liquid in (word or) morpheme initial position as in list (13)

A PHONOLOGY OF ITALIAN

61

fluto flaccido in+frazione Franco, etc. but non-initially only /r/ may follow /f/ (as well as /v/). In addition /f/ may precede a liquid or a vowel, but not a consonant. Therefore we formulate the following rule: MS19

[+cns]

Γ+vocl [_—cntj

—voc +grv +cnt —nas —cmp

in env.

The longest consonantal cluster has three segments. The set of consonantal clusters has the following subsets: CL, CCL, CC, LC, LL, LCL. Consonantal clusters include vocalic (L) and non-vocalic segments (C). We observe that the length of immediately adjacent vocalic or non-vocalic consonantal clusters is two segments. The set is reduced to its first three members (CL, CCL, CC) morpheme initially. Since the third member in a consonantal cluster is always a liquid we have introduced MS 15. MS 16 constrains the first member of a three-term consonantal cluster. The elements of non-vocalic clusters are also severely constrained. In fact the only one that can enter into combination with palatals and affricates is a nasal, and this can only precede the palatal or affricate. Vowels in word final position (or immediately before word boundary # ) are short. This phonetic observation about Italian is formulated in the following MS rule: MS20

[-ens]

[-Igt]

in env. X_

- #

A further constraint is observed on the tenseness feature of midvowels when they occur before /u/ and /a/. In other words only the non-tense member of the midvowels occur in my dialect. Thus a non-diffuse, non-compact vowel is non-tense before grave and diffuse or compact vowels as for example in list (15) Lea boa dea MS21

[—ens] -*• [—tns]

reo Meo (Patacca)

in env. -dif —grv —cpt

—ens +grv +dif _+cpt_

The final set of MS rules concerns mostly the prediction of the features vocality and consonantality on the basis of adjacent segments. The vowel quality (i.e., [—ens]) of segments is predictable in the following environments: (1) before word boundary # in most words, (2) after a word initial /v/, (3) before /f/ or /v/ preceding a consonantal

62

A PHONOLOGY OF ITALIAN

segment, except where (2) applies, (4) before two or one word-final consonantal segments in N, (a restricted number of barbarisms) and N y (per, con, ...) respectively, (5) after palatals and affricates, and (6) immediate to a nasal-in-cluster. The following rule accounts for environments (1), (2), (3), (4), and (6). Environment (5) has already been taken care of in MS 14. '

MS22

[X]

[-ens]

in env.

#

—voc [+cns] +grv —cmp +cnt —nas [+cns]2'# (NJ [+cns] # (N y ) [ + n a s ] [+cns] #[~-voc +grv +cnt —nas +voi [+cns] [ + n a s ]

In Italian morphemes the longest sequence of consonantal and non-consonantal segments has three members, thus MS23

[X] -* [ x e n s ]

in env. [xcns] 3

and therefore one may state that Italian exhibits most regularly an alternation of plus and minus consonantal segments with a maximum of three segments. Within morpheme boundary one finds no lower bound to the number of consonantal segments (for example s+radicare, and a+ccogliere). Aiol-\-a, ai+a, aut+o, alt+o, ghiai+a, etc., illustrate that if there is a segment (and not a boundary) following a string of three nonconsonantal segments (i.e., vowels), that segment will be a consonantal one (i.e., a consonant or a liquid). MS23 captures in a compact way this most general regularity of the consonantal feature, framing the most basic sequentially recurrent unit. The specification of other features by rules provides a network of constraints or sequential one-choice possibilities tracing the feature sound pattern of Italian or the phonotactics of its phonological matrices. The strongly justified simplicity criterion demands that the number of distinctive feature specifications be kept to a minimum. The task is partly carried out in the phonological matrix of the maximal segment-types. There that hierarchical arrangement is chosen which yields the most symmetrical distribution of features (minimum

A PHONOLOGY OF ITALIAN

63

number of nodes: minimum ratio of decisions per segment) and affords the system of subsets most convenient for the statements in the phonological (P) rules, yet preserving pairwise distinction in the decodable set of maximal segment-types. Additional feature economy is introduced in the underlying phonological spelling of the lexical matrix D x by omitting those features which are observed to occur only and always in just certain environments delimited by features, bundles of features and/or boundaries. The MS morpheme structure rules contain the information about those omitted features. They are feasible if the claim which they make is an observed generality in the language, that is, if the number of features which the rule allows us to omit from the lexicon is greater than the number of features which are needed to state it. However to maintain this application of a simplicity criterion one assumes that feature-specification economy in the lexicon can always be traded against that in the rules. This assumption appears to be doubtful.

3.3. PRECYCLIC RULES

In this section we deal with phonological specification and characterization of the processes of inflection and derivation. Nominal inflection is discussed in section 3.3.1 and verbal inflection in 3.3.2. More specifically 3.3.1 treats the inflection of nouns and 3.3.1.1 accounts for inflection of adjectives. 3.3.1. Inflection of Nouns

Hall (1948, p. 21) displays six different singular endings for nouns, regardless of their gender class: /a/, jo/, /e/, /i/, /u/, and 0 . I present a similar arrangement of the facts for analysis. In addition to Hall's data I include singular endings in /ε/ and /o/. (1)

(i) (ii) (iii) (iv) (ν) (vi) (vii) (viii)

I a/: ferita (f.), telegramma (m.), boia (m.)... /o/: delitto (m.), dito (m.), auto (f., m.), mano(f.), braccio (m.),... /o/: oblo (m.), como (m.), ... /e/: monte(m.), fonte(f.), mille, ... /ε/: caffe(n.), bebe(m., f.), ... /i/: analisi(f-), harakiri (m.), vatussi(m.), ... /u/: domu(f.), menü(m.), ragii(m.), ... /s, m, η .../: lapis (m.), gas (m.), film (n.), camion (m.), recläm (f.), barman (m.), tost (m.), ...

The plural endings of the nouns in (1) are given in (2) in correspondence with singular endings. (2)

(i) / a / - / e / : ferite (f.), gatte (f.), fette (f.), ... (ii) I a/—/i/: telegrammi (m.), drammi (m.), temi (m.), ... (iii)/a/—I a/: boia(m.),...

64

A PHONOLOGY OF ITALIAN

(iv) (ν) (vi) (vii) (viii) (ix) (x) (xi) (xii) (xiii)

joj—Ii/: delitti (m.), auti (m.), raani (f.), ... joj—/a/: dita (f.), braccia (f.), miglia (f.), migliaia (f.),... /ο/—/ο/: auto (f.), moto (f.), foto (f.), ... /o/—jo/: oblo (m.), como (m.), ... /e/—/i/: monti(m.), fonti(f.), ... /ε/—/ε/: caffe (m.), bebe (m., f.),... /i/—ßl: analisi (f.), harakiri (m.), vatussi (m.), ... /u/—/u/: ragü(m.), menü(m.), ... /u/—/us/: domus (f.), ... /s, m, η .../—/s, m, η .../: lapis (m.), camion (m.), film (m.), recläm (f.), ...

The apparent chaos of singular ending (with respect to gender class) shows regularities if we consider that the majority of the items in question can be labelled non-native, including learned and foreign. If this distinction is accepted as a grammatically valid one, then we would state that regularly masculine singular items end either in /o/: delitto, auto, ... or in /e/: monte, conte, .... Similarly regularly feminine singular items end in /a/: ferita, carta, ... or in /e/: morte, corte, .... Furthermore non-native words maintain the original final segment: camion, comö, menü, bebe, ... (from French); telegramma, tema, analisi,... (from Greek); barman, tost,... (from English); etc. ... Irregularly mano violates the regular feminine ending. The corresponding plural endings in (2) likewise show irregular correspondence solely on gender distinction. In this case the process seems to depend on the origin of the word, the phonetic shape of its singular form (if one assumes derivation of plurals from singulars), as well as the gender class. For native regular items as in (2) (i), (iv), (viii), representing perhaps the bulk of the Italian vocabulary we can say that if the singular ends in /a/, then the plural is /e/ and if the singular is joj or /e/, then the plural is /i/, without any reference to gender. Native items like dito, braccio, miglio, migliaio, ... are deviant in that they are masculine in the singular and feminine in the plural, mano peculiarly is feminine but follows a masculine inflectional process. The item mille in (1) (iv) is a numeral, as such inherently distinct from other nouns in that it does not undergo inflection and should not be considered for that process. Hall includes mille in his data. It seems that English one thousand, two thousands, ..., or some other consideration has led linguists to construe mille as the singular of mila. On the contrary I believe that mila should be viewed as a conditioned alternant of migliaio, and mille should be considered purely a numeral like uno, due, ... which do not inflect for gender and number. This suggestion is motivated by the fact that there is already a class of words following a kind of Latin neuter plural inflection: ditoldita, miglio/miglia, ... Non-native items of the Greek type (2) (i) have a plural ending in /i/. For the rest of non-native items the plural ending is the same as the singular. In my variety of Italian the plural o f f i l m , album is not distinct from the singular. However if one were to account for such plurals in /s/ (including Sardinian words) a further subclassification in the lexicon would be required.

65

A PHONOLOGY OF ITALIAN

In a transformational grammar of the kind proposed in Chomsky (1965) inflection can be assigned either morphemically or in terms of syntactic distinctive features. Morphemic inflection involves earlier branching rules. These would rewrite, say, a NOUN approximately as follows: (3) (i) N O U N (ii) AFFIX

Ν AFFIX jFEM SING [MASC PLUR

The syntactic component (including the Lexicon) would then yield structures like NOUN

[ + N , +Native, —Neutralizing,...]

FEM

PLUR

/käs/' to which the phonological rules of inflection would apply in the following approximate formulation: (5)

(i) FEM

"/aI /e/

I [+Native, —Neutral., ...] [+Neutral., ...]

"/o/ /e/

/

0

(ii) MASC

0

(iii)

[+Native, —Neutral., ...] [+Neutral., ...]

[ + G r e e k * ...] /a/ /e/ /o/ /a/ [4-Latin-neuter]

Ν Μ

/

/a/

(iv) PLUR -> 0 and generate the following derivation: (6) (i) # k ä s + F E M + P L U R # #käs+a+PLUR#

(5) (i) (5) (iii)

.PLUR

66

A PHONOLOGY OF ITALIAN

#käs+e+PLUR# (5) (iv) #käs+e# We assume that D matrices of non-native items have been entered in full in the lexicon #NOUN#

[ + N , +Fem. +Plur., +Native, -Neutral.^...]

and the rules of inflection would be the following: (8)

[+Neutral., +P1,...] [+Masc., —Neutral., +P1, ...] [+Greek χ, +P1, ...]

insert Ν

[X]

#

/ο I

{

[+Masc., —Neutral.,...] [+Pseudo masculine ...]

J [ + F e m , —Neutral.,...]

/a/

{[+Greekx, ...] [+Fem, -Neutral., +P1, ...]

Μ

Rule (8) has four unordered, context-sensitive subrules. It inserts /i/, /e/, /o/, /a/ in the environment X # where X is a phonological matrix and the whole structure is a [ + N , +Native] except where otherwise specified. The segments in question are to be inserted only if the parallel syntactic condition holds. In this grammar we shall adopt the syntactic-feature suggestion for inflection of nouns and consequently the phonological rule (8). The motivation for adopting (8) instead of (5) is patent. The morphemic solution (5) requires additional rules, such as (3), earlier in the grammar, whereas the syntactic feature solution (8) does not. Furthermore (8) is only one insertion rule with four unordered outputs, whereas (5) consists of four transformations consisting of eight inputs and ten outputs. Moreover (8) uses for segment-types in toto, whereas (5) uses eleven segment-types (specified to some degree). Finally technical rules deleting the symbols MASC, FEM, and PLUR are needed in (5), which render the description clumsy without any linguistic contribution at all. 3.3.1.1. Inflection of

Adjectives

Italian adjective inflection is simpler than, but similar to, noun inflection. The difficulties caused by unassimilated borrowings, handled by partitioning the lexicon,

67

A PHONOLOGY OF ITALIAN

does not arise for adjectives. I shall use again a display of the facts in question similar to Hall's (1948, p. 21) presentation. Singular adjective endings are found in (9). (9)

(i) /a/: buona, cara, bianca, ... (f.) (ii) jo/: buono, caro, bianco, ... (m.) (iii) /e/: fertile, forte, verde, ... (m., f.)

Like nouns Italian has a class of adjectives that mark gender ((9) (i) are feminine and (9) (ii) are masculine) and a class that ambiguously uses /e/ for both masculine and feminine gender. The plural endings of the adjectives in (9) are given in correspondence with the respective singulars in (10). (10)

(i) /a/—/e/: buone, care, bianche, ... (f.) (ii) /o/—/i/: buoni, cari, bianchi,... (m.) (iii) /e/—/i/: fertili, forti, verdi,... (m., f.)

The data in (9) and (10) do not include numerals nor the type /a/—/i/, which Hall exemplifies with idiota—idioti. Numerals are excluded because they do not inflect. idiota is considered a [+Greek x, ...]. For the same reasons given for nouns we adopt the syntactic feature presentation for phonological specification of inflection. On substructures like #ADJ# [+Adj, +Masc., —Neutral

]

biank the following rule is defined (12)

insert

"/i / /e/

[+Neutr., +Plur. ...] [+Masc. + P l u r . . . . ] ^Neutral.,...] . # , where [+Masc.,...] [+Fem, +Sing.,...]

/ο/ I /a/ Rule (12) is identical with rule (8) except for the syntactic features involved. The two rules can then be profitably collapsed into one. In 3.3.1.2 we have assumed an agreement rule which has extended the complex symbol of the adjective to include the concord features of the noun it modifies. 3.3.2. Inflection of the Verb

In this section we seek, for the regular predicated verb, a characterization of the

68

A PHONOLOGY OF ITALIAN

inflectional processes of the phonological component. Predication involves the following data, representing the phonemic reflections of theme vowel, person and number. The rows represent first, second and third person singular and plural. The columns give the endings of the verbs classes are, ere and ire respectively. (1)

(i) Present Indicative: -i-0 -i ^a -iämo -äte -ano

-ί-Ο -i -e -iämo -ete -ono

-iämo -ite -ono

(ii) Imperfect Indicative: -ävo -ävi -äva -avämo -aväte -ävano

-evo -evi -eva -evämo -eväte -evano

-ivo -ivi -iva -ivämo -iväte -ivano

-ei -esti -e -emmo -este -ero(no)

-ii -isti -i -immo -iste -irono

(iv) Present Subjunctive: -i -4 -i -iämo -iäte -ino

-a -a -a -iämo -iäte -ano

-a -a -iämo -iäte -ano

(v) Imperfect Subjunctive: -ässi -ässi -ässe -ässimo -äste -ässero

-essi -essi -esse -essimo -este -essero

—issi -issi -isse -issimo -iste -issero

±0 -i

(iii) Preterite: -äi -ästi —0 -ämmo -äste -ärono

A PHONOLOGY OF ITALIAN

69

Predication excludes non-finite and compound verb forms. Compound forms in this analysis are considered derived forms consisting of a predicated form and other elements. Compound forms are verb forms consisting of a predicated auxiliary verb plus a past participle or a gerund such as: sono + arrivato, aveva + mangiato,parti + piangendo,.... Characteristically we assign to them also the future and the conditional. These are considered to be syntactic compounds analyzed as a particular form of the predicated auxiliary verb avere (present for the future and preterite for the conditional) plus an infinite. Unlike traditional compounds and like predicted forms these undergo a change in sentential order: the auxiliary and the infinitive exchange place. Like predicated verbs and unlike traditional compounds the auxiliary forms one stress-word with the main verb: the infinitive. Special morphophonemics adjust the phonetic shape of the future and conditional. The data in (1) is arranged in three columns representing the affixal ending of the three inherent subclassifications of the verbs: -are, -ere, -ire, respectively. Notice that -ere subsumes both the paroxiton and the proparoxiton class. As mentioned in the case of nominal inflection a transformational grammar allows for both morphemic verbal inflection as well as inflection in terms of syntactic feature (paradigmatic inflection). A morphemic analysis would require an analysis of the symbol VERB as Base and inflexional Affix, and require the following (or similar) categorial rules. (2) (i) VERB - • BASE + AFFIX (ii) AFFIX THEME VOWEL + TENSE + ENDING Imperfect (iii) TNS Past Present and the following rules of concord. (3) (i) TV

(ii) Ε

TV-1 TV-2 TV-3

/

[+Base-a, ...] [+Base-e,...]. [+Base-i,...] W

Ei E2

E3 where the thematic vowel is subcategorized on the basis of the inherent feature of its base and the ending is subcategorized for person. The following rules of predication also assume that subjunctive and number markers have been assigned. (4) (i) Pst (ii) Plur.

Ε Psti Pst 2 Pli Pl 2

Pl 3

/ Ei

E2 E3

70

A PHONOLOGY OF ITALIAN (iii) T V

TVi

TV3-

TV2>3

TVj.

->

_TV3 (iv) I m p e r f .

J pres.+subjunctive

TV2

[+stress]

pres.+E3+subj.

E1.2 + P I +

Subj.

Pst

( v ) T V -»· [ + s t r e s s ]

Impf., Subj. Pres., E1>2, Plur. (vi) M A I N S T R E S S R U L E J Pst + E 3 + (vii) T V -»· [ — l e n g t h ]

PI

[Impf.+Subj. i T V + Pres. + E x + PI ( + S u b j . )

(viii) i n s e r t

/i/

[ T V + Pres. + E 2 + PI + S u b j . flmpf. +

(ix) i n s e r t -»· / r o /

E3

+P1 +Subj.

|Pst + E3

+P1

{Pres.+E1+Pl (x) T V

/a/

[Pres.+E1>2+Pl+Subj. Pst+E2

(xi) i n s e r t (xii) P s t x - •

/s/ /te/

(xüi) T V 2 . 3 (xiv) T V !

Prs+E3+Pl

lol /a/

(xv) T V 2 -

/e/

(xvi) T V 3 -

m

(xvii) P l 3 - •

0

[impf. +

E3.

[BaseI +

Pst.

(xviii) I m p f . -> /s/ /va / (xix) [ — c o n s o n a n t ]

+Subj.

0

ΕΛ E

(XX) / a / (xxi) Ε

,+Subj.

+P1

4

/ I 0

Pres.

+Subj. +PI+Subj.

(xxii) E x

iPst

'/i/

j/s/

+Subj.

(xxiii) E 2 - » / i / ( x x i v ) E 3 -»• Γ / β /

L

0

(xxv) [ — c o n s o n a n t ] (xxvi) P U

/mo/

(xxvii) P l 2

/te/

(xxviii) P l 3

/no/

/ 8/ + S u b j .

0

f/s/

+PI+Subj.

]/t/

+P1

71

A PHONOLOGY OF ITALIAN

(xxix) fPst 2 I -»· 0 (Pres.] The ordered set of rules above plus the cyclic rules of ellipsis, consonant lengthening, and glide, account for predicated verb inflection. They operate on strings like (5) which are derived with the following superstructure. VERB

[ + V , Gend-a,...]

Rules (4) apply one at a time to the string (5) in the manner illustrated by the derivation (6): pari pari pari pari pari pari pari pari

+ + + + + + + +

TVX + pres. + E t + PI TV^ + pres. + E, + PI t TVi + pres. + E t + Pli i TVX + pres. + Ej + P^ iä + pres. + Ex + Plx iä + pres. + Pli iä + pres. + mo iä + mo

Rule (ii) (v) (viii) (x) (xxi) (xxvi) (xxix)

Rules (4) assign unambiguously and in a mechanical manner the correct ending (presented in the paradigm (1)) to all possible strings of the form (5) generated by the syntax. The statements (4), although complicated and perhaps a bit awkward, are maximally explicit, and in their explicitness they meet one of the basic aims of transformational grammar. They are motivated by the simplicity criterion assumed for phonological descriptions. In fact a surprisingly small number of segments (actually their features) is employed to attach automatically the correct endings to verb bases. Now, assuming explicitness for grammatical presentation, let us consider an approach to inflection of the predicated verb based on syntactic features. Rules (9) mechanically and unambiguously assign each ending in (1) to phonological lexical matrices on the basis of its extended complex symbol. We assume that this symbol has features developed in the branching rules, subcategorization rules, lexicon, and transformational rules. In order to comply with the simplicity criterion that phonological specification

72

A PHONOLOGY OF ITALIAN

must be kept at a minimum, rules (9) are given in a form which I discuss directly. Rules are of two general forms: (7) (i) insert -»· A/X (ii) A [ χ F]/X

#, #,

(a) Complex Symbol (aa) Complex Symbol

(7) (i) Reads: insert a segment A in the context X # (X = Phonological matrix and # is the word boundary), if the structure (X # ) is a given Complex Symbol. (ii) Reads: assign the feature x F t o a segment A in the context X # , if the whole structure is a given Complex Symbol. In other words, form (i) annexes a new segment to the right of a phonological matrix, whereas (ii) assigns a given feature to the last segment of a phonological matrix. We label the context of form (i) with single small letters of the English alphabet (a) and form (ii) with double letters (aa). On the basis of this notational convention we simplify (7) as (8). (8) (i) /A/ I (a) Complex Symbol (ii) [ χ F] / (aa) Complex Symbol One last remark about the notation of rules (9). They are unordered with respect to the labels (i)-(xii). That is one may start with any one of them. However they are ordered with respect to the English letters of the alphabet: (a, aa, ..., b, ...). To apply the rules one begins with context (a) and makes one pass through rules (i)-(xii) in any order, than goes to context (aa) and again makes one pass through the rules in any order, and so forth for (b),..., (c), ... until one pass ha& been made for each lettered context. Finally in entering the complex symbol we have assumed a redundancy criterion used for lexical entries: if either Plur. or Subj. is not specified in an item of (9) that item is Singular or Indicative respectively. (9)

(i) /I/

/

(ii) /a/

/

(a) Present, Person 1-2, Plur., Subjunctive, Present, Person 1, Plur. (b) V-i, Imperfect, V-i, Preterit, (e) Imperfect, Person 1, Plur., Subj., (h) V-a, Present, Person 2, Subj., Person 2, ± Subj., — [V-e-i, Pres., Subj.] Person 1, Preterit, Person 1, Imperfect, Subj., V-i, Person 3, Preterit, V-i, Present, Person 3, Subj. (b) Present, Person 1, Plur., V-a, Imperfect, ± Subj., V-a, Preterit, —[Person 3],

A PHONOLOGY OF ITALIAN

(iii) /el

I

(iv) /ο/

I

(v) /r/

I

(vi) /sI

I

(vii) HI

I

(viii) (ix) (χ) (xi)

/ν/ I /m/ I /η/ I [—length] /

(xii) [+stress]

/

Present, Person 2, Plur., Subj. (d) Imperfect, Plur., (h) V-e-i, Present, Subj. V-a, Present, Person 3, Imperfect, Person 3, (b) V-e, Imperfect, ± Subj. V-e, Preterit (e) Imperfect, Person 3, Plur., Subj., (h) V-e-i, Present, Person 3, V-e, Preterit, Person 3, Imperfect, Person 3, Subj., Person 2, Plur., (b) V-e-i, Present, Plur., V-a, Preterit, Person 3, (h) Present, Person 1, Imperfect, Person 1, (j) Person 1, Plur., ± Subj., Person 3, Plur., ± Subj., (f) Imperfect, Person 3, Plur., Subj., Preterit, Person 3, Plur., (c) Imperfect, Subj., Preterit, Person 2, (g) Preterit, Person 2, Person 2, Plur. ± Subj., (c) Imperfect, ± Plur., (i) Person 1, Plur., ± Subj., (i) Person, Plur., ± Subj., —[Preterit], (bb) Imperfect, Subj. Preterit, Person 1, Plur., (bb) Present, Person l-(2, Subj.,) Plur. Preterit, ±Plur., Imperfect, + Person 3, Plur., Imperfect, ±Plur., Subj. (dd) Imperfect, Person 1-2, Plur.,

Rules (9) apply to substructures in the form of (10), generating derivation (11)

74 (11)

A PHONOLOGY OF ITALIAN

(i) (ii) (iii) (iv) (v)

parl+i parl+ia parl+iä parl+iäm parl+iämo

Rules (9) (i) (a) (9) (ii) (b) (9) (xii) (bb) (9) (ix) (i) (9) (iv) (j)

We have considered two approaches to verbal inflection in this section. Both approaches assume that the process must be generative and maximally explicit. Rules (4) have a morphemic basis as illustrated in (5), whereas rules (9) have as basis a complex symbol as illustrated in (10). The morphemic approach employes 29 rules and the syntactic feature view uses only 12. Furthermore, and more significantly, the syntactic feature view uses fewer feature specifications than the morphemic approach. The former is adopted on a crude simplicity: fewer rules and phonological features. However neither the paradigmatic nor the morphemic solution yields linguistically explanatory results. It might be that such goals cannot be achieved in principle in describing Italian verb inflection. 3.3.3. Word Formation A discussion of word formation is relevant for a precise statement of the stress rules. The general linguistic processes involved are not quite clear. The following lists illustrate some of the problems. (1)

(a) telefono telefonico telefonia (c) precetto concetto preconcetto (e) bomba bombola bomboletta

(b) perimetro perimetrico perimetria (d) superuomo ultraforte arcivecchio (f) servitu cittä cittadino

The above words (with primary stress indicated) illustrate the following superficial information: (2)

(i) A stem can have as many as two prefixes in pre+con+cetto and two suffixes in bamb+ol+ett+a. (ii) The two affixes must occur in a fixed order. (iii) In (a) and (b) the prefixes tele-, peri, carry the primary stress only when no suffix is present. (iv) Although tele- is interchangeable with peri- in the context +metro, this does not hold for peri- in the context +fono. (v) Likewise the suffix -ol- of bombola does not go with any other word in the list.

A PHONOLOGY OF ITALIAN

75

(vi) Finally there is cittd, civile, civiltd, cittadino, and servitü, servile, but no serviltd, servitMine and so forth. The problem seems to be one of selection within word boundary, apparently similar to selection at the sentence level in syntactic description. The following is suggested as a reasonable analytical approach. As proposed by Chomsky (1965, p. 188) we allow branching rules to operate on the complex symbol. The following rules: (3)

(i) (ii) (iii) (iv) (v) (vi)

[+N] PREF SUF ÄFF [+F] STEM

(PREFIX) + STEM + (SUFFIX) (Pfl) + (Pf2) (Sfl) + (Sf2) + (Sf3) [ + F , ...] [ + G , ...] Complex Symbol

yield substructures like the following:

Rule (3) (vi) subcategorizes the stem for each possible combination of prefixes and suffixes (actually the features of their complex symbols). Rules (3) characterize the structural information (2). Specifically sequential order is insured by rules (3) (i)-(iii), and selectional frames by (iv)-(v). Rules (iv) and (v) introduce inherent affixal features F and G. We might define F as the feature which is common to tele-, peri-, alti, ... in the environment +metro, say the feature MEASURE, and G as the feature which is common to tele-, grammo-, in the environment +fono, say the feature SOUND. Rule (vi) subcategorizes STEM on the basis of the frames in which it appears: one subcategory for each possible frame. The general lexical rule here chooses the stem also on the basis of the particular affixal structure of the word. For example, given strings:

76

A PHONOLOGY OF ITALIAN

(5) (i) #[Prefix, + F , + G ...] + [ + S t e m + , [+Pref + F , + G , ] ...] . . . # (ii) #[Prefix, + F , - G , ...] + [+Stem, [+Pref + F , - G , ] ...] . . . # and the lexical entries: (6)

(i) (ii) (iii) (iv)

(/metr/, [+stem, +[Pref + F , - G ] ...]) (/fon/, [+stem, +[Pref +G],...]) (/tele/, [+Pref, + F , + G ...]) (/peri/, [+Pref, + F , —G, ...])

the lexical rules assign either (6) (i) or (ii) as stems for (5) (i), but only (6) (i) for (5) (ii). In other words they will insert either /metr/ or /fon/ in the context /tele/+ , but only /metr/ in the context /peri/+ . Lexical entries of affixes are minor features of phonological relevance. In Italian we distinguish a class of suffixes which always carry stress: (7)

(i) (ii) (iii) (iv) (v) (vi) (vii) (viii)

-in: pallina, piedino, motorino, ... -on: pallone, piedone, cannone, ... -εΐΐ: cannelle, fratello, anello, ... -ol: figliolo, spagnolo, cannolo, ... -i: pazzia, razzia, democrazia,... -il: servile, porcile, senile,... -ett: carretto, amichetto, boschetto, ... -ta,-tu: civiltä, servitu, ...

Another class requires stress on the preceding vowel (8)

(i) -in: (ii) -ol: (iii) -ik: (iv)-il:

rüggine, frässino, lamina,... bämbola, gondola, donnola,... medico, scientifico, mägico, ... äbile, facile, nobile,...

Prefixes unlike suffixes appear to have no effect on stress. The stress assigning rule is not limited by prefixation. In our theory based on vocalic length we need not give syntactic features to predict affixal stress. We shall enter the phonological matrices of affixal entries of type (7) with a [+length] vowel, as opposed to those of type (8) whose vowel is spelled with [—length]. Only the exceptions (7) (iii) and (vii) which must be set up as /εΐ, et,.../ for other reasons will be introduced in the lexicon with the syntactic feature [takes primary stress]. For assigning stress we shall refer to affixal information only in the exceptional cases of /εΐ, et/ as formulated by the following rule: (9) [—cons]

[+stress]

/

(a) [+Suffix, + takes primary stress, ...] (b) General Stress Rule (3.4.2.)

77

A PHONOLOGY OF ITALIAN

where the environments are disjoint: (a) accounts for residual exceptions and the general stress rule for the rest. This section of word formation has dealt specifically with non-productive derivational processes. Productive processes are of course simpler to handle. They involve mainly suffixes, namely, diminutive, augmentative, pejorative, superlative, etc., whose phonological shape is respectively /in, et, εΐ; ön; aö; isim; .../. These suffixes seem best introduced by a process transforming, say, Ν -» N-d as N[ ] N.d[N[ ]+suffix]. This solution is phonologically motivated in that the occurrences of alternation (10) (i) [pedone]/[pyed0ne] are explained as (ü) N[pyed+ön+e]/ N .,i[ N [py8d+e]+ön] that is, the rule in 3.4.3 ('dittonghi mobili') applies to nouns but not to derived nouns. Thus the glide is deleted if the diphthong is not stressed yielding the first alternant. The hypothesized derivation also explains such extended constructions like bambolettina as N . d [ N . d [ N [bamb+ol+a]+et]In]. 3.3.4. Alternations In this section we formulate rules predicting a number of restricted (morphophonemic) alternations. In 3.3.4.1 we deal with palatalization. 3.3.4.1. Palatalization In distinctive features we find it consistent the use of the term SHARPING to cover both palatalization and affrication, since we have found it feasible to account for both processes by operating on the feature sharp, as we shall see in 3.4.4. In this section however we deal with palatalization proper, the phenomenon which results in alternant forms like [amiko] 'friend' versus [amiöi] 'friends' in the case of nouns and [lego] Ί read' versus [legi] 'you read' in the case of verbs. The phenomenon of palatalization proper consists generally in transforming the segments /k, g/ into /δ, g/ before front vowels. Specifically we distinguish affixal palatalization for nouns and inflectional palatalization for verbs. Rule PI deals with the former and P2 with the latter. PI

[—shp]

[+shp]

/

X —voc +grv -dif —voi

—ens where X —grv +dif

i.e., /k/ ->• /δ/ before /i/, if the substructure X

Y = [+suffix,...]

Y is a suffix.

78

A PHONOLOGY OF ITALIAN

PI transforms items with suffixes like: /mon+ak+i/ -» /mon-f aö+i/, /med+ik+i/ -»· /med+iö+i/, / m e d + i k + i n + a / / m s d + i ö + i n + a / , ... but not those without suffix like: / f u o k + i / , / e k + i / , . . . Notice that this phenomenon occurs also with the segment /g/ for some speakers, in which case we have [antropo+log+i], ··· In my variety of Italian the phenomenon may or may not occur in the case of the item above, in others definitely does not occur. This puts in doubt the predictability of g/g for suffixal palatalization. There is however the suffix / + i a / before which palatalization always occurs: [antropo+log+ia],... If I had consistent palatalization in the case of /g/, PI would be generalized by omitting reference to the feature voice, which distinguishes /k/ from /g/. P2 [ - s h p ]

[+shp] I —voc +grv -dif

+ Γ—cnsl [ + V , -[Gend-a], ...] grvj

i.e., /k/ and /g/ are palatalized respectively to /δ/ and /g/ when they occur before front vowels /ε, e, i/ and the item is a Verb (except -are verbs). P2 transforms /dik+i/ /diö+i/, /dik+e/ /diö+e/, /leg+e/ -»· /leg+e/, etc. but will not apply on /seg+i/, /pag+i/,... which are -are types of verbs. There is one more rule which we must add in order to adjust the constructs resulting from palatalization of verbs with the suffix / + i s k + / like [finisko]/ [fini§i], [finiSe], [kapisko]/ [kapi§i], [kapiSe], .... We formulate the following rule. P3

—voc +grv -dif —voi +shp

i.e., /δ/

[—cnt] I

—voc —grv —shp +cnt —nas

/§/ when it follows /s/.

Rules P3 supplemented by the independent rules of Ellipsis and Consonant lengthen ing (3.4.5.2, 3.4.6) yields the following derivation

(1) k a p + i s k + i kap+is5+i kap+is§+i kap+i§+i kap+i§+i

P2 P3 (Stress), Ellipsis Consonant lengthening

3.4. PHONOLOGICAL CYCLE

The rules of the syntax have provided the structural information with respect tc

A PHONOLOGY OF ITALIAN

79

sentential order, categories, subcategories, and relations, all summarized in the form of complex symbols. The lexicon has provided phonological matrices of stems and affixes through the lexical rules. The phonological rules of inflection have attached endings marking gender, number, and predication. At this stage of the grammar we have strings of incompletely specified segment-types in concatenation and bounded by phrase, word, and morphemic (derivational, inflectional) markers. The strings still retain their syntactic superstructure. The task left to us is that of specifying the coefficient of every as yet unspecified feature. This is partly done by the phonological redundancy rules (MS rules above) and by the phonological rules. We distinguish the latter into Phonological (P) rules which perform phonological operations of deletion, transformation, etc., (partly on the basis of its syntactic superstructure) and Phonetic (ph) rules which FILL IN non-significant features. The latter would also change the binary coefficients of features into specific indices approximating the various degrees of voicing, tenseness, ..., found in actual speech. A phonological cycle is a series of rules applying repeatedly on strings in a given order. The application is based on the constituent structure of the string. The rules apply successively once to each constituent. After each pass the constituent bracketing structure considered is erased. The phonological cycle of Italian presented in this section applies first to the smallest constituents, i.e., within morpheme boundary ( + + ) , performs certain specific operations, erases the innermost, morphemic or inflectional/derivational bracketing (or parentheses) and reapplies to the whole string which is now within word boundaries (# # ) . The cycle is repeated until the bracketing is used up. The operations performed by the phonological cycle of Italian include stress assignment, characterization of the DURATION RHYTHM of length between a vowel and the following consonant, glide (semi vowels), and ellipsis. Stress is assigned by the same set of rules to monosyllables and longer words and phrases. The rules assign primary, secondary and tertiary stress. The DURATION RHYTHM accounts for all long (double, tense) consonants on the basis of the length of its preceding vowel, within word boundary. The same rule in its cyclic reapplication across word boundary accounts for the so-called rafforzamento. The same glide (semi-vowels) rule accounts for the phenomenon across word boundary. Moreover dittonghi mobili, ellipsis, and apocope are included. The rules are presented in extended form for easier discussion. In order to achieve as simple distinctive-feature statements as possible I introduce here some phonetic (ph) rules filling in non-distinctive specifications. The need for such advanced specification has been discussed in dealing with phonological redundancy. The following phonetic rules specify the features tense, diffuse, and compact for vowels, and nasality and continuancy for stops. phi Γ—cnsl -*• [—cmp]

80

A PHONOLOGY OF ITALIAN

i.e., the vowels /ε, e, i, έ, e, 1/ are non-compact. ph2

Γ-cnsl

[—dif]

i.e., /a, ä/ are non-diffuse. ph3

["[—ens]!

[—tns]

i.e., the vowels /i, u, a, I, ü, ä/ are non-tense. ph4

—vocl -> [—cnt] -dif +grv

i.e., the consonants /k, g, δ, g/ are non-continuant. ph5

Γ—vocl -> [—nas] L-cntJ

i.e., stops /t, d, c,^, p, b, k, g, δ, g/ are non-nasal. Notice that in order for ph5 to apply generally to all stops ph4 must precede. The following rules (6-26) constitute the phonological cycle. 3.4.1. Glides P6

I

[-ens].

i.e., the high vowels /i, u/ become respectively glides (semi-vowel/-consonant) [y, w] when they are adjacent to a vowel. In distinctive features, rule P6 applies to all segments with the feature composition at the left of the arrow and in the sequential context indicated by the blank at the right of the slant line. The rule consists of changing the coefficient of the feature vocality from + to —. Notice that it is not necessary to express [+voc] at the left of the arrow, since all [—ens] segmente have been marked [+voc]. This change in the feature vocality forms a new set of segments uniquely identifiable as [—ens, — voc, ...]. The only elements of the set will be the two glides, of course. Thus on the basis of the features vocality and consonantality with their binary values four subsets are specified, namely [+voc, —ens, ...]: the vowels, [+voc, +cns, ...]: the liquids, [—voc, +cns]: the consonants, and [—voc, —ens,...]: the glides. The following partial derivations result from the application of P6.

A PHONOLOGY OF ITALIAN

(1)

#man+i+#e#pied+i# #man+i#e#pyed+i# #man+y#e#pyed+i#

P6 cycle I P6 cycle II P6 cycle III

81

#klaudi+o# #klawdi+o#

3.4.2. Constituent Stress A detailed statement on Italian stress is given in Chapter 2, section 4. Here we state the rules in distinctive features and illustrate how they work through derivations. Because of the nature of cyclic ordered rules we present the rules in parts for discussion. The stress-assigning rules are divided into PRIMARY STRESS presented in 3.4.2.1, SECONDARY STRESS in 3.4.9., and PHRASE STRESS in 3.4.10. 3.4.2.1. Primary Stress The following rules assign primary stress within word boundaries. Stress is assigned in two consecutive and overriding cycles. The following are the rules assigning primary stress to stem formatives and words: P7

Γ-cnsl [_+vocJ

[+str]

/

[+cns]

T[+cns] Ί [[-Suffix]]

i.e., assign stress to all one-vowel (monosyllabic) stem formative. P8 P9 ΡΙΟ

Verb Stress Rules (cf. 3.3.2) Affixal Stress Rules (cf. 3.3.3) Γ—cns~| [+str] / Γ Ί L+vocJ

b

n S

J

#

i.e., stress the ultimate segment of words ending in /ε, ο, έ, 5/ if the domain is a word. Pll

Γ-cnsl L+vocJ

[+str]

/

. # , [ + N , - N a t i v e , , ...]

i.e., stress that ultimate vowel of words of the subclass [—Native,]: astracan, nadir, bazar, etc. P12

-cnsl +vocJ

[+str]

/

X —ens X Γ - c n s l X # X = [+cns], 0 +voc L+vocJ + -Igt

i.e., in all other words stress the antepenultimate vowel if the penultimate is short. P13

Γ-cnsl |_+vocJ

[+str]

/

X Γ - c n s l X # , X = [+cns], 0 |_+vocj

i.e., stress the penultimate vowel of the remaining words. Briefly P7 is needed to assign stress to monosyllabic words and to provide the basis for lengthening a consonant following a stressed short vowel when this situation occurs

82

A PHONOLOGY OF ITALIAN

within morpheme boundary, i.e., during the first cycle of the rules. The relevance of P7 will become clearer when we return to it in connection with the consonant lengthening rule PI8 which relies on P7. Rules P8 and P9 concern the morphological bases of stress. The former refers to the predicated verb and the latter accounts for residual affixal exceptions. Both have been discussed in details in 3.3.2 and 3.3.3. Rules Ρ10 and P l l take into consideration lexical exceptions. The former, based on the special phonological shape of substantial number of French borrowings, predicts the oxytonic stress patterns of words like /oblo, bine, gabare, gigold, rokokd, gile,.../. The latter makes use of syntactic features marking the true exceptions: Iastrakan, nadir, bazar/ and perhaps a few more. This handful of words are finally stressed with no apparent conditioning factor. In our theory we could assign stress phonologically by marking their last segment long, and state an appropriate rule. This would be adequate but certainly no less ad hoc than rule P l l . For more information on the marginality of the domain of PI 1 see 2.4. Rules P12 and P13 account for the great majority of Italian words. They account for most paroxytons and proparoxytons. PI2 and PI3 are stated basically as the stress rule for Classical Latin. The only difference is that stress is assigned by P12 to the antepenultimate vowel if the penultimate is short, whereas in the Latin rule brevity of the penult was defined on both brevity of the vowel and closeness of syllable. The reason for the difference in our rule is motivated by the fact that a number of words have been acquired in Italian with a proparoxytonic pattern in which the penult is in a closed syllable. I am referring mostly to place names like /otranto, ofanto, lepanto, .../. Some partial derivations illustrating rule P6-P13 follow. The index 1 refers to primary stress (2)

(i) # k l a u d i + o # klawdi+o

#käpu+än+o#

P6

Cycle I Cycle I

P7, 13

Cycle I

käpu+än+o

P6

Cycle II

käpw+än+o

P13

Cycle II Cycle I

kapw+än+o

P13

Cycle I

P12

Cycle II

1

klawdi+o

1 1

1

klawdy+o

1

1

klawdy+o # k 5mod+o# (ϋ) 1

k5mod+o 1

k5mod+o (iü) # r u g + i n + e # X

rug+in+e

P7

1

rug+in+e

P12

A PHONOLOGY OF ITALIAN

3.4.3. 'Dittonghi

83

MobilV

The following rule PI5 predicts the alternation of [o/wa] and [e/ye]: as illustrated in [py8de]/[pedone], [bwono]/[bonta], .... We set up every occurrence as /ie/ and /UD/. Rule P6 will turn /i, u/ into glides, PI5 deletes the glides before long, lax, unstressed midvowels, and a later general rule will tense unstressed midvowels. P15

]

—voc —ens

0

— ens +voc —tns +lgt —str

/

[—[+Derivative]]

i.e., delete a glide before an unstressed long, open vowei. (3) (i) # k u 5 r + e # 1 kwör+e kw5r+e

Cycle I

(ii) # k u ö r + ä g + o # 1 kwör+äg+o

P6, 7 Cycle I P13

Cycle II

P15

Cycle II

kwör+äg+o 1 k5r+äg+o

Rule PI5 excludes Noun-derivative; in fact there is [pedone] as well as [pyedone]. The latter form is considered to belong to a subclass of nouns with the structure Nd[N[pyede]+suffix] (cf. 3.3.3). 3.4.4. Ajfrication This phenomenon is subsumed under the term SHARPING to include also Palatalization (3.3.4.1). Both processes in fact consist basically in changing the value of the feature sharp from negative to positive. Affrication proper predicts a large number of alternants like t/c, d/ξ, in [grata, gracya ...]. P15' [ - s h p ]

[+shp]

+

/ +cns —voc —cnt

i.e., /t, d/ become [c,

—voc —ens +grv

[

—cns~| +vocJ

before yod followed immediately by a vowel.

The rule above predicts a large number of alternants like forte/forza, grata/grazia, media/mezza, .... A sample derivation follows: (1)

(i) / mad+ia

(ϋ)

fort+ia

(iii) grät+ia /

84

A PHONOLOGY OF ITALIAN

med+ya me^+ya me^+ya

fort+ya förc+ya

grät+ya gräc+ya

Glide, Stress Affrication Cons, length

The rule of affrication must be complemented by one which predicts when the yod is retained and when it is deleted. In fact we want deletion in (i) and (ii) but not in (iii). Rule P20 predicts the glide deletion. 3.4.5. Ellipsis Ellipsis concerns both consonants and vowels. Consonant ellipsis corresponds to what has been generally treated as assimilation. A detailed statement of the motivations for this analysis is in 2.3. Vowel ellipsis is treated in 3.4.5.1 and consonant ellipsis in 3.4.5.2. The following independently motivated rule is needed before ellipsis and consonant tensing rules P16

+voc~| —ens J

[-igt]

J[+cns] :

I

l# i.e., a vowel is short before two consonants or word boundary. 3.4.5.1. Vowel Ellipsis P17

"—ens

-> 0

/

+

—ens +VOC

+VOC

#

—str X

V

—ens +VOC

i.e., unstressed vowels are elided in two environments: (a) before an identical vowel across morpheme boundary (within a word), and (b) before any vowel across word boundary (this occurs in free variation with gliding for high vowels characterized by P6). (4)

(i)

#prinäpi+i#

C.I C.I

P17

prinöipi+i ι prinöipi+i ι ρπηδϊρ+i

P16

ρπηδϊρ+i

C.II

P12 P16

C.I

C.I

(ϋ) P7 P16 P6

tf+i # 1 ^i+i 1 zii+y

C.I C.I C.II

85

A PHONOLOGY OF ITALIAN

(iii) # l a # am + ik + a # # l a # am + Ik + a #

C.I, P7, C.II, P16,

# 1 # am + Ik + a #

P17.

It seems interesting that Italian orthography uses in some cases the phonological rather than the phonetic representation. For example /prinöipi+i/ is often written prirtcipi and one finds commercial sings for the same establishment in the forms "VINI Ε OLII/VINI Ε OLI". The latter form is often home made, a less official spelling. 3.4.5.2. Consonant Ellipsis

Consonant elision eliminates impossible clusters of consonants brought in contact by derivation within words and by lexical choice in longer sequences. Readjustment of an impossible cluster is carried out in two steps by consonant ellipsis and lengthening of the remaining consonant, to reestablish the balance within Porena's (1942) METRIC SYLLABLE.

P18

~[+cns],

0

/

. ( + ) [+cns] x .

[—voc] Γ+grv" +dif +cnt —nas [—cnt]

[—voc]

—voc —grv —shp +cnt —nas

[+shp]

r —voc~| 1 +nasj

T+vocI [_+cns J

i.e., a consonantal segment is elided (at boundary) in the following contexts: (a) it it precedes an identical consonant, (b) if it is a /v/ or /f/ or any stop before any consonant, (c) if it is an /s/ and precedes a sharp (palatal/affricate) segment, and finally (d) if it is a nasal before a liquid. The following are some derivations: (5) (i) # d e k + t + o # dek+t+o

(ii) # ö i v + t ä d + e # C.I, P7

öiv+täd+e

C.I, P7

86

A PHONOLOGY OF ITALIAN

1 dek+t+o

P16

1 öiv+täd+e

P16

de+t+o

P18

δΐ+täd+e

P18

3.4.6. Consonant Length and'Rqffbrzamento' Extensive treatment of this topic is given in Chapter 1, and 2.2. The following general cyclic rule predicts all long consonants both within a word and word initial (rqffbrzamento). P19 [+cns] - [+lgt] I

+voc —ens +str -Igt

_[+voc]

i.e., a consonantal segment is long if preceded by a short, stressed vowel, but only if followed by a vowel or a liquid (in open syllable). (6) (i) # d e k + t + o # de+t+o

C.I, P7, 16, 18

de+t+o P19 (ii) # 6 i v + t ä d + e # δί+täd+e

C.I, P7, 16, 18

δί+täd+e

P19

δί+täd+e C.II, P13 ι ι (iii) # p a r l + o # f o r t + e # C.I, II, Verb Stress, #parl+o#fort+e# 1 1 (iv) # t r a m # e l E t r + i k + o # #tram#elEtr+ik+o#

C.III, P19 C.I, II C.III, P19 3.4.7. Apocope

This rule discussed in connection with oxytonic stress patterns and its independent motivation, is found in 2.4. Apocopation in Italian accounts for words like cittd, servitü, civilta, ... the phonological spelling of which is given as /civ+täd+e, .../ on the basis of words like cittadino,.... Thus these words are regularized as paroxytonic. Apocope, the following rule, applies after stress has been assigned.

A PHONOLOGY OF ITALIAN

P20 [X] -»· 0

/

—voc — ens — grv —shp —cnt —voi

+VOC

.#

87

Χ φ [+Suffix,...]

—ens +grv t+cptjl [+dif]J

i.e., delete the slice between ta/u and the word boundary, if that slice X does not contain a suffix. (7) (i) # Ü + t ä d + e #

C.I, II

Ci-ftä

C.II, P20

(ii) # 5 i + t a # m o d e r n + a #

C.I, II

δί+ta # m o d e r n + a #

C.III, P16

5i+ia#modern+a#

Pl9 3.4.8. Glide Residue

Here we predict deletion of yod after affricates, to complement the rule of affrication formulated in 3.4.4. P21

—voc —ens +grv

0

/

[+cns]

+cns +shp = +cns +igt +shp

i.e., jod ([y]) is deleted following a long, sharp segment or a sharp segment preceded by a consonant. The following derivation (2) continues derivation (1) in 3.4.4 to complete adjustment of glides on the gasis of P21 (2) (i) me^+ya

(ϋ) forc+ya

(iii) gräc+ya P20

me^+a

forc+a P30, 31 grac+ya 3.4.9. Secondary

Stress

Within word boundaries three main stress levels are recognized. The primary stress is

88

A PHONOLOGY OF ITALIAN

assigned by general rules, mostly based on vowel length, to the phonetically loudest most intense) vowel in the word. Secondary stress includes two additional levels. That of vowels preceding primary stress is less intense than the highest but more intense than the lowest. We term the middle level secondary stress proper. Tertiary stress will be assigned to the vowels following primary stress, and this is the lowest of the three. The following rules assigning secondary stress are phonetic rules. It is important to give them at this place for the correct formulation of secondary phrasal intensity peaks which coincide with pitch levels. ph22

r+cns~l [_+vocJ

[+(2)str]

/

#X

Y # , Χ, Υ Φ #

+ VOC

+cns +str

i.e., assign secondary (2) stress to vowels preceding primary stressed vowels. ph23

r + c n s " | -» [+(3)str] |_+vocJ

/

#X

_ # , Χ, Υ Φ #

+VOC

+cns +str

i.e., assign tertiary (3) stress to vowels following primary stressed vowels. 3.4.10. Phrase Stress The stress pattern of words repeats itself in the phrase. There is one highest pitch (intensity in words) corresponding to the rightmost primary stress. As in words vowels to the left of the highest pitch are lower than this but higher than vowels at the right of the highest pitch vowel. The following rules applying within phrase boundary (% %) characterize the phrase pattern by readjusting stress indices. ph24

—ens

[+(l)str]

I

%

X%, Χ Φ

+VOC

+str

—ens +voc +str

i.e., assign the highest pitch to the rightmost primary stress in the phrase. ph25

η -»· η—1 /

%X [+(3)str] Y

+VOC

Z%

—ens +(l)str i.e., lower by 1 the index of a tertiary stress, if it occurs before the highest pitch in a phrase. ph26

η -> n + 1

/

—voc , n>l +cns +( )str

i.e., raise all other stress indices by 1.

A PHONOLOGY OF ITALIAN

89

The application of rules ph24, 25, 26 is illustrated by the following derivation. (8)

(i) % # ö i + t a # m o d e r n + a # % 2

1

2

1

3

öi+ta#modern+a 3

2

3

1

C.I, ΙΓ, III ph22,23

4

öi+ta#modern+a

ph24,26

3.4.11. Cyclic Convention Rules The following rules are part of the general theory of linguistics adopted. They give instructions to erase the innermost structure at the end of each cycle, and repeat certain rules just in case there is more structure. They are part of the theory of Italian in that they specify just which rules must be repeated. P27 [X] -

X,

[

i.e., remove innermost structure. P28

Do P6-P28

/

[X]

i.e., repeat rules P6-P26 (where they apply) if there is more structure, otherwise go to the next rule. 3.5. POSTCYCLIC RULES

After the phonological cycle we deal with strings composed solely of segments of distinctive features an boundaries. Post cyclic rules include additional Phonological (P) rules which predict allophonic variations strictly phonetically conditioned and phonetic (ph) rules filling in the value of those features which are still unspecified. P29

+VOC

—ens -dif +tns

[

[—tns] I Γ— —voc~J. —ens J

i.e., all midvowels are open following a glide. P30

Γ + s h p l -> [+lgt] L-grvJ

i.e., the segments /η, 1, c, ξ, s/ are inherently long. P31

Γ+vocl L-cnsJ

j[+lgt] I [+cns] l l Igt] elsewhere J

i.e., vowels are long before one consonant or a vowel (in open syllable). In the complementary environment vowels are short.

90 P32

A PHONOLOGY OF ITALIAN +VOC

[+tns]

/

—ens -dif —tns

[+(n+l)str],n=l

i.e., open midvowels are pronounced closed when not stressed. ph33

f[+voc]~| —ens —voc +cns —voc +nas

[+voi]

i.e., vowels, liquids, glides, and nasals are voiced. ph34

voc~| -»· [—voi] +cns +cnt —nas —grv

i.e., Is/ and /§/ are voiceless. P35

—voc —ens —grv —shp +ent —nas

[+voi]

I

[—voi]

r+ens! (_+voiJ

elsewhere

i.e., /s/ is voiced before a voiced consonantal segment. P36

—voc —ens —grv

0

/

—ens +shp +cnt

i.e., the glide [y] is deleted if preceded by a palatal (and followed by a vowel). The last set of phonetic rules complete specification of all unspecified features, ph37

Γ+vocl L+ens J

x dif] [-dif]

ph38

Γ + v o c ] -> Γ + c n t l [_—ensj [_—nas J

/ j^xshpj

A PHONOLOGY OF ITALIAN

[

ph39 Γ— —voc~| + cnsj

ph40

[+cns]

91

[-cnt]

[—nas]

/ [xdif]

I Γ—grv~j [_ χ s h p j

[ X cpt]

[Xdif]

/ [xtns] P41

[-nas]

Γ x grv~j j_ χ cpt J

[xlgt] +cns xgrv xcpt

Rule P41 states that nasals are homorganic with the following consonantal segment. This rule is set up to account for the phenomenon across boundaries. Morpheme Structure rules characterize the same occurrence within boundaries. The phonemic matrix of Italian introduced in section 3.1.1, (4) appears as a phonetic one in the form (9) on the following page. If we assume all the rules developed in this dissertation we have at this stage utterances represented in terms of the above mentioned matrix: a relatively broad phonetic transcription. The task of our phonology is thus ended. Followup study would consist of a phonetic study for the purpose of replacing the binary coefficients of the features with numerical indices stating the various degrees of voicing, tenseness, etc.

92

A PHONOLOGY OF ITALIAN



4

+

I

4

-

4

4

I

I

I

4

(-1

+ + I + I +

I

I

4

I

4

l«J

I

I

4

4

4



4



4

41



+

+1

4

4

4

+

4

4

4

4

I

α

4

I

4

I

I

•ο

4

I

4

I

I

I

4

4

4

+

ΙΟ

4

I

4

I

I

I

4

4

4

I

ι**

4

I

I

I

I

4

4

4

4

ιω

+ +

I

I

I

I

I

4

4

4

4

I

I

I

I

I

4

4

4

I

4

I

4

4

4

I

4

4 4

•w

4

I

4

I

+

4

4



I

4

I

I

Ο

4

I

4

I

I

I

+

I

4

ο

4

I

4

I

I

!

4

I

4

I

4

4



I

I

I

I

4>

4

I

I

I

I

I

+ +

ω

4

I

I

I

I

I

4

I

4

I

I 4

β

I

4

4

I

I

4

4



>

I

4

4

I

I

4

4



«Μ

I

+

4

I

I

4

+

+ + + +



I

I

χ>

I

4

4

I

I

4

I



I



Οι

I

4

I

I

+

I



I

I



>60

I

+ +

+

I

I

4

I

+1

>ο

I

4

4

I

I



I



60

I

4

I

4

I

I



I



I

4

+ +

I

4

I

I



I



>0

I

4

I

4

4

I

+

4

ΧΛ

I

4

I

4

4

I

4

4

I

I



NT-

I

4

I

4

I

4

I



+

I



ο

I

4

I

4

I

4

I



I

I



C

I

4

I

I

CO

I

4

I

I

I

4

I

I ο ο >

4 Μ ö •-Η ο

Μ m

I ««-» Α

αϊ

•ο

+ιΗ •> Β υ

Ρ.

I

&

νο

4 I 4 I

4 I 4 I 4

4

-Η -Η



+

41



I

41



4

I





I

I t» cd α ο »-Η

41



4

-Μ ο Ö4Q > r— οο ο\

CO α 1—1 Ψ-1

es

BIBLIOGRAPHY

Applegate, J., "Phonological Rules of a Subdialect of English", Word, 17 (1961). Arce, J., "II Numero dei Fonemi in Italiano in Confronto con lo Spagnolo", Lingua Nostra, 23 (1962). Bach, E., An Introduction to Transformational Grammars (New York, 1964). Bartoli, G., Degli Elementi del Parlar Toscano (Firenze, 1584). Battaglia, S., and V. Pernicone, La Grammatica Italiana (Torino, 1951). Bembo, P., Prose delta Volgar Lingua [1549\, Ed. Mario Marti (Padova, 1955). Bever, T. G., "Formal Justification of Variables in Phonemic Cross-Classifying Systems", Quarterly Progress Report, 69 (1963), (Research Laboratory of Electronics, M.I.T.). , "Theoretical Implications of Bloomfield's 'Menomini Morphophonemics'", Quarterly Progress Report, 69 (1963), (Research Laboratory of Electronics, M.I.T.). , "The Reciprocating Cycle of the Indo-European e-o Ablaut", Quarterly Progress Report, 69 (1963), (Research Laboratory of Electronics, M.I.T.). , "The e-o Ablaut in Old English", Quarterly Progress Report, 69 (1963), (Research Laboratory of Electronics, M.I.T.). Bianchi, E., "Alcune Osservazioni sul Rafforzamento Consonantico nel Parlar Fiorentino", Lingua Nostra, 9 (1948). Bloomfield, L., Language (New York, 1933). Buscaino-Campo, Α., Regole per la Pronunzia della Lingua Italiana (Trapani, 1873). Camilli, Α., "I Rafforzamenti Iniziali", Lingua Nostra, 3 (1941). , Pronuncia e Grafia dell'Italiano (Firenze, 1947). , "Spostamento di Tono nella Pronuncia Italiana", Lingua Nostra, 11 (1950). , "Dal Ritmo Latino al Ritmo Italiano", Lingua Nostra, 15 (1954). , "Π Ritmo Intensivo Italiano", Lingua Nostra, 17 (1956). , Grammatica Italiana (No date). Castellani, Α., "Fonotipi e Fonemi in Italiano", Studie di Filologia Italiana, 14 (1956). Castiglione, P., Italian Phonetics, Diction and Intonation (New York, 1957). Chomsky, N., Morphophonemics of Modern Hebrew (Master's Thesis, University of Pennsylvania, 1951). , The Logical Structure of Linguistic Theory (Mimeo, Cambridge, Mass., 1955). , "Semantic Considerations in Grammar", in Monograph Series on Languages and Linguistics, Vol. 8 (Georgetown University, 1955). , "Three Models for the Description of Language", I. R. E. Transactions on Information Theory, IT-2 (1956). , Syntactic Structures (The Hague, 1957). , Rev. of R. Jakobson and M. Halle Fundamentals of Language, International Journal of American Linguistics, 23 (1957). , Rev. of C. F. Hockett A Manual of Phonology, International Journal of American Linguistics, 23 (1957). , "A Transformational Approach to Syntax", in Third Texas Conference on Problems of Linguistic Analysis of English [1958] (University of Texas, 1962). , Rev. of B. F. Skinner Verbal Behavior, Language, 35 (1959).

94

BIBLIOGRAPHY

, "Some Methodological Remarks on Generative Grammars", Word, 17 (1961). , "On the Notion 'Rule of Grammar'", Proceedings of Symposia on Applied Mathematics, Vol. 12 (American Mathematical Society, 1961). , "Explanatory Models in Linguistics", in Logic, Methodology and the Philosophy of Science (Stanford University, 1962). , "Formal Properties of Grammar", in Handbook of Mathematical Psychology, Vol. 2. (New York, 1963). , Current Issues in Linguistic Theory (The Hague, 1964). , Aspects of the Theory of Syntax (Cambridge, Mass., 1965). , "Topics in the Theory of Generative Grammar", in Current Trends in Linguistics, Vol. 3. (The Hague, 1966). , Cartesian Linguistics (New York, 1967). and G. A. Miller, "Introduction to the Formal Analysis of Natural Languages", in Handbook of Mathematical Psychology, Vol. 2. (New York, 1963). and M. Halle, "The Morphophonemics of English", Quarterly Progress Report, 58 (1960). , "Some Controversial Questions on Phonological Theory", Journal of Linguistics, 1 (1965). and F. Lukoff, "On Accent and Juncture in English", in For Roman Jakobson (The Hague, 1956). Chomsky, N. and M. Halle, The Sound Pattern of English (New York, 1968). Davis, Ε. B., "Italian e's and o's", Italica, 14 (1937). Davis, M., Computability and Unsolvability (New York, 1958). Demattio, F., Fonologia Italiana (Innsbruck, 1875). Fillmore, C.,A System of Characterizing Phonological Theories (Ph.D. Dissertation, Ann Arbor, 1962). , "The Position of Embedding Transformations in a Grammar", Word, 19 (1963). Fiorelli, P., "Una Sibilante e Due Campane", Lingua Nostra, 12 (1951). , "Degli Elementi del Parlar Toscano", Lingua Nostra, 17 (1957). , "Del Raddoppiamento da Parola a Parola", Lingua Nostra, 19 (1958). Fodor, J. A. and J. J. Katz, eds., The Structure of Language (Englewood Cliffs, N.J., 1964). Foley, J., "Prothesis in the Latin Verb sum", Language, 41 (1965). Gleason, H. A., An Introduction to Descriptive Linguistics, Revised Edition (New York, 1961). Grandgent, C. H., From Latin to Italian (Cambridge, Mass., 1927). and Ε. H. Wilkins, Italian Grammar (Boston, 1904). Hall, R. Α., Jr., "Italian Phonemes and Orthography", Italica, 21 (1944). , Descriptive Italian Grammar (Ithaca, New York, 1948). , "The Reconstruction of Proto-Romance", Language, 26 (1950). , Leave Your Language Alone! (Ithaca, New York, 1950). , Bibliografia deila Linguistica Italiana (Firenze, 1958). , Italian for Modern Living (Ithaca, New York, 1959). , "Developments in Modern Italian", Modern Language Journal, 44 (1960). , "Italian [z] and the Converse of the Archiphoneme", Lingua, 9 (1960). , Applied Linguistics: Italian (Boston, 1961). , "Doubling in Western Romance", Language, 40 (1964). Halle, M., "The Strategy of Phonemics", Word, 10 (1954). , "Why and How Do We Study the Sounds of Speech?" in Monograph Series on Languages and Linguistics, Vol. 7 (Georgetown University, 1954). , "In Defence of the Number Two", in Studies Presented to J. Whatmough (The Hague, 1957). , The Sound Pattern of Russian (The Hague, 1959). , "Questions of Linguistics", Nuovo Cimento, 13 (1959). , "On the Role of Simplicity in Linguistic Description", Proceedings of Symposia on Applied Mathematics, Vol. 12. (American Mathematical Society, 1961). , "A Descriptive Convention for Treating Assimilation and Dissimilation", Quarterly Progress Report, 66 (1962), (Research Laboratory of Electronics, M.I.T.). , "Proto-Slavic Diphthongs", Quarterly Progress Report, 66 (1962), (Research Laboratory of Electronics, M.I.T.). , "Phonology in Generative Grammar", Word, 18 (1962). , "On Cyclically Ordered Rules in Russian", in American Contribution to the Fifth International Congress of Slavists, Vol. 1. (The Hague, 1963).

BIBLIOGRAPHY

95

and Ν. Chomsky, The Sound Pattern of English. (To appear). , "On the Bases of Phonology", in The Structure of Language (Englewood Cliffs, 1964). Halle, M. and Κ. N. Stevens, "Analysis by Synthesis", in Proceedings of Seminar on Speech Compression and Processing (1959). Halle, M. and V. J. Zeps, Latvian Morphology (To appear). Halle, M., C. Cherry and R. Jakobson, "Logic and Phonemics", Language, 29 (1953). Halliday, Μ. A. K., "Categories of the Theory of Grammar", Word, 17 (1961). Harris, Z., Methods in Structural Linguistics (Chicago, 1951). , "Co-occurrence and Transformation in Linguistic Structure", Language, 33 (1957). Hjelmslev, L., Prolegomena to a Theory of Language, transl. by F. J. Whitfield (Madison, Wise., 1963). Hockett, C. F., "Two Models of Grammatical Description", Word, 10 (1954). , A Manual of Phonology (= Indiana University Publications in Anthropology and Linguistics, Vol. 11), (1955). , A Course in Modern Linguistics (New York, 1958). , "Linguistic Elements and Their Relations", Language, 37 (1961). Householder, F. W., "On Some Recent Claims in Phonological Theory", Journal of Linguistics, 1 (1965). Jakobson, R., "The Prosodic Questions of Slavic Historical Phonology Restated", Quarterly Progress Report, 72 (1964), (Research Laboratory of Electronics, M.I.T.). and M. Halle, Fundamentals of Language (The Hague, 1956). , C. G. M. Fant, and M. Halle, Preliminaries to Speech Analysis, Revised Edition (Cambridge, Mass., 1963). Josselyn, F. M., Etude sur la phonetique italienne (Paris, 1900). Katz, J. J., "Mentalism in Linguistics", Language, 40 (1964). and J. Fodor, "The Structure of a Semantic Theory", Language, 39 (1963). and P. Postal, An Integrated Theory of Linguistic Description (Cambridge, Mass., 1964). Klima, Ε., "Relatedness Between Grammatical Systems", Language, 40 (1964). , "Negation in English", in J. Fodor and J. J. Katz, eds., The Structure of Language (Englewood Cliffs, N.J., 1964). , "Current Development in Generative Grammar", Kybernetica, 1 (to appear). Lamb, S., Outline of Stratificational Grammar, Revised Edition (Berkeley, 1966). Langandoen, D. T., "A Restriction on Grassman's Law in Greek", Language, 42 (1966). Lausberg, Η., Romanische Sprachwissenschaft, Vol. 1 (Berlin, 1956). Lees, R. Β., Rev. of Ν. Chomsky Syntactic Structures, Language, 33 (1957). , "A Multiply Ambiguous Construction in English", Language, 36 (1960). , The Grammar of English Nominalization (= Indiana University Publications in Anthropology and Linguistics, Vol. 26), (1960). , The Phonology of Modern Standard Turkish (= Uralic and Altaic Series, Vol. 6), (Indiana University, 1961). , "Grammatical Analysis of the English Comparative Construction", Word, 17 (1961). , "Turkish Nominalization and a Problem of Ellipsis", Foundations of Language, 1 (1965). and E. Klima, "Rules for English Pronominalization", Language, 39 (1965). Leone, Α., "A Proposito del Raddoppiamento Sintattico", Bollettino del Centro di Studi Filologici e Linguistici Siciliani, 7 (1962). , "Elisione e Troncamento", Lingua Nostra, 24 (1963). Lieberman, P., "On the Acoustic Basis of the Perception of Intonation by Linguists", Word, 21 (1965). Ligthner, Τ. M., "Vowel Harmony in Classical (Literary) Mongolian", Quarterly Progress Report, 68 (1963), (Research Laboratory of Electronics, M.I.T.). , "A Note on the Formulation of Phonological Rules", Quarterly Progress Report, 68 (1963). , "Remarks on the Morphophonemic Component of Russian", Quarterly Progress Report, 69 (1963). , "Nasal Diphthongs in Russian", Quarterly Progress Report, 70 (1963). , "Preliminary Remarks on the Morphophonemic Component of Polish", Quarterly Progress Report, 71 (1963). , "A Note on the Motivation for Using Transformational Rules in Phonology", Quarterly Progress Report, 76 (1965).

96

BIBLIOGRAPHY

Longacre, R. Ε., "String Constituent Analysis", Language, 36 (1960). Malagoli, R., VAccentazione Italiana (Biblioteca di Lingua Nostra, Vol. 7, 1946). Malmberg, B., "A propos du systfeme phonologique de l'italien", Acta Linguistica, 3 (1942). Manna, F., II Puro e Dolce Idioma (Bologna, 1951). Matthews, G. H., "Morphophonemic Change in Crow and Hidatsa", Quarterly Progress Report, 59 (1960). , Hidatsa Syntax (The Hague, 1964). McCawley, J., "Stress and Pitch in the Serbo-Croatian Verb", Quarterly Progress Report, 70 (1963). Merlo, C., "Del Rafforzamento Sintattico", Lingua Nostra, 10 (1949). Metz, C., "Ein experimentell-phonetischer Beitrag zur Untersuchung der italienischen Konsonantengemination", Vox, 24 (1914). Meyer-Lübke, W., Italienische Grammatik (Leipzig, 1890). Migliorini, B., Pronunzia Fiorentina ο Pronunzia Romana (Firenze, 1945). , "Prontuari Ortografici e Ortofonici", Lingua Nostra, 10 (1949). Muljaöic, ί., Fonologia Generale e Fonologia della Lingua Italiana (Bologna, 1969). Norman, H. L., "Reduplication of Consonants in Italian Pronunciation", Italica, 14 (1937). Panconcelli-Calzia, G., De la nasalite en italien (Paris, 1904). , Italiano (Leipzig and Berlin, 1911). , "Zur Frage der Frequenz-bewegungen im Italienischen", Casopis Pro Moderni Filologi, 17 (1921). Panzini, A. and R. Allulli, Grammatica Italiana (Milano, 1953). Parmenter, C. E. and J. N. Carman, "Some Remarks on Italian Quantity", Italica, 9 (1932). and S. N. Trevino, "Italian Intonation", Italica, 7 (1930). Peruzzi, E., Problemi di Grammatica Italiana (Torino, 1959). Porena, M., "Sillabe Brevi e Sillabe Lunghe nella Poesia Italiana", in Note di Lingua e Stile (Napoli, 1908). , "Per una Piü Esatta Descrizione dei Suoni Consonantici Italiani", Lingua Nostra, 4 (1942). Porru, G., "Anmerkungen über die Phonologie des Italienischen", in Etudes phonologiques dediees a la memoire de Ν. S. Troubetzkoy (= Travaux du Cercle Linguistique de Prague, Vol. 8, 1939). Postal, P., Some Syntactic Rules in Mohawk (Ph.D. Dissertation, Yale University, 1962). , Constituent Structured Publication Thirty of the Indiana University Research Center in Anthropology, Folklore, and Linguistics, 1964). Rohlfs, G., Historische Grammatik der italienische Sprache und ihrer Mundarten, Vol. 1 (Berne, 1949-1954). Sapir, E., "The Psychological Reality of Phonemes", in Selected Writings of E. Sapir (Berkeley, 1949). Saussure, F. de, Cours de linguistique generale (Paris, 1922). Schane, S., French Morphology and Phonology (Ph.D. Dissertation, M.I.T., 1965). Stoppato, L., Fonologia Italiana (Milano, 1887). Tolomei, C., II Cesano [1547] (Milano, 1864). Trager, G. L., "Comment on Italian e's and o's", Italica, 16 (1939). Trissino, G., II Castellano [1528] (Milano, 1864). Vanzon, C., Grammatica Ragionata della Lingua Italiana (Livorno, 1834). Wells, R., "Immediate Constituents", Language, 23 (1947). Zeps, V. J., "Optional Rules in the Formation of the Old church Slavonic Aorist", Language, 40 (1964).