274 89 9MB
English Pages [252] Year 1977
INFORMATION T O USERS
This material was produced from a microfilm copy of the original document. While the most advanced technological meant to photograph and reproduce this document have been used, the quality is heavily dependent upon the quality of the original submitted. The following explanation o f techniques is provided to help you understand markings or patterns which m a y appear on this reproduction. 1. The sign or "target" for pages apparently lacking from the document photographed is "Missing Paga(s)". I f it was possible to obtain the mining page(s) or section, they are spliced into the film along with adjacent pages. This may have necessitated cutting thru an image and duplicating adjacent pages to insure you complete continuity. 2. When an image on the film it obliterated with a large round black mark, it is an indication that the photographer suspected that the copy may have moved during exposure and thus cause a blurred image. You will find a good image of the page in the adjacent frame. 3. When a map, drawing or chart, etc., was part of the material being photographed the photographer followed a definite method in "sectioning" the material. It is customary to begin photoing at the upper left hand corner of a large sheet and to continue photoing from left to right in equal sections with a small overlap. If necessary, sectioning is continued again — beginning below the first row and continuing on until complete. 4. The majority of users indicate that the textual content is of greatest value, however, a somewhat higher quality reproduction could be made from "photographs" if essential to the understanding of the dissertation. Silver prints of "photographs" may be ordered at additional charge by writing the Order Department, giving the catalog number, title, author and specific pages you wish reproduced. 5. PLEASE NOTE: Some pages may have indistinct print. Filmed as received.
University Microfilms International 300 North Zeeb Road Ann Arbor, Michigan 4B106 USA St John's Road, Tyler's Green High Wycombe, Bucks, England HP10 8HR
77-14,934 CERRQM-PALCMINO, Rodolfo Marcial, 1940HUANCA-QUECHUA DIALECTOLOGY. University of I l l i n o i s a t Urbana-Chanroaign, Ph.D., 1977 Language, l i n g u i s t i c s
XerOX University MICrOfilm8 r Ann Arbor, Michigan 48106 . }
© Copyright by Rodolfo Marcial Cerro'n-Palomino 1977
HUANCA-QUECHUA DIALECTOLOGY
BY RODOLFO MARCIAL CERRON-PALOMINO B.Hum., Universidad Nacional Mayor de San Marcos, 1967 M.A., Cornell University, 1969
THESIS Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Linguistics in the Graduate College of the University of Illinois at Urbana-Champaign, 1977
Urbana, Illinois
UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN THE GRADUATE COLLEGE
December, 1976
W E HEREBY RECOMMEND THAT THE THESIS BY RODOLFO MARCIAL CERRON-PALOMINO „ ™ T ~ T __
EN I I I L E D
HUANCA-QUECHUA DIALECTOLOGY
BE ACCEPTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE O F
DOCTOR OF PHILOSOPHY
\yu J*^j£u3*0Lw\K D\ST«\\^>VmoM OF PG9AJV\M4 a*JECV\V)K
auECv\vi /k/ is hardly a natural one (as far as I know, such a change has 2 never been recorded),
it is reasonable to assume that the merger between
/q/ and /k/ was, among the varieties mentioned, due to a substratum phenomenon. mind easily.
At least two arguments in favor of this hypothesis come to One is that the area where the two sounds merge is contiguous
(and the varieties therein spoken also share a voicing rule, whereby /p, t, k/ go to [b, d, g] after a nasal consonant).
The second argument
is that speakers who do not have the velar-postvelar distinction, tend to underdifferentiate the contrast, say /k/:/q/, in favor of the former. Such was--and still is—the case of the Spanish speakers who, from the time they came into contact with Quechua, rendered automatically the Quechua sound /q/ as Spanish /k/ in loan-words.
Thus, Quechua words with
an original /q/—especially those referring to institutions and place names—incorporated into Peruvian Spanish have /k/ invariably (written, according to the Spanish spelling, with c or with /?/ that is, postvelar */q/ became a glottal stop without any exception. Subsequently, however, /?/ dropped word initially; thus, the next state
41 of the change can be captured as follows:
0 / ## (8b)
*/q/ > J
IV \
Unfortunately, from a synchronic point of view, i t i s no longer p o s s i b l e t o know whether t h e change was, from i t s inception, as r u l e (8b) s t a t e s . I have failed in t r y i n g to find a v a r i e t y t h a t preserved /*?/ in i n i t i a l position. (8a).
I t seems reasonable, however, t o postulate an e a r l i e r stage as
Later, / ' / dropped word i n i t i a l l y as i t was probably i n t e r p r e t e d as
a mere prolongation of the preceding pause.
Thus, since there seems to
be no synchronic evidence whatsoever for p o s t u l a t i n g i n i t i a l /•?/, I conclude t h a t the r e f l e x e s of PQ t h a t contained */q/ in i n i t i a l p o s i t i o n must be entered in the Yaqa-Huanca lexicon as ##V(C)..» r a t h e r than *## ? V(C)... sequences.
In t h e following l i s t I i l l u s t r a t e the a c t u a l s i t u a t i o n as
compared to an e a r l i e r stage which, following Torero (1970: 245), I c a l l Huancay: (9) Huancay
Yaqa-Huanca
*qipi
ipi
'bundle'
*qutu
utu
'goiter'
*qalu
alu
'tongue'
*uqa
u?a
'oca (Oxalis tuberosa)'
*muqu-
mu'u-
•to knob'
*atuq
atu?
•fox'
*musuq
musu?
'new'
*wiqtl
wi'ti
(gummy secretion of the eyes)
*paqwa-
pa*?wa-
'to finish'
*pirqa
pil*?a
•wall'
*ayqi-
ay?i-
'to flee'
42 Thus, I assume that the second column contains the underlying form of the lexemes, as they must be introduced in the lexicon.
From this basic
form, the phonetic realization of /'/ is subject to a series of ordered rules which will constitute a topic to be discussed extensively In Chapter 6. What has been said in relation to the lowering effect of /q/, lost whenever this phoneme is no longer a postvelar, is also valid here. Thus, as in NH, in this variety we do not find derived mid vowels at all. Therefore, it must be clear that to postulate an underlying /q/ for this dialect has no basis at all. Once rule (8a) became consummated, the YH speaker was simply unable to recover the postvelar /q/, and therefore, faced with this phoneme through the contact with other varieties, especially Cuzco-Quechua (a variety that acted as the superstratum language for at least seventy years), he "identified" /q/ with his /k/, as the Spaniards did.
As a result of this, words such as:
(10) qacwa
(a kind of dance)
qasya-
'to boil'
qata
'to cover1
luql
(a variety of tree)
muql-
'to spoil'
waqta-
'to hit'
were borrowed into YH as: (10a) kacwa kasyakata luki mukiwakta-
and as such they must be entered in the lexicon. Sometimes we doublets.
Thus, aside from inherited forms such as: (11) caw'i
(a variety of potage)
a la
'nude, skin'
wal'l
'bag'
aspa-
'to singe'
aytu-
•to ball'
iwi-
•to lift'
u"?ra-
•to hug'
lu?lu
'spoiled egg*
which originally come from: (11a) *cawqi *qara *walqi *qaspa*qaytu*qiwi*uqTa*luqlu we also find the following borrowings: (lib) cawkl kala
'nude'
walk! kaspakaytu kiwi-
'thread'
44 uktaluklu As we see, only some of the borrowed forms became semantically specialized. Finally, words such as the following betray their loanword status: kari
'male; sane'
irki
'dwarf, feeble'
kiswar
(a kind of tree)
kurmay
(a big potato)
since they not only correspond to original words with /q/ instead of /k/, but also keep the /r/ phoneme that regularly, as we already know, changed to /l/ in Huanca. Before we finish this section, it is convenient that we discuss, as was done in Section 2.2, the underlying forms of the reflexes of *-rqu ~ *-rqa, *-rqa, and *-sqa.
Contrary to what happens in the other
branch, in YH we find the following forms: respectively.
-?lu ~ -91a. -l?a, and -sa.
Thus we see that, at least in reference to the first two
suffixes, there is evidence that shows that in this dialect the reflex of */q/ is preserved.
The only change being that the eductive suffix
was metathesized from *-lqu ~ *-lqa (assuming that the lambdacism was prior) to -?lu ~ -?la, whereas the quasi-homophonous past tense marker remained -l?a. As for the past participle, I was simply unable to find any trace whatsoever of a form *-s?a.
There is evidence, however, to the
effect that this dialect, as NH, inherited an earlier form *-sa, rather than *-sqa.
The evidence I would like to offer is the following.
As we
shall see (cf. Chapter 7, Section 7.7), the syllable lu ~ JLa of the eductive modal drops entirely when the next suffix has the shape -CV(C)» but not -C(CV(Q).
This is because otherwise the syncope would yield
45 an intolerable syllabic structure.
Thus, the entire syllable falls in
[l&'+sV] from /li+qlu+sa'/ 'I will go right away', but not in [li+'luVqki] from /li+*?lu+nki/ 'you will go right away'.
Now, in an
expression such as [kamaka+'+sa] from /kamaka+'lu+sa/ 'recently finished', we see that the syllable drop takes place; this would have not been possible had the past participle been -s?a (from *-sqa), for otherwise we would have got *[kamakaV?+s?a]. For these reasons I conclude that the underlying form of this modal should be simply -la. In reference to the basic form of the eductive -*?lu ~ -?lu, I must mention, however, that in most varieties of YH I have found alternants of the same morpheme in the shape of -lu ~ -la, that is without any trace of /'/•
This happens actually elsewhere, except in the southwestern
region of the valley (from Tres de Diciembre, Chongos Bajo and Huayicachi down to the South).
Thus, whereas in this area we only find forms such
as [liV+sa*?], [caVHmu+T]], etc., from /li+^lu+sa*?/ 'I will go right away' and /ca+?lu+mu+n/ 'he/she just arrived', respectively, in the remaining territory (that is, the major section) we find [li+lu+sa*?] and [ca+la+mu+T]], respectively.
In this latter case, one might be tempted
to say that, as in most of this area the glottal stop phoneme drops after a consonant (cf. Chapter 6), perhaps *-l?u - *-19a did not undergo metathesis, and therefore /?/ disappears by means of a general rule of glottal elision.
However, the fact is that there are places—such as
Sapallanga and Acobamba—where /•?/ is firmly preserved in that environment.
Thus, one cannot simply invoke such a rule, and yet the eductive is
sometimes -lu — l a .
In view of these facts, I must assume that whereas
all the varieties of Yaqa-Huanca inherited *-rqu ~ *-rqa, some others kept the simplified form, that is *-ru ~ *-ra, which probably existed as an
46 alternate form.
In any case, I am perfectly aware of the fact that thiu
situation needs to be carefully checked, for which more detailed fieldwork is needed.
For the purposes of the present dissertation I shall assume
that the underlying form of the eductive in Yaqa-Huanca is -?lu ~ -?la. Finally, in regard to the past tense marker, the evidence of an inherited *-rqa is conclusive, as it is attested in all varieties as -l?a (with subsequent loss of /'/ in some varieties).
The fact that -l?a did
not metatheslze as the eductive did, might be explained by the very sporadic nature of the process of metathesis itself.
This, incidentally,
proves that it would be totally wrong to postulate an underlying eductive *-19u ~ *-l?a subject to a later rule of metathesis. From the above, it should be clear by now how the YH variety differs drastically from any other Quechua dialect.
In fact, by eliminating
initial /?/, this variety has disfigured its lexicon:
this one change was
responsible for what we may call the "dequechuization" of YH. As a result of that, there are now more entries in the lexicon beginning with the three types of vowels.
As we shall see, that is not the whole history,
for the rule of glottal elision is still at work, and more "deformation" of the lexicon is taking place at the phonetic level.
The drastic
erosion of the YH lexicon as a result of this change led most laymen to the belief that this variety is not a Quechua dialect at all. The only reason for this, I must assume, are the changes which have operated on */q/, since the other variety, that is NH, has never been considered a non-Quechua language.
Since both varieties share the historical rule
*/r/ > /l/, it is clear that the only thing responsible for the g "dequechuization" of YH is the special treatment of */q/.
47
FOOTNOTES FOR CHAPTER 2
It seems probable that the Central coastal dialect described by the Dominican friar Domingo de Santo Tomas (1560), now extinct, had the same change. See Torero (1964: 451) and Parker (1969b: 162-168). 2 To be true, however, Campbell (1971: 300) cites /q/ > /k/ as an innovation defining the subgroup Cholan-Tzotzilan of the Mayan languages family, also shared by Huastecan. The author however merely lists the change, without even providing us with examples. Al-Ani (1973), on the other hand, cites the same type of change for the Baghdad! dialect of Arabic. Curiously enough, however, the occurrence of /k/ as a reflex of */q/ "is quite limited, appearing in only four items I" To this we may also add the merger of /q/ and /k/ in Hebrew. But note that in these two cases we are dealing with diglottic situations, and in the case of Hebrew we should not forget that the language is spoken today as a result of its revival. In both cases the change seems unnatural. Compare this situation to the more natural weakening processes that affect /q/ in other languages; cf. footnotes 5 and 8. 3 The substitution of /k/ for /q/ seems to be a universal one. Thus, Arabic borrowings in Swahili have /k/ instead of the original /q/. Here are some examples: /qawl/ 'to say1, /qalam/ 'pen', /qawm/ 'people', /maqas/ 'scissors', etc. were rendered as /ko:li/, /kalajma/, /ko:mu/, and /maka:si/, respectively. I am grateful to Mohammad I. Abasheikh for providing me with these examples. / . The only case I know that in spite of the .total loss of /q/ the lowering effect still persists is the treatment of past tense suffix *-rqa in Argentinian Quechua. In this variety, /q/ has been lost only in this suffix—a universal tendency in Quechua, as we will see elsewhere. However, here we find the following alternation: /zuqsi-nki/ 'you leave' versus /zuqse-ra-nki/ 'you left', 'punu-nki/ 'you sleep' versus /puno-ra-nki/ 'you slept', etc. In this case, I think, it is more reasonable to refer to it as a vowel split phenomenon (or, better, restructuring), rather than lowering process attributable to a segment that no longer is there. For a discussion of the same topic, see Parker (1969b: 182-183). 4
Incidentally, note that the splrantization of /q/ as a natural process of weakening is well-attested among the Malayo-Polynesian languages, where Proto-Malayo-Polynesian */q/ changed to /h/. Cf. Dyen (1953). Also, Campbell (1971) cites the change /q/ > /x/ among the Quichean linguistic family.
48
It seems safe to say that the change was already consummated before the arrival of the Incas. In the earliest document I was able to find (dated in 1582), one can read the name of an Indian chief named canea guala. both forms traceable to a prior *kanka+q waqra. The fact that in the same document we find the same name transcribed as cancac guacra, shows clearly that the name was retranscribed according to the Cuzco pronounciatlon, whichever since was considered as the best. Incidentally, the name, univerbated as Cangahuala. is a common last name in the region. In some communities, especially those of the central-northern section of the valley, I have found forms such as /ha'a/ 'cliff and /hun?ul/ 'knee', instead of the normal correlates, that is /a?a/ and /un'ul/, respectively. There seems to be no doubt that in this case such forms reflect an imitation of the more prestigious pronunciation (that of Southern Quechua), since they correspond to PQ forms *qaqa and *qunqur, respectively. The same can be said about doublets such as /panha/ 'husk', /nangal/ 'in vain', /walanga/ 'one thousand', and /ulunguy/ 'bee', found in certain areas, instead of the normal correlates /pan'a/, /nan'al/, /walan'a/, and /ulun'uy/, respectively; these forms go back to *panqa. *nanqa(l)» *waranqa and *urunquy, respectively. The fact that we also find in the same northern area forms such as /ha?a-/ 'to grind' and /hu'u/ 'wet', instead of /a?a-/ and /u'u/, respectively, is a clear demonstration of the efforts made by those Huanca-speakers who tried hard to imitate a southern pronunciation. Since these forms come from original *aqa- and *uqu, respectively, we have here a nice example of hyperurbanism. Cf. also Ecuadorian /huku/, instead of the expected */uku/, for the second form. Finally, in the southwestern mountainsJ[ have also found the variant /halu/ 'tongue' instead of the common /alu/, coming from *qalu. o
Note that the change /q/ > /*?/ seems to be another natural process by which the postvelar stop becomes weakened. Thus, Campbell (1971: 118, 135) cites a similar process taking place among certain Quichean dialects. Moreover, the process is also a common one among certain Arabic dialects. Thus, Al-Ani (1973) mentions that the change /q/ > /'/ takes place in the Baghdad! dialect, and also "in Cairo Arabic, SyrianLebanese Arabic and in some dialects in North Africa."
49
CHAPTER 3 PHONOLOGICAL SKETCH
On the basis of the general configuration of Huanca-Quechua stated in the foregoing chapter, I shall present here a brief phonological description of YH.
3.1. Consonants
The consonantal phonemes are the following: Voiceless stops
P
k
t A
Affricates
V
c c
Voiced stops
(b)
(d)
Fricatives
(f)
s
s
m
n
n
1
r
Nasals Laterals Non-Lat. liquids Glides
9
(8) V
h
r (r)
w
y
The parenthesized consonants are borrowings from Spanish.
Of them,
/b, d, g/ are realized, as in most Quechua dialects, invariably as stops, even in intervocalic position; the spirantization rule common to Spanish is absent here,
/f/ is realized, depending on the degree of bilingualism
of the speaker, either as a bilabial fricative [cp] (as is the case in Peruvian Andean Spanish) or as a velar fricative [x ] somewhat labialized. /?/ corresponds to standard Spanish trill /?/, and it is realized (again as in most Andean Spanish) as a retroflexed apical spirant. Among the stops, /?/ Is realized as a glottal catch, and, as we already know, Is the reflex of */q/.
It does not occur word initially,
50 and, as i t w i l l be seen (cf. Chapter 6 ) , i t s realization varies from a true g l o t t a l to zero.
Here are some examples that i l l u s t r a t e s the
occurrence of / ' / : wa'a+y
'to cry'
a?u+y
•to rubb'
hi'a+y
•to climb'
al'u
•dog'
cu*?lu
'corn'
pis'u
•bird'
atu?
•fox'
The contrast between non-retroflex / c / and retroflex / c / can be seen in the following l i s t of minimal pairs (for further d e t a i l s , see Chapter 4 ) : (2)
2aki
'dry'
caki
'foot'
caka+y
'to bridge'
caka+y
'to serve from a pot'
cakca4y
'to grasp'
cakca+y
'to bitter'
cinpa+y
'to cross a river'
cinpa+y
'to kneel'
cukcu+y
'to tremble'
cukcu+y
'to be asquath'
puc'u
'sour'
puc'u
'dwarf
The phonetic realization of /s/ is [s], that is a retroflexed apicoalveolar sibilant, except in the following environments:
before and after
/!/ and also after /y/. Here are some examples (for further discussion concerning the process involved, see Chapter 5): (3)
[saywa]
'scarecrow'
[asyafy]
'to tear'
[masa]
'son-in-law'
[simi]
'mouth'
[$u*ti]
'name'
[iskay]
'two'
[affas]
'skunk'
[uysu+y]
'to bow down'
51 /h/ is a segment that is realized sometimes as a velar spirant.
In
native words, it occurs only word-initially, with the exception of /muhu/ 'seed'.
In other positions it occurs only in loanwords from Spanish, as
a rendering of Spanish /x/. Nasal /n/ is realized as [t]] in final-syllable position, except before obstruents, where the following assimilatory process takes place:
[
a Ant 1
(4) /
8 HighJ
/
I
- Son "|
a Ant I B HighJ
Thus, according to this rule, underlying forms such as the following: (5) panpa
'flat ground'
muyun+pa
'during day'
panta-
'to make a mistake'
pi+wan+tan
'with whom?'
cinka-
'to get lost'
liyun+kuna
•lions'
are realized as follows: (5a)
[pampa] [muyumpa] [panta] [piwantaT]] [ciTflca-] [liyufpcdna]
(where we see that the rule applies vacuosly before /t/). In connection with the nasals, it should be mentioned also that there is a dlssimilatory rule by which /m/ becomes /n/ before a grave sonorant.
The rule however applies only to the root /am/ 'you', since
there is no other morpheme ending in that consonant. Examples:
(6) [aljmi] [a'TJwaT]]
'you yourself 'with you'
As mentioned in Chapter 1, /r/ is a borrowing either from other Quechua varieties or from Spanish.
It contrasts with /r/ only in inter-
vocalic position. I have found only one illustration: the pair /ka:ru/ 'expensive' and /ka:ru/ 'car', where both words are of clear Spanish origin. IT I neutralizes with /?/ in the environments mentioned in the following rule: (7) Ixl - [r] j
[+ Cor] \
Here are some illustrations: (8) [yawar]
'blood'
[ri:ku]
'rich'
[kafta]
'letter'
[birnis]
'Friday'
[kurcu] It should be mentioned in passing that the rule is also valid for the realization of Spanish /r/ in certain Andean varieties. To be precise, however, the rule also applies after /n, s/; but accidentally this only happens in these two proper names:
[i7|riki] 'Henry' and [israyil]
'Israel'. 3.2. Vowels The vocalic system of Yaqa-Huanca consists of the following vowels:
53 Front
Central
Short
Long
1
i:
High
Short
Long
a
Low
Back Short
Long
u
u:
a:
As can be seen, there are no mid vowels, not even phonetically, unlike those Quechua dialects that retain /q/.
Spanish loanwords that
contain /e, o/ are incorporated with /i, u/, respectively; e.g.: (9) inirru
< Sp. enero
'January'
bini:nu
< Sp. veneno
'poison'
piluta
< Sp. pelota
'ball'
Thus, we see tha. whereas the consonantal system has incorporated lew phonemes through borrowing, the situation is different with respect to the treatment of Spanish mid vowels. Depending on the degree of bi
igualism, however, speakers will tend to assimilate them into their
system. The systematic opposition between short and long vowels is apparent from the following contrasting pairs: (10)
taku+y
'to blend'
ta:ku+y
'to sit down'
puka
'red'
pu:ka+y
'to blow'
ha:pa+ra+mi
'it's just a desert'
hapa+la+n 'alone' There is no contrast between /!/
and /!:/ within a morpheme; however,
since the first person marker either actor or allocator is -j_, we get these types of contrasts: (11) malka
'town'
malka+:
'my town'
awsa-
'to play'
awsa+:
'I play'
cuku
'hat'
cuku+:
'my hat'
54 punu-
'to sleep'
puffu+:
'I sleep'
maki
'hand*
maki+:
'my hand'
puli-
'to walk'
puli+:
'I walk'
Aside from these cases, there is in Huanca-Quechua a number of suffixes that underlyingly have long vowels.
These are:
-Fa: 'affective',
-ma: 'cislocative', -na: 'desiderative', -pa: 'continuative', -la: 'stative', -pa:ku 'pluralizer', -lka: 'pluralizer', -yka: 'durative', -cka: 'simultative-durative', -ya: 'transformative1, and -ta: 'involvent'. Furthermore, the long-versus-short opposition has become enriched due to the massive incorporation of Spanish loanwords. Thus, native words such as: (12) ilu
'log'
kilu
•tooth'
ucu
'hot pepper1i
usu
'cough'
pala-
'to Insult*
lasu
1
snow'
contrast now with the following items, all of them completely assimilated: (13)
i:lu
< Sp. hilo
•thread'
ki:lu
< Sp. kilo
'kilogram'
u:cu
< Sp. ocho
'eight'
u:su
< Sp. oso
'bear'
pa: la < Sp. pala la:su
< Sp. lazo
'hoe' 'rope'
I must mention in passing that since Spanish does not make phonemic distinction of vowel length, it is reasonable to ask how is that this dialect treats those borrowings with a long vowel.
It is not the case
55 that all Spanish words are entered with long vowels; nor is it the case that the vowel lengthening in the nativized words can be predicted by means of a rule intrinsic to the phonological component of the target language.
This being so, it would seem that the Quechua-speaker somehow
"hears" low level lengthening distinctions made originally in the source language.
However, since Quechua vowels are somewhat lax as compared to
those of Spanish, it may also be the case that the Huanca-Quechua speaker is interpreting as long the Spanish tense vowels. Finally, as it will be shown in Chapter 6, there are derived long vowels that arise at the phonetic level due to the dropping of certain segments. In regard to the phonetic realization of short vowels, it must be said that the only significant change that was observed affects /a/. The rule is abbreviated as follows:
(14) /a/ - [»] /
y
Thus, /a/ is fronted before /y/.
Some examples that illustrate this
assimilatory process are: (15)
[aeyka]
'how much/many'
[alkaeya:]
'I am waiting'
[limaeyaf]]
'he/she is talking'
This seems to be a fairly widespread rule; its application however depends on the style of speech used, being favoured in careless speech. Aside from this, also a process of weakening that affects again only short vowels was observed.
Thus, /i, u, a/ tend to neutralize in word-
final postion, as seen in the following examples:
(16)
[paeysa]
'(it is said that) he is'
[fiika'toV
'clay'
[lu:ku'sd]
'desperately'
These forms correspond, respectively, to the following lexical representations: (16a) pay+si nikatu lu:ku+*?lu+sa Another phenomenon that merits special mention is the pan-Quechua alternation between /u/ and /a/ among certain suffixes. Thus, the following suffixes:
-caku 'multiplicative', -ku 'pseudo-reflexive',
-mu 'translocative', -JM 'benefactive', -lku 'upward directional', -lpu 'downward directional', -yku 'adductive', -*?lu 'eductive', and -pa:ku 'pluralizer*, substitute their vowel /u/ for /a/ when followed by the "controller" suffixes, that is -ci 'causative', -ma: 'cislocative', and the pluralizer -all.
Of the former group of suffixes, -mu and -pu have
also the controlling effect upon the others.
Illustrations of this type
of alternations will appear elsewhere, especially in Chapters 6 and 7. Furthermore, there are instances in which a short vowel drops entirely, but in this case the process involved is an idiosyncratic behaviour of the morpheme within which it is contained and not a rule governed phenomenon.
Thus, for example, the vowel of the suffixes -pa
'genitive', -mi 'first hand information', -si 'second hand information', and -ca 'conjectural', drops when the suffixes are attached to a stem that ends in a short vowel. (17)
Thus, we have:
[nuna+p]
'of a person*
[nuna+ra]
'(I know) it is a person'
[nuna+s]
'(it is said) it is a person'
[nuna+c]
'may be it is a person'
[wala's+pa]
'of a boy'
[walas+mi]
' (I know) it is a boy'
[waLas+si]
'(it is said) it is a boy'
[walas+ca]
'may be it is a boy'
as compared to: (18)
As for the long vowels, it must be said that they are automatically shortened in checked syllable. (19)
Thus, consider the following examples:
[miku+ya+T)]
'he is eating*
[wata+la+y]
'to be tied up'
[punu+na+y]
'to be sleepy'
[muyu+pa+T|ki]
'you are turning around again and again*
These forms come, respectively, from: (19a) miku+yka:+n wata+la:+y punu+na:+y muyu+pa:+nki where we see that the suffixes immediately following the roots have long vowels.
Otherwise, the vowel lengthening is preserved, as in: (20)
[miku+ya':+na'?]
'he/she was eating'
[wata+la^+cuf]]
'let him/her be tied up'
[punu+na':+ku+y]
'the action of being sleepy'
[muyu+pa':+sa?]
'I'll be turning around again and again'
Table 1 shows the systematic phonemes of Yaqa-Huanca in terms of distinctive features (redundant values are represented by parenthesized plus sign if plus, and by blank space if minus).
Table 1 Distinctive Features
p t B c k * (b) (d) (g) (f) s s h m n n 1 I r (r) w y i i: u u: a a: Syllabic
- - - - - -
-
-
-
Consonantal + + + + + + +
+
+
++
Sonorant
-
-
-
-
-
-
-
.
_
Continuant
-
-
-
-
-
-
+
+
+
Grave
+ - - -
+ + +
-
+
+ - - + + - - - - - - + - - - + + +
Anterior
+ + + - - - +
+
-
+ + - - + + - + - + -
Voiced
-
+
Retroflex
- - +
_
-
-
-
-
-
-
-
-
-
-
+
- - - - - - - - + + + + + + + + _ +
+
(
+
(+
+
Nasal
Low Long
+
+
+ +
+
+ +
+
+ +
+
+ +
+
+
_ -
+
- -
(
+
+
+ + +
+
+
+
+
+
+ +
+
+
+ +
+
+
+ +
+
+
+ +
+
+
+ +
+
)
+ +
)
+
+
)
+ + + - - -
Lateral High
+
-
+ + (+) + (+)
(+)
(+) -
(+)(+)
+ + + + - -
(+)
+ + - +
- +
- +
00
59 3.3.
Syllabic Structure
Assuming that C represents any consonant and V any vowel, we can have the following types of syllables: V, CV, VC, and CVC. general formula would be:
(C)V(C).
Thus the
Below I illustrate these combinations
(where a period indicates syllable division): (21) a.pay
'to carry'
ka.lu
'far'
al.'u
'dog'
huk.nin
'one of them'
A sequence of two or more syllables forming a morpheme or even a word can be symbolized as:
(C)V(Q.CV(C)... , where we see a restriction
according to which any sequence of two vowels is prohibited.
This
contraint is very important, for, as we shall see, rules that create intermediate sequences of vowels are counterparted by rules that destroy the prohibited sequence. This also explains why Spanish diphthongs are either broken down, as the following examples illustrates: (22) tiyu
y
+
+
©
+
+
+
+
+
-
+
+
+
+
+
+
+
-
65 (33)
[muna:] [awsal'a': ]
'I want' 'I played'
[linli:]
'my ear'
[upya'lu:]
'I just drank'
Thus, as we see, the long vowel attracts the stress. This shift can be best regarded perhaps as a rule per se, rather than a mere exceptional phenomenon. The other exception has to do with a set of exclamatory words, such as: (34)
[alalaw]
'what a coldf'
[atataw]
'what a lousy thing!1
[acalaw]
'what a beauty!1
[anataw]
'what a sweet thing!'
[acacay]
'what a fearful thing!'
These words can also be used with the normal stress; thus the shift is clearly an expressive device used by the speakers. The predictability of the stress placement rule in Huanca is best corroborated by the treatment of Spanish loanwords.
Since in this
language the stress can move from one syllable to another, the different inputs, once they enter into the borrower language, are all leveled into penultimate accented words. Thus, according to this, we have: (35)
[mi:diku]
< Sp. medico
'doctor'
[sa:badu]
< Sp. sabado
'Saturday'
[papil]
< Sp. pap[ell
'paper'
[mfhur]
< Sp. melfo'lr
'better'
66 Note that the first two words have a lengthening compensation that witnesses the former Spanish stress; this occurs only when the syllable is free, otherwise, as expected, the lengthening disappears, as in [mlrkulis]
It should be clear that without the "minus next rule" convention, we cannot formulate the process involved, for it would be cumbersome to list all the environments where the rule applied, leaving aside a predictable context where it did not apply.
Rule (5) then predicts correctly when
/c/ became a retroflex and when it did not.
71 The most dramatic consequence of the operation of rule (5) was the neutralization of oppositions such as the following: (6) caki
•dry'
caki
'foot'
caka
'bridge'
caka-
'to serve from a pot'
maca-
'to irrigate'
maca-
•to get full'
pa?ca
'waterfall'
pa'ca
'forehead'
puc'u
'acid*
puc'u
•dwarf
since they became homophonous, that is: (7) caki caka maca pa'ca puc'u Also, as a result of the process, the distribution of /c/ was restricted to its occurrence in front of /i/, whereas the frequency of /c/ increased, for we have now original and historically derived affricate retroflexes. This is true, however, only for the facts described thus far. A closer inspection of the data gives us this new situation. retroflexion applied to forms such as the following: (8) cukla
'hut'
cuklus
'cricket'
cu'lu
'corn'
culu-
'to melt'
we would have got surface forms, such as: (8a) *cukla *cuklus
Had the process of
72 *cu?lu *culuinstead of what we currently find, namely: (8b) cukea cukeus cu'cu V
V
CUCU-
As we see, whereas initial /c/ does not change, we get a derived /c/ from an original /l/, Leaving aside the latter change, which will be taken into account in the next section, we have to explain why initial /c/ did not become retroflexed in these forms. The explanation for this apparent irregularity lies in one of the morpheme structure conditions I have introduced in Chapter 3 (Section 3.4), namely the "palatal harmony" principle.
As it was mentioned, this condition says that a retroflex
cannot co-occur with a non-retroflex palatal consonant within the same morpheme.
Note that the last qualification is important, for across
morpheme boundary it is perfectly normal to have non-harmonizing palatals within the same word.
Thus, we easily find forms such as:
(9) cakca+ykaca-
'to be grasping things here and there'
alpa+ca
'it is dust probably'
nawi+caw
'in the eyes'
caki+la+pa
'by foot only'
cukcu+yka:+sa+n+caw
'when he/she was asquath'
The only exception, as I have mentioned in Chapter 3, is the word /cas?a-/ 'to ripe', realized phonetically as [cas'a-].
Here, however, it seems
that, historically, the root goes back to */ca+sqa-/ (cf. /cawa/ 'unripe'), where *-sqa could be the past participle morpheme.
73 Thus, turning to the items of (8a), it is clear that these forms are violating the palatal harmony principle.
In other words, the fact that
we get (8b) instead of (8a) is explained naturally as a result of the controlling role of that principle, and therefore rule (5) needs no modification.
Consider, however, the following items:
(10) yumcuy yucka-
'daughter-in-law' 'to slip'
These forms correspond to the following found in Huailla-Huanca: (10a) lumcuy luckathat is, the historically basic forms. What we see in (10) is a derived retroflex that does not violate the principle, since, in view of (10a), it can be assumed that at the time rule (5) applied, the first segment was already /y/ and not a palatal lateral. What this implies, however, is that the yodlzation of III was a kind of "feeding" rule, since it prepared the appropriate environment for rule (5) to apply. this be maintained?
To what extent can
After a close look at the lexicon of YH, we find,
mostly for the area where retroflexion takes place, the following items: (11) yupi
'track'
yuta-
'to paster*
yawi-
'to mess up'
ya*?ti-
'to roll a rope'
corresponding, respectively, to the ones we find elsewhere in Huanca: (11a) lupi futa-
lawila^ti-
74 In view of this situation, we might be tempted to say that, in fact, such a rule of yodlzation existed.
And we may even say that it applied only in
initial position, for elsewhere we do not find /I7 - /y/ alternations at all (except for /liwla/ ~ /liwya/ 'clear', where the first form is found only in NH, thus being irrelevant for the present discussion). However, the situation is not that neat, for we immediately find a host of "exceptions": (12) lamka-
'to touch'
lanpu
'soft'
la*?wa-
'to lick'
lumlu-
'to crunch'
lusi-
'to paint one's face'
These forms never show alternation of their initial /!/.
(Incidentally,
I must note that a sequence *yi from /li/ would have been prevented by the morpheme structure rule mentioned in Chapter 3). This being so, the only reasonable alternative is that we cannot postulate a yodlzation rule as a feeding process, for it would not capture any regularity.
We certainly
cannot deny the possibility for it of having been a historical rule, but there is no evidence whatsoever for postulating its synchronic status at the time the retroflexion rule was operating.
Therefore, as for the items
of (10) and (11), I assume that they were relexified well before the operation of rule (5). Furthermore, it should be noted that the /l/ ** /y/ alternation is found not only where retroflexion takes place, but also in other communolects where that process is totally absent, thus showing that it was not functionally related to that rule.
If this conclusion is
correct, then the basic forms of (10) were, in Huaicha-Huanca, as follows:
75 (13) yumcuy yuckaforms upon which rule (5) applied without any conflict, giving us the correct output, that is (10).
4.2.
Delateralizatlon
One of the features that characterize northern QII is the change of */l/ to /j/ (such as in Cajamarca) or to /z/ (as in San Martin). type of change I will hereafter call delateralizatlon.
This
Among QI
dialects, only the Huanca-Quechua variety knows a similar process, as I have shown in the preceding section.
Thus, exactly in the same area where
retroflexion takes place, we find the following forms: (14) aca-
'to begin'
icay
'money'
kucu
'wood'
ayci-
'to dawn'
micwa
•wool'
wacpa
'chicken'
corresponding, respec tively, to these Huailla cognates: (14a) afailay kufu ayUmitwa walpa
76
As we see, the change took place everywhere, including the only two suffixes that carry /t7, namely the affective -la: and the limitative -la. which became, respectively, -ca: and -ca.
However, there Is one
environment where the process was invariably blocked, and this is in initial position after pause. (15)
Some examples are:
lapsa
'thin'
Tapi-
'to squeeze'
lantu
'shadow*
lipta
(a kind of elaborated eatable ush)
la'wa-
'to lick'
In view of this we must conclude then that the change did not operate word initially.
Thus, the process of delateralizatlon can be formulated
in such a way that it gives us the correct synchronic forms from their corresponding historically basic forms.
Since the rule did not take place
when the input was not preceded at least by one segment, we can formulate it as follows: (16) */!/ > /c/
/ [+ Segment]
Assuming now that this rule makes the correct predictions, our next step is to look at the interactions that existed between this process and that described by rule (5).
4.3.
Rule Interactions
As was shown in Section 4.1, there is evidence that shows that rules (5) and (16) were functionally related.
To convince ourselves that such
was the case, we only need to consider the following earlier forms:
77 (17) pala-
'to pick up'
alpa
'earth, dust'
wayla
'meadow*
watwa
(a kind of medicinal plant)
These forms appear in Huaicha-Huanca as follows: (17a) pacaacpa wayca wacwa As we can see, rule (5) did not operate, and consequently the items of (17a) constitute opaque cases in relation to it.
In view of this situation,
the only reasonable alternative is to call for rule ordering.
If (17a)
was not affected by rule (5), then obviously it applied prior to the operation of rule (16). Thus, to derive (17) correctly we have to apply (5) before (16); the reverse order would give us a wrong output. The following derivations illustrate the proper application of the rules: (18) *caka
'bridge' *laca 'frog'
caka
*tufu 'bone' *luna 'flood'
lac a
Rule (5) tucu
lu*?ca Rule (16)
As it is clear enough, rule (16) does not feed the application of rule (5). Now, for historically underlying forms such as: (19) cu'lu
'corn'
cukla
'hut'
cula
'odd'
culu-
'to melt'
we have to explain why rule (5) did not operate in order to give us forms such as the following:
78
(19a)
*cVlu *cukla *cula *culu
for instead of t h a t , we encounter, respectively: (19b)
cu*cu cukca V
V
cuca V
V
cucuThe reason, of course, is the preventative action of the palatal harmony principle; consequently, (19a) probably never arose. This situation illustrates then the powerful capacity of blockage inherent to that principle.
If, however, rule (16) had applied before (5), then an
"intermediate" stage such as (19b) could have subsequently changed to (20): (20) *cu?cu *cukca . A
A
*cuca . A
A
*cucu since, as we have seen in Section 4.1, forms such as the following: (21) ca?cu
'to sprinkle'
cakca-
'to grasp'
cukcu-
'to tremble'
became, respectively, by rule (5): (21a) ca?cuA
. A
cakcaA
. A
cukcu-
79 thus showing that, if there is more than one affricate non-retroflex within a sequence, both of them can change to their corresponding retroflexed forms.
This proves, once more, that if we do not get (20) but
(19b) it is because at the time (16) converts the following: (22) cu'lu cukla cula culu into (19b), rule (5) was no longer operative.
It is clear then that,
from the facts observed, rule (5) has precedence over rule (16). The historical implication is also clear: obviously, the change described by the former rule preceded the latter.
4.4.
General Discussion
As we have seen, rule (5) affected drastically the Huaicha-Huanca lexicon, since the opposition retroflexed-non-retroflexed became neutralized in most environments.
Furthermore, the only original /c/
that remains is that which we find in front of /!/, and also the /c/ immunized by the palatal harmony principle, that is in a very limited number of roots and in two suffixes. Elsewhere we get only historically derived /c/ merged with original /c/, with the only difference that whereas the latter occurs in any environment (including in front of / ! / ) , the former is absent in one context, namely preceding the vowel /!/. This severe change had its consequences, of course.
Since, as a result
of the almost complete loss of the non-retroflex affricate, the functional load of /c/ is limited to its appearance before /!/, the phonological system of the language became, so to speak, unbalanced.
This would be
80 particularly true, for it seems that on implicatlonal terms the existence of an affricate retroflex in a language generally implies the existence of its non-retroflex correlate.
(Note, incidentally, that the predominance
of /c/ over /c/ is no-iced by the Huailla-speakers who claim that the Huaicha-speakers "pronounce everything" with a [c].) Now, the unbalanced situation just depicted was precisely remedied by the operation of the other rule, that is delateralizatlon.
In fact, what
this rule did was to fill the empty hole left by /c/. Thanks to this process, the phonological system has now more occurrences of /c/. The interplay of the two rules is visualized as follows: (5)
III
(23)
-
l£l
(16) t
III We have thus here a nice example of the "drag chain" type of change 2 studied by Martinet.
v
~
As we see, /c/ drags /!/ into its former slot. By
doing this, the system recovers its equilibrium, since the hole that is vacated by the retroflexion of /c/ is now occupied by /c/, historically derived from /l/.
In this way, I think, the facts illustrated here
support the claim made in the sense that, if in a language there is a retroflex consonant then there should be also a corresponding nonretroflex. As we have just seen, rule (5) acted against this tendency, but then rule (16) came in to help, and the (universal?) principle was not contradicted. Having determined the type of relationship that existed among the rules studied, we must now direct our attention to the processes involved in them.
Thus, let us consider rule (5). Note that the rule, as stated,
apparently does not show any process whatsoever. And this is because the
81 change became generalized, at the present state it being impossible to find the conditioning factor.
However, once we realize that the process
did not take place in front of /i/, then we can at least say that the rule got inhibited in a naturally (articulatorily) blocking environment.
In
other words, what I am implying is that a sequence ci is, in articulatory terms, heterorganlc, whereas ci, on the other hand, is fully homorganic. Thus the natural tendency would be for a language to avoid such a sequence. In fact, in a cross-linguistic research on retroflexion, Bhat (1973: 4850) finds that, especially for fricatives and affricates, the process of retroflexion does not take place in front of /i/ (and he cites eight cases illustrating this situation); back vowels, on the contrary, appear as one of the most prominent conditioning environment that induce retroflexion (the same author lists at least nine cases where this happens).
The
blockage, as we shall see, is also true for the process that affects /s/ in Huanca-Quechua (see Chapter 5).
This is not to say, of course, that
a sequence such as ci is completely impossible.
As the data presented
illustrates, such sequences do occur in the language.
It must be clear,
however, that this fact does not contradict the natural tendency mentioned, and which is apparent in the inhibition of rule (5) in front of /i/. As for the original inducing environment for retroflexion, one can perhaps hypothesize that it was a following back vowel (including of course /a/); then the next step would have been its full propagation across the lexicon. This is a quite plausible hypothesis; unfortunately the synchronic situation does not help, since it does not provide us with morphophonemic alternations or dialectal evidence. But even if our hypothesis turns out to be true, it remains to be explained why it is that back vowels favor retroflexion, what kind of process (assimilation
82 perhaps?) is involved in it. Unfortunately, the present distinctive feature theory cannot give us any satisfactory answer to this problem, especially because retroflex sounds have not been carefully studied. Rather than treating retroflexion as a concomitant feature only, it is important to realize that, as Bhat (1974) says, "the consonants that have a retroflexed tongue-position show a number of interesting phonological characteristics that cannot be explained or taken care of without referring to their specific articulations." As for the formulation of rule (5), I must say that, to my knowledge, rules of this type cannot be successfully stated within the realm of generative phonology.
A formulation such as:
(24) */c/ > Icl I
~ III
does exactly the same as the "minus next rule" convention used in the formulation of rule (5). Thus, in view of this, I think what is more important here is to know that the process was naturally blocked in an environment articulatorily determined.
Therefore, I shall not discuss
further the correctness of rule (5); for our present purposes, it should be clear that the blockage involved is apparent. Before we discuss rule (16), it would be worthwhile to inquire whether perhaps rule (5) can be more naturally described as a case of rule inversion.
In fact, it is tempting to formulate (5) as, say:
in
(25) Icl - Icl I
This r u l e , of course, looks more natural, i l l u s t r a t i n g a nice assimilatory deretroflexion by p a l a t a l i z a t i o n . a s e r i e s of d i f f i c u l t i e s .
However, such a formulation runs into
In the f i r s t place, now that h i s t o r i c a l l y
derived / c / i s taken as basic underlying segment at the same l e v e l as o r i g i n a l / c / , we have to mark a r b i t r a r i l y which / c / — t h e inherited or the
83 historically derived—goes to Icl,
since, as we know, there are sequences
of ci that do not change, and therefore would constitute counterevidence—opaque cases—to rule (25). Secondly, as Vennemann (1972) and Schuh (1972) would point out, an inverse rule is crucially motivated when a historically derived variant appears in the lexically basic forms (such as, singular versus plural, indicative as opposed to subjunctive forms, etc.), so that the new underlying form can be taken as the norm, thus inducing leveling.
In other words, inverse rules become productive
phonological processes.
In our case, there is no such morphosyntactic
predominance, for what we find is the operation of rule (5) across the lexicon, irrespective of grammatical distinctions.
Thus, we do not have
cases of allomorphy, and therefore no isomorphism needs to be induced. For these reasons, I believe, we cannot postulate seriously an inverse rule.
Rather, I would prefer to maintain (5) as the historical process
that affected /c/ in the subvariety studied. In regard to rule (16), what is intriguing is its non-operation wordinitially.
As rule (5), this process is also completed; that is, no
traces of alternations could be found, either internally or externally. Whatever its conditioning factor, the only thing we know is that initial position after pause proved to be the most resistant environment for delateralizatlon. As I have said in Section 4.2, a similar process took place in Northern QII; but here the change was generalized across the lexicon, leaving no traces of environmental reluctance.
In other words, here total
restructuring took place, thus the rule being eliminated from the grammar. Now, as there is ethnohistoric evidence of the presence of mitmas--ethnic groups moved from their original home-land to new hostile territories--
84 3 from one of the northern delateralizing provinces, specifically Canaris, in Lulin-Huanca, that is precisely where rule (16) takes place, one may wonder whether such a process was induced through borrowing.
This may
explain perhaps its irregular propagation, not affecting initial /!/. However, I am reluctant to accept this type of explanation for three main reasons.
First of all, it is unusual for a borrowed rule to have a
systematic (regular) exception, and no cases of hypercorrections and doublets.
Secondly, the mitmas were always a minority group that, rather
than influencing the local language (or dialect), soon became assimilated, at least linguistically.
And thirdly, we have seen that rule (16), if
correctly interpreted, was internally motivated in order to restore the equilibrium of the phonological system put in serious danger by the operation of rule (5). For these reasons, while the non-application of the rule word-initially remains a mystery, I would prefer to subscribe to the internal motivation I tried to show, rather than to an external explanation, no matter how tempting it would be. As in the case of rule (5), here also we cannot postulate an inverse rule, say:
(26) Icl - III I ## for even though the rule looks simple (although not necessarily natural), it would wrongly predict the change of the £i sequences and the original Icl
that does not change because of the controlling effect of the palatal
harmony principle. III
To avoid this, we would have to mark which /c/ goes to
and which remains unchanged.
For these reasons, again, I prefer to
consider rule (16) as a historically completed process. As I have said in discussing both rules, the two processes are no longer operative.
There is evidence that shows that, at the time the
Spaniards came, both rules were no longer productive.
In Chapter 8 (cf.
Section 8.3), I shall provide more evidences to show that the two processes led to restructuring in Huaicha-Huanca. To end this chapter, I must add that the reason I named the two subvarieties as Huailla and Huaicha is because on the west bank of the Mantaro river, near the border between the two areas, there is a place called Huailla (that is /wayla/ 'meadow') by the Huailla-speakers, but Huaicha (i.e. /wayca/) by the speakers of the other subbranch.
FOOTNOTES FOR CHAPTER 4
The only exception I found is the change of the PQ suffix *-cka: into -cka: 'durative-simultative'. I have found isolated relic areas where that morpheme is still pronounced with the retroflex (in Comas, for example). Cf. also with the modal -cka found in Cajamarca (QII). 2
Cf. Martinet (1955: 48-62).
3 According to Espinosa Soriano (1973: 4 6 ) .
87
CHAPTER 5 RETROFLEXION OF /s/
In view of the evidence shown by modern Quechua varieties, there is v
1
no doubt that the PQ phonemic system had a contrast between /s/ and /s/. Whereas in most QII dialects both phonemes merged into Is/,
in QI we find
the original opposition, although in some dialects /s/ went to /h/, and even disappeared.
Among the QI varieties, Huanca-Quechua not only main-
tains such distinction, but has developed a third sibilant, namely the retroflex [s].
Aside from Huanca, only Cajamarca-Quechua (QII) is known 2
to have a similar sibilant, and also an apical non-retroflex.
In
relation to the situation in Huanca-Quechua, as we shall see, [s] comes into existence as a result of a process of retroflexion that affects /s/. The appearance of this new segment is common to most Huanca communolects, 3 including some other neighbouring dialects. Impressionistically, this process makes Huanca-Quechua sound very peculiar among the other varieties, comparable to the peculiarity of peninsular Spanish—which has an apical [s]—as opposed to American Spanish.
In the following I shall study this
innovation, trying to argue in favor of its synchronic status. To account for the process involved, two main alternatives will be presented: flexion and palatalization.
retro-
I will try to support the first alternative.
The study will be based entirely on data corresponding to the YH variety; but, as I have said, the process covers the NH branch also, with some conservative areas. Occasionally, however, we will make some references to this subbranch in support of our claims.
88 5.1.
The Process
To see how retroflexion operates, let us consider the following phonetic forms: (1)
[sasa-]
'to drop (gral
[simi]
'mouth'
[suti]
'name'
[asi-]
'to seek'
[sansa]
•coal'
[ispa-]
'to urinate'
[sukuy]
'sandal'
[iskay]
•two'
[asta-]
'to carry out'
[wispa]
'mouth-twisted'
[tusu-]
'to dance'
[is-?uy]
'pus'
[kaspa-]
'to roast'
[isana]
(a variety of herb)
[asna-
•to stink'
[aysa]
'wasteland'
[asya-]
'to tear'
[uysu-]
'to bow down'
[tupsu]
•beck'
[ulmis]
(a variety of plant)
[liwlis]
'seagull'
[anas]
skunk'
A simple glance at the data offered is enough for delimiting the complementary distribution of [s] and [s]. The latter allophone occurs in these two environments:
either before or after /i/, and after /y/ (cf.
[asya-] but [aysa]; a form such as [isya-] 'to be sick* shows the "redundant" conditioning of /y/, since here the principal inducing factor is, obviously, vowel / ! / ) .
Retroflex [s] occurs elsewhere.
Thus, to
account for the data so far presented, we can formulate our rule of retroflexion in the abbreviated following form:
(2)
Isl
[-next rule] / { 0A]
}
89
As stated, the rule does not apply either before or after /i/ (and that is why we appeal to the "neighbourhood convention"), and also after semivowel /y/. Notice also that, as in the case of rule (5) in the preceding chapter, we have to make use of the "minus next rule" convention, for otherwise we would have to list all the environments where the retroflex appears, leaving the elsewhere statement for a predictable and limited context.
Assuming then that rule (2) is correct, let us examine somewhat
more complex examples. The items in (1) have in common one thing:
they are all mono-
morphemic stems. This is true at least synchronically, for historically, roots such as [anas], [ulmis] and [liwlis] are bimorphemic in origin (going back to *ana+s,ftqulmi+sand *liwli+s, respectively). Before we introduce polymorphemic stems, it should be mentioned that in Huanca v 4 there is at least seven suffixes that contain /s/. (3) -sa?
These suffixes are:
'1st. p. sg. future'
-sa
'past participle*
-su
'3rd. p. actor (always followed by the 2nd. p. object)'
-sun
'1st. p. pi. future'
-Stiri
' subordinator'
-js
'attributive'
-si
'hearsay reportative'
Of these, the last two show automatic alternation, as is apparent from the following. Let us consider some examples with the attributive first. Assuming that the forms listed in (4) constitute the underlying basic form:
90 (4) Stem + Attributive utu+s
(one who has goiter)
uysu+s
(one who is bowing down repeatedly)
cina+s
'womanly'
uma+s
'big-headed'
ffuti+s
'sniffling'
linli+s
'big-eared*
we get (4a), after the application of rule (2): (4a)
[utu+s] [uysu+s] [cina+s] [uma+s] [nuti+s] [linli+s]
that is, the correct surface forms predicted by the rule. Now, let us consider some examples with the reportative.
Assume, again, that (5)
constitutes the underlying forms: (5) Stem + Reportative nuna+si
'(it is said) it is a person'
lu:ku+si
'(it is said) it is an insane person'
walmi+si
'(it is said) it is a woman*
tukl+si
'(it is said) it is beautiful'
ulpay+si
'(it is said) it is a dove'
mama+:+si
'(it is said) it is my mother'
Here, as we can see, rule (2) cannot apply, since the environment where /§/ appears is precisely the one that forbids retroflexion. Nevertheless, we get the following surface forms:
91 (5a)
[nuna+s] [lu:ku+s] [walmi+s] [tuki+s] [ulpay+si] [mama+:+si]
As it should be apparent, to account for (5a) we have to postulate a new rule, call it "vowel deletion" (cf. Chapter 3, Section 3.2).
This rule,
which will be treated more extensively in Section 5.5, eliminates final vowel /!/.
Once it applies, then rule (2) comes into play, giving us the
correct output as in (5a). Note that this implies that vowel deletion is a "feeding" rule in relation to retroflexion. Thus, so far we have seen that in stems with the attributive and reportative, rule (2) applies regularly, with the only proviso that in the second case a new rule has to prepare the correct input.
This latter
will be handled in Section 5.5. Now, let us consider stems with the remaining five suffixes listed in (3). The situation we find is the following. For underlying forms such as: (6) a)
b)
Stem + 1st, p. future li+sa?
'I will go'
puli+sa'?
'I will walk'
Stem + Past participle asi+sa+yki
'what you sought'
lanti+sa+n
'what he/she bought'
92 c)
Stem + 3rd, p. actor asi+su+nki
'he/she seeks (or will seek) you'
wanu+ci+su+nki 'he/she will kill you' d)
e)
Stem + 1st, p. pi. future puli+sun
'we will walk'
paki+sun
'we will break (it)1
Stem + Subordinator asi+stin
'laughing'
mici+stin
'pasturing'
we find, respectively, the following alternations: (6a) a)
[li+sa'?]
- [li+sa*?]
[puli+sa*?] ~ [puli+sa*?] b)
[asi+sa+yki] ~ [asi+sa+yki] [lanti+sa+f]] ~ [1 anti+sa+T]]
c)
[asi+su+nki] ~ [asi+su+T]ki] [wanu+ci+su+T]ki] ~ [wanu+ci+su+flki]
d)
[puli+suT]] ~ [puli+suT]] [paki+sut]] ~ [paki+su*!)]
e)
[asl+stiT|] ~ [asi+stil]] [mici+stiT]] ~ [mici+sti-n]
Thus, we see that whereas in the first column rule(2) is blocked, as expected, in the second column the same rule is apparently violated, since it applies after vowel III.
The same could be said about the
following underlying constructions: (7) paki+'lu+sa alwi+?lu+sa
'recently broken' 'recently entangled'
93 realized in the most conservative areas, as follows: (7a)
[paki+'+sa] ~ [paki+-?+sa] [alwi+?+sa] - [alwi+'+sa]
(where, as I have advanced in Chapter 2, the syllable lu of the eductive -?lu drops entirely leaving alone the glottal stop).
However, note that
in (7a) we are actually dealing with a different situation, since now it is the first column that violates rule (2). As we see, whereas the second column is correct, the first one shows [s] occurring in an environment not provided by rule (2). In other words, the items of the first column are opaque cases for rule (2), since the contiguous protecting environment is absent.
The apparent contradiction does not occur when we
take into consideration the same forms of (7), but this time as rendered in the most innovative areas, that is: (7b)
[paki+:+sa] ~ [paki+:+sa] [alwi+:+sa] ~ [alwi+:+sa]
(where /?/ drops leaving as a witness the vowel lengthening upon the preceding vowel).
As we see, now it is the second column that violates
rule (2), as it does in (6a). Comparing (7a) with (7b), it is clear that the first column in the former is not a counterevidence to rule (2), but rather a regular derivation, once we realize that a segment such as /?/ does not constitute a barrier for /s/ to be preserved as a non-retroflex. As it is known, there is the general tendency for laryngeal segments to be "transparent" to certain phonological processes, especially nasalization and palatalization.
In the first column of (7a) then /s/ is well-
protected since /?/ does not interfere.
Accordingly, the fact that in
the second column retroflexion does take place is obviously due to the puzzling alternations we have found in (6a), and not because /?/ leaves
94 I si unprotected.
In sum, the presence of the glottal stop does not change
the nature of the alternations brought forward. As we have just seen, however, it is apparent that rule (2) needs to be reformulated.
For, as stated, the prohibiting environment does not
include /!'/ sequences, such as those of (7a). Since a sequence in the form of */s?i/ never occurs in Quechua I (thus being an accidental gap; cf., on the other hand, lis?!-
'to know'), we can reformulate (2) as
follows:
( [-next rule] / { i ( ? ) } (
(8) in ->{
y
( W
(
)
where /?/ can be expanded optionally.
This rule takes care for the
regular derivations presented so far.
However, we have yet to explain
the irregular appearance of the retroflex after /i/ in the second column of examples (6a), (7a) or (7b). 5.2.
Morpheme Boundary Amnesty and Rule Propagation
Before we enter into discussing the matter, let me point out that the alternation does not seem to be areally determined or stylistically governed.
This appears to be true, because, as I have carefully observed,
such variation takes place within the same community—and even among the members of the same family.
On the other hand, one and the same speaker
cannot alternate [s] and [s] in that environment; he uses either the palatal non-retroflex—which is what we expect—or the retroflex, but not both of them.
Furthermore, the speaker does not seem to be aware of the
existence of such alternation, unless we provide him patiently with "contrastive" examples.
In this case, a speaker who always says
[li+'N-sa?] 'I will go right away', confronted with another who has the
95
"irregular" form, that is [li+'+sa?], would say that his pronunciation is also fine. Now, to understand the alternations found in (6a), let us consider the following examples: (9) isu
'illegitimate son (or daughter)'
isana
(a variety of herb)
wisa
'harelip'
tis*u
'island'
wistu
'limp'
If rule (8) had not incorporated its negative statement, then we would get the following incorrect forms: (9a) *[isu] *[isana] *[wisa] *[tis?u] *[wistu] As we can see, it is clear that alternations such as (6a) occur only across morpheme boundary.
In other words, the presence of a morpheme
boundary provides a sort of amnesty for retroflexion to take place in an environment otherwise prohibited.
Since the items of (9) are mono-
morphemic words, there is no place for alternation.
In the vicinity of
a morpheme boundary, on the contrary, such variation is possible. Thus, even for an idiom such as /ima+y+su+n?a+la/ 'what a mourn!', we find both, the regular, and the "irregular" derivation, that is [ima+y+su+Tpa+la], But a form like *[aysa] never occurs. In order to formulate this facultative process of retroflexion we have to appeal to some sort of "sub-rule," of the form:
96
(10)
ill
K?) -([5]) / {y }+
where, as can be seen, the environment is actually a subset of the negative context stated in rule (8). Since, by convention, an environment that does not mention explicitly a morpheme boundary applies across it, whereas the reverse is not true, (10) correctly specifies that /s/ becomes optionally a retroflex after /i, y/, only when these segments are followed by a morpheme boundary. With regard to the plausibility of conflating rules (8) and (10), it seems clear that any attempt in that direction would lead to an ad-hoc formulation. Thus, one possible abbreviation could bt: [-next rule] / { (11) III - {
i(9)
)
I i (?) )
l \ y
{+
[3] but then the first statement enters into contradiction with the second, for the first prohibits the application of the rule, and the second opens the possibility of breaking down a subset of that environment. Clearly then, a rule such as (11) cannot; have the least psychological plausibility, aside from being too complex and, worse than that, contradictory. These defects cannot be attributed, however, to the infelicitious formulation of the rule, but to the fact that the present theory of generative phonology is not able to handle processes such as the one we are describing. And yet the process itself is, at least partially, a very natural one. In fact, as we have seen in the preceding chapter, it seems natural that a process of retroflexion did not take place in contact with a front vowel, for otherwise we would get a heterorganlc sequence such as *[si]
97 or *[is].
In other words, the vowel /!/ protects /s/ against the
propagation of retroflexion, in the same manner as the same vowel prevents Icl
from becoming [c], with the only difference that in the latter the
protective power of /i/ is felt only when the affected segment precedes it. Rule inhibitions because of protecting environments are not hard to find.
One such case is the Spanish spirantization rule. Details aside,
in Spanish /b, d, g/ are realized as stops only initially after pause and also after nasals; elsewhere they are realized as [6, 6, v], respectively. What is important for our present purposes is that a nasal segment prevents the stops from becoming spirants.
In other words, a nasal consonant
protects /b, d, g/ from becoming softened.
It is clear then that the
motivation for the inhibition of the two processes—retroflexion in Huanca and spirantization in Spanish—is one of articulatory nature. Therefore, even if rule (8) does not show explicitly any process, nevertheless it captures the inhibitory character of /s/ in contact with /i/. As for rule (10), I prefer to maintain it as a second stage of rule (8), therefore not collapsible with it. With regard to the apparent violation of rule (10) with respect to (8), I must say that the process involved is also natural, but this time not in language universal terms, but language-specifically. rule (10) shows two things:
In fact,
(a) that the process of retroflexion is in
progress, this time affecting "protected" /s/'s; and (b) that the change is being implemented, as in most changes in Quechua, first of all across morpheme boundary only.
When the process gets consolidated in that
position, it will eventually propagate morpheme-internally.
This is
precisely what is happening in localities such as Marco and Millpun
98 (NH), where rule (10) is no longer optional and, more than that, /!/ is realized variably as [s] in front of /i/, which proved to be the last resistant environment for retroflexion.
In view of this, it is likely
that in Yaqa-Huanca we will eventually have the same outcome, as the last step towards the total generalization of the rule, that is restructuring. In Chapter 7 we will try to suggest why it is that certain changes begin in the environment of a morpheme boundary.
In the meantime, however, I
would like to discuss a possible alternative motivation for such a gradation. Thus, if we consider the following paradigms: (12) a)
b)
c)
d)
e)
asi+sa'] ~ [asi+sa*?]
'I will seek (it)'
miku+sa-?]
' I will eat (it) '
lika+sa?]
'I will see (it)»
asi+su+Tlki] ~ [asi+su+Tjki]
'he/she will seek you'
niiku+su+T|ki]
'he/she will eat you'
lika+su+Tlkl]
'he/she will see you'
asi+sul]] ~ [asi+sul]]
•we will seek (it)'
miku+suT)]
'we will eat (it)'
lika+suT|]
•we will see (it)'
asi+stin] ~ [asi+stin]
'laughing'
iniku+stiT)]
'eating'
Uka+stiT)]
'seeing'
asi+sa+T]] ~ [asi+sa+7]]
'what he/she sought'
miku+sa+T]]
'what he/she ate'
lika+sa+Tj]
'what he/she saw'
99 we note immediately that the suffixes carrying /s/ have two different realizations depending on the nature of the preceding vowel; if it is /i/ they appear with the palatal non-retroflex, elsewhere /s/ is realized as a retroflex, as expected.
What this means is that the allomorphs with the
retroflex are predominant, whereas the allomorph with the non-retroflex occurs only once. This situation, I think, might prove to be a powerful reason for the leveling of the allomorphs, reshaped uniquely with an [s].
In other words, the appearance of alternants such as the following: (13)
[asi+sa?] [asi+su+7|ki] [asi+sul]] [asi+sti-n] [asi+sa+'n]
would be an indication of a movement in the direction of a paradigmatic coherence, reducing or minimizing the surface phonetic differences. This is, I think, a cogent reason that would explain the inception of forms such as (12), a common morphological process found elsewhere.
As it
usually happens, the process of retroflexion, being a regular change, created morphological irregularities in the language; these irregularities, when sufficiently pressed by a paradigm, tend to be minimized.
This
implies, psychologically, that a speaker feels uncomfortable in having to deal with two or more different morphs to convey the same meaning and the same function.
Compare now the situation in (12) with that in (9);
in the latter, obviously there is no such pressure since those forms do not alternate at all. Thus, as we have seen, the notion of paradigmatic uniformity can explain more or less convincingly the apparent "irregularity" of forms
100 such as (13). However, in spite of its seductiveness, I do not think that in this case the paradigmatic coherence principle is playing any role at all.
For this would mean that forms such as those of (9) will never, at
least theoretically, end up with the retroflex.
And yet, as I have said,
in certain areas of NH [s] ~ [s] alternations can be found morphemeinternally.
For this reason, I would prefer to see in forms such as (12)
a further step in the process of total generalization of rule (8).
5.3.
Palatalization:
An Alternative
As in the cases of the retroflexion of /c/ and delateralizatlon of /l/, it would be worthwhile to explore whether rule (8) is amenable of being reinterpreted as a typical example of rule inversion.
In this case,
the rule involved would be palatalization instead of retroflexion. Thus, the rule would be formulated as follows:
(14) /s/ - [s] /
r
where the underlying segment is the historically derived retroflex sibilant.
Accordingly, we would have a natural assimilatory process of
palatalization (an unusual situation in cases of rule inversion), and the rule would look very simple.
If this is so, then forms such as (13)
would appear as reluctant to undergo the process.
Since this resistance
occurs in the environment of a morpheme boundary, then rule (14) runs into the same difficulties we have found before, for now it has to have a prohibitive environment which, again, is a subset of the second context. In any case, rather than losing protective environments, according to this new interpretation, the palatal allophone would appear as propagating to new contexts.
This situation, while theoretically possible,
101 does not seem to be the case in Huanca.
As I have said, there i s evidence
showing that the opposite process i s true; that i s , the retroflex i s becoming more and more generalized at the cost of the non-retroflex. Thus, a formulation such as (14) would not be capturing any actual process, but merely giving account of the situation, irrespective of the rule propagation that alternations such as (6a) clearly show.
If however
one has to choose between a simple rule that correctly predicts the facts and another more complex that, nevertheless, seems to capture a sound change in progress, I would certainly prefer the l a t t e r . 5.4.
Comments The facts observed so far clearly show that the retroflex sibilant
will soon become the basic underlying form.
In this way, the retroflexion
rule will get lost, and restructuring will take place instead. Note that one of the consequences of this fact affects the palatal harmony principle, for underlying forms such as: (15) cusnu-
'to disinflate'
sanu
'clay'
sulu
'miscarriage'
suli
'mist'
are realized, in Huailla-Huanca, as follows: (15a)
[cusnu] [sanu] [sulu] [suli]
As we see, we now have occurrences of [s] and a non-retroflex palatal within the same morpheme.
In this case, the palatal segments did not
102 prevent /s/ from becoming [s], as was the case in the retroflexion of /c/ in Huaicha-Huanca.
This is, I think, a drastic change in the underlying
morpheme structure of Quechua. At this point, a word must be said in relation to the sameness of the retroflexion of /c/ in Huaicha-Huanca and of /s/ elsewhere. As far as I can see, I do not think the two processes are related.
In terms of
diffusion, the first process is restricted to a very limited area, surrounded by Huailla and NH; the second process, on the other hand, covers almost the whole territory of Huanca, and even extends beyond it. The fact that the two found in /!/ a resistant environment can surely be attributed to the natural tendency among retroflexes mentioned in the preceding chapter.
Aside from this fact, there is evidence showing that
whereas the retroflexion of /c/ was completed at the time the Spaniards arrived, the retroflexion of /s/ was fully productive. Thus, an early borrowing such as paja *[pasa] 'straw' was incorporated as [pa:sa]; other loanwords maintain Spanish medieval /s/, because the target words contain a favourable environment as in [sintil] from gent11
*[sentil]
'profane', [uwisa] from oveja *[obesa] 'sheep', [silgis] from jilguero *[silge'ro] 'linnet', etc.6 As for the original inducing environment for the retroflexion of III,
we may postulate that, as in the case of the retroflexion of /c/,
perhaps it was a following back vowel. However, in the few areas where this process is almost absent (Paca and Yanamarca), I have found that retroflexion takes place only before alveolars I pi.
We thus have forms such as the following: (16)
[a'sta-]
'to carry'
[la'sta]
'snow'
and, sometimes, before
[cusnu-]
'to disinflate'
[ka'spi]
'stick'
[puspa-]
'to soak'
Since /p, t, n/ are anterior consonants one may wonder what kind of process is involved in here.
And even if we know that there is an
assimilatory process, could this be considered as the initial stage of the change, and therefore could we postulate this as the original situation for the remaining area where retroflexion is almost completed? Or could it be that retroflexion begun syllable-finally, as a kind of weakening process?
The answer surely requires a more careful study of
the process in its very beginning, as in (16), as well as a more refined comprehension of the articulatory nature of the sounds involved.
5.5.
Rule Ordering and Dialect Differentiation
In a relatively vast area that includes the northern region of Huailla and covers the whole Huaicha branch, we find an apparent case of counterexample to rule (8). Thus, within this area, which I call hereafter Northern Yaqa-Huanca (NYH), we find consistently the following surface forms: (17)
[wamla+s]
'(it is said) it is a girl'
[ka:ru+s]
'(it is said) it is a car'
[luntu+s]
'(it is said) it Is an egg'
[rasum+pa+s]
'(it is said) it is true'
As we see, the items of (17) constitute clear violations of rule (8), since there is nothing like HI
or /y/ to protect /s/ from becoming [a],
and yet it does not change. This situation does not arise in the rest of the Huanca area, which I call hereafter Southern Yaqa-Huanca (SYH),
104 for instead of (17) we find the normal output after the application of rule (8), that is: (17a)
[wamla+s] [ka:ru+s] [luntu+s] [rasum+pa+s]
The picture is more complicated for in the same NYH-area, as well as in the rest of Huanca-Quechua (with the exception of the few conservative spots I have mentioned earlier), we find the following forms: (18)
[utu+s]
(one who has goiter)
[uma+s]
'big-headed'
[c1na+s]
'coward'
[silj'a+s]
'big nose'
that is, correctly predicted surface forms. Comparing (17) with (18) we immediately notice that (17) cannot be a systematic violation of rule (8), for otherwise why the same violation does not occur in the items of (18)? Could it be that (17) is an exception to the process of retroflexion? If that is the case, then we would be in face of a morphologically conditioned exception:
the rule is blocked when /s/ marks the reportative.
Although morphological exceptions to a rule are not unusual, I think there is evidence to show that in this case we are in face of a more interesting phenomenon. In order to provide a more convincing explanation for the apparent violation of rule (8) in (17) we must recall what we said in discussing the items of (5) in Section 5.1. Thus, in (17) the ending is one of the realizations of the reportative suffix whose basic form is -si. In other words, the items of (17) are derived from (19):
(19)
wamla+si ka:ru+su luntu+si rasun+pa+si
To t h e s e forms we apply a pan-Quechua minor r u l e of vowel d e l e t i o n , a process t h a t d e l e t e s the vowel of t h e three r e p o r t a t i v e suffixes, namely -mi ' f i r s t hand information', - s i 'hearsay i n f o r m a t i o n ' , and -ca 'conjectural'
(which sometimes appear as - c i i n c e r t a i n communolects).
The r u l e applies only when the stem to which t h e suffix i s added ends in a short vowel (thus cf.
(5) and ( 5 a ) ) .
The r u l e involved can thus be
formulated as follows: (20)
Reportative Vowel Deletion (RVD) C V - [0] / XV + [+ Report]
Turning now to (17), and comparing it to (18), it is clear why the former shows [s] instead of the expected [s]: somehow the underlying /!/ of the reportative, although absent on surface, is responsible for its preservation as a palatal non-retroflex. Now, with regard to the different treatment of the reportative allomorph -jf among NYH and SYH, we can say that one way in which such discrepancy can be satisfactorily explained is by postulating different rule orderings.
(21)
Accordingly, we would have:
For NYH 1.
Retroflexion
2. RVD For SYH (including NH) 1.
2.
RVD
Retroflexion
106 We can illustrate the two different orderings by deriving /wamla+si/ '(it is said) it is a girl' and /walas+si/ '(it is said) it is a boy', as follows: (22) For NYH wamla+si
walas+si Retroflexion
RVD
wamla+s For SYH wamla+si
walas+si
wamla+s
RVD
wamla+s
Retroflexion
As can be seen, the different treatment of the reportative among NYH and SYH is nicely accounted for by the postulation of different rule orderings.
5.6.
Semantic Blocking
In this last section, I would like to discuss the case of some residual aspects that could be taken as counterevidence to our rule (8). Thus, let us consider the following examples: (23)
sapu
'fuzzy dog'
asuti
•doggy'
sanay
'small'
lapsa
'light'
v~
'bit'
as la 'small' taksa If we apply rule (8) to these forms we would get the following incorrect derivations:
107
(23a) *[sapu] *[asuti] *[sanay] *[fapsa] We have to explain then why (23a) is an incorrect output. What is the reason for rule (8) to get blocked in the forms of (23)? Notice that in a traditional phonemic analysis (23) would constitute enough evidence for postulating two separate phonemes—/s/ and / s / — , since they now appear in the same environment, that is other than next to /i/. This implies, accordingly, that neither III
or /s/ are predictable. Rule
(8), for instance, would wrongly predict (23a). true?
To what extent is this
A closer inspection of the data will give us the answer.
First, we must recall that one of the most controversial questions raised by the first grammarians and philosophers (at least in the Western culture) was the relationship between sound and meaning.
While in
linguistics the idea prevailed that such a relation is entirely arbitrary (especially after the consecration of the Saussurean dichotomy signifiant/signifie). literary theorists and poets always maintained that, at least some type of sounds, are iconic par excellence. That this is not an impressionistic idea, is shown by the inhibition of rule (8) to applying to (23). As it is known, not infrequently palatal sounds convey some sort of affectionate meaning.
In the present case, I would like to prove that the
blockage of (8) in forms such as those of (23) is due to the fact that the palatal sibilant not only has a distinctive function (as opposed to /s/) but also an expressive one, namely affectivity.
It is not difficult to
realize that the items listed in (23) share the common feature
I
108 [+ Affectivity], Thus, accordingly, the rule does not apply, for otherwise that would imply a partial change in the meaning of the items, since the palatal nature of /s/ is part of the semantic content of them.
It is
clear then that the apparent counterexamples of (21a) are semantically conditioned, and therefore predictable. To prove that, in fact, /s/ conveys affectivity, I would like to offer some evidence:
1. When adult speakers talk with children, trying to
imitate them, they would use invariably /c/ and /s/ instead of /c/ and [s], respectively (incidentally, note that children find very difficult the articulation of these latter two sounds).
2. When a Huanca-speaker
listens to another Quechua speaker who only has /!/ (e.g. a Tarma-Quechua speaker), he would say that his accent is "childish."
3.
It is usually
the case that female story-tellers, especially when they narrate children tales, use only [s] instead of [s] in the dialogs. 4. Nicknames, based on current words that after rule (8) have [s] instead of [s], only have the palatal non-retroflex sibilant; e.g. words such as /akas/ 'guinea pig', /aksu/ 'potato', /kasa/ 'thorn', etc., are retained with [s] in the following nicknames:
[akas dimitru] 'Demetrio, the guinea pig',
[aksu kilku] 'Gregory, the potato', [kasa hisu:sa] 'Jesusa, the thorn', etc.
5. Finally, note that the affectivity attributed to /s/ is also
shared by the palatals /l, n/; thus, as in most Quechua dialects, proper names of Spanish origin that carry/s, 1, n/ are pronounced with /s, 1, n/ when used as nicknames in Quechua.
Some examples: /aliku/ 'Alejandro',
/pull/ 'Apolinario', /iffu/ 'Inocente', /bini/ 'Benedicto', /santi/ 'Santiago', /usta/ 'Eustaquio', etc.
In the latter two cases, as we see,
we get more "opaque" cases for rule (8), but easily predicted as exceptions.
109 Thus, it seems to me that the affectionate meaning conveyed by I si (and in general of any palatal sound) is out of question at least in Quechua.
In relation to the apparent counterexamples of (23a), I think
we can safely conclude that they are satisfactorily accounted for by the affectionate nuance of /s/. The application of (8) would have turned the items less affectionate, crude. Thus, in the face of this, we must conclude by saying that rule (8), as stated, correctly predicts the facts discussed in this chapter.
110
FOOTNOTES FOR CHAPTER 5
A PQ */s/ (Isi in my notation), in addition to */s/ and */s/, has been proposed by Orr and Longacre in their article on "Quechumara" (that is, Quechua-Aimara). Cf. Orr and Longacre (1968). Such a postulation-and in general what they propose as the proto-phonemic system of Quechumara--is questionable in most respects. Their reconstruction fails almost entirely to consider QI data, and, what is more important, the authors do not consider at all two modern languages related to Aimara, namely Cauqui and Jacaro (spoken in the province of Yauyos, Lima). 2 Cf. Quesada (1976: 41-42). From the data the author presents, it is not entirely clear whether apical [s] is different substantially from the retroflex or not. 3 These are Huancavelica (especially in Moya and Vilca) to the South, and some districts in the province of Yauyos (Lima). The Huanca-Quechua areas where the process is almost absent are the microvalleys of Paca and Yanamarca (Jauja). 4 v Leaving aside the assistive -ysi. which is always realized with the palatal non-retroflex, since it is surrounded by the^two conditioning segments. Besides, this morpheme is lacking in the NH variety (Jauja), as well as in the Huaicha subdialect. 5
Cf. Kiparsky (1972: 208).
Since at the time of the Conquest, Spanish /s/ and /z/ had merged into ill, I assume that the target words had only the latter, which also was on its way to disappearing. Furthermore, there is evidence to the effect that Spanish apical [si was also assimilated as Huanca /s/, not as /s/, at least during the early period of the Conquest. Some examples are: [asnu] < asno 'donkey', [kanasta] < canasta 'basket', but [martis] < martes 'Tuesday', [birnis] < viernes 'Friday', etc. I have found some spots in NH where the vowel never drops (for example, in Huertas, Molinos and Yanamarca). Outside QI, the only QII dialects that lack that rule are San Martin (Peru) and Ecuadorian Quechua.
Ill CHAPTER 6 GLOTTAL ELISION
In the present chapter I shall be concerned with the phonetic realizations of the glottal stop /?/ among the YH varieties. As I have mentioned in Chapter 2, this phoneme is subject to different surface manifestations, going from its thorough preservation to its almost total loss.
The different treatments of HI
are responsible for the extreme
fragmentation of Yaqa-Huanca. As we shall see, the general trend is towards the complete elision of the glottal phoneme. This process is taking place gradually, depending on region, age and style variables. Therefore, a rule that is optional for a certain area becomes categorical (= obligatory) for another; by the same token, the speech of old speakers may have optionally a rule that deletes /*?/ in certain positions, which rule is however obligatory in the speech of the younger generations; furthermore, careful speech would tend to avoid the application of a rule that drops /*?/, whereas in a free conversation such a rule is obligatory, and so on.
In the present chapter, however, mention will be made of the
geographic distribution of the rules mainly.
A more detailed study of
this change in progress would have been more rewarding had a more sophisticated descriptive device been used such as that developed by Labov. In any case, the present description, despite its obvious shortcomings, provides a good illustration of a gradual spatial generalization of a rule. In general, as we shall see, the glottal stop phoneme has two different outcomes:
after a consonant it disappears without leaving any trace, and
before a consonant or pause it lengthens the preceding vowel. Aside from this, its elision produces side effects, such as vowel contraction and subsequent vowel lengthening.
112 6.1.
Rule Inception:
Morphological Conditioning
Among the different YH speaking communities we find only one that preserves HI
in all positions, except initially after pause, where it
disappears according to the historical rule presented in Chapter 2 (cf. Section 2.3). This community is Huacrapuquio (natively called /wa'la pukyu/ 'horn-shaped spring'), located on the Central-Southern section of the valley and bordering on QII varieties. Because of their preservation of HI,
the speakers are famous as people who speak
"groaningly" (/ama+stin/).
In fact, this "accent" is one of the most
noticeable features that catches the foreigner's attention. Thus, as far as the preservation of HI
goes, the Huacrapuqio variety is the most con-
servative. The same could be said about the community of Carhuacallanga, on the Southwestern mountainous region; however, here, as we shall see (cf. Section 6.4.1), there occur some side-effects on the nasal /n/ in front of HI
that force us to consider this variety a separate dialect.
Thus, leaving aside Carhuacallanga, Huacrapuquio stands alone with HI everywhere. For ease of reference let us call this variety "Dialect A." It should be noted, however, that young speakers, especially in fast speech, tend to suppress HI
in the environments to be described shortly.
In the communities neighbouring Huacrapuquio, especially the districts of Viques, Sapallanga, and Huayucachi, conveniently called together, "Dialect B," we find the situation illustrated by the following items: (1)
[atu]
*fox'
[yula]
'white'
[cu'lu+ka]
'the corn'
[mama+yki+pa]
'for your mother'
[suftrf+la]
'still six'
113 These forms are derived from the following underlying representations (which, in this case, correspond closely to what we find in A ) : (la)
atu? yula? cu'lu+ka* mama+yki+pa? su'ta+la?
In view of this situation we may perhaps formulate a rule by which HI
is lost before pause.
However, the incorrectness of this analysis is
shown by the fact that once we attach suffixes to the items of (1) we get the following results: (2)
[atu":+mi]
'(it is) a fox'
[yula:+ka]
'the white one'
[cu'lu+ka^+ta]
'to the corn (ace.)'
[mama+yki+pa:+waT]]
'for your mother also'
[su?ta+la:+si]
'(it is said) it is still six'
that is, the roots and suffixes that used to carry HI with a long vowel instead.
in (la) appear now
In articulatory terms, this shows that the
timing employed in the articulation of the glottal stop is added to the preceding vowel, a natural compensatory process found in a number of languages.
In face of this, a rule of glottal loss before word boundary,
such as the data in (1) suggest has no basis, since otherwise we would need an extra rule of vowel lengthening to account for the cases of (2), a rule whose structural description however would be hard to determine; in other words, a perfectly predictable situation would become desperately unpredictable.
For this reason, the more natural solution is to postulate
an intermediate stage in the derivation of (1) from (la), something like:
114 (3)
[&u:] [yula:] [cu'Lu+ka: ] [mama+yki+pa: ] [su?ta+la: ]
which would be the output of the general rule of Vowel Lengthening (VL), statable as (4): (4) Vowel Lengthening IV?I - [v] and to derive (1) from the intermediate stage (3), we must postulate another rule, Vowel Shortening (VS): (5) Vowel Shortening [v] - [v] /
##
that is, a long vowel becomes short before pause. Note, however, that rule (5) applies to the output of the VL rule only (and also, as we will see in Chapter 7, to the output of an idiosyncratic rule). It does not apply therefore to non-derived underlying long vowels, as in: (6) puli+:
'I walk'
miku+:
'I eat'
lika+:
'I see'
wasi+:
'my house'
cuku+:
'my hat'
uma+:
'my head'
for if it applied, then VS would be eliminating a morphological distinction. However, as it usually happens, phonetic processes tend not to destroy grammatical signals, thus preventing uncomfortable homonyms (but cf. our remarks in Section 6.4.2 below).
Anyhow, the fact is that VS has to "know"
115 which vowels are derived and which ones are not; that is to say that, in order to apply correctly, the VS rule has to be sensible to the derivational history of its input. From the facts observed so far, it must be clear that the VL rule is too general since, as formulated, it is a context free rule; yet, as we see in (1) and (3), it does not apply to the roots [cu'lu] 'corn' and [su'ta] 'six'. The same is true for the following examples: (7)
[wa^la^:lu+T]]
'it hooked (to someone)'
[pa?a+:+cul]]
'let him/her wash (it)'
that are derived, respectively, from these following basic forms: (7a) wa'la+'lu+n pa'a+'lu+cun As we can see, whereas VL applies in the vicinity of a morpheme boundary, it does not apply morpheme internally; thus, forms such as [wa'la-] 'to hook' and [pa"?a-] 'to wash' are immune to the process.
In view of this
situation, the only way to restrict the application of rule (4) is by stating its proper environment, namely before pause and in contact with a morpheme boundary.
Such a formulation would be as follows:
(8) Vowel Lengthening
rr\
IW - [v] /
Rule (8) is then, if correctly interpreted, an example of a phonetic rule morphologically conditioned.
Even if we say that it is not unusual
for a phonetic rule to be implemented by word boundaries, the fact is that our VL rule does not affect the roots. This means that, to derive correctly (7), the rule has to be restricted to the environment of a morpheme boundary.
As in Chapter 5, we find again a process that originates
116
across morpheme boundary first.
Since phonetic rules morphologically
conditioned are almost non-existent in the literature, especially when they originate as such, the case we are discussing is certainly a very interesting one. As we shall see in the next chapter, changes of this sort seem to be universal among the Quechua varieties. So far I have presented cases of glottal elision in syllable-final position.
In syllable-initial position after a consonant we find, within
the same dialect B, the following situation. (9) mal9a+19a hul'u+n'a
Forms such as:
'he/she carried (it)' 'he/she will remove (it)'
are realized, respectively, as (9a): (9a)
[mal'a'+la] [hul'u+'na]
that is, the HI
of the suffixes is dropped, especially in fast speech.
Nevertheless, note that within the roots HI remains unaffected. Thus, again, the glottal elision process takes place only among the suffixes. The examples are limited in this case, since in this variety of Quechua there are only two suffixes of the form -(C)9V; -the past tense -19a, and the third-person-future -n£a.
The topic marker -9a, a unique suffix of
the -9y shape, has a special treatment (cf. Section 6.4.3 below). Thus, leaving aside this suffix, we can formulate the process involved in the derivation of (9a) from (9), as follows: (10) Glottal Elision
HI - [0] / [+ suffix] As can be seen in (9a), one of the consequences of the application of GE is that it induces automatically a syllabic restructuring of the consonants:
from final they come to be in initial position.
In this
117
way, we get an [T]] in syllable-initial position, an environment completely prohibited by the sequence redundancy rules seen in Chapter 3 (for further discussion about the velar nasal, see Section 6.4.1 below). Let me mention in passing that these rules—VL and G E — , almost obligatory in normal speech, are systematically avoided by the older generations, especially when trying to speak carefully.
6.2.
Syntactic Interlude
So far we have seen that the VS rule applies correctly, in the environment specified.
However, from the examples listed below it should be
apparent that (8) is somehow incorrect, since here in spite of the appropriate input the rule does not apply.
Thus, consider the following
cases: (11)
[hampi+: nuna]
'healer (lit. 'one who heals')'
[atu: bartulu]
'Bartholome, the Fox'
[wasi+yu: nuna]
'house owner (lit. 'person with house')'
[musu: yula: wasi]
'new white house'
[pusa: cuT]ka]
'eighty (lit. 'eight tens')'
These forms come, through the application of the VL rule, from underlying representation such as those of (11a): (11a) hanpi+9 nuna atu9 bartulu wasi+yu9 nuna musu9 yula9 wasi pusa9 cunka Now, as we know, VS applies obligatorily to the output of VL. done, however, we get incorrect forms such as:
If this Is
118 (lib) *[hampi nuna] *[a*tu bartulu] *[wasi+yu nuna] *[musu yula wjfsi] *[pusa cuTlka] Thus the question arises as to what the conditioning environment is for the preservation of the vowel length In (11). We may perhaps say that VS operates only before an actual (physical) pause, so that (11), showing sequences of words, blocks the rule. This is a reasonable explanation, since there are languages where such a conditioning factor is not uncommon. But in the present case that explanation simply collapses in view of the following set of examples: (12)
[wala's+ka l&ku+'n]
'the boy left'
[li+sa ni+ya+:]
'I'm planning to go'
[mama+yki+pa apa+y]
'take (it) for your mother!'
which come from the underlying forms listed in (12a): (12a) walas+ka*? li+ku+n li+sa9 ni+yka:+: mama+yki+pa9 apa+y As we can see, after the VL rule, VS operates giving (12) as output. Notice furthermore that we cannot say that in (11) the vowel lengthening is the only overt marker of a morphological distinction, and therefore preserved. Whereas this can be true for the first item, where [hampi+:] comes from /hanpi+9/ (that is the root hanpi- 'to heal' plus the agentive "2)»
in
tne o t n e r
cases the lengthening does not signal anything, or it
is part of a morpheme at best.
119 Other cases where VS is blocked are illustrated by the following: (13)
[tusu+: li+ya/+T|ki]
'you are going to dance*
[tak^+: ka*+la]
'he/she used to sing'
[mana+la: sa+mu+n+cu]
'he/she doesn't come yet'
These forms come, respectively, from: (13a)
tusu+9 li+yka:+nki taki+9 ka+19a mana+la9 sa+mu+n+cu
Here again, one cannot say that the vowel lengthening in the first two items of (13) is not shortened because it is marking the purposive and the habitual morphemes, respectively (both, underlyingly -9, and therefore homophonous), since in the last item the lengthening does not signal anything, being only a part of the stative suffix -la?. Compare this latter situation to that of the items of (12) where the lengthening is also a part of the suffixes involved, but nevertheless it is affected by the VS rule. Yet more compelling evidence showing that the vowel length in the first item of (13) is not preserved for morphological reasons is the fact that, if we move the goal /tusu+9/ to the right, we get optionally [li+ya'+T|ki tusu], that is without vowel lengthening. If it is the case that VS is blocked because of the avoidance of homonymy, then one should expect at least the last item of (13) to behave like the items of (12). Moreover, as we have seen, all the items of (11), except the first, maintain their vowel lengthening apparently for euphonic reasons only, since it does not reflect any categorical distinction whatsoever (at least synchronically). As we saw, however, when vowel lengthening is the only overt morphological signal, the VS rule is blocked. This is true also for the following situation.
In some varieties of YH,
120 the determiner -ka9 is realized sometimes as [-a:] or even as a vowel lengthening only, after the vowel becomes eliminated in favor of the stem vowel (for all this, see Chapter 7, Section 7.1.2). Now, we have seen already that [-ka:] is subject to the VS rule, as we can see in (12). However, among the varieties where -ka9 has drastically changed we have the following situation: (14)
[wamla+: sa+mu+na]
'the girl had come1
[walpa+: ciTlka+ku+t]]
'the hen got lost*
both forms coming from underlying: (14a) wamla+ka9 sa+mu+na9 walpa+ka9 cinka+ku+n Thus, we see that VS is blocked again. The only reason I can think of for this fact is avoidance of homonymy.
Here, if we eliminate the
lengthening we would be suppressing a whole morpheme, namely the determiner. The same is not true, however, for the items presented in (11) and (13), where a putative application of VS certainly could not create any conflicting situation.
This is further confirmed by the fact that VS
applies to all of them in final position before actual pause; thus, in this environment, we get forms such as: (15)
[hampi]
'the one who heals'
[a'tu]
'fox*
[wasif+yu] 'house owner' [mifsu]
'new'
[yifla]
'white*
[pusa]
'eight'
121 and also, considering the examples of (13), if we ask 'Where are you going?', the answer would be [tusu] 'to dance'; the same is true for a question such as 'Doesn't he/she come yet?*, since the answer is simply [mana+la].
As we see, in both cases the VS rule applies (the second
example of (13) is impossible to test, but this is because of the idiosyncratic property of the plusquamperfect constructions). This is not the case, however, with the lengthening that signals morphological distinctions, such as the first-person marker and the determiner.
The
latter fact is especially important because the lengthening is a compensatory phenomenon.
In these cases VS does not apply even before an
actual phonetic pause. Once more, in this case we have a clear illustration of semantic blockage. If the above conclusion is correct, then we must try to explain the different behavior of the VS rule in (11) and (13) as opposed to that of (12).
Our next step will be directed to this attempt.
For that purpose, I would like to claim that the blockage of VS In cases such as (11) and (13) is due to non-phonetically conditioned factors. This is so because, as far as I can see, there is no reason on purely phonetic terms why VS should not apply among those items as it does in (12). Therefore, I would say, the explanation lies in the syntax of those expressions.
Let us then reconsider the examples of (11) and (12). On
close inspection, it is not difficult to realize that the items of (11) form a noun phrase composed by a modifier and head.
Thus [hampi+:] is an
adjective that modifies nuna; in the same way, [atu:] modifies bartulu. and so on.
Note also that the preadjectival [musu:] modifies the phrasal
construction [yula*: wa'si] 'white house', where [yula:] itself modifies wasi.
Now, if we consider the examples of (13), we see that in each case
122 we are confronted with a verb phrase, where the preceding element constitutes a complement, such as [tusu+:] and [taki+:], or an adverbial modifier such as [mana+la:]. Now, if we take into consideration examples such as those of (12a), we see that in both cases each chunk constitutes a phrase per se; thus /walas+ka9/ is a NP, and /li+ku+n/ a VP; in the same way, /li+sa9/ is a sentential complement with respect to the matrix /ni+yka:+:/, and finally /mama+yki+pa9/ is a purposive clause that functions as a complement of the imperative form /apa+y/.
Notice that in these as in the other cases
we are referring to the surface configuration of the sentences and not to their underlying representations, for if they were taken as a reference for differentiating the first group of sentences as opposed to the latter, then a case may be made to the effect that even the sentences of (11) and (13) do not contain single phrases, but some of these phrases are sentential in origin also. From the above it should be clear that the notion of phrase that I am trying to characterize in this context is a surface-structural configuration.
With this in mind, we can now say that VS operates differently
within a phrase and outside a phrase.
In the first case the rule is
blocked, whereas in the latter it operates. This indirectly shows that the cohesion among the elements that form a phrase is stronger than the cohesion among different phrases within the same sentence.
In a way, the
cohesiveness within the elements of the same constituent resembles that which exists among roots and suffixes, as seen in (12), for example. Granted that the observations made above are correct, then our VS rule must be restricted in such a way that its application be blocked within a phrase. Whereas at present it is difficult to imagine a
123 non-ad hoc formalization of such a restriction, we can perhaps paraphrase the rule as follows: (16) "Shorten long vowels before phrase boundary" According to this restriction on the VS rule, a sentence such as: (17)
lasa9 lumi+ka9 palpu+91u+na9
'The heavy stone had just fallen down*.
will end up, after VL and VS rules, as: (17a)
[rasa*: lumi+ka palpu+:+na]
where we see that VS is blocked within the NP [fasa": luml+ka], but at the same time applies shortening the long vowel of the determiner -ka9 (through an intermediate stage [-ka:]), as well as that of the narrative -na9 (which after VL becomes [-na:]); in the latter two cases the vowel lengthening is outside a phrase. As I have said before, phonetically there is no reason for the long vowel of [lasa':] not to become shortened by VS, or alternatively, there is no reason why the long vowel of the determiner -ka9 after VL has applied does not remain as such but rather shortens. The only reasonable answer which I can come by is that VS is sensitive to the syntactic notion of phrase.
If this analysis turns out to be true, then here we
have an illustration of the interaction between syntax and phonology.
As
we saw, VS does not apply blindly irrespective of the syntactic configuration of an utterance, but rather selects its domain of application by distinguishing different types of structures. Cases such as this, sometimes referred to as "external sandhi" phenomena, have been found in a variety of languages, of which Chi-Mwi:ni (cf. Kisseberth, 1973) and French (cf. Selkirk, 1974) have received recent treatment.
The problem
124 remains however as to what constitutes a phrase, since this notion is crucial for the operation of VS in the case discussed here.
6.3.
Rule Propagation:
Phonetic Conditioning
In Section 6.1 we have seen the operation of VL in a morphologically conditioned environment.
In this section, we shall see how this rule gets
generalized to the roots, irrespective or morpheme boundaries.
In other
words, we shall see the propagation of VL from a morphologically determined environment to a phonetically conditioned context. As we know, this type of rule generalization is unusual, for the reverse is what we normally find in most languages. The case presented here, if correctly described, could be taken as a good illustration that the other direction is also plausible.
In what follows, I shall try to delimit the gradual propagation
of VL and also that of GE.
6.3.1.
HI Before a Consonant
In places such as Huamancaca Grande and Santo Domingo de Acobamba, one can notice that VL not only applies in the environment stated in (8), but also to the roots. Thus, within this area, called henceforth "Dialect C," we find the following phonetic realizations: (18)
[cif:Iu]
'corn'
[tu:pa-]
'to form a bunch'
[wa^ta-]
'to love'
[we*: la] [pd*:ca]
'horn' 'waterfall'
These forms are derived, respectively, from the following lexical representations:
125
(18a) cu9lu tu9pawa9tawa91a pa9ca forms that, in careful speech, can be recovered, especially among older people.
Since, as we see, VL has become generalized from a morphologically
conditioned to a phonetically conditioned environment (now it applies in syllable-final position), to account for the data in (18) we can formulate this second stage of the process, as follows: (19) Vowel Lengthening (2nd. stage) /V9/ - [v] / —
$
As it should be noted, this rule presupposes the first one (that is, (8)), since it is a more generalized version of it. Notice that even when HI disappears phonetically, the opposition of (18) as against (20) is carefully preserved, although now the "contrast" would seem to be carried by the short/long vowel distinction. (20) culu-
Thus, compare (18) with the following:
'to melt'
tupa-
'to run into'
wata
'year'
wala-
'to lighten'
paca
'earth'
So far, we have seen one instance of a rule becoming generalized from a derived to a non-derived form.
Now, recall that within this
dialect the GE rule (10) is still morphologically conditioned, since it applies to suffixes only, leaving the roots immune.
However, in the
speech of younger generations I was able to detect an ongoing
126
generalization of the rule.
This rule, as we shall see in the next
section, becomes obligatory in normal speech.
6.3.2.
HI After a Consonant
In places such as the Northern communities of Sapallanga, Ocopilla (Huancayo), and the districts of Comas and Cochas (Concepcion), the GE rule has become generalized to phonetic environments. Let us call this area "Dialect D." Within D, lexical representations such as: (21) puc9u
'sour'
sit9a
(a variety of edible plant)
was9i-
'to wean*
mal9a-
*to carry in one's arms'
pal9a-
'to bifurcate'
sun9un
'heart'
are realized, respectively, as in (21a): (21a)
[pucu] [s&a] [wasi-] [mala-] [ptfa-] [suT)uT]]
that is, HI drops after a consonant.
To account for this fact, we must
formulate a second stage in the process of glottal elision. would be then: (22) Glottal Elision (2nd. stage)
IV - [0] / C
The rule
127 This rule is now a phonetically conditioned one. As in the preceding case, the rule becomes generalized and therefore simpler.
Note, however,
that the effects caused by the application of GE (22) are in no way comparable to what we have found in the above section, since here, after the glottal consonant drops, surface neutralization takes place. Thus (21a) has become homophonous with: (23) pucu
'left over'
sita-
'to get wet'
wasi
'house'
malay
'bottom millstone*
pala-
'to pick up'
It is not surprising however that in careful speech the distinction is still preserved, especially among older generations.
It has been observed
also that even though a speaker may not be able to actually pronounce the glottal stop consonant, he would however easily recognize that [puc9u] can only be 'sour' and not 'left over' (cf. Chapter 8, Section 8.4, for a more detailed discussion of this topic).
6.3.3.
HI in Heterorganlc Intervocalic Position
As it should be realized, the only environment where HI
is main-
tained is in intervocalic position; elsewhere, as we saw, it is gradually disappearing.
The next step in the propagation of the elision of HI is
precisely that intervocalic context.
In an area located on the Southwest
side of the valley, around Chongos Bajo and Chongos Alto (the latter in the mountainous region), we find that, whereas GE (22) is categorical (and no traces of postconsonantic HI
is found on the surface, especially
in careless speech), the glottal consonant drops in intervocalic position,
128 but only when the vowels are of different quality.
Let us call this area
"Dialect E." Thus, here, roots such as the following: (24) u9i
'grey*
a9 u -
'to rub'
tu9a-
'to spit'
hi9a-
'to climb up'
wi9aw
'waist'
are realized, respectively, as the following alternating forms: (24a)
[0*1]
~ [dwi]
[au-]
~ [awu-]
[tua-] ~ [tuwa-] [hia-] ~ [hiya-] [wiaw] ~ [wiyaw] In careful speech (actually in word-by-word elicitation), I have noticed that the speaker can recover the glottal stop that in normal pronunciation is absent.
Since we are interested in the latter, to account for the
items of the first column in (24a) we must formulate the new stage in the propagation of GE, as follows: (25) Glottal Elision (3rd. stage)
IV - [0] I v±
Vj
As shown in (24a), whereas the output of GE is [0] in the first column, in the second we find the semivowels /w/ and /y/ instead of underlying HI.
What is intriguing here is of course the appearance of
these semivowels.
One may ask if there is a natural link, acoustically
or articulatorily determined, between HI
and the semivowels. The answer
seems to be no, in face of the overwhelming evidences showing that the
129 semivowels in (24a) are introduced by a low level rule, Semivowel Insertion (SI). To see more clearly the role played by SI, let us consider again the items of the first column in (24), repeated here as (26): (26) [u'i] [au-] [tifa-] [hia-] [wiaw] As I have said in Chapter 3 (cf. Section 3.3), the syllabic structure of Quechua prohibits any sequence of vowels like those illustrated by (26). The application of GE, however, produces an output that blatantly violates this morpheme structure condition.
Thus, to rule out this kind of
violation of a surface phonetic constraint, a low level rule is called in which inserts a semivowel within the same space left by the glottal stop, thus eliminating the anomalous sequence of two vowels. The alternations in (24a), however, show that, at least transitorily, this restriction can be violated on the surface. show also that SI is optional.
But on the other hand, they
In any case, as I have advanced in
Chapter 3, this rule applies not only to eliminate outputs such as those of (26), but also applies to Spanish borrowings with vowel sequences. Some examples are: (27)
tlya
< Sp. tj£a
'aunt'
liyun
< Sp. ledn
'lion'
luwis
< Sp. LuLs
'Louis'
bawul
< Sp. baul
'trunk'
130 In both cases, it seems clear that the basic function of SI is to implement the restriction on vowel sequences. Our next step is then to formulate the SI rule. To this effect, let me mention that there are no sequences of the type *a9i in YH (cf., on the other hand, with NH cahi < *caql 'barley soup').
Sequences of the
type 19u are also very rare; I have found only one lexical item containing it:
sj9uta 'windy rain', but, on the other hand, Spanish loanwords
carrying vowel sequences such as eo and diphthongs like io yield lu sequences that provide input to the SI rule. Taking into account these observations, we may formulate tentatively SI as follows: (28) Semivowel Insertion
(0)
-}
w
' i-Z—i — v
W / [«*„.] — v As we can see, the vowel that actually determines the nature of the semivowel is the preceding one.
If this is so, then both processes can
be collapsed into one, namely (28a): (28a) Semivowel Insertion
T-Sylll
»- -- ' fed I oGrave 1
U
-»
Thus, to sum up: After the application of GE on items such as (24), yielding (26)—a violation of the morpheme structure rules—SI optionally comes into play, shaping them up to give an output in accordance with the syllabic canon of Quechua. rules apply in that order.
Therefore, it is clear that the GE and SI
131 6.3.4.
HI
in Homorganic Intervocalic Position
In Northern Yaqa-Huanca (beginning in the districts of Iscos and Chupaca on the right bank of the river, and in Hualhuas on the left bank), including the whole Huaicha-Huanca variety, we find what appears to be the ultimate stage of the glottal elision process, this time dropping HI in intervocalic position in general, irrespective of the restriction we have found in dialect E.
Thus, in this variety, called "Dialect F,"
lexical items such as: (29) si9i-
'to strangulate'
wi9i
'tear'
mu9u-
'to slaughter'
su9u-
'to suck'
ma9a-
'to beat*
la9a-
'to paste'
are realized, respectively, as follows: (29a)
[si*:-] [wi:] [mif:-] [s\f:-] [ma*:-] [la:-]
In face of this, we may perhaps say that it is the byproduct of VL. However, that this is not so is evident from the fact that in (29) the glottal consonant, according to the syllabic structure of Quechua (and I would say, universally), is in syllable-initial position, therefore VL is not applicable. Moreover, if that were the case, then we would have expected from (24) an output similar to that of (29a), that is with vowel
132 lengthening.
For these reasons, the most reasonable alternative is to say
that in the passage from (29) to (29a) there exists an intermediate stage such as the following: (29b)
[sii-]
[muu-] [su'u-] [maa-] [Laa-] forms which indeed appear in careful speech, especially in those of the older people.
These forms are then the output of the application of a
more simplified version of GE, that is: (30) Glottal Elision (4th. stage) HI
- [0] / V
V
To derive (29a) from (29b) we must assume the operation of another rule, "Vowel Contraction" (VC), statable as follows: (31)
Vowel Contraction
/V
iV - w
that is, two vowels of the same quality contract in a single long vowel. From what has been observed, it is clear that, to derive (29a) from (29), we must apply GE (30) and VC in that order. As we can note, the drop of HI
in intervocalic position, provided the surrounding vowels are equal,
gives us more instances of surface long vowels. As we shall see, the new rule, whose role is very similar to that of SI, is independently motivated and has a wider application.
Besides, in face of the occurrence of (29b),
we can say that also VC is optional.
133 6.4.
Collateral Effects
In this section we shall see some collateral effects produced by the loss of the glottal stop. Within this context, the sequence of nasal plus glottal stop merits special treatment.
Furthermore, as I have
mentioned in Section 6.1, there is in Huanca-Quechua one suffix that presents this sequence:
the third-person-future -n9a, realized as
[-119a] in dialect A, and as [-T]a] in dialects B through E. The sideeffect then also operates on this morpheme.
These phenomena only occur
within the most innovative area, namely F.
6,4.1.
Preglottalic Nasal Absorption
In a relatively wide area, which I call dialect F, the following underlying forms: (32) pan9a
'husk*
tan9a-
' t o push'
un9ul
'knee'
sin9a
'nose'
un9a-
' t o forget'
sa+mu+n9a
'he/she will come*
correspond respectively to the following alternating forms: (32a)
[p«fa] ~ [pa':] [ta'a-] ~ [ t a V ] [iful] ~ [if:l] [ s i a ] ~ [siya] [ua-] ~ [uwa-] [sa+md+a] ~ [sa+mu+wa]
134 To account for the disappearance of /n/ in these forms, we may think of three alternatives: before HI
(a) /n/ dropped after HI
did; (b) /n/ dropped
did; and (c) they both dropped simultaneously.
The first alternative is in accordance with the facts observed in Section 6.1 (cf. stage one of GE) and Section 6.3 (cf. second stage of GE), since after the application of glottal elision in postconsonantic position, we get forms such as [stfl|uT|] 'heart' and [hulii+T|a] 'he/she will remove (it)' from /sun'un/ and /hul9u+n9a/, respectively.
In other words,
there are dialects that illustrate the occurrence of the velar nasal after the glottal consonant dropped.
As was noted, the appearance of
syllable-initial [l]] violates one of the sequential constraints of Quechua. This would then be a powerful reason why that segment is finally elided. Compare this with a form such as /cam?a-/ 'to smash'; after GE we get [cama-], but here /m/ does not drop since this segment can occur in syllable-initial position as well.
Incidentally, a rather nice example
that shows the struggle in maintaining the nasal consonant is given by the semilexicalized form umul from /un9ul/ 'knee', found in San Jerdmino (dialect F ) .
Here the nasal was preserved by being interpreted as /m/.
At this point we may ask why the velar nasal, being an allophone of Id,
was not reinterpreted as an alveolar. One reason for that could be
perhaps the fact that a reinterpretation in that direction would have produced homonyms. Thus, compare the intermediate forms of (33): (33)
[pa>]
'husk'
[ta*7|a]
'to push'
[ifT]a+y]
'to forget'
[sa+mu+Tja]
'he/she will come'
with the following items:
135 (34) pana
'scattered'
tana
'entangled'
unay
'long time ago'
sa+mu+na
* arrival'
The last item in (33) is especially important because the "alveolarization" of the velar nasal in the third-person-future would have derived in a sequence similar to that of the nominalizer -na in (34). However, as is apparent elsewhere, the process of homonym avoidance does not always apply. Therefore, the evaporation of [T|] is better explained in terms of sequential segment constraints. If the above is true, then our next step is to formulate the process involved, namely "Nasal Evanescence" (NE). To this effect, it is important to note that this rule applies only to velar nasal, since we do not get a form such as *[cla] from cina 'female'. The rule then can be stated as follows: (35) Nasal Evanescence
[ £ 3 - EH / - » This rule applies after GE, yielding a vocalic sequence that, again, violates one of the morpheme structure rules of Quechua. As we know, this situation could be optionally remedied by applying the low-level rules of the preceding section, that is SI and VC. Thus, to change (32) to (32a) we have to apply the following rules, in that order: 1.
GE (22)
2.
NE (35) SI (28) 1
3.
vc
(3i)J (Optional)
136 • where the last two rules apply optionally to different inputs. To illustrate the operation of these rules, let us derive the items of (32). For this, we have to assume that the nasal assimilation rule mentioned in Chapter 3 (cf. Section 3.1) has a prior application, providing the right input. Thus, accordingly, we would have: (36) pan9a
tan9a-
un9ul
sln9a
un9a-
sa+mu+n9a
Underlying Forms
paf|9a
taT)9a
uf]9ul
siTj'a
uTj9a
sa+mu+fpa
Nasal Assimilation
paT|a
taT]a
ufjul
siT|a
uT(a
sa+mu+^a
Glottal Elision
paa
taa
uul
sia
ua
sa+mu+a
Nasal Evanescence
siya
uwa _
sa+mu+wa
Si) y c j (Optional)
pa:
ta:-
u:l
As we see, the output is in accordance with (32a), with the exception of the item in the last column which needs, in order to be fully correct, the application of the /If
retroflexion rule. Notice that the optional
application of VC produces long vowels in checked syllables, as In [u:l] in clear violation of the surface phonetic constraint prohibiting them (cf. Chapter 3, Section 3.3).
Other such forms are [na:l] 'in vain' from
/ffan9al/, [su:T]] 'heart' from /sun9un/, etc.
This violation, however,
seems to be corrected in fast speech, long vowels becoming short.
As we
have seen elsewhere, fast speech usually indicates the building up of new processes. Thus far I have tried to show that the first alternative suggested, namely that [f|] dropped after GE, is well-attested.
This is further
confirmed by the fact that within the bordering areas between dialects E and F there is apparently free variation between forms such as [taf]a-]
137 •to push', [na*T]al] 'in vain' from /tan9a-/ and /na9al/, respectively, and [uwa-] 'to forget', [piya+ku-] 'to be ashamed', etc. from /un9a-/ and /pin9a+ku-/, respectively.
What this demonstrates is that, at least
transitionally, the velar nasal can survive after GE takes place. Note, incidentally, that one of the consequences of the drop of n9 is the appearance of surface homonyms. Thus in F, a word such as [pa:] can be either 'to wash' (from /pa9a-/) or 'husk' (from /pan9a/); in the same way, [uwa] can mean either 'to forget' (from /un9a-/) or a variety of sweet potato (from /u9a/).
In either case, however, the surface leveling does
not seem to cause serious problems. According to the second alternative I mentioned, nasal absorption took place before GE.
The evidence to the contrary we have just seen
would have been enough to dispose of this second alternative, had it not been for the fact that within a tiny area located in the Southwest mountainous region, specifically in the district of Carhuacallanga, we find, corresponding to the forms listed in (32), these surface manifestations: (37)
[pa'9a] [ta9a-] [u9ul] [s!9a] [u'9a-] [sa+mu+9a]
that is, a clear indication that the nasal was the first to go. As I have said in Section 6.1, had it not been for this fact, this variety would stand at the same conservative level of dialect A (that is, Huacrapuquio), in reference to the preservation of the glottal stop.
138 This special treatment of the preglottalic nasal, however, forces us to consider Carhuacallanga as a separate dialect, surrounded by dialect E. The situation illustrated by (37) indicates that the second alternative is a fact.
It is true that the items of (37) alternate some-
times with their corresponding forms showing n_9, but this can be due to borrowing from the surrounding varieties.
If this is true, then to
account for (37) we have to postulate the following rule: (38) Nasal Absorption
\zi] -OT/ - /