Printer Friendly

Coarticulations and coronals in Malayalam.

Recently, in this journal, Murray Emeneau (1995) significantly clarified the phonological conditions required for palatalization in Tamil-Malayalam. Generally, initial k [greater than] c [tC] before a front vowel. This change was blocked by a retroflex consonant immediately following the front vowel in Tamil-Malayalam, but not in Telugu where the same change occurred independently and somewhat later.(1) This change was completed by the time of the earliest records in Old Tamil, about two thousand years ago, and has not operated in the historic philological record. Neighboring Kannada did not palatalize in this environment; loan words back and forth have produced numerous counter-examples. Emeneau demonstrated that this palatalization could also be blocked by a following alveolar in Tamil-Malayalam, but not in Telugu. In discussing problem cases, several instances occurred where Malayalam did not show the expected result and often had r/r doublets. While he was certain that shifts between alveolar /r/ and dental /r/ provided the explanation, Emeneau was unable to provide an underlying mechanism. On reading this, I realized that I had the explanation from my own research, which I had never previously considered important enough to elaborate on.

The phonology of Malayalam can only be described as a tour de force. It has successfully combined a very conservative Dravidian six-stop phonology with a conservative Sanskrit phonology; adding voicing, aspiration, and a complete set of Sanskritic sibilants, at least in educated styles. It has a near world-maximum contrast in nasals (m, n, n, n, n, n), six contrasting stop positions in deep phonology (labial, dental, alveolar, retroflex, palatoalveolar, and velar), to which two more are added in surface phonology (retroflex affricates and palatals). However, a critical problem, and the one relevant here, concerns the two contrasting taps, "dental" /r/ and alveolar /r/.(2) In terms of surface phonemics, they clearly contrast initially and intervocalically, while the contrast is neutralized finally and in consonant clusters. In deep phonology both taps exist as liquids, although at this level /r/ comes only from loanwords: e.g., rotti 'flat bread' from Urdu. The deep alveolar single stop joins /r/ when intervocalic. In terms of phonetics, three clear entities exist which are reflected in the script: a single tap r, a single tap r, and single (or double) tap r, which is the pronunciation used in the final position. The contrast is an active one and very important for Malayalam's over thirty-four million speakers, but the precise phonetic nature of the contrast was unclear to me at the time.

I trained as a phonologist in graduate school. Although my dissertation concerned other aspects of Malayalam, during my field work in Kerala I set out to resolve this issue. While the Malayalam I had learned in the United States equipped me with an acceptable contrast in slow speech, I still was not sure what that contrast really was. From my experience with Tamil and the instruction in Malayalam, I had a good framework on which to begin. I had learned that Malayalam's dental /n/ is strongly velarized. On the basis of that distinctive articulation, I knew that Malayalam had at least some coarticulations, which had not been considered usual in South India.(3) The alveolar tap /r/ seemed to be just that, a single quick tap on the alveolar ridge [r]. The final tap /r/ was a single (or double) tap on the alveolar ridge [r]. The more common /r/ was a single tap, slightly fronted on the alveolar ridge [r[less than]]. I was fairly certain that neither point of articulation nor manner of articulation was the contrastive element here. In slow speech both taps could be double. Careful examination revealed that /r/ was palatalized with no off-glide. Since the dental /n/ was already known to be velarized, and the retroflexed (or retracted) series could readily be analyzed as pharyngealized, Malayalam seemed to have coarticulation added as a feature to some of its phonemes. In these, the tongue body took vowel-like positions: i-like for palatalized, w-like (with no lip rounding) for velarized, and a-like for pharyngealized. The alveolar /r/ seemed pharyngealized, the common /r/ was palatalized, and the final, noncontrastive /r/ was neither.

I was able to develop this view considerably, basing myself on a remark in the Keralapaniniyam, one of the best of the traditional grammars of Malayalam. The aryyalipi script of Malayalam, while very good at handling these contrasts (much better than the IPA), uses two graphemes to indicate an r in a consonant cluster - one for before and one for after the stop. However, it does not indicate which r (/r/ or /r/) was present in the cluster. To clarify this, Rajaraja Varma analyzed in detail which r was present with which stops (1895 [1969: 106]).(4) What had started as an inquiry into differentiating two taps was turning out to reveal the organizing principle for the surface phonology.

The consonants of Malayalam are organized into several series on the basis of the position of the root of the tongue. Velars [k, n] and the dental [n] are velarized [+high, +back]. All retroflexed consonants ([t, d, n, ???, ???, ???, ???] and alveolar [r] are pharyngealized [+low, +back]. All palatals [[k.sup.j], [n.sup.j], j], all palatoalveolars [tC, dz, n, ???], "dental" [r], alveolar [1], and all remaining voiced stops [b, d, g] are palatalized [+high, -back]. The remaining consonants [p, f, m, v, s], dental [t], and alveolar [n] are neutral -high, -back, -low]. The aspirated stops are not relevant to this discussion, and I have ignored them here.

The voiced stops, since they are palatalized, require further comment. In the original Dravidian system, voicing is natural, i.e., sonorants are voiced while obstruents are voiceless, with the further condition that single lax obstruents become sonorants (and hence voiced) intervocalically. Malayalam added its contrastive voicing from loans. This is distinct initially or when the stops are geminate, but the lax voiced stops tend to fall together with the lax voiceless ones in intervocalic position; this is the normal case in colloquial Tamil. Malayalam has added palatalization to keep the voiced series distinct in all positions. Likewise, the front sonorants /r/ and /l/ are palatalized to increase the contrast with their backed counterparts /r/ and /???/. This is clearly a surface rule since in deep phonology alveolar /1/ functions as [-high]. Also the velarization of dental /n/ maximizes the contrast with other nasals, particularly with alveolar /n/, which has no coarticulation.(5)

After returning from my field work, I prepared sound spectrograms of the contrasting taps using minimal pairs. The definitional case used ara 'half' and ara 'room'. On the spectograms the contrast was very clear.(6) The second formant for /r/ rose with a constant slope to the gap, after which it fell along the same steady slope. The second formant for /r/ was level throughout. Using a very narrow transcription showing vowel coloring, ara could be represented as ???, while ara was ???. In the contrasting pair, kiri ??? 'mongoose' vs. ??? ??? 'rag', the changes were reversed. The second formant of /r/ remained steady throughout, while ??? had a noticeable dip in its second formant before the gap and a partial recovery after. The gaps on the spectrograms, i.e., the taps, were identical in nature. This continually adjusting vowel (sloping second formant) can also be seen next to the palatalized consonants. The short /a/ is normally ???. However, if it is surrounded by palatalized consonants, the second formant does not have time to recover and gets pulled up at both ends to the position for short /e/ [[Epsilon]]. This only happens if there are no counterbalancing influences. The word balam 'strength' is normally pronounced [b[Epsilon]l[Epsilon]m], where the palatalized /b/ and /1/ jointly raise the second formant in the first vowel, while the second vowel follows the first in subphonemic vowel harmony. This is only a phonetic norm with dialect variation in the details and can be overridden by this same vowel harmony. The pronunciation of the poetic word abala 'woman' ([less than] 'not strong') is normally ???. Here the vowel harmony set by the initial ??? cancels out much of the shift seen in the formants. While of interest, this was only one aspect of a very complex phonology. Also, it seemed to be a fairly recent innovation in Malayalam.

Emeneau's paper jarred the relevance of these data back into a much larger perspective. It made it clear that this r/r contrast conditioned a historic change which took place more than two millennia ago and has not operated since. The palatalization of /r/ kept the front vowels high, allowing the k [greater than] c shift. On the other hand, alveolar /r/ slightly lowered the preceding vowel, blocking the change. Any variation between /r/ and /r/ is easily explainable given the coarticulations in rapid speech. While a personal observation, it seems to me to take more effort to pronounce the low /r/ in a high environment than to pronounce the high r/r in a low one.

The history of the dental and alveolar nasals in South Dravidian is complex. Tamil normally has them as allophonic variants (dental initially and in dental clusters, alveolar elsewhere) although Old Tamil(-Malayalam) has a handful of reported contrasts. In Malayalam, the original alveolar nasal-stop clusters (Ta. nr) became geminate dental nasals, although later assimilation rules recreated contrasting alveolar geminate nasals from the final alveolar /n/ plus dental or retroflex stops. These changes effectively wiped out any information Malayalain might have shed on the initial situation. The philological record as reported by Emeneau (1995: 407) is ambiguous showing one case with the blocking of palatalization (DEDR 2021), one case without (DEDR 1600), and a third case mixed (DEDR 1989).(7) However, the alveolar /1/ clearly shows the conditioning factor. Emeneau (1995: 407) is stumped for an explanation: two cases with alveolar /1/ block palatalization, and three cases do not. The two cases blocking palatalization, Ta. kil (kirp-, kirr) 'to be able', etc. (DEDR 1570), and Ka. gel- (geld-) 'to win', etc. (DEDR 1972), are both monosyllabic verbs. As such, the alveolar /1/ enters directly into verbal morphology and is regularly assimilated. Here, the deep phonology form of alveolar /1/ as [-high] comes into play, the resulting clusters remain [-high], and palatalization is blocked here. The verb forms without the alveolar clusters, which are much less common, were analogized from the forms with them. The three sets showing palatalization, Ta. cil 'some, few', etc. (DEDR 1571), Ta. cil 'sound, noise', etc. (DEDR 1574), and Ta. cil 'small piece', etc. (DEDR 1577), are either not verbs or have verbs derived with an added syllable. As a result, the alveolar /1/ does not enter verb morphology assimilations, remains an [1] in surface phonology, and is changed to [+high] by an alveolar raising rule. As such, it allows the palatalization of the velars. The contemporary phonology that I have described for Malayalam still matches the historic change perfectly.

As I reexamined my data on alveolar /r/, however, several details refused to fall into place. First, alveolar /r/ does not have the pervasive influence on surrounding vowels and consonants that the retroflexes have. There is no lowering of the vowel /a/ in ara; the second formant was perfectly level in the environment of ???. Second, while /r/ contrasted with /r/, it also had a more significant contrast with retroflex ???, especially so in intervocalic position where it is a single flap ???.

It then became clear that alveolar /r/ was uvularized rather than pharyngealized, i.e., [+back] but [-low, -high] instead of [+low]. Restated phonetically, the alveolar tap was colored by ??? rather than by [a]. However, to my surprise, a quick check in The Sound Pattern of English revealed that while uvularized dentals were provided for by the theory (Chomsky and Halle 1968: 306), none were known actually to exist! Something was either very wrong or very right.

The Sound Pattern of English has become quite dated in many of its particulars. However, it is still the starting point for generative phonology based on articulatory features; the arguments here can readily be restated in more current terms. One of the most powerful and successful generalizations of the SPE features is the handling of all palatal, velar, and post-velar consonantal positions. Here, the same set of tongue-body features is used to handle both the vowels and these consonants. The three binary features of [[+ or -]high], [[+ or -]low], and [[+ or -]back] not only generate a basic vowel frame of high/mid/low with front/ back, but can also handle the point of articulation for these back consonants. Palatals (i.e., front k) are [+high, -low, -back], velars are [+high, -low, +back], uvulars are [-high, -low, +back], and pharyngeals are [-high, +low, +back]. The same three features also handle coarticulations for consonants in the front part of the mouth. While the features coronal (blade of the tongue raised) and anterior (construction in front of the palatoalveolar region, i.e., ???) provide a base that is intended to be universal, the consonants at the prepalatal region of the mouth (the t's) are not given as firm a foundation in SPE phonology.

Chomsky and Halle (1968: 312-13) attempted to handle these ts, i.e., prepalatal coronal contrasts - dental, alveolar, and retroflex - with the feature distributed where, simplifying somewhat, apicals (articulated with the tip of the tongue) and retroflexes are [-dist], while laminals (articulated with the blade of the tongue) are [+dist]. Either dentals or alveolars can be [+dist], but Chomsky and Halle claimed that any given language that has both will contrast them with this feature. This strong statement was provided with a clearly laid out counterposition: "It would be controverted if, for example, a given language were shown to have dental and alveolar consonants which both had apical articulations" (Chomsky and Halle 1968: 312).

Malayalam is such a language. It has contrasting dental and alveolar stops with apical articulations. So do Irula, Toda, and other Dravidian languages. Needless to say, SPE phonology has not been very good at handling this aspect of Dravidian phonology and certainly has not captured the proper distinctions in its phonology and morphology.(8) The fact that Malayalam on its face destroys Chomsky and Halle's strong case for the feature distributed has been pointed out often enough.(9) The problem is that no universal replacement has been accepted.(10) Ad hoc replacements maintain the critical apical versus laminal distinction, usually with a feature apical or its equivalent.

When seen through the lens of generative phonology, Emeneau's observations on historical developments involving alveolars, as well as my own field and laboratory data, take on added significance. In the Dravidian languages of south India, the tongue-body features of [[+ or -]high], [[+ or -]low], and [[+ or -]back] play a major role in coronal contrasts, in addition to their roles in post-alveolar contrasts and coarticulations. Part of this is not new (the retroflexes have been recognized as [+low] by numerous sources, either explicitly or by implication). I am claiming that the strong retroflexion in the Dravidian languages is created by the combination of [-ant, +cor, +low].(11) Retroflex stops are inherently [+apical], but this does not apply to all retroflex articulations; Malayalam's retroflex approximant ??? has no point of contact or constriction. All retroflexes are redundantly [-high, +back]. Dentals [+ant, +cor, -high, -low, -back] remain the unmarked case. Alveolars are [+ant, +cor, -high, -low, +back]. Therefore, in terms of SPE phonology, the contrasting alveolars are the missing uvularized dentals. It does not conflict with, and in fact augments, the use of tongue-body features to describe coarticulations. In physiological terms, I recognize that while the tongue is a remarkably flexible organ, the front of the tongue cannot move independently of its root. The height-neutral feature back will not form a constriction by itself in the mouth, except for the far-back uvular articulation. However, it can readily impact the location of a constriction in the front of the mouth. The strongest case for this is that the vowels are directly influenced by these features, as shown by Emeneau's data.

The previous description of Malayalam's phonology needs to be amended. The listing for the velarized and palatalized consonants remains the same. All retroflexes are pharyngealized [+back, +low]. However, the remaining coronals are handled as follows in the deep phonology. Dental/t/,/r/, and/s/are [+cor, -high, -low, -back]. Alveolar /t/, /r/, /n/, and /l/are [+cor, -high, -low, +back]. However, alveolar /l/ and /r/ are palatalized [+high, -back] by surface phonology rules (except final /r/). Thus, alveolar /r/ and /n/ remain uvularized in the surface phonology, providing for a fourth series of coarticulations in addition to the unmarked series.

This formulation provides for easy description of the natural classes. Alveolars and retroflexes are [+cor, +back]; dentals and palatoalveolars [t??] are [+cor, -back]. It provides a natural environment for dentals (but not alveolars and retroflexes) palatalizing to palatoalveolar before front vowels, another major change in Tamil-Malayalam.

Returning to Emeneau's statement on the effect of alveolars on the k [greater than] c palatalization rule, it is clear that the actual conditioning factor is [-back]. The general rule can now be restated as follows. In Pre-Tamil-Malayalam, initial k [greater than] c when followed by a front vowel and when the immediately following consonant was [-back], i.e., not a retroflex or uvularized alveolar. Some seeming exceptions are due to an analogical regularization. This could happen when only some conjugational forms occurred in the palatalizing environment.(12)

The coarticulations that had seemed to be a phonological aberrancy in Malayalam can now be seen to be preservations, although it seems highly likely that Malayalam extended its inherited system for the dental /n/ and the palatalized voiced stops. From inscriptional evidence, Kannada and most Tamil dialects seem to have lost r/r contrasts by the twelfth century. This presumably included any coarticulations, except for the pharyngealization of retroflexes, which seems ubiquitous in south India. This concept also gives valuable insight into how phonological systems with numerous points of articulation are maintained in everyday speech. While I often disagree with him on the details of interpretation, Murray Emeneau has an unerring ability to point out critical sets of data. It has been a real pleasure to be able to support one of his observations forcefully.


1 There were other limiting factors to this sound change, such as expressives (onomatopoetics) in Tamil-Malayalam (see Annamalai 1968) and other changes in Telugu. The present note deals only with the phonologically conditioned changes in Tamil-Malayalam.

2 The terminology and symbolism can be confusing. Both /r/ and /r/ are alveolar taps; see Ladefoged (1971: 50-51) for a description. The more common and unmarked /r/, which is fronted, is called dental to maintain a distinction with /r/ which is retracted and a major allophone of the alveolar stop. There is no standard symbol for this contrast. In general, the standard symbolism for maintaining dental and alveolar contrasts requires a forest of diacritics, appropriate for phonological discussions, but not running text. In print, I use a typeface with marked serifs for dentals and an austerely sanserif one for alveolars. Wherever possible, I have included the label with the symbol. In handwriting, I have adapted letters from the Gaelic miniscule for dentals (i.e., ???, ???, ???) with good effect. They are clear, unambiguous, and easy to write.

In this note I have used detailed phonetic symbols and terminology rather than the more usual ones to help clarify the phonology involved. I have used ambiguous terms and symbols as little as possible.

3 This perspective is changing. Note that there was a recent conference on coarticulations in India (International Journal of Dravidian Linguistics 24: 149-54).

4 See Prabhakar Variar (1976: 163) for a translation of a more complete description given by Sukumaran (1974) in a traditional Malayalam framework.

5 In pronouncing the dental /n/ the whole body of the tongue, front and back, is raised and level. Contact is made just inside the lower edge of the teeth with the sides as well as the tip of the tongue. The dental contact is molar to molar. The strong, and clearly audible, velarization seems to be an inherent part of the general lifting of the tongue. While the dental /n/ sounds complicated, it is easy to learn and easy to make. This articulation seems to be an innovation in Malayalam.

6 Spectrograms of this contrast have been published by Ladefoged, Cochran, and Disner (1977: 51). While not mentioned in the text, they show the second formant shift, although not as clearly as between two identical vowel phonemes.

7 References are to the second edition of A Dravidian Etymological Dictionary (Burrow and Emeneau 1984). All language abbreviations follow its usage.

8 See the papers and discussion in Schiffman and Eastman (1975) for numerous attempts to address this issue. In the consensus of the discussions no approach presented was satisfactory.

9 The most authoritative citation is Ladefoged (1971: 102).

10 While SPE's claims to be universal were certainly overblown, such universal arguments are of great interest in comparative linguistics. Universals and near universals put needed structure into historical contexts and allow meaningful statements about long-dead phonologies.

11 This meaning of the term strong retroflexion follows Schiffman (1975: 82-83) and refers to the fact that assimilation in Dravidian languages is almost always to the retroflex articulation. I know of only one example of a retroflex assimilating to anything else and that is an alveolar: in Malayalam pronominal case morphology, en + (u)te [approaches] enre [[Epsilon]nd[Epsilon]] 'my'.

12 The resistance of the expressives probably has the same explanation since they are commonly reduplicated, i.e., syllables three and four repeat syllables one and two.


Annamalai, E. 1968. Onomatopoetic Resistance to Sound Change in Dravidian. In Studies in Indian Linguistics (Professor M. B. Emeneau Sastipurti Volume). Pp. 15-19. Poona: Deccan College and Annamalai University.

Burrow, T., and M. B. Emeneau. 1984. A Dravidian Etymological Dictionary, second edition. Oxford: Clarendon Press. Referred to as DEDR.

Chomsky, N., and M. Halle. 1968. The Sound Pattern of English. New York: Harper & Row.

Emeneau, M. B. 1995. The Palatalizing Rule in Tamil-Malayalam and Telugu. Journal of the American Oriental Society 115: 401-9.

Ladefoged, P. 1971. Preliminaries to Linguistic Phonetics. Chicago: Univ. of Chicago Press.

Ladefoged, P.; A. Cochran; and S. Disner. 1977. Laterals and Trills. Journal of the International Phonetic Association 7.2: 46-54.

Prabhakara Variar, K. M. 1976. Phonological Theories in Malayalam Grammars. In Dravidian Linguistics-V, ed. S. Agesthialingom and P. S. Subrahmanyam. Pp. 143-65. Annamalainagar: Annamalai University.

Rajaraja Varma, A. R. [1895] 1969. Keralapaniniyam. Kottayam: National Book Stall.

Schiffman, H. F. 1975. On the Ternary Contrast in Dravidian Coronal Stops. In Dravidian Phonological Systems. Pp. 69-85.

Schiffman, H. F., and C. M. Eastman, eds. 1975. Dravidian Phonological Systems. Seattle: South Asian Studies Program, University of Washington.

Sukumaran, V. 1974. Malayalattile svanimannalum aksarannalum. In Malayalabhasapathanannal. Trivandrum: State Institute of Languages.
COPYRIGHT 1998 American Oriental Society
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1998 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:McAlpin, David W.
Publication:The Journal of the American Oriental Society
Date:Jul 1, 1998
Previous Article:The industrial park of Girsu in the year 2042 B.C. interpretation of an archive assembled by P. Mander.
Next Article:Matsyendra Samhita, Ascribed to Matsyenranatha: Part I.

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters