On the natural phonology of vowels (original) (raw)

Thus; whereas the .American .child everttually overcomes her substi-_tut.ion of Co.:! for C'PJ, the Nupe chiid confronts no foJ in his mother tongue, and the (Q.J-substitution manifests, itself wheen, in the inf'lex:.;. ibility of ~dulthood, be enc_ounters (:,J in :a foreign vord. ') If one could collect all the_sound-subst;ituticins of children, the dfacbr_onist Grammont (1965} _said, one would hf.I.Ye a. sort of grammar of all the possible sound-changes. The synchronist Baudouin de Courtenay {1895} would have added, also all the possible phonological alt_erna.-tic>ns. And, as the panchronist Jakobson argued in his monumental Child Lan@ase, Aphasia 1 and Phonological Universals (1968}, _also all the• • possible phoneme systems. Such systems are determined, a.s Greenberg (1966) concluded in his study of phonological universals, by the collective effect of phonological processes (particularly the 'unconditioned' processes vhich, since they affect a. sound)n all its oc,currences, thereby affect the inventory as well a.s the distribution of phonemes). But, pace Greenberg, such processes are not mere historical events: they are the living expression of the phonetic capacity of the individual. This dissertation presents my explorations in the world of vowels. Although I have occasion to mention aesimilatory processes, my focus is on those processes traditionall.y labeled 'unconditioned I or 1 sponta.-' neous 1-the ones pessimistically cWed unexplainable because they apply to segments regardless of their contexts (or worse, in dissimila:.;. tion, despite their contexts), and because, as is implied by their classification as 'strengthening' processes, they seem to defy the law of least effort. These are processes which, in the earliest speech of a child, ca.n scramble all his vowels into one, and yet can, in a language like Faroese, Juggle two dozen vowels with hardly a merger. A full account of the nature, operation, and causality of these processes is not in sight, Thorough and useful descriptions are unavailable for many languages: the brief phonemic sketches of vowel systems that appear in many surveys and in some monographs often fail to provide sufficient phonetic information on vowel quality; and conversely, some descriptions vhich do provide such phonetic information lack the phonological data that is crucial to the sort of analysis that the study of processes requires. For many little-know languages, historical studies, which provide much usef'ul data, are unavailable-and indeed impossible, given linguists' current knowledge of such languages. ' (Corresponding to this difference in the nature of processes and rules is a difference in the order of application in speech proce~sing: processes apply after the applications of secret-1anguage rules, and after unintentional slips of the tongue (Stampe 1973a:45; Donegan and Stampe 1978a, Sec. 2. 5). • (1,7) For example, the process that palatalizes [kJ to [cJ before a palatal vowel applies after the secret-language 9 Another way :irt ~hich children: •sometimes limit the surface effects of a process is by constrairting its natural iterative application (Stampe 1973,: 59-68). The fUnction of each process is to Efubstitute •. a'iess difficult class of sounds or sound-sequences for a ~ore difficult cla.ss~ But once a process has applied~ sveeping away~. as it vere, a.• certain class 'or difficulties, another process may (subsequently or simultaneously), in removing some entirely different difficulty, create new me~ rs of the very class the fir~t process got rid of. /' (1.14) For example, for a child who substitutes zero for CJJ, this CJJ-deletion eliminates a difficult segment. But suppose another process-delateralization, as in (1,10)-simultaneously substitutes CJJ !or CIJ. Unless the first process is alloved to apply again, the child vill have to produce CJJ's-for ti l's. If each process is to accomplish its function on the surface forms (the forms that are actually pronounced), then the (JJ-to-¢ process should apply again, after the CIJ-to-[JJ process, and again after any other process that creates [JJ's. Such absence of ordering restrictions-unconstrained iteration-is the natural state of process application. But there is a catch to this free-handed elimination of difficulties: the processes thus applied merge, in actual pronunciation, the distinction between between /1/ and /j/ in the child's underlying representations (corresponding to adult tlJ and CjJ); both become zero, so that e.g. less and yes would both be pronounced [€SJ. One way for the child to maintain a distinction without having to suppress either process is to restrict the iteration of [JJ-to-¢,.so that it may not apply again after [IJ-to-(JJ. Thus the child says [€SJ for~ but [JESJ for less-not the underlying or adult distinction, to be sure, but a distinction nevertheless. This seems to be the situation which holds vith children who initially substitute zero for both (jJ and ti] but who later produce zero for [JJ and (JJ for.ti] (cf, Jakobson 1968:15, and Donegan and Stampe 1978a, Sec. 3.3), Such apparently paradoxical sets of substitutions, described by Jakobson as 'sound shifts', may persist into adulthood. Thus, such constraints may account for some of the peculiar situations in language in which a speaker cannot pronounce a segm~nt or sequence when he tries to produce it, but produces that very segment or sequence when he is trying to pronounce something else. (1.15) For exe.mple, many speakers of English find it difficult or impos'sible to produce the seq'l,lence ta.9J in phrases like How now, brown cow? or in bo!rowed or foreign words like He.usfrau or Laut (cf, (1.6)). Instead they substitute Cais!J quite automatically. But many or the same speakers in whom the ta.2J-to-(6!12J process is active also make a substitution which produces the phonetic sequence Ca2J: de.rk, syllable-final C:lJ optionally becomes tyJ"'[S?J so that doll, Sol, etc. a.re pronounced (dC1.2J, Isa.2J, etc.