Desktop version

Home arrow Language & Literature

Discriminative and categorical variation

A discriminative perspective also clarifies why recurrence at a phonemic level provides an imprecise measure of identity and therefore redundancy. This is not ultimately a deficiency of phonemic analysis but a design feature of such analyses. The hierarchies of linguistic levels inherited from descriptivist approaches are often interpreted as encapsulating all of the variation at each level that is relevant for analyses at the next level up. On this view, the phonetic properties relevant to phonemic analysis are encapsulated in phones, the phonemic properties relevant to morphemic analysis in phonemes, etc.11 [1]

However, this conception misconstrues the organization of linguistic levels, which in fact categorize patterns of continuous variation in terms of discrete units. This categorization does not aim to approximate the aggregate properties that distinguish units, but instead to isolate a single dimension of categorical contrast. This objective is explicit in standard definitions of phonemes. A pair of phones ф1 and ф2 are assigned to different phonemes if they are themselves discriminable, and if they suffice to distinguish forms (usually words) with a discriminable difference in meaning. The phonemes of a language isolate contrasts with a high individual functional load, since each phonemic contrast discriminates at least some minimal pair. But there is no principled reason to expect that the contrasts with a high individual functional load will have a high collective functional load within a system. No contrast at the level of sound patterns, however salient or discriminable it may be, will be phonemic if it is invariably triggered or reinforced by some other contrast. In either case, the co-occurrence of the other contrast will ensure that there are no minimal pairs that are solely distinguished by the original contrast.

In sum, phonemic analyses locate discrete units within a system of categorical contrasts. The notion of ‘phonemic identity’ does not characterize the general notion of ‘perceptual identity’ relevant to speakers because it isolates one dimension of discriminable variation within a multi-dimensional space.

Many of the contrasts investigated in studies of sub-phonemic differences are concomitants of some independent source of variation. If these sources are completely outside the phonological or morphological system, the sub-phonemic variation can in principle be treated as externally-conditioned ‘micro-variation’ exhibited by a single categorical unit. The case for abstracting away from subphonemic variation is strongest if the variation is governed by factors such as frequency, syntactic position, lexical neighbourhood density, etc. For example, the observation that forms may have different duration and prosodic properties at the beginning or ends of phrases or clauses does not impinge directly on the phonological or morphological analysis of the forms.

Conversely, the impact of sub-phonemic constrasts is strongest if the contexts that influence variation are paradigmatic or grammatical. In the case of the Dutch stems discussed by Baayen et al. (2003) and Kemps et al. (2005), the prosodic contrast between geit when it occurs as the singular and when it occurs as the stem of the plural geiten appears to be conditioned by general prosodic differences between monosyllabic and disyllabic forms. However, the comprehension studies in Kemps et al. (2005) clearly indicate that Dutch speakers learn to recognize these differences as marking number contrasts.

Since it is the retention of the conditioning context that makes the prosodic difference sub-phonemic, the loss of this context can create a phonemic opposition within a language. This appears to be the origin of the grade alternations in the Estonian declensionsand conjugations in Chapter4, and thesourceofsimilar alternations in other Finnic and Uralic languages. From a discriminative perspective, it is significant that analyses of the evolution of these grade alternations treat the discriminability of stem variation as a prerequisite for the grammaticalization of a sub-phonemic contrast. The account of the origin of Finnic grade in Viitso (2003) is particularly clear on this point: there are no other known phonetic preconditions for the rise of quantity correlation of long syllables than the phonetic alternation of first syllables depending on the openness vs. closedness of the following unstressed syllable in Finnish... In order to develop into a morphophonological gradation, this automatic alternation needed only to be perceived. Only a perceived alternation could be accepted as a linguistic norm and subjected to further polarization and reinterpretation of sound patterns caused by openness vs. closedness of the second syllable of a foot as a function of syllabic accentuation in initial syllables... This polarization as a norm caused the quantity alternation of stems in their morphological paradigms so that the stressed syllable was longer before an open syllable and shorter before a closed syllable. This quantity alternation was retained, in principle, also after the former conditioning environments changed. (Viitso 2003:164b)

Viitso (2003:163) illustrates these developments by contrasting “the nominative and genitive forms of the stem for ‘bushel’, cf. *vakka : *vakan” in which “the geminate stop was shortened before a closed second syllable”. As in Modern Dutch, prosodic differences conditioned by syllable structure in the Finnic languages must have been discriminable, functional load-bearing contrasts at the point where the conditioning environments were still present (as they are, for the most part in Modern Finnish). Hence an analysis of* vakan that associated genitive case features solely with the ending -n would be as misleading as an analysis of geiten that associated plurality exclusively with -en.

  • [1] As expressed in the quotation from Bloomer et al. (2004:180) in footnote 5: “each sentenceconsists of clauses, each clause consists of phrases, each phrase consists of words, each word consists ofmorphemes and each morpheme consists of phonemes”.
 
Source
< Prev   CONTENTS   Source   Next >

Related topics