Linguistics Seminar Talk - Rob Mailhammer

Title: Minimal pairs without segmental contrasts: velars consonants in Iwaidja and vowels in German

Abstract: 

The topic of this talk is the question what the basis is for the diagnosis of a phonological contrast in a language. The main point the talk makes that a lack of consistency in the articulation of a segment may not be a problem for a phonological opposition, if some consistent property differentiating minimal pairs can be established. This talk will discuss the case of the velars in Iwaidja and then draw a comparison to the vowel opposition in Modern Standard German. The Australian language Iwaidja is traditionally assumed to have a contrast between a stop [g] and an approximant [ɰ] (Evans 2000; Mailhammer & Harvey 2018). However, there is only one confirmed minimal pair, /ɻaga/ ‘he throws it’ vs. /ɻaɰa/ ‘3sg lies upside down’, and transcriptions confuse the two phonemes from time to time. Moreover, a targeted articulatory and acoustic analysis of the hypothetical contrast in an a_a environment found that there was no consistent phonetic property that would differentiate /g/ and /ɰ/ on a lexical and an intra- and inter-speaker level (Shaw et al. 2020). The conclusion the authors draw is that this does not support a phonemic opposition (Shaw et al. 2020: 609), and that the confirmed minimal pair is evidence of a marginal contrast at best. Despite this evidence, this conclusion sits somewhat uncomfortably with fieldworkers with first-hand knowledge of the language, who have been corrected for their mispronunciations. First, the results in Shaw (2020) only account for a subset of the data. Second, recent perception data supports a phonemic opposition between /g/ and /ɰ/ for Iwaidja (Bundgaard-Nielsen et al. submitted). Third, Shaw et al. (2020: 608) say that the range of realisations of the two velars may be different. In this connection, an informal impression is noteworthy. It appears that the first vowel in /aɰa/ sequences appears to be longer than that in /aga/ sequences, so that /raɰa/ ‘3sg is upside down’ is more like /ra:.a/ as opposed to /raga/ ‘he throws it’. Thus, although the total length of an /aga/ sequence is higher than that of /aɰa/ (Shaw et al. 2020: 604), it could be that the first vowel is shorter in the former than in the latter. If that proved to be correct, then it is possible that Iwaidja in fact has a vowel length opposition resulting from something like compensatory lengthening and that the consonantal segment that follows is variable in realization and not contrastive by itself. This possibility would explain the facts, and it would be a case where an opposition was dependent on the consistent realization of a vocalic segment and the variable realization of a following consonantal segment. It would explain why looking for a consistent realization of a consonantal opposition has not yielded any results and why there would be variable ranges of phonetic realisations. The case of Modern Standard German is instructive in this regard. Whereas traditional approaches assume a segmental vocalic contrast between e.g. Komma ‘comma’ and Koma ‘coma’, an alternative view is that there is a prosodic contrast between the stressed vowel and the following consonant where the phonetic properties are interdependent (see e.g. Vennemann 1991; Restle 2003). The interaction of vowel and consonant creates the contrast rather than one segment by itself. Languages of this type typically show a variation of phonetic properties. The combinatorics of these properties create an opposition that can look segmental, but which shows various unmotivated restrictions and asymmetries that are concomitants of a more underlying prosodic contrast. The implication for the minimal pair method to diagnose phonological contrasts is that contrasts may not necessarily be segmental, which has practical implications for linguistic fieldwork.