Synergy II: Linguistics

Morphology, Phonology, and the Interaction

This study investigates morphophonological alternants, namely, the output forms of the interacting processes between morphology and phonology and suggests that such alternants cannot be explained by pure morphology or pure phonology, but morphophonology. In the current study, morphophonology is not a separate submodule of the grammar, but comprises interactive processes of an autonomous phonology and an autonomous morphology in line with Pounder (2000), Sadock (2012), and Kunduracı (2013). The study evaluates (i) the view that morphophonology falls within morphology (e.g., Bochner, 1993), (ii) the view that it falls within phonology (e.g., Kiparsky, 1996), (iii) the option that it is another, intersecting submodule between morphology and phonology, but concludes that morphophonological formations must be the outputs of the interactions between phonology and morphology. With a variety of cases in Turkish, the study claims that whereas morphophonology is an interactive ←13 | 14→phenomenon relevant to both morphology and phonology, it is controlled by morphology.¹

Morphophonological Alternations

This section is dedicated to a distinction between morphological alternations and phonological alternations: The formal outputs of autonomously morphological processes must be distinguished from the formal outputs of purely phonological processes, for which Turkish provides certain data (see below). Haspelmath and Sims (2010) also draw a line between morphologically conditioned alternations and phonologically conditioned (automatic) alternations based on a number of characteristics such as phonological vs. lexical conditioning, application to loan words, yielding new segments in the language, formal complexity, phonetic coherence and distance, application across word boundaries, and obligation (p. 214–217). The first three of these are considered in terms of the Turkish data, in this section.

When we consider the expression of causative verb forms in Turkish, for example, which involves the suffixation of -DIr, -Ir, -(I)t, and -Ar(t), we see that whereas some alternants can be explained based on simply phonological environment, others are not so, and that a reference to lexical conditioning is also needed for these. Phonological conditioning can explain why multisyllabic verb stems ending in a consonant, such as gül ‘smile’ > gül-dür ‘make smile/laugh’, are always suffixed the -DIr form, and multisyllabic verb stems ending in a vowel or a liquid, such as uyu ‘sleep’ > uyu-t ‘make sleep’, karar ‘get/be dark/black’ > karar-t ‘make dark/black’, are always suffixed the -t form. In other cases, which involve monosyllabic stems, however, phonological conditioning itself does not suffice to account for the relevant marker of the causative in Turkish. For instance, the causative form of the verb stem ak- ‘run (water)’ is ak-ıt, whereas the causative form of another monosyllabic stem, tak- ‘wear/affix’ is tak-tır.² Clearly, morphophonological alternants cannot be limited to and explained by merely phonology.

Another example is word-final voicing (of /p, t, k, ʧ/) in some noun stems in the environment of vowel-initial suffixations in Turkish (see also § 3.4). For ←14 | 15→instance, the noun stem havuç /havuʧ/ ‘carrot’ involves a voiceless consonant which does not undergo any processes in the word-final position and in the environment of consonant-initial suffixations, but which undergoes voicing, ʧ > ʤ, in the environment of vowel-initial suffixations as seen in havuc-um (carrot-1.poss)³ /havuʤum/ ‘my carrot’. This process can be explained by phonological environment, i.e., the vowel trigger in the suffix, of course. However, the same [-voiced] segment, /ʧ/, does not undergo any voicing within similar environments (vowel-initial suffixations) when the stem that it occurs in is not a nominal unit, but verbal. For example, the final consonant in the verb stem kaç- /kɑʧ/ ‘escape’ does not change to /ʤ/ in the inflected and derived forms which include vowel-initial suffixations, as seen in kaç-ar-ım (escape-aor-1.sg) /kɑʧɑrɨm/ ‘I (will) escape’, kaç-acak (escape-fut) /kɑʧɑʤɑk/ ‘(S)he/it will escape’, and kaç-ak /kɑʧɑk/ ‘escaper’. The fact that voicing applies only with nominal stems and excludes non-nominals, which means morphological sensitivity to stem category, is another clear case which shows that only phonological conditioning cannot suffice for morphophonological outputs.

Another criterion to consider is the application of the processes to loan words. Haspelmath and Sims (2010) show that automatic (phonological) alternations apply both in native and loan words, whereas morphologically conditioned ones do not apply in loan words, which creates another contrast between phonological and morphological alternants. They exemplify such cases from German, where the word-final /d/ undergoes devoicing and becomes /t/ in both native and non-native expressions (p. 216). Turning to Turkish, which also displays distinctions between morphological and phonological conditioning, as shown above, it is realised that the loan word criterion indeed does not work in Turkish: Both phonological and morphological processes apply to loan stems in Turkish. For instance, the loan word kraliçe /kɨɾaliʧe/ ‘queen’ is pronounced with an epenthetic /ɨ/ in the first syllable as Turkish does not allow consonant clusters word-initially. This process, which applies to a loan stem, is clearly phonological. Nonetheless, a native morphological process may also apply to loan stems in Turkish as seen in the expression pas-la-mak (pass-vrb-ınft) ‘to hand ←15 | 16→sth off’, which involves a foreign root, pas ‘pass’ with the native -lA suffixation applying on it. This case, among others, exemplifies that not only purely phonological but also autonomously morphological processes may apply to both native and foreign stems in Turkish, unlike other languages which show contrasts between native and foreign stems morphologically and phonologically.⁴

Finally, let us consider the possibility of deriving new segments in a language at the end of formal processes. Haspelmath and Sims (2010) discuss that new segments may be created at the end of phonological but not morphological processes usually (p. 216), which is used as one of the tools to discriminate automatic and morphological alternations. As an example, they refer to the [ɾ] in American English, which is limited to flapping, i.e., when /t/ and /d/ are positioned between vowels, e.g., butterfly [bʌɾə(r)flaɪ]. Turning to Turkish, it is observed that similar to the above case regarding loan stems, derivation of new segments may also be involved in both phonological and morphological processes. For instance, the expression oku-yacağ-ım (read-fut-1.sg) ‘I am going to read’ can be pronounced as [okʰujɑʤɑɨm], [okʰujɑʤɑm] or [okʰi:ʤɑm] due to the allomorphic variations of the future marker -(y)AcAk. The last pronunciation, [okʰi:ʤɑm], which does not involve either the palatal /j/ or the preceding and following vowels /u/ and /ɑ/, but involves /i:/ rather, which is a long vowel, is important for the purposes of the current section: Native roots in Modern Turkish do not involve long vowels apart from the velar glide /ɣ/ (so-called “soft g”) effect.⁵ However, forms including the future suffix such as oku-yacağ-ım ←16 | 17→[okʰi:ʤɑm] allow a long vowel, /i:/, in many speakers, including myself, despite the absence of /ɣ/, the trigger for vowel lengthening. Then, not only phonological but also autonomously morphological processes may result in new segments in a language.