The Nature of Gradualness for Changes in Place of Articulation Features of Vowels in Serial Optimality Theory: Evidence from Three Iranian Languages

In the earliest version of Optimality Theory put forth by Prince and Smolensky (1993/2004), the Generator is allowed to apply more than one process on the input. This version of OT, which is known as “Parallel OT” is not able to identify derivations and intermediate forms. The requirement of the recognition of intermediate forms paved the way for the newer model of Serial OT, where the Generator is only allowed to apply one single process on the input. By analyzing the assimilation of the vowel before the glide /w/ in Kalhori Kurdish, Sorani Kurdish, and Bakhtiari Luri, this article aims to investigate whether this single process must be on each feature, or whether the node of place of articulation must be targeted. The Kurdish data were collected from 10 speakers of Kalhori and Sorani, and the Biakhiari examples were collected from Anonby and Asadi (2014). The analysis of the present study shows, in order to explain the assimilation of the vowel in Serial OT, the Generator must target the node of feature geometry rather than each phonological feature.
Introduction
First introduced by Prince and Smolensky in 1993/2004, Optimality Theory does not perceive phonological alternations as a result of rules application – like what is believed in the derivational phonological theory of Chomsky and Halle (1968) – but rather believes in the interaction of universal violable constraints to be at play. Despite the explanatory power that this theory offers, especially when rules conspire to achieve a common aim (McCarthy, 2008a:2), it fails where there has to be a fixed order of application of processes, e.g. the requirement of stress application before the deletion of the unstressed vowel in the world’s languages (McCarthy, 2008b) or in the cases of phonological opacity resulting from counter-feeding or counter-bleeding interactions (Jam et. al. 2020, Fattahi and Javaheri 2022, Jam 2023).
     As a response, this earlier version of OT, also known as Parallel OT, was modified in such a way to validate steps of derivation in a serial fashion (McCarthy 2008b). One well-known model of Serial OT is Harmonic Serialism, put forth by McCarthy (2010). Unlike in Parallel OT, the Generator in this approach is allowed to apply only one single process to the input – a characteristic known as Gradualness. This article aims to investigate the nature of this gradualness when the unfaithful mapping involves the Place of Articulation features of vowels. The question is whether to consider the change in only one phonological feature as the single change of each step or to take the feature geometry into account and consider the change in a node as a single change, in which case more than one feature may be affected under one single node.
Materials and Methods
The data in this study were collected through interviewing and recording 10 Kalhori speakers from Eslamabad-e Gharb and 10 Ardalani (Sorani) speakers from Sanandaj. The Bakhtiari data, were, however, collected from Anonby and Asadi (2014). The collected data were then transcribed, and the underlying vowels were identified by comparing the alternating forms.
Results and Discussion
Kalhori Kurdish has an inventory of 8 vowels, with /i, y, e:, e[1], a/ being non-back and /u, o, ɑ/ being back. With this in mind, the subjunctive/imperative prefix in this dialect is /be-/ underlyingly. That is while when this prefix is attached to a verbal root initiating with the glide [w], the subjunctive/imperative prefix is phonetically realized as [bu-], as in the word [bu.wɑ.ɾe:d] (if it rains). The derivation of [bu] from [be] entails changes in features of tongue height and position, and lip rounding. In Parallel OT, all these changes could be seen in the winning candidate in one go as the Generator is restricted to the application of one process only. However, in the serial approach to Optimality Theory, (such as that in Harmonic Serialism), This derivation could pose challenges. There are two hypotheses. One is to consider separate steps for the change in each of the features [high], [back], and [round], and one is to have the feature geometry in mind and apply the sole permitted process to the Place of Articulation node, which dominates the aforementioned features.
     In the first case scenario, the only logical path to have one feature change each time in Kalhori would be eàiàyàu. Each of these vowels in this chain is different in only one feature to its adjacent step. The following tableaus illustrate this gradual change in the word [bu.wa.tɑd] (if s/he had said - /be-watɑd/ underlyingly)