آیا پردازش ادراکی خوشه‌های همخوانی آغازی در زبان فارسی منجر به درج واکه خیالی می‌شود؟

در این مقاله، نقش واج‌آرایی زبان فارسی در درک خوشه‌های آغازی (غیرمجاز) فارسی در چارچوب مطالعۀ آزمایشگاهی بررسی شده است. هدف از انجام این پژوهش، پاسخ‌گویی به این پرسش است که آیا با توجه به محدودیت‌های ساخت هجایی زبان فارسی مبنی بر وجود تنها یک همخوان در آغازۀ هجا، فارسی‌زبانان در رشته‌های آوایی دست‌کاری‌شدۀ بی‌معنیِ c1v1c2 واکه می‌شنوند که در آنها v1 به درجه‌های مختلف از سیگنال آوایی گفتار حذف شده است. یا آنکه به درجه‌های مختلفِ حذف v1 در سیگنال گفتار حساسیت شنیداری نشان می‌دهند؛ بنابراین، بین صورت‌های آوایی که در آنها v1 حذف شده و صورت‌هایی که در آنها v1 حذف نشده است، تمایز می‌گذارند. برای انجام این پژوهش، تعداد 20 ناواژه با الگوی هجایی c1v1c2v2c3 طراحی شد؛ به‌طوری که ناواژه‌ها هر 6 واکۀ زبان فارسی را در خود داشتند. ناواژه‌ها پس از تولید و ضبط در نرم‌افزار پرت بازسازی شدند؛ به این صورت که واکۀ اول درون هر ناواژه، به‌طور تدریجی طی چندین گام از سیگنال آوایی ناواژه حذف شد. نتایج به‌دست‌آمده نشان داد درک واکه‌های خیالی به‌طور نظام‌مند برای تمامی توالی‌های c1c2 و تمامی واکه‌های فارسی روی نمی‌دهد؛ بنابراین، این فرضیه که فارسی‌زبانان هجاهای حاوی خوشه‌های آغازی c1c2 در آغاز ناواژه‌ها را با توجه به محدودیت‌های نظام آوایی زبان فارسی به‌صورت توالی‌های آوایی c1vc2 درک می‌کنند، فرضیۀ معتبری نیست.

Does Perceptual Processing of Initial Consonant Clusters Lead to the Perception of Illusive Vowels in Persian?

Abstract In this article, we investigated the role of Persian phonotactic parameters on the perception of Persian initial consonant clusters (ungrammatical) in the context of an experimental study. Given the limitations of the Persian syllable construction of having only one consonant at the beginning of the syllable, the purpose of this study was to answer the question whether Persian speakers perceive a vowel in nonsensically manipulated C1V1C2 sequences in which V1 is deleted in different degrees from the speech signal or they are found to be sensitive to the magnitude of vowel deletion in the sequences in which they differentiate between sequences in which V1 is deleted and those in which V1 is preserved in the speech signal. To conduct the study, twenty nonwords with C1V1C2V2C3 syllable pattern were designed so that they contained one of the six vowels of the Persian sound system. The target nonwords were re synthesized in the Praat software after being produced and recorded in such manner that V1 in each nonword was deleted gradually from the speech signal in several steps. The results showed that the perception of an illusive vowel does not occur systematically for all C1C2 sequences and all Persian vowels. Thus, the hypothesis that Persian speakers perceive C1C2 sequences in initial consonant clusters as C1VC2 at the beginning of nonwords, due to phonotactic constraints in the Persian sound system, is not valid.  Keywords: Initial Clusters, Illusive Vowel, Speech Perception, Nonwords, Auditory Sensitivity.   Introduction The phenomenon of perceptual illusions particularly that of illusory vowels has become a very useful probe to understand both loan word patterns and the speech perception mechanism. Previous research has unearthed a variety of factors that affect where illusory vowels are perceived. They are perceived in stimuli that contain: (a) consonantal sequences that are phonotactically illicit in the native language of the speaker (Dupoux et al., 1999; Durvasula and Kahng, 2016); (b) consonantal sequences that violate putative language universals such as the Sonority Sequencing Principle (Davidson, 2007), and (c) specific acoustic cues in the stimuli, such as strong consonantal release bursts, increased voicing amplitudes, etc. (Davidson, 2007). What evidence exists that phonotactic constraints play a role in perception? Adults have rather clear intuitions about permissible sequences. For example, English speakers know that “mba” is not a possible English word. Davidson (2007) has shown that nine month old infants are sensitive to the phonotactic patterns of the words in their language and some researchers have argued that such regularities could be useful in helping the child to discover words. Mattingley et al. (2015) investigated the influence of phonotactic constraints on phoneme percep tion. They used the fact that /sri/ and /Sli/ are not allowed in English while /sli/ and /Sli/ are allowed. They synthesized a series of stimuli ranging from [s] to [S] and presented them to participants in the / li/ and / ri/ context. There was a significant shift in the identification functions between the two contexts, demonstrating that participants tend to hear segments that respect the phonotactics of their language. This paper addressed the role of Persian phonotactic constraints on the perception of initial consonant clusters. Given that initial consonant clusters are not permitted in the Persian sound system, we investigated whether Persian listeners perceive an illusive vowel in C1V1C2 sequences in which V1 is deleted in different degrees from the speech signal, or they are found to be sensitive to the magnitude of vowel deletion in the sequences. Materials and Methods Twenty nonwords with C1V1C2V2C3 structure were designed such that they contained one of the six vowels of the Persian sound system. The target nonwords were resynthesized in the praat software such that V1 was deleted from the speech tokens in magnitude degrees. All acoustic measurements were made using Praat acoustic software (Boersma and Weenink, 2010). The stimuli were digitized on a PC Compatible computer using an OROS AU22 A/D board. Five different files were then created from each original item by splicing out pitch periods of the target vowel, V1, at zero crossings. Stimulus 1 contained little or no vowel (most of the transitions in and out of the vowel were also removed). Stimulus 2 contained the two most extreme pitch periods of the vowel (i.e., one from the transition of the first consonant to the vowel, and another from the end part of V1 into the following consonant). Stimulus 3 contained the four most extreme pitch periods (two on each side), and similarly, Stimulus 4 six pitch periods, and Stimulus 5 eight pitch periods. Stimulus 6 was the original stimulus in which the number of pitch periods varied from 10 to 13 across items (10.7 periods in average.) The average overall duration of one pitch period in V1 in each item was 9.06 ms. There were a total of 60 stimuli in one session.   Discussion of Results & Conclusions Results suggested that listeners do not consistently perceive an illusive vowel in C1C2 sequences.  Thus, our study does not provide sufficient evidence for confirming the hypothesis generally assumed in the phonetic literature that C1C2 sequences in languages which do not permit initial consonant clusters are perceived as C1VC2.