Academic journal article Perception and Psychophysics

On the Causes of Compensation for Coarticulation: Evidence for Phonological Mediation

Academic journal article Perception and Psychophysics

On the Causes of Compensation for Coarticulation: Evidence for Phonological Mediation

Article excerpt

This study examined whether compensation for coarticulation in fricative-vowel syllables is phonologically mediated or a consequence of auditory processes. Smits (2001a) had shown that compensation occurs for anticipatory lip rounding in a fricative caused by a following rounded vowel in Dutch. In a first experiment, the possibility that compensation is due to general auditory processing was investigated using nonspeech sounds. These did not cause context effects akin to compensation for coarticulation, although nonspeech sounds influenced speech sound identification in an integrative fashion. In a second experiment, a possible phonological basis for compensation for coarticulation was assessed by using audiovisual speech. Visual displays, which induced the perception of a rounded vowel, also influenced compensation for anticipatory lip rounding in the fricative. These results indicate that compensation for anticipatory lip rounding in fricative-vowel syllables is phonologically mediated. This result is discussed in the light of other compensation-for-coarticulation findings and general theories of speech perception.

Hockett (1955) once described the problem of speech recognition with the metaphor of recognizing colored eggs on a conveyor belt after they have been crushed by a wringer. After the wringer, the different colors run into each other, and it is difficult to say where a given color starts and ends. This metaphor correctly indicates that speech sounds are neither separable-that is, there is no point in time at which the speech signal is influenced by only one phoneme-nor invariant, because the acoustic form for a given phoneme will be influenced by the surrounding phonemes. One example is the case of fricative-vowel syllables in American English. Although /s/ is supposed to be pronounced with unrounded lips, a following rounded vowel leads to some anticipatory lip rounding, making the fricative more /∫/-like. How are such coarticulated phonemes then recognized? A number of studies have shown that coarticulation is compensated for in perception (e.g., Beddor& Krakow, 1999; Fowler & Brown, 2000; Liberman, Delattre, & Cooper, 1952; Mann, 1980; Mann & Repp, 1981). Mann and Repp (1980), for instance, showed that listeners will still accept a fricative with some cues for lip rounding as /s/ if it is followed by a rounded vowel, whereas the same fricative is interpreted as /∫/ if followed by an unrounded vowel. Listeners thus take the context into account in a compensatory way when making phonetic decisions. The underlying mechanisms that cause compensation for coarticulation are hotly debated. In the present article, I will contrast an auditory account with phonological accounts (see Figure 1).

Auditory accounts hold that context sensitivity is pervasive in perception in general (KJuender, Coady, & Kiefte, 2001; Warren, 1999) and that perceptual systems have evolved to cope with the lack of invariance in the environment. Lack of invariance in the environment is caused by, among other forces, inertia. Inertia is also one cause of coarticulation (Farnetani, 1997; Whalen, 1990). Given that the auditory system has evolved to deal with inertia, it can compensate for coarticulation through general auditory perception. One particular mechanism that has been put forward is spectral contrast, which arises from adaptation at different levels along the auditory-processing chain (see, e.g., Holt, 2005; Holt, Lotto, & Kluender, 2000). It is assumed that although coarticulation with a preceding /1/ leads to a higher F3 in a velar stop /g/-making it more /d/-like-listeners adapt to the high F3 in the preceding /1/. This adaptation decreases the sensitivity for frequencies in the higher part of the F3 skirt for the [g], thereby decreasing the perceived F3 center frequency and canceling out the /l/'s coarticulatory influence. Evidence for spectral contrast stems from, for instance, an experiment by Lotto and Kluender (1998), who used stimuli with the structure "sine wave sound + /{d, g}V/. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.