![]() ![]() But when a cue is missing in a signal, listeners may compensate for it by assigning greater weight to other cues. Ambiguous cues may be evaluated as less reliable and assigned smaller weight. However, it is not clear how compensation affects cue weighting. ![]() Research on compensation ( Fowler, 2006 Gaskell & Marslen-Wilson, 1996 Lotto et al., 1997 Lotto & Kluender, 1998 Viswanathan et al., 2010) suggests that listeners are likely to recover the intended category by assigning ambiguity in an acoustic signal to coarticulation. Second, listeners are to deal with ambiguity and compensate for coarticulation or absence of cues ( Mann, 1980). Although researchers largely agree that the weight of a cue is a function of its reliability to predict a category (e.g., Nearey, 1990 Toscano & McMurray, 2010, among others), particular mechanisms of cue weighting in categorization are a matter of debate. Since each cue provides an estimate of the relevant part of the acoustic signal, listeners assign some weight or importance to each cue in order to get an accurate estimate of the whole category. First, they are to parse an acoustic signal ( Fowler, 1984 Gow, 2003) and attribute combined acoustic cues to sources, such as gestures ( Fowler, 1984 Fowler & Brown, 2000) or phonetic features ( Gow, 2003 Cole et al., 2010). In the process of categorization, listeners usually have to deal with two issues. Features are typically encoded by several cues, and cues are often used to encode more than one feature ( Repp, 1983 Lisker, 1986 Nearey, 1989). Mapping acoustic cues to phonological categories is an important and complicated area in phonology and speech perception because links between cues and features are usually multidimensional. The findings suggest that the model with phonological compensation performed most similar to human listeners both in terms of accuracy rate and error pattern. The listeners’ results were used to evaluate three categorization models to predict the intended category of a coronal stop: a model with unweighted and unadjusted cues, a model with weighted cues compensating for phonetic context, and a model with weighted cues compensating for the voicing and emphasis contrasts. ![]() The perception experiment revealed that listeners were able to categorize ambiguous tokens correctly and compensate for phonological contrasts. VOT was most relevant for voicing, but F2 was mostly associated with emphasis. Each cue influences production of coronal stops while their relevance to phonological contrasts varies. The contrast is also maintained in spectral cues. The analysis of the acoustic data collected from eight native speakers of the Qatari dialect showed that the three stops form three distinct modes on the VOT scale: is (pre)voiced, voiceless is aspirated, and emphatic is voiceless unaspirated. The current study investigates multiple acoustic cues–voice onset time (VOT), spectral center of gravity (SCG) of burst, pitch (F0), and frequencies of the first (F1) and second (F2) formants at vowel onset-associated with phonological contrasts of voicing and emphasis in production of Arabic coronal stops. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |