Consonant categorization exhibits a graded influence of surrounding spectral context

Abstract

When spectral properties differ across successive sounds, this difference is perceptually magnified, resulting in spectral contrast effects (SCEs). Recently, Stilp, Anderson, and Winn [(2015) J. Acoust. Soc. Am. 137(6), 3466–3476] revealed that SCEs are graded: more prominent spectral peaks in preceding sounds produced larger SCEs (i.e., category boundary shifts) in categorization of subsequent vowels. Here, a similar relationship between spectral context and SCEs was replicated in categorization of voiced stop consonants. By generalizing this relationship across consonants and vowels, different spectral cues, and different frequency regions, acute and graded sensitivity to spectral context appears to be pervasive in speech perception.

Musical instrument categorization is highly sensitive to spectral properties of earlier sounds

“…Not only were SCEs observed in nearly every condition tested, but their magnitudes varied continuously: As larger filter gains were tested, spectral differences between the context sentence and target vowel progressively increased, and SCE magnitudes increased in kind (see also Stilp & Alexander, 2016). This relationship was later replicated and extended in consonant categorization (Stilp & Assgari, 2017). Each of these studies reported strong linear relationships between filter gain (which introduced spectral differences between context and target sounds) and biases in speech categorization (i.e., SCEs magnitudes), supporting acute sensitivity to context when perceiving and categorizing speech sounds.…”

“…The magnitudes of SCEs vary continuously in speech categorization Stilp & Alexander, 2016;Stilp & Assgari, 2017), but it is unknown whether the same is true for SCEs influencing musical instrument categorization (as in Stilp et al, 2010). One possibility is that continuous variation in SCE magnitudes only occurs for highly familiar stimuli (e.g., speech).…”

Assessing the relationship between talker normalization and spectral contrast effects in speech perception.

“…Modest peaks produce modest SCEs. When there is a small change from context to the target, there is a smaller shift in categorization (Stilp, Anderson & Winn, 2015;Stilp & Assgari, 2017a). Detecting a decrease in an already modest effect requires more power.…”

“…It has been established that SCEs are linear in nature: when the size of the change between the context and target increases, so does the shift in categorization Stilp & Alexander, 2016;Stilp & Assgari, 2017a). In these studies, the magnitude of the change from context to target to sentences was varied.…”

“…Mixed effects models were conducted in R (R Development Core Team, 2016) using the lme4 package (Bates et al, 2014) to analyze the results for each study. Previous work in our lab has utilized mixed effect models to analyze results from SCE experiments Stilp & Assgari, 2017a). The models conducted here were maximal following the suggestions of Barr, Levy, Scheepers and Tily (2013).…”