Answers are grounded in this paper and cite their sources.
An experiment was devised to test whether compensation for coarticulation could be motivated by nonspeech for which gestural recovery is impossible. Subjects were presented with CV stimuli formed by concatenating an /s ~ ∫/ continuum fricative with an /i/ or /o/ from three types of synthesized vocalic nuclei: full spectral vowels (Set A), vowels with F2 but no other formants (B), and pure sine tones at F2 (C). F2 was chosen as the common parameter between sets because extremely high or low F2 is enough information for English speakers to judge vowel roundedness (and rounding lowers centroid frequency, a key difference between /s/ and /∫/). Comparing fricative identification boundaries between the vowels within each set, compensation occurred reliably in Sets A (t = 4.6, p < 0.01) and B (t = 3.7, p < 0.01) but not C. Comparing conditions against each other, Set A showed significantly higher vowel-triggered boundary shift than B (F = 8.18, p < 0.05). Results support the conclusion that relevant acoustic cues might not trigger compensation if they cannot be associated with speech; additionally, sounds reminiscent of speech do not generate as strong an effect as sounds firmly identifiable as speech.
No takes yet. Share an insight, caveat, or question.
Gregory Finley (2012) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: