The role of rhythm class, speaking rate, and F0 in language discrimination

Abstract

Abstract:The division of languages into stress-, syllable-, and mora-timing is said to be supported by experiments showing that languages are discriminated only if they belong to different rhythm classes, a distinction said to be reflected in the duration and variability of consonantal and vocalic intervals (timing). The role of rhythm classes in discrimination is tested here along with the alternative that discrimination is due to speaking rate and F 0 differences and is independent of rhythm class. Five AAX experiments with English as context (AA) and Polish, Danish, Spanish, Greek, or Korean as test (X) were conducted using the sasasa transform and modifying F 0 and speaking rate, so as to compare responses to stimuli that retained the original speaking rate and F 0 of each language and stimuli that were stripped of this information (speaking rate, F 0 , or both) while retaining their timing. Discrimination was possible both across and within rhythm classes when speaking rates differed between context and test but was largely impossible once speaking rate differences were eliminated. F 0 also played a significant if less consistent role in discrimination. The changes in responses associated with speaking rate and F 0 indicate that language discrimination arises from interactions between prosodic factors and that timing contributes but little. Consequently the results cast doubt both on the ecological validity of the sasasa transform, which brings timing to the fore while eliminating F 0 modulation, and on the rhythm class typology said to be reflected in timing distinctions.

Segments, syllables and speech tempo perception

“…Several tempo perception studies report correlations of listeners' judgements and syllable rate measurements in the region of r=0.80 [4,5]. Research into rhythm perception highlights listeners' attention to syllable rate when judging whether utterances are rhythmically alike or distinct [6,7]. To date only one study has included segment rate measurements in the comparison [8].…”

Five-month-old infants’ discrimination of unfamiliar languages does not accord with "rhythm class"

“…In all cases, values for Finnish were higher or lower than both those for French and Spanish, suggesting that no simple set of timing features can account for the observed behaviour. There is some evidence pointing to the role of f0 variation in adult discrimination tasks, though secondary to that of speech rate [20]. It should also be noted that intact natural speech was used here, as in previous discrimination experiments with fivemonth olds [17], and so segmental cuessuch as the sharply contrasting vowel inventories of French and Spanishare also available.…”

“…Thus, adult discrimination experiments have exposed listeners to stimuli modified to eliminate segmental information and focus solely on timing (e.g., sasasa speech, created by replacing all vowels with /a/ and all consonants with /s/, but retaining the durational values of the original intervals). Initial results suggested that adult discrimination was also governed by categorical rhythm class distinctions [18], but this interpretation was challenged by studies showing within-class and within-language discrimination on the basis of timing cues alone [19,20]. Such results suggest thatrather than discrete categoriesadult language discrimination relies on gradient variation along prosodic dimensions, with differences in speech rate and utterance-final lengthening shown to be particularly salient [19, see also 21 for infant discrimination between accents of a familiar language].…”

Intonation Plays a Role in Language Discrimination by Infants

“…They have proven less successful, however, when a wider range of materials, speakers, and a larger set of languages are considered (e.g., Arvaniti, , ; Loukina, Kochanski, Rosner, Keane, & Shih, ; White & Mattys, ; Wiget, White, Schuppler, Grenon, Rauch & Mattys, 2010). Further, perception studies have failed to find consistent evidence for such classes (e.g., Arvaniti, ; Arvaniti & Rodriquez, ; White, Delle Luche, & Floccia, ).…”