Rhythm Types and the Speech of Working-Class Youth in a Banlieue of Paris 97e.g., Valdman 1993), one could hypothesize that contact with Semitic lan-
guages of north-west Africa, languages with a strong tendency towards vowel
reduction, could alter some of these well-known characteristics of French.
The goal of this paper is to measure these effects empirically.
1.4 Empirical measurements of speech rhythm
Empirical studies of speech rhythm have a tumultuous history marked by
attempts at ¿ nding the best acoustic phonetic measures, allowing the classi¿ -
cation of languages in distinct rhythm types. The most recent approaches to
rhythmic typology have focused on perception. Psycholinguists observed that
young infants could discriminate between their mother tongue and another
language before even developing the ability to segment speech. Infants’ dis-
crimination patterns closely matched dichotomous distinctions proposed
earlier between so-called syllable-timed and stressed-timed languages. The
former were characterized by syllables that “tend to come at more-or-less
evenly recurring intervals so that, as a result, phrases with extra-syllables
take proportionately more time” (Pike 1945: 35), while the latter were thought
to display uniform spacing of metrically strong, accented syllables. Mora-
timing, with Japanese as the most well-known representative of this third
rhythm class, was later added to this dichotomy.
Approaching the issue from its perceptual underpinnings, recent psycho-
linguistic experiments have shown that infants can successfully discriminate
between a stress-timed and a mora-timed language, e.g., English and Japa-
nese, but are less able to discriminate between two stressed-timed languages,
such as English and Dutch (Mehler et al. 1996, Nazzi, Bertoncini, and Mehler
1998). Taking these studies as their input, Ramus, Nespor, and Mehler (1999)
(henceforth, RNM) hypothesized that infants’ perception of rhythm types is
centered on the alternation of vocalic intervals of variable length with “noisy”
portions of the speech signal. However, rather than computing a raw measure
of sonority derived from spectral information, RNM resorted to identifying
and then collapsing into longer stretches of vocalic and consonantal intervals
discrete phonological units, i.e., vowels and consonants.^21 They claimed that
“a simple segmentation of speech into consonants and vowels” is all that is
needed to arrive at language-speci¿ c auditory patterns reminiscent of the syl-
lable- vs. stress-timed distinction that forms the basis of infants’ successful
discrimination between various languages. Rhythm types were conceived as
a continuum. It was hypothesized that languages with predominantly simple
CV-type syllable structure, absence of vowel reduction, and relatively little