Defining Fundamental Beats in Linguistic Models

insights will be reviewed in the following section2.2.4.

Figure 2.9: The distinction between tatum and tactus in musical rhythm is analogous to the one between mora and syllable in language rhythm. Mora and tatum define the minimum quantity units possible, while the tactus and syllable define the beat. The number of tatums or mora contained in a beat defines the beat’s strength or weight. Naturally, music has more degrees of freedom concerning the number of tatums contained in a beat, while in language, the number of morae contained in a syllable is —language dependently— restricted to one, two or three.

consisting of the final consonant(s) (cf. Figure 2.10, van der Hulst (1984); Selkirk (1982)). The notion of syllable weightwas introduced into phonological models for several reasons. . In so-called quantity-sensitive languages, syllables having com-paratively much weight attract phonological accents. Besides, syllable structure constrains type, number and order of the phonological segments contained in it, e.g. the syllable nucleus consists of the segment(s) with the highest sonority, i.e.

vowels or syllabified consonants like nasals or laterals. Clements(1990) defends the position that a segment’s level of sonority goes hand in hand with its perceptual prominence. This connection between phonological sonority and perceptual promi-nence fits in nicely with the p-center concept introduced earlier, since it is useful to assume that the perceptual onset of a beat is perceptually salient. Due to the fact that the syllable rhyme determines the perceptual weight and marks the beginning of its most salient part, phonologists introduced a hierarchical structure to a syl-lable with a dominant sylsyl-lable rhyme and nucleus (cf. Figure 2.10). This concept again goes hand in hand with another phonological prosodic unit introduced ear-lier, the mora, since the number of morae in a syllable is defined by the structure of the syllable rhyme (cf. Figure 2.11). Usually, an open syllable with a short vowel is regarded as alight syllable while a syllable with a long vowel is regarded as heavy.

Closed syllables can be either light or heavy, depending on language specific con-straints, i.e. in some languages, closed syllables may be heavy, in other languages, their weight may depend only on vowel quantity and the coda may be irrelevant for syllable weight. In some languages, even “superheavy” syllables do occur. A light syllable corresponds to one mora, a heavy one to two, a superheavy one to three morae. However, the lacking impact of the onset on syllable weight appears to be a universal phenomenon⁹. The segment’s sonority seems to have an impact on the number of morae contained in a syllable as well: Zec(1995) showed that in some languages, short-vowel syllables closed by sonorants are heavy, while short-vowel syllables closed by obstruents are light. It is obvious that all of the syllable related effects mentioned above have an impact on the respective languages’ rhythm.

Figure 2.10: In standard phonological approaches, a syllable is hierachically structured, consisting of theonset, the initial consonant(s), anucleus, consisting of the vowel or most sonorous segment and thecoda, consisting of the final consonant(s). Nucleus and coda comprise the syllablerhyme(orrime) which determines the syllable weight.

The insight that the syllable makes up a phonologically relevant and indepen-dent level of phonological descriptions has —among other reasons— lead to the development of the so-called Nonlinear Phonologies in the 1980s. The most rel-evant phonological theory for the description of rhythmic phenomena has become the theory of (Autosegmental-)Metrical Phonology, e.g.Liberman and Prince(1977),

9However, seeMengel(2000) for a phonetic study that found a small effect of onset complexity on perceptual prominence in German.

Figure 2.11: Syllable structure defines whether a syllable is light, heavy or even superheavy. The connections between syllable structure and syllable weight are language dependent, but syllable weight and number of morae contained in a syllable stand in a linear relationship. I.e. a light syllable comprises one mora, a heavy one two and a superheavy one three morae.

Selkirk (1984), Hayes (1994). In these phonological theories, or sometimes rather, descriptive models, the various descriptive levels, such as syllable, segments, tones etc. are notated on more or less independenttiers. Between those tiers exist associa-tion linesindicating temporal overlap in the phonetic realization, e.g. segments that are associated with a syllable temporally co-occur. If association lines are deleted or changed, phonological processes like reduction, elision etc. can be modelled.

In nonlinear descriptions, the mora plays an important role as well. Moraic se-quences form the lowest tier within theprosodic hierarchy. There exists no one-to-one-mapping between different tiers, i.e. a mora may comprise one or more segments and one segment may be even associated with two morae (Cohn(2003)). It is impor-tant to understand that the moraic structure is independent of the so-calledskeletal tier, which organizes the linear order of segments. We have discussed above (cf. sec-tion2.1.2), that the perception of linear order does not imply a cognitive estimation of the pertaining durations. Thus, the cognitive processing of segmental order — illustrated by the skeletal tier in phonological models — is largely independent of the estimation of phonological durations or phonological weight. The phonological models imply that the mora is the smallest meaningful quantitative unit in speech.

In line with this reasoning,Goedemans(1998) argues that syllable weight (or num-ber of morae) is a better phonological equivalent ofperceptualsyllable duration than the number of segments, based on findings that the JND is much higher in onset segments than in rhyme segments.

One could argue that the assumption of the mora as minimal timing unit does challenge the idea of equating p-centers and syllable rhymes as fundamental psy-cholinguistic rhythmical entitites in speech. Such a view receives further support from so-called mora timing languages (c.f. section 1.1). We know from these lan-guages (e.g. Japanese) that their entire poetry is organized based on morae. How-ever, since the p-center is defined relative to the syllable, the mora does not have a 1:1-correspondance to our previously defined fundamental beats: One syllable may consist of one or two morae. Obviously, a speaker of Japanese may “count” two fundamental rhythmical entities (=beats?) when listening to a bimoraic syllable.

This superficial contradiction can be solved if we take into account the notion of the tatum. Language specifically, listeners inner clocks “tick” more fine-grained in listeners of mora timed languages (also see Sagisaka(1999), quoted afterWarner and Arai (2001a)). Thus, their windows of temporal resolution may be more fine-grained as well and react more sensitive if a syllable is significantly longer than normally the case¹⁰. Thus, a listener of a mora-timed language may start a new window of temporal integration during a bimoraic syllable, thus counting one more fundamental event without this being a beat in the traditional sense. Findings by Kato et al. (1997) suggest that the psychoacoustic boundary between vocalic and consonantal portions across morae play the most important role for opening such a new window of temporal integration, e.g. the area of strongest syllable internal spectral change. Thus, an alternative approach towards explaining the relationship between morae and beats/syllables might be that in mora timed languages, beats are not restricted to syllable rhymes, but can also be realized as consonants. The syllable itself may simply play a less central role as a perceptual unit, but remains to have an impact on segmental durations (Campbell(1999);Warner and Arai(2001b)).

10For further evidence concerning this point, refer to section2.3.1. Here, it will be shown that lis-tener’s temporal resolution becomed more fine grained the longer they listen to a sequence of stimuli of equal length. Thus, a language which has less variability due to less phonotactic complexity make listeners more sensitive of temporal distinctions.

It is likely that the perception of morae as fundamental rhythmical units are expli-cable by a combinatory effect of the higher perceptual temporal resolution and the classification of syllable final consonants as additional minimal rhythmical units or additional rhymes.

Contrary to mora timed languages, listener of a syllable or stress timed lan-guage may have more flexible windows of temporal integration. Thus, listeners of such languages will perceive the bimoraic syllable as being longer, more prominent and having more phonological weight, but will not distribute the bimoraic syllable across two windows of temporal integration. The listener perceives its weight, but does not count this weight as two separate temporal entities. In musical terminol-ogy, we might say that a listener of a stress timed language perceived a quarter note where the listener of the mora timed language perceived two eighth notes. Sum-ming up, we conclude that in mora timed languages, the windows of temporal in-tegration are less flexible. This can be explained with the relatively uniform syllable structure exposing listeners to more similar and overall shorter syllable durations (Dauer(1983)).

Still, speakers of stress timed languages may use syllable weight for rhythmical structuring, thus making an indirect use of the mora. In these languages, moraic weight may be a good indicator to locate rhythmical accents forming rhythmical groups (cf. section2.3.2).

It is likely that the connection between perceptual processing and the different levels of phonological structure behaves differently, depending on the respective language.