What's the evidence for and against isochrony?

Question

The question

What evidence is currently known that favors or disfavors the hypothesis that a regular beat of some kind—that is, an “isochrony”—plays some important role in languages?

I've run across some claims that there are three main ways that languages have regular beats:

• an equal time per syllable,

• an equal time between stresses,

• and an equal time per mora (a beat that takes less time than a syllable, where a typical syllable takes a small-integer number of morae: one, two, or three).

On that hypothesis, different languages or different speakers may depend on different "isochronies" to different degrees, but people rely on some kind of isochrony for comprehension of most or perhaps all languages. My question is: what evidence is there about whether this hypothesis is true—do (many) languages depend on some sort of regular beat at the same order of magnitude as the length of a word or syllable?

Not categorization

I am not asking whether languages can be unambiguously categorized as stress-timed, syllable-timed, or mora-timed. This question is about whether a hypothesized mechanism of language exists, not about "typology". If isochrony exists, it might have multiple forms, which might or might not provide a basis for defining categories into which some languages could fit, but I'm not asking about ways to categorize languages. I'm including this paragraph because nearly all the writing about isochrony that I've found so far is arguments about whether languages can be unambiguously categorized in this manner—which is not what I'm interested in. It's OK with me if plenty of languages make use of two or three (or more) isochronies simultaneously. I just want to know if isochrony really matters at all.

Some information I've found so far

One argument that different kinds of isochrony exist is that the variance in syllable lengths is greater in English, which supposedly depends more on regular stress-to-stress intervals, than in Spanish, which supposedly depends more on regular syllable-to-syllable intervals. However, as Mark Liberman says in this blog post (which is mostly about the categorization argument), that doesn't imply anything about isochrony, because syllables in English have a much greater range of complexity than syllables in Spanish (due to the variety of consonant clusters that English allows before and after the vowel). Liberman says that people have been trying for a long time to save the isochrony hypothesis by postulating various ways to measure the assumed regularity (for examples, runs of successive syllables of the same length), and claims that no one has ever found an objective measure in the sound of speech itself. I don't know if that's true, though.

A weird argument is that the perception of isochrony disagrees with the timings actually found in the sound signal, and that people adjust the real timings of syllables and stresses to create the perception of isochrony. There must be some interesting evidence for or against this, but I haven't found it yet.

This paper by Peter Roach attacks the categorization theory, but also has a note from the author saying that it's out of date.

So, what evidence is up to date?

Answer 1

Does isochrony of syllables or feet exist in any language?

As user6726 pointed out, answering this question is impossible as the vast majority of the currently spoken languages are undocumented. So let us restrict the question to languages for which it can be (and has been) answered. Commonly cited examples of stress-timed languages (where feet were thought to have the same durations) are English and German, and, of syllable-timed languages, Spanish and French. Dauer (1983), for example, showed that in some of the usual suspects, syllables and feet do not have equal durations, not even relatively speaking.

Does this mean the concept of syllable-timed and stress-timed languages should be abandoned?

It appears that listeners are relatively reliable in identifying so-called syllable-timed and stress-timed languages. Specifically, even infants can distinguish syllable-timed from stress-timed languages even if only rhythmic cues are provided (i.e. other sources of information, such as intonation, are removed) (Nazzi et al. 1998; Ramus and Mehler 1999; Ramus et al. 2003).

How can we then distinguish syllable-timed from stress-timed languages?

In other words, what is the acoustic reality of differences in speech rhythm? A number of authors have suggested so-called rhythm metrics, which measure for example the degree of variation in the duration of vocalic intervals (sequences of vowels uninterrupted by consonants). One such metric is VarcoV, the standard deviation of the durations of vocalic intervals divided by the mean duration of vocalic intervals. Another is called nPVI-V (the normalised pairwise variability index of vocalic intervals), which is the mean difference between the durations of successive vocalic intervals, divided by the sum of their durations. You get the picture ;)

Basically syllable-timed languages have vocalic intervals relatively similar in duration to each other, while in stress-timed languages there is greater variation in the durations of vocalic intervals. One important reason for this is that stress-timed languages tend to have vowel reduction in unstressed syllables, unlike syllable-timed languages.

Both these rhythm metrics are normalised for speech rate, and a comparison of a larger number of such metrics showed that these two are the most reliable ones (White and Mattys 2007a,b). If you want a relatively non-technical introduction into how these rhythm metrics work, and why they were proposed, I would suggest the book chapter You got the beat: Rhythm and timing (disclaimer: This was written by me, but I'm providing this link because most of the literature on the topic can be fairly technical).

P.S.: Syllable-timed languages vs. stress-timed languages?

The earlier literature on the topic assumed there are rhythm classes, i.e. discreet sets of languages, either syllable-timed or stress-timed (and mora-timing fit in a bit awkwardly as a third group). The more recent literature has abandoned this view and often assumes a continuum with a prototypical stress-timed and a prototypical syllable-timed pole, which are not actually reached by any languages (see references above).

References

• Dauer, R. M. (1983). Stress-timing and syllable-timing reanalyzed. Journal of Phonetics 11, 51–62.
• Nazzi, Thierry, Josiane Bertoncini and Jacques Mehler (1998). Language discrimination by newborns: Towards an understanding of the role of rhythm.​​​​ Journal of Experimental Psychology: Human Perception and Performance 24, 756–66.
• Ramus, Franck and Jacques Mehler (1999). Language identification with suprasegmental cues: A study based on speech resynthesis. Journal of the Acoustical Society of America 105.1, 512–521.
• Ramus, Franck, Emmanuel Dupoux and Jacques Mehler (2003). The psychological reality of rhythm classes: Perceptual studies. Proceedings of the 15th International Congress of Phonetic Sciences (ICPhS 2003). Ed. by Daniel Recasens, Maria-Josep Sole and Joaquin Romero. Barcelona: Universitat Autonoma de Barcelona, 337–342.
• White, Laurence and Sven L. Mattys (2007a). Calibrating rhythm: First language and second language studies. Journal of Phonetics 35.4, 501–522.
• — (2007b). Rhythmic typology and variation in first and second languages. Segmental and Prosodic Issues in Romance Phonology 282, 237–257.

Answer 2

My "answer" is more of a comment, but it's too long for a comment. I see at least three distinct questions here. One is whether any language has isochrony; the second is whether every languages has some form of isochrony; the third is whether many languages have isochrony. Taken in reverse order, I would say that the third question is unanswerable for the foreseeable future. The main problem there is that I don't know what number is in the ballpark of "many". If I had 3 purported cases, that wouldn't normally be "many". However, there isn't a huge amount of in-depth phonetic research on a variety of languages, and this is a question that requires a lot of work. There are studies like Lehtonen 1970 Aspects of quantity in Standard Finnish which give a lot of detail of duration in Finnish, but I can't say that there is a single comparable phonetic study of any of the 2,000 languages of Africa, or 1,200 Austronesian languages.

For the first question, you may have better luck getting an answer. Estonian (whose durational properties have been extensively studied: Lehiste 2003, "Prosodic change in progress: from quantity language to accent" may be the best "hook" article) is claimed to have "foot isochrony". However, not all feet in Estonian have exactly the same duration, regardless of segmental content. I think the logic of the argument is based on the major incorrectness of the alternative, that duration within the foot is a simple additive function of a constant duration value for each segment in the foot. So IMO, Estonian is your best best for isochrony, and if you find it wanting, that means something.

Answer 3

Cauldwell's paper (Functional Irrythmicality of English) argues that naturally occurring speech is not rhythmic, offering data to support this view.