339. Tonoexodus 2

In a recent comment, Maju called our attention to a very interesting new article in which a possible connection between tonal and non-tonal languages is tested and discussed: Real-Time Correlates of Phonological Quantity Reveal Unity of Tonal and Non-Tonal Languages, by Juhani Jarvikivi1, Martti Vainio and Daniel Aalto. (How does he find this stuff?) The authors point specifically to the influence of tone in certain non-tonal "quantity languages," (i.e. languages in which differences in syllable length have phonemic import), citing evidence suggesting

that in non-tonal quantity languages such as, Estonian,
Finnish, Japanese, and Serbo-Croatian, tonal differences affect
speakers’ judgments of vowel length, in so far
as the available evidence can be taken to suggest that the speakers
of these languages tend to categorize syllables or words as long
more often than short when the target syllable has a falling rather
than a level tone. (p.2)

To further test this hypothesis, the authors performed experiments with Finnish speaking subjects, to determine the effects of certain tonal configurations on their perception of lexical difference. For them, the results of these experiments "are clear: whether the first syllable has a falling or a level (high) tone is a robust online cue to . . . lexical identity in Finnish" (p. 4).

In a Discussion section, they elaborate on the meaning of their results:

In contrast to the usual assumption that there is a clear-cut
conceptual distinction between tone and non-tonal quantity
languages, we have put forth the idea that, cognitively, these two phonological systems could perhaps be seen as two variants of . . . the same underlying mechanisms. In addition to reviewing the available evidence that we thought would point this way, we carried out two experiments investigating whether pitch information would affect perception of length and thus word recognition in a language with a par excellence example of a quantity-based lexical-phonological system. The answer based on the two experiments was a clear affirmative (p.4).

In short, "our results showed that pitch information is an important co-index of the quantity opposition in Finnish." On this basis, they make a rather startling claim: "Consequently, . . . our results imply that in terms of the production and perception mechanisms, pitch in Finnish is probably in all respects like pitch in any prototypical tone language, e.g., Mandarin Chinese" (p. 5).

In more general terms,

we would like to argue that rather than a
discrete categorical classification of languages into tone languages
and non-tone languages, a more fine-grained account is needed
that takes into account the extent to which (in this case) pitch
information is actually used to distinguish phonological categories
in processing. This would not only sharpen our criteria of tone
languages, but would also provide a more realistic, more refined,
explanandum for studies of linguistic evolution. (p. 6)

Moreover,

With regard to tonogenesis - at least in some cases - it
may be that tone in the phonetic sense has been present all along
and only surfaces phonologically when other linguistic factors force
the change. Importantly, our results suggest that there is no
unidirectional link from perceptual sensitivity to pitch information
to the emergence of a tone language. (p. 6)

The authors never go far as to question the tonogenesis dogma per se, but their work certainly raises many questions regarding its validity as a "unidirectional link" in linguistic evolution.

What I find especially intriguing in this research is the fact that two of the three European languages they cite as typical "quantity languages," Finnish and Estonian, are Uralic languages, thus among the very few non-Indoeuropean languages on that continent. Since the establishment of Indoeuropean throughout almost all of Europe appears to be a relatively recent phenomenon, and Uralic is widespread among languages spoken by indigenous peoples scattered through vast regions of northern Europe and Asia, it seems likely that the Uralic complex could predate Indoeuropean and thus might represent an earlier stage of lingustic evolution.

Indeed, according to a very interesting paper by Mario Alinei (Interdisciplinary and linguistic evidence for Palaeolithic continuity of Indo-European, Uralic and Altaic populations in Eurasia, with an excursus on Slavic ethnogenesis, 2003), a new theory of Uralic origins

was advanced about thirty years ago and is now universally recognized by linguists as well as archaeologists: it is called the Uralic Continuity Theory (UCT) and claims an
uninterrupted continuity of Uralic populations and languages from [the] Paleolithic (Meinander 1973, Nuñez 1987, 1989, 1996, 1997, 1998)

According to this theory, which historically represents the first claim of uninterrupted continuity of a European people from [the] Paleolithic, Uralic people must belong to the populations of Homo sapiens sapiens coming from Africa, who occupied mid-eastern Europe in Paleolithic glacial times . . . and followed the retreating icecap in [the] Mesolithic, eventually settling in their present territories . . . (pp. 12-13)

I don't want to pursue my speculations too far, since my knowledge of historical linguistics is very limited and I might well be on the wrong track entirely. Nor are such speculations really necessary with regard to the overall argument I've been presenting over the last few posts. Nevertheless, I do find the link between tonal languages and non-tonal quantity languages very interesting and definitely worthy of further investigation. As I wrote in my response to Maju's comment,

If the earliest language was indeed tonal, as I strongly suspect (due to the saturation of tone languages in Africa, and the lack of evidence for "tonogenesis" on that continent), then the association these linguists found between tone and quantity could represent a first step in an evolution from tonal to non-tonal language. . . . I'm now wondering whether Uralic languages such as Finnish, Estonian and Saami were among the "native European" language families displaced by the advent of Indoeuropean. If so, then the close association with tone language demonstrated in this paper would make a great deal of sense. . . The evolution from a tone to a quantity language would have been the exact opposite of the "tonogenesis" so confidently assumed by so many linguists.

338. Tonoexodus

As I see it, there is little question that the earliest languages must have been tone languages. Since modern humans are almost universally thought to have originated in Africa, and since the great majority of African languages are tonal, it would be extremely difficult to explain how an originary non-tonal language could have produced so many tone languages on the continent of its birth. The hypothesis is reinforced by the fact that "there are no documented cases of tonogenesis in Africa, despite the wide variety of languages . . . and the widespread presence of tone on the continent" (George Tucker Childs, An Introduction to African Languages, 2003, p. 86.).

Since linguists are in agreement that tonogenesis represents some sort of universal process through which all tonal languages are generated from non-tonal ones, the abundance of tone languages in Africa, plus the lack of evidence for tonogenesis anywhere on that continent, should represent something of an embarrassment -- but apparently not. From what I've read in the surprisingly extensive literature on tonogenesis (not to mention many other topics in linguistics), linguists seem much too preoccupied with the discovery of universally valid principles and far too little concerned with the messy contingencies of history, as reflected in the worldwide distribution of the traits they study (the WALS project being a notable, and very welcome, exception).

Given the preponderance of tone language in Africa, it seems likely that the original Out-of-Africa migrants must have also spoken a tone language. And since this is generally understood as the founding group, both genetically and culturally, for all peoples outside of Africa, it seems likely that non-tonal languages could only have arisen via a process that must be regarded as the reverse of tonogenesis, i.e.: tonoexodus.

When I "coined" this term in a tongue-in-cheek comment on the previous post, I wasn't aware that it was already in circulation. And, yes, some linguists have considered the possibility of what they too have named (with a straight face, apparently) "tonoexodus":

Tone systems are not static. A language can acquire tones and then increase the complexity of this tone system but it can also decrease the number of its tones and ultimately become non-tonal. These two processes, acquisition and recession of tones, have been termed tonogenesis [Matisoff 1970, 1973) and tonoexodus [Lea 1973). Cases of tonoexodus are rare and it is not clear what the intermediate historical stages between the tonal and non-tonal stages are. (CONSONANT TYPES, VOWEL HEIGHT AND TONE IN YORUBA, by Jean-Marie Rombert, 1977, p. 174.)

I suspect that "cases of tonoexodus are rare" only because 1. linguists aren't looking for them; and 2. they tend to focus on very specific processes within specific languages, rather than taking the big picture into account. I've seen countless studies of "tonogenesis" as it appears to have developed in a single language, but have noticed not one study of the topic as applied to the worldwide distribution of tone.

But the (apparently revolutionary) notion that tone language came first, is only part of the story. Because if the first language was a tone language, then it seems only logical to go a step farther to consider whether it might have consisted exclusively of tones. Or, to be more accurate, specific tones presented in specific rhythms, which also happens to be a way of defining music. In a comment on the previous post, Marnie reminds us that a great deal of content in a great many African languages can be conveyed by the "talking drum," limited exclusively to differences of tone and rhythm. She asks the very sensible question, "is it possible that pitch and rhythm developed together in our earliest languages?"

In response to my previous post, I received an email from a very perceptive reader, Alex Petrov, who provided a link to this extremely interesting Wikipedia article on Whistled Language. I had always assumed that so-called whistled "languages" were merely elaborate signalling systems, but there is clearly more to it than that:

A whistled language is a system of whistled communication which allows fluent whistlers to transmit and comprehend a potentially unlimited number of messages over long distances. Whistled languages are different in this respect from the restricted codes sometimes used by herders or animal trainers to transmit simple messages or instructions. Generally, whistled languages emulate the tones or vowel formants of a natural spoken language, as well as aspects of its intonation and prosody, so that trained listeners who speak that language can understand the encoded message.

Whistled language is rare compared to spoken language, but it is found in cultures around the world. It is especially common in tone languages where the whistled tones transmit the tones of the syllables (tone melodies of the words). This might be because in tone languages the tone melody carries more of the "functional load" of communication while non-tonal phonology carries proportionally less. The genesis of a whistled language has never been recorded in either case and has not yet received much productive study.

Especially interesting is the observation that "In continental Africa, speech may be conveyed by a whistle or other musical instrument, most famously the "talking drums . . . As two people approach each other, one may even switch from whistled to spoken speech in mid-sentence." If so much in so many African tone languages can be communicated by tone and rhythm alone, then it is only logical to wonder whether any of the other features of such languages are necessary -- and whether their existence could be undersood as the initial stages of a progression from a language of pure tones to a tonal language, and from there to a non-tonal language -- i.e.: tonoexodus.

337. Did Music Originate as a Behavioral Adaptation? -- 8: Speech

Most linguists have managed to convince themselves that tone languages must have derived from non-tone languages, under the assumption that the earliest languages must have been non-tonal. The process through which a non-tone language evolves into a tonal one is known as "tonogenesis." Very strangely, however, almost all the research on tonogenesis has been centered in either East Asia or the Americas. Africa, the continent with the largest number of tone languages by far, has been all but ignored -- and for good reason, apparently:

What is quite surprising . . . is that there are no documented cases of tonogenesis in Africa, despite the wide variety of languages . . . and the widespread presence of tone on the continent. (George Tucker Childs, An Introduction to African Languages, 2003, p. 86.)

Since almost every single language in sub-Saharan Africa is tonal, "widespread presence" is something of an understatement. To illustrate, let's take a look at the world map of tone languages produced by WALS, the World Atlas of Language Structures:

The red and pink dots represent tone languages, the white dots non-tone languages. As is clearly evident, Sub-Saharan Africa is simply saturated with tone languages, with only two or three exceptions represented in the enormous WALS sample. It's interesting to note that a similar degree of tonal saturation is depicted for Southeast Asia and Melanesia. I've discussed the possible meaning of this very odd distribution in an earlier post, but it need not concern us here.

What does concern us at this point is the overwhelming genetic and archaeological evidence that's developed over the last 20 or 30 years pointing to Sub-Saharan Africa as the locus for the development of "modern" humans (homo sapiens sapiens), who are thought to have migrated from there to the rest of the world roughly 60,000 to 80,000 years ago. Since most historical linguists now agree that all human languages must have had a common ancestor, then, if the Out of Africa model is correct, that ancestor could only have originated in Africa. And since just about every language in Africa (including Khoisan, considered by many to be the oldest surviving language) is a tone language, then there is clearly something very wrong with the widespread assumption that the earliest languages must have been non-tonal, and linguistic tone could only have been produced via "tonogenesis."

Which returns us to the experiments by Diana Deutsch (see previous posts), and the surprisingly strong correlations she found between tone language and absolute pitch. Unlike some of the other common features of language and music, such as interactivity, cooperation, phrasing, etc., the use of discrete pitches is the only one generally regarded as uniquely musical. And the puzzle we've been considering, of how such tones could have developed, and, more important in the context of the present discussion, what sort of adaptational advantage they might have posed, can now be seen in an entirely new light.

Based on the evidence presented above, the following sequence may now be considered:

1. Interactive "hooted" vocalizations of early primates and pre-humans, along the lines of the "duetting" and "chorusing" of certain contemporary ape and gibbon populations. The adaptational advantage of such behavior would most likely be the facilitation of both long distance communication and cooperation.

2. The development from the above, among early humans, of precisely pitched vocalizations. Among the various means by which this may have come about, one stands out as particularly suggestive as far as adaptation is concerned. Since many birds sing using discrete pitches, there would have been an advantage for humans in learning how to imitate bird songs as a lure. This could have been accomplished through the morphing of pre-human "hooting" into precisely pitched yodeling. Since yodeling involves a process akin to the "overblowing" of wind instruments (such as pipes, flutes, etc.) to produce discrete overtones, it might have been the simplest means by which humans would have become aware of certain basic pitch relationships. Another possibility might have been the discovery that simple reed pipes or hollow bones could be blown into in such a way as to produce discrete pitches that in many cases could be used as bird-call imitations. Since each reed or bone could only play a single note, it would require close cooperation on the part of a group to imitate multi-pitched bird songs. Reed ensembles of this type are still widely found in Africa and elsewhere among indigenous peoples, and such performances are in many cases associated with birds and their calls. Vocal ensembles organized along similar lines may have developed either independently or in imitation of the wind ensembles.

3. Since bird songs are precisely pitched, hunters with absolute pitch would have been more effective than those without it, giving a selective advantage to those with absolute pitch.

4. On the basis of the above, admittedly speculative, sequence, it's not difficult to see how both vocalizing and playing with discrete pitches could have led to the development of a language of sorts, based exclusively on tonal relations. For one thing, each such musical sequence would have symbolized a specific species of bird. For another, it's possible to see how, for those with perfect pitch, each pitch could have been perceived as an easily identified semiotic "module," very close, in fact, to a linguistic phoneme, which it could have anticipated.

5. If the earliest "language" consisted essentially of discrete pitches, then we can see how, for early humans, the development of musical awareness would have had a powerful adaptational advantage (now lost, of course). This would also explain the widespread presence of tone languages in the continent where early humans developed, since the use of tonal phonemes would have persisted even after non-tonal elements were added.

The above is highly speculative of course. A great deal depends on whether or not Deutsch's results, based on research among East Asians, can be replicated with African subjects.

336. Did Music Originate as a Behavioral Adaptation? -- 7: Speech

In her recent Scientific American article, Speaking in Tones, psychologist Diana Deutsch describes some remarkable research, by herself and others, revealing some unexpected and very exciting links between speech and music. For example, despite many years in which it was assumed they were controlled by two completely different regions of the brain, "Psychologists, linguists and neuroscientists have recently changed their tune . . . as sophisticated nueroimaging techniques have helped amass evidence that the brain areas governing music and language overlap." The two regions are so interconnected that "an awareness of music is critical to a baby's language development and even helps to cement the bond between infant and mother" (p. 37).

This overlap makes sense, because language and music have a lot in common. They are both governed by a grammar, in which basic elements are organized hierarchically into sequences according to established rules. In language, words combine to form phrases, which join to form larger phrases, which in turn combine to make sentences. Similarly, in music, notes combine to form phrases, which connect to form larger phrases, and so on. (pp. 38-39)

I'm a bit skeptical regarding the many examples of baby-mother interaction she provides, because, like so many others in her field, and in cognitive science generally, she assumes that all babies and mothers interact similarly, based on research typically limited to American and European subjects. Before attempting to universalize such evidence, it's important to compare it with evidence from non-Western societies, as well as various indigenous groups from a wide range of different world areas.

The above reservations do not apply to her most remarkable and exciting results, regarding a completely unexpected and indeed very surprising correlation between absolute (or "perfect") pitch and tone language. She made the astonishing discovery that among students who had received musical training by the age of five, fluent speakers of Mandarin, a tone language, were far more likely to have absolute pitch than a comparable group of students who grew up with English or some other nontone language. We're talking a huge difference, of 92% of "very fluent tone language speakers," as opposed to only 8% of English speakers. To determine whether the correlation were primarily genetic rather than linguistic, she tested East Asian students who grew up speaking a non-tone language and discovered that they too scored only about 8%. The correlation seems definitely associated with tone language rather than genetic inheritance.

Another important discovery concerns the pitch sensitivity of tone language speakers generally. It's always been assumed that the pitches of tone language are relative and not absolute, yet Deutsch learned that

not only were Vietnamese and Mandarin speakers very sensitive to the pitches that they hear, but they can produce words at a consistent absolute pitch. . . We found that their pitches were remarkably consistent: when compared across days, half of the participants showed pitch differences of less than half a semitone (p. 42).

In the next post, I'll explain why I attach such importance to these results.