Speech adaptation and situational optimality.

Abstract

Language serves as a primary means of human adaptation. This paper critically examines attempts to identify certain phonological aspects of language systems as particularly adaptive. These postulate that the CV (consonant plus vowel) syllable is an optimum in speech adaptation to warm climates that encourage "distal communication." These attempts are shown to be over-simplistic, motivated by a model that stresses transmission but disregards speech differentiation. An alternative analysis is proposed that recognizes that speech adaptation must negotiate between transmission and differentiation factors.

"What is decisive, in language, is achieving communication, and this is secured if, at every point in the utterance, the unit chosen is kept distinct from the ones that could have been used, in the very same context, in order to make a different message.--Andre Martinet (1962:8)

Introduction

Language represents a most important contribution to human adaptability, and the capability to use an open and productive communication system provides a significant advantage for humans over all other life forms. This paper will examine whether and how the concept of adaptation is relevant to the structural properties of speech communication. More specifically, is it possible for a particular type of language structure to serve as an optimum in speech adaptation cross-linguistically? A series of papers by Robert Munroe et al. (1996, 1999, 2000) and Melvin and Carol Ember (1999, 2000) have addressed this question by attempting to demonstrate the adaptive value of a consonant-vowel (CV) syllable structure. The latest in this series, by John Fought et al (2004), refines this approach by calculating a mean sonority score (MSS) instead of CV Score (proportion of CV syllables) for an utterance. In fact, this is not much of a change since utterances with a high MSS must also necessarily be utterances with high proportions of CV syllables (Fought et al: 34)[1]. In this paper we will simply refer to these approaches as CV studies and their dependent variable as the CV syllable.

This syllable type is found to carry the maximal sonority and sound contrast at the minimal cost of vocal effort. Based on their findings that it attains greater frequency of use at the expense of other syllable types in warmer climates, the CV syllable structure is assigned the significance of an optimal strategy in speech adaptation to "distal communication." This line of investigation exhibits both promises and problems. On the one hand it may stimulate research on patterns of speech adaptation. On the other hand, however, it employs an overly simplistic interpretation of optimality. We intend to show that assigning adaptive value to CV syllabic structures indiscriminately can be misleading and, contrary to what is concluded by Munroe st al and others, is especially problematic in distal communication.

Our argument derives primarily from the functionality of speech. The locus of speech is in communication. Its effectiveness cannot be evaluated solely in terms of speech production; consideration must also be given to its ability to ensure effective speech reception and comprehension. We will argue that the complex challenges for speech recognition and sound differentiation render the "optimality" of CV syllables situational. What is adaptive or optimal in one situation may not be adaptive or optimal in others. More importantly, the adaptive value that may appear operative for some speech component can be negated by the maladaptive results of this same component. Borrowing a term from psychology, one may say that in speech adaptation, humans show considerable "choice behavior." If this is true, then a single measure such as speech production will not necessarily pinpoint the "best" language property (Aitchison 2001:252). Nor, in this case, will it reveal the true impact of CV syllabic structures. This paper will argue that other factors must be taken into account.

Linguistic Adaptation

Language is a specifically human adaptation (Jackendoff 2002). Not only is language a product of innate genetic programs (e.g. Universal Grammar) that are peculiar to the human mind, but it also creates rich associations that produce metaphor and theory, facilitates communication, and gives humans an evolutionary advantage. In our use, "linguistic adaptation" is a part of cultural adaptation and reflects the workings of selection that occur in the interaction between the language system as an open and productive system and the environment in which human societies must survive. It is one thing to recognize the value of linguistic adaptation in general but quite another to attribute specific features of language (e.g. phonology) it to natural selection to an external environment. According to the CV studies, linguistic adaptation is subject to the same type of natural selection as biological adaptation is. This is reminiscent of Max Muller (1861) flatly asserting, "The better, the shorter, the easier forms are constantly gaining the upper hand." Contemporary linguists, however, have found evidence to the contrary. A phonological innovation, for instance, is always a source of confusion until the phonological space is rearranged to accommodate it (Labov 1972:178-80

Adaptive Value of CV Syllables

If one language uses utterances with a 39% proportion of CV syllables (or an MSS of 34.3), as in Blackfoot, whereas another employs 71% CV syllables (or an MSS of 45.2), as in Dogon (Munroe, et al 1996:78; Fought et al 2004:35-36), CV studies link this difference to the two group's "external environments"--that the Blackfoot live in a "cold" climate whereas the Dogon live in a "warm/hot" climate" (Munroe et al. 1996:63). Instead of explaining the difference in structural terms (i.e. internal processes within language, such as simplification, coarticulation, and assimilation), this approach argues that cross-linguistic variation in CV usage is, above all, environmentally sensitive. It follows that the analysis of CV distribution in the world's languages is reduced to asking one crucial question, namely, under what environmental conditions CV usage should be most adaptive and thus highly encoded in language (Munroe et al. 1996:63). CV studies call attention to the distinctive phonetic property of CV syllable that produces a maximum acoustic contrast and sonority. According to the authors of CV studies (Munroe et al 1996:62, Ember & Ember 1999:731), a high CV percentage is an optimal choice for speakers in distal communication because of its greater transmission strength. And this would be especially the case in warm and moderate climate where people converse more frequently across a distance in the open.

The Situational Nature of Optimality

There are two different optimality approaches in evolutionary studies. Under one approach, evolution is held to be a process searching for optima. This assumption is controversial (Wolfgang and Eeckman 1995:2). For one thing, there tend to be multiple solutions available for a population in the face of environmental pressure, which is true in both biological and cultural evolution. Even more problematic is the assumption about the existence of an optimum, for it assigns an intentional agency to adaptation. Gould and Lewontin (1979) are the most outspoken critics of such a notion. According to Lewontin (1987:157-158), the optimal organism does not exist; since the optimal state always involves a specific set of alternatives, it is never the optimum.

Under the second optimality approach, more consideration is given to the coexistence of alternatives and the constraints of what is optimal. Optimality, therefore, involves adopting the best strategy available from a set of alternatives to deal with the exigencies of environment. Conceived this way, it avoids many of the difficulties inherent in the more straightforward optimality approach (Dupre 1987:4). It has also found increasing use in anthropology, where optimality analysis is employed to generate explanatory hypotheses regarding cross-cultural and intra-cultural variation in divers aspects of human behavior (Smith 1987:202). It is the contention of this paper that what is optimal is relative. We will follow the second approach to optimality outlined above: Optimality is a dynamic state that favors different alternative strategies in different situations. This relative nature of optimality is particularly true of language-based communication, which necessarily involves the effects of context and is highly subject to situational constraints. Let us consider distal oral communication, which is the focus of CV studies, where the speaker and hearer and removed from each other. Optimal transmittal of messages in language-based communication must involve the adoption of more than one strategy. In addition to "signal strength," speakers' speech must be able to be differentiated to facilitate the effective transmittal of messages.

Issues of Signal Differentiation

When communication occurs distally against even a mildly noisy background, inter-signal differentiation is just as important as signal strength. Noise and frequency distortion can seriously undercut the intelligibility of human speech even though its audibility is good. In the case of English, the culprits are usually found among a set of 16 consonants that are notoriously confusable: (represented here using English letters) "p, t, k, f, th, s, sh, b, d, g, v, dh, z, zh, m, n" (Miller and Nicely 1961:153). They make up almost 75% of all the consonants and 40% of the all phonemes uttered in everyday English, thus making their differentiation quite important for the intelligibility of speech.

A series of experiments were conducted to measure the "confusion matrices" of these English consonants (Miller and Nicely 1961:154-155). Each of the consonants was spoken initially before the vowel/a/(father) to yield sixteen nonsense CV syllables, which were made to recur in a random order to form a 200-item list. With each of the five subjects taking mms to be the talker, the list was administered to the four nontalkers under various conditions. It was found that under the condition of a full response range (200-6500 cps) and a speech-to-noise ratio of 0 (noise and speech being equal in volume), the phoneme "k" was perceived correctly only 50% of the time, "p" 51% of the time, and "t" 77% of the time. A reduction of the background noise enhanced correct perception of all the sixteen consonants. But even with a speech-tonoise ratio of +6 db, there was a 31% to 20% chance that "p" and "k" were confused with a different consonant respectively.

Note that these experiments all simulated speech being generated by a source 15 inches away from the subjects, with various degrees of noise in the background. In situations involving distal communication, the statistics presented above would undoubtedly further deteriorate. It is known that high-frequency components of speech are most susceptible to masking by noise, and that acting as a low-pass filter, noise masking would truncate the high frequencies of distal communication. According to Miller and Nicely (1961:174-175), when frequencies above 1000 cps are filtered out of the speech, the features of place of articulation and duration are effectively absent, leaving nasality, affrication, and voicing to do all the work. As a result, the sixteen English consonants are collapsed into only five distinguishable phonemes: "p, t, k", "f, tb, s, sh", "b, d, g", "v, db, z, zh", and "m, n". Within the phoneme set "f, tb, s, sh", for instance, the sounds "f', "sh", "s", and "sh" become, in effect, indistinguishable allophones, and each of them can pass ihr all the four, causing perceptive confusions. Therefore, signal differentiation would be a major challenge in distal communication where the only choice made was for greater proportion of CV syllables.

What is the solution to the interplay between transmission and differentiation? One strategy that increases signal differentiation is to provide redundancy in the input to help people distinguish among confusable signals and reduce perceptive errors. Linguistically, redundancy is synonymous with predictability due to an abundance of specification provided by the speech situation--more than enough to decode the message (Crystal 1987:146). The syllable "pa", for instance, was misinterpreted as "ta", "ka", "fa", and "tha" in the project conducted by Miller and Nicely (1961:157-158). But if a word-initial/s/was added to the tested syllable, turning it into "spa", the listeners would have made fewer errors. This is because the sequences "s-f-a" and "s-th-a" are not phonologically permissible in English, and both would have been flagged down by the speech situation. Because phonological redundancy involves adding consonants, it must work against the prevalence of CV syllables.

We cannot argue with the transmission benefits of CV structures. The effectiveness of CV structure, however, becomes questionable once the distance or background noise is such that onset consonants are indistinctive from one another. The lack of intra-word differentiation that arises from this "fusion" of consonants will cause a breakdown in the reception of messages by the hearer and, therefore, communication. This entails consideration of occasions in which inter-signal differentiation overrides signal strength as the primary concern in distal communication.

Conclusion

To sum up, when intra-word differentiation is fair (i.e. not a salient challenge), CV structure can be optimal. When intra-word differentiation becomes problematic, however, it is necessary to resort to alternative strategies to achieve effective communication. Phonological redundancy contributes to greater communicative effectiveness by generating more complex syllables. In doing so it necessarily works against any predominance of CV syllables.

This paper is an attempt to refine this new attempt at the study of speech adaptation by offering two criticisms. First, it is too single-minded to capture the more flexible system of priorities that speakers apply in real life. This paper calls for consideration of alternative strategies, arguing that optimality is a dynamic state. It is proposed that such alternative strategies include choices that feature situational enhancement of signal differentiation. Indeed, as social behavior, speech is so complex that reducing its systemic adaptation to a consideration of transmission strength or rhythm alone is seriously limited. Second, there is reason to believe that studies examining phonological adaptation will be more fruitful when their investigation takes account of contextualization and speaker-listener interaction. It will open up a productive area of research for examination of the roles that such aspects as CV structures play in linguistic and social behavior alongside other features of speech interaction such as gestures, deixis, indexicality, communicative routines, language socialization, and so on, including phonological redundancy.

References

Aitchison, Jean 2001 Language Change: Progress or Decay? 3rd ed. Cambridge: Cambridge University Press 

Crystal, David 1987 The Cambridge Encyclopedia of Language. New York: The Cambridge University Press.

Dupre, John 1987 Introduction. In The Latest on the Best: Essays on Evolution and Optimality. John Dupre, ed. Pp. 1-24. Cambridge, MA: The MIT Press.

Ember, Melvin, and Carol Ember 2000 High CV Score: Regular Rhythm or Sonority?" American Anthropologist 102(4): 848-851.

Ember, Melvin, and Carol Ember 1999 Cross-Language Predictors of Consonant-Vowel Syllables. American Anthropologist 101 (4):730-742.

Fletcher, H. 1929/1953 Speech and Hearing in Communication. Toronto: D. Van Nostrand

Fought, John, Robert Munroe, Carmen Fought, Erin Good 2004 Sonority and Climate in a World Sample of Languages: Findings and Prospects. Cross- Cultural Research 38(1):27-51

Gould, Stephen Jay, and Richard C. Lewontin 1979 The Spandrels of San Marco and the Panglossian Paradigm: A Critique of the Adaptionist Programme." Proceedings of the Royal Society of London B205:581-598.

Jackendoff, Ray 2002 Foundations of Language. New York: Oxford University Press.

Lakoff, George and Mark Johnson 1980 Metaphors We Live By. Chicargo, IL: The University of Chicago Press.

Labov, William 1972 Sociolinguistic Patterns. Philadelphia, PA: University of Pennsylvania Press.

Lewontin, Richard C. 1987 The Shape of Optimality. In The Latest on the Best: Essays on Evolution and Optimality. John Dupre, ed. Pp. 151-159. Cambridge, MA: The MIT Press.

Martinet, Andre 1962 A Functional View of Language. Oxford: Clarendon Press

Miller, George A., and Patricia E. Nicely 1961 An Analysis of Perceptual Confusions among Some English Consonants. In Psycholinguistics: A Book of Readings. Soi Saporta, ed. Pp. 153-175. New York: Holt, Rinehart and Winston. First published in 1955.

Muller, Max 1861 Lectures on the Science of Language." Delivered at the Royal Institute of Great Britain. First Series. London: Longman, Green, Longman & Roberts.

Munroe, Robert, Carmen Fought, and John Fought 2000 Rhythmicity or Sonority: Response to Ember and Ember's 'Cross-Cultural Predictors of the Consonant-Vowel Syllables.' American Anthropologist 102(4): 844-848.

Munroe, Robert, Ruth Munroe, and Steve Winter 1996 Cross-Cultural Correlates of the Consonant-Vowel (CV) Syllable. Cross-Cultural Research 30:60-83.

Munroe, Robert, and Megan Silander 1999 Climate and the Consonant-Vowel (CV) Syllable: A Replication within Language Families. Cross-Cultural Research 33:43-62.

Smith, Eric Alden 1987 Optimization Theory in Anthropology: Applications and Critiques. In The Latest on the Best: Essays on Evolution and Optimality. John Dupre, ed. Pp. 20150. Cambridge, MA: The MIT Press.

Wolfgang, Banzhaf, and Frank H. Eeckman 1995 Editor's Introduction. In Evolution and Biocomputation. Banzhaf Wolfgang and Frank H Eeckman, eds. Pp. 1-4. New York: Springer-Verlag.

John Rhoades, St. John Fisher College, NY

Zhiming Zhao, SUNY College at Geneseo, NY

Endnotes

[1] The computation of "mean sonority score" is based on the work of Fletcher (1929/53), which assigns a numerical value to the "relative phonetic power" of each English phone, ranging from 100 (for example "a") to 2 (for example "p", "t"). The mean sonority score of a word is the result of adding up the scores for individual phones and dividing the total by the number of phones. Since the minimum sonority value for vowels is 41 ("i"), and the maximum sonority value for consonants is only 36 ("r"), an utterance of CVCVCV is bound to score higher than an utterance of CCVCVC or CVCCVC or CCVCCV, etc. Emphasizing "mean sonority score" is identical to emphasizing CV syllables.

Rhoades, Ph.D, is Associate Professor of Anthropology and Chair of the Anthropology Department, and Zhao, Ph.D, is Assistant Professor of Anthropology