Evolving speech

In the previous chapters I first concluded that the vocal periphery, long emphasized in discussions of language evolution, played a minor role in the evolution of speech. A language-equipped brain could master, and communicate using, the vocal tract of a chimpanzee, or indeed a dog, quite adequately. Furthermore, the flexibility of the vocal apparatus attested in living mammals indicates that there is little hope of reconstructing the speech abilities of extinct hominids from their fossil remains. Finally, the discovery that several animal species possess a reconfigured vocal tract similar to our own, but do not use it in speech production, means that even if we could use fossils to determine when the larynx descended in hominid evolution, we could not necessarily deduce whether those hominids spoke. These relatively negative conclusions supported the positive conclusion that changes in the brain were crucial for the evolution of speech, and we then explored the neural and genetic bases that currently seem to be critical for vocal learning and imitation.

In this chapter, I will attempt to synthesize these diverse strands of evidence, and begin to investigate theories of the evolution of speech and basic phonology. I begin by considering models of speech evolution (often termed “theories,” despite lacking the gravitas normally associated with this term in science). My goal will be to evaluate the central innovations and insights in each model.

On a fine spring afternoon in 1828, a shoemaker in the German town of Nuremberg encountered a strange sight: a young man waddling down the street with his arms outstretched, squinting and yelling incomprehensibly. When asked to explain himself, he stared uncomprehendingly, and only repeated his name – Kaspar Hauser – and the phrase “I want to be a horseman like my father.” He vomited when fed anything other than bread or water, but was taken in by a kindly scholar, Georg Daumer, in the town. Kaspar Hauser's saviors quickly learned that, despite his lack of knowledge of German or any other language, and a total inability to clothe or feed himself, Hauser was perceptive and intelligent, and a remarkably fast learner. Over the several years that Hauser lived with Daumer, he acquired a considerable vocabulary, some basic syntax, and the ability to read and write – enough to write a crude autobiography – along with the basic skills necessary to survive in normal society. As Hauser's language abilities grew, the horrible story of his life became clear in outline. From a young age, for as long as he could remember, Kaspar Hauser had been kept in a dark dungeon, with absolutely no social contact. His entire existence was confined to a windowless room with an earthen floor and a few wooden toys, with no verbal or physical contact with other humans.

Introduction

In this chapter I will discuss contemporary evolutionary theory as it relates to models of language evolution. I first attempt to give a brief, balanced overview of contemporary evolutionary theory, a condensed non-mathematical version of what one might receive in an undergraduate class on evolution, but tailored to questions concerning language evolution. I adopt a historical viewpoint because the history of evolutionary theory provides several nice examples of past success in the unification of warring disciplines – precisely what is needed, in the future, in biolinguistics. For those interested in the fascinating history of evolutionary theory, see Mayr (1982), Ruse (1986), Gould (2002), and Bowler (2003). For more detailed introductions to modern evolutionary theory, Dawkins (1987) gives an excellent popular introduction, and a good textbook is Ridley (1997), which also provides balanced overviews and detailed historical perspectives on some of the ongoing debates discussed below.

Second, I survey areas of controversy in language evolution. My guiding assumption is that evolutionary theory is general, applying to all living organisms: we neither need nor want a version of evolutionary theory specific to humans (Darwin, 1871; Hockett and Ascher, 1964; Pinker and Bloom, 1990). Language evolution is just another question for mainstream evolutionary theory, and special pleading about human evolution must be resisted unless clearly required by the data. Several major points of contention have yet to be resolved one way or another, and these will be my focus.

Introduction

This chapter will discuss what is, to many, the most intuitive class of models for the evolution of language. Such models posit a “lexical” protolanguage, with a large learned lexicon of meaningful words, but no complex syntax. Words in a lexical protolanguage are not combined into complex syntactic structures, leaving modern syntax as the final step in language evolution. This “syntax-final” model of language evolution is shared by a diverse group of scholars who disagree about almost everything else (e.g. Lieberman, 1984; Bickerton, 1990; Givón, 1995; Jackendoff, 2002). A lexical protolanguage assumes, as prerequisites, an ability for vocal imitation (necessary to develop a shared spoken vocabulary) and a capacity and drive for referential communication. Although the idea of a lexical protolanguage is implicit in a number of distinct hypotheses about language evolution, it has been defended most explicitly by Derek Bickerton, especially in his book Language and Species (Bickerton, 1990). Bickerton's notion of protolanguage has since been adopted by various other theorists (Jackendoff, 1999, 2002) and criticized by others (Deacon, 1997; Lieberman, 2000). Bickerton's treatment draws on a wide range of data, and is relatively explicit, giving reasons (both data and argument) for most of his assumptions. He explicitly aims to build bridges between linguistics and evolutionary theory, taking both Chomsky and Darwin seriously, and thus provides a reasonable starting point for discussions of contemporary theories of language evolution. Bickerton's model of lexical protolanguage solves a number of problems quite neatly, while leaving several important evolutionary problems open.

Speech is not language, but is important nonetheless

Speech (complex, articulated vocalization) is the default linguistic signaling mode for all human cultures, except when the audio-vocal modality is unavailable, as for the deaf. Though speech is the default linguistic modality, it is not the only possibility. Signed languages of the deaf are full, complex, grammatical languages, independent of but equivalent to spoken languages (Stokoe, 1960; Klima and Bellugi, 1979), and this demonstrates that speech is not the only signaling system adequate to convey language. Writing is another example of a visuo-manual system of linguistic communication, but since writing systems are typically “parasitic” representations of spoken systems, they illustrate the multi-modality of language less convincingly. One of the first distinctions to be made in studying language evolution is therefore that between speech (a signaling system) and language (a system for expressing thoughts, which can incorporate any one of several signaling systems). Many languages use the same word to designate both speech and language (e.g. German Sprache), and in English one often finds the word speech used as an exact synonym for language, showing how close these concepts are in everyday use. Throughout this book, I will use “speech” only in the narrow sense of complex articulated vocalizations, where “articulated” implies tight coordination between the supralaryngeal vocal tract and the larynx. Besides linguistic speech, other examples of articulated vocalization include infant babbling, “speaking in tongues,” nonsense speech, jazz scat singing, or Asian formant singing.

Introduction: phonology remains puzzling

In the previous two chapters we have discussed models of the origin of several key components of language. We saw how a lexical protolanguage could provide a scaffolding for complex syntax, derived from pre-existing conceptual primitives, and we saw how gesturally supported protolanguage could provide one route to open-ended reference via iconic, intentional pantomime. However, both of these models have major difficulties explaining human phonological competence. While Bickerton's model takes a capacity to intentionally generate complex vocalizations for granted, based on the notion that chimpanzees have some cortical control over vocalizations, we have seen that complex vocal control is not present in chimpanzees or other apes, nor therefore was it present in the LCA, and it does not evolve automatically as organisms get “smarter.” Gestural theorists have grappled more earnestly with this problem, but most, including both Hewes and Arbib, have had to postulate some separate selective story in order to achieve the full vocal generativity of modern humans. This suggests that the evolution of vocal control and phonology remains a major open issue, and in this chapter we will discuss the ideas of theorists who have grappled directly with this problem.

At the turn of the twentieth century, a remarkable horse named Hans was paraded through Germany by his owner Wilhelm von Osten, a horse trainer and high-school mathematics teacher. Not only could “Clever Hans” understand complex questions put to him in plain German – “If Tuesday falls on the eighth of the month, what date is the following Friday?” – but he could answer them by tapping out the correct number with his hoof. Using this simple response, it appeared that Hans could add, subtract, multiply, and divide, tell the time, understand the calendar, and both read and spell words. Suspicious, the German board of education appointed a commission, including circus trainers, veterinarians, teachers, and psychologists, to investigate the situation. Surprisingly, they concluded in 1904 that no trickery was involved. This did not satisfy the board, and the case was passed to psychologist Oskar Pfungst for experimental investigation. Braving both the horse's and owner's notoriously bad tempers, Pfungst finally was able to demonstrate that Hans was no mathematician, but rather a fine observer of human behavior (Pfungst, 1911). In a story now told to countless “Intro Psych” students, Pfungst demonstrated that Hans could only answer questions correctly when: (1) the questioner knew the answer; and (2) Hans could see the questioner.

On an autumn day in 1947, much like any other, Cathy and Keith Hayes returned to their suburban American ranch house with their newborn infant girl, Viki (Hayes, 1951). After a few difficult days, Viki began feeding well and growing rapidly. She was a very quiet baby, sweet and affectionate, and loved to be held and tickled. She learned to walk early, entering a rambunctious phase and breaking numerous household objects, but eventually her loving parents' gentle discipline bore fruit and she developed into a playful, obedient little girl. By the age of three, Viki could feed and bathe herself, eat with a spoon and drink through a straw, and help with cleaning. She was fond of looking at herself in the mirror, and loved assembling jigsaw puzzles. She enjoyed playing on the backyard swing, climbing trees, and playing peekaboo with the neighborhood children. She was in many ways a normal young girl, with one major exception: Viki did not speak. Not a word. She was able to grunt, scream, and laugh, so her problem was not with vocalization in general; instead it seemed to stem from a neural difficulty specific to spoken language. After consultation with experts, Cathy Hayes instituted a speech training regime, manipulating her young pupil's lips manually and rewarding her with treats whenever she approximated a word.