This week’s posts are trying to make sense of what this blog has learned since it first appeared last September.
Language is usually defined as a system of communication, which seems obvious enough until you think about it in evolutionary terms. Ape communications consist of involuntary vocalizations. Among humans such vocalizations include things like laughter and sobbing, activities no one is likely to confuse with speech, precisely because they are involuntary. It would be hopelessly paradoxical to deny that language is used to communicate, but at least we can deny that it is anything like ape or computer communications.
This week’s posts are trying to make sense of what this blog has learned since it first appeared last September.
Language is usually defined as a system of communication, which seems obvious enough until you think about it in evolutionary terms. Ape communications consist of involuntary vocalizations. Among humans such vocalizations include things like laughter and sobbing, activities no one is likely to confuse with speech, precisely because they are involuntary. It would be hopelessly paradoxical to deny that language is used to communicate, but at least we can deny that it is anything like ape or computer communications.
Further evidence against the engineering understanding of communications has been provided by Jean-Louis Dessalles who points out that if we were really engaged in the kind of communication exchanges that computers use, we should be speaking a lot faster. Moore’s Law famously notes that in the world of microchips communication capacity doubles every 12 to 18 months. Yet humans worldwide speak at a rate of about 50 bits per second. Surely we could have evolved a more efficient system if downloading data were the point of speech.
A better understanding of speech comes from seeing it as an expansion of powers of joint attention, a voluntary activity that, in human infants, precedes the beginnings of speech. Michael Tomasello at the Max Planck Institute for Evolutionary Anthropology has been a leading proponent of this idea. It’s chief theoretical strength is the way it makes good evolutionary sense.
- It links a unique human trait (speaking) with a behavior (paying joint attention) sometimes observed in apes, so we don't have to imagine speech evolving out of nothing.
- Attention is a function of perception, so the evolution of speech based on joint attention would not require the creation of a completely new intellectual system such as a symbol proecessor.
There are also some good empirical reasons for linking joint attention and speech:
- Listeners must pay attention to speakers in order to understand what is said.
- A common problem in speech is for speakers to lose their “train of thought,” that is their own attention shifts as they speak and they forget what they were saying.
- Speech is full of perceptual metaphors.
- Before children begin speaking they engage in joint attention (typically mother and child share attention); meanwhile, autistic children who cannot engage in joint attention have severe problems learning to speak.
- Humans have a unique trait in the whites of their eyes (primates and other mammals typically have back sclera) indicating that there has been strong evolutionary pressure to let others see where you are looking.
There are also some strong pragmatic reasons for linking joint attention with sppech, especially the way it makes it easier to think usefully about meaning. Words, in this view, are pilots of attention (to use Giorgio Marchetti’s definition). They are not symbols or representations and they do not rest on formal definitions. This approach cuts through a fat Gordian knot of confusion about the distinction between representations and rules, by throwing them all out. It also eliminates the idea of semantic “content” that leaves everybody arguing over where the content is stored.
Dessalles distinguishes between protosemantics and semantics proper. Protosemantics pilots attention by, so to speak, pointing. “Lemon” or “lemon good” are examples of verbal pointing. The first draws attention to the lemon, the second to the response to the lemon. To understand the first the listener must look. To understand the second, the listener must imagine feeling good in response to a lemon. That second understanding seems problematical since the listener is perceiving something internal that may have nothing to do with the feeling that inspired the speaker’s remark. But speech works by directing attention and there is no reason to assume that everyone in a conversation attends to the very same perceptions.
Semantics proper pilots attention to a perception’s gestalt of both figure and ground, and requires both words and syntax. “A lemon slice dropped into her glass of diet cola.” That sentence pilots attention to the both the lemon and its setting, creating a real scene.
This kind of analysis makes very clear the difference between a sentence and an equation. A sentence directs either the eye or the imagination, while an equation calls forth a procedure. L=4p+35 can be solved without having any idea what L or p represent. The procedure would be exactly the same for M=2r-35. A variation in a sentence, however, can create a very different scene—e.g., “A golf slice dropped her out of contention for queen of the country club.” The difference between the two sentences is not in their interpretive procedures but in where they direct attention.
This approach to meaning can also help clarify probably the biggest debate over meaning of the past quarter century. That dispute has focused on John Searle’s thought experiment about a Chinese Room. It imagines an employee working alone in a room. From time to time a pneumatic tube delivers questions written in Chinese characters. The employee does not speak Chinese but responds by looking up the characters in rule books and returning whatever the rule book says. The answer might be good, but the employee has no idea what anything means and cannot be said to understand Chinese. The point of the thought experiment is to argue that a computer which takes input and produces output according to formal rules cannot be said to understand the language it uses. The resultant brouhaha over this thought experiment persists to this day, but has been inconclusive because nobody has been able to say what meaning is.
Searle guaranteed continuing confusion by calling meaning “mental content” without defining or locating this content. But the idea that words pilot attention bypasses the whole issue of content. A word like “lemon” does not summon a definition from a database, but directs attention to a perception, either real or imagined, of a lemon. Thus, we can see that while Searle was correct in his intuition that the employee in the Chinese room would not understand Chinese. The employee’s task is hopeless, as far as understanding Chinese characters are concerned, because the messages coming into the room are incapable of directing the employee’s attention anywhere.
There are some other pragmatic reasons for linking joint attention with speech, but this post is becoming long, so I will save them for tomorrow.
Comments