Which model does language follow? (1-5)
Does linguistic communication follow the code model? The common intuition says yes: there is a firm association between linguistic signals (words) and meanings, such as “cat” and the animal (p. 33).
The author says this is not enough. Language is full of metaphor and allusion and conveys meanings well beyond the literal. Intonation and body language also cue meaning. Even a simple utterance like “See you later” can be affection, a threat, a request for a meeting, or a warning; the hearer works out which by reasoning about the speaker’s mental state (p. 33).
So both code and ostensive-inferential communication are involved. There are two ways to picture their relation (p. 34).
- The standard view: the linguistic code is central, and ostension and inference are add-ons.
- The author’s view: ostension and inference come first, and the linguistic code became powerful on top of them.
The standard view shows up in mainstream linguistics, where syntax and semantics are central and pragmatics is peripheral (p. 34). Much computational and mathematical modelling also assumes a code linking signals and meanings (pp. 35–37).
The code is “weak”
The author says this assumption is “wrong” and returns to underdetermination (p. 38).
- The literal meaning of language does not fix the speaker’s meaning. Indexicals (pronouns), metaphor, sarcasm, understatement, and indirect expressions are all examples.
- “It’s raining” can carry endless readings beyond the literal, such as “at least the weather could have been nice” (p. 39).
- One cannot in principle list the speaker’s meanings, because one can always add another to any list of possible meanings (p. 39).
So the linguistic code alone cannot make communication work. The author gives this exchange (p. 40; paraphrased by ToL):
Mary: Would you like to have dinner with me? Peter: I’ve just eaten.
Literally the reply says only that he has eaten. The meaning “no thanks” arises only when context and shared knowledge (people don’t eat twice in a night) are added. If the robot Hymie could process only literal meaning, this conversation would fail. As a code, language is weak.
And yet language is powerful
On the other hand, when a linguistic code is laid over ostensive-inferential communication, expressive power rises greatly (p. 41). While tilting a coffee cup one can say “more coffee, please.” One can say “Delicious!” Language makes ostensive communication accurate and rich.
Here the author reverses a common assumption. Ambiguity (one expression having several meanings) is a great advantage, not a defect, because one word can express many meanings efficiently (p. 41). This links to the discussion of ambiguity in article 281.
He sets the code of the code model against the linguistic code (p. 42).
| Code of the code model | Linguistic code | |
|---|---|---|
| Number of meanings | Finite, roughly fixed | Ambiguous, unresolvable |
| Things it can refer to | Limited | Unrestricted |
| Works alone? | Yes | Needs an ostensive-inferential base |
The two are thus ontologically different (p. 42). He warns that conflating them misleads, and gives them separate names.
Natural code and conventional code
He distinguishes two terms (p. 44).
- Natural code: code that works purely according to the code model. Bacterial quorum sensing, chemical signals between mother and infant, locust calls, gesture-calls, and so on.
- Conventional code: code that strengthens a pre-existing system of ostensive-inferential communication. Language, and also Morse code.
Whether a non-human species’ code is natural or conventional is an empirical question. What matters is that the two have different origins.
The condition for a conventional code is agreement in a community. The author gives the example of switching from driving on the left to the right. In principle, if everyone wants it, it can be changed (Sweden did in 1967 and Iceland in 1968; p. 45). Word-meaning pairings are the same. “Dog” refers to a dog because we agree that it does, not because there was an association between word and meaning to begin with.
What is language? (pp. 45–48)
The author gives his answer. An individual language is a structure made of many conventional codes that strengthen ostensive-inferential communication within a community (p. 46). Language is a kind of ostensive-inferential communication with linguistic conventions on top.
As an example he briefly takes up the debate over Pirahã, spoken by Amazonian hunter-gatherers. It is claimed that Pirahã lacks recursion (p. 46). He says that even if this is settled, it does not affect the definition of language. The lack of any one property does not make a language cease to be one, because language is a collection of conventional codes (p. 47).
Natural and conventional codes combine
Natural codes and ostension can combine. Laughter is the example. Involuntary laughter (Duchenne type) is a natural code, but one can also laugh on purpose. Laughing theatrically to mock a friend’s joke is the latter (p. 47).
This matters for thinking about evolution (p. 48). Natural and conventional codes differ decisively. The former is the ostensive use of behaviours like laughing or groaning; language arises with the latter. So trying to derive the origin of language directly from the evolution of animal natural codes is, he says, a category error.
Two senses of “meaning” (1-6)
Last, the author notes that “meaning” is used in two ways (p. 49).
(1) Meaning in pragmatics (Grice). Grice distinguished natural meaning and non-natural meaning (p. 50).
- Natural meaning: a causal link in the world, as when dark clouds mean rain.
- Non-natural meaning: what a speaker intends to convey.
In Grice’s analysis, non-natural meaning arises in three stages (pp. 50–52).
- The speaker intends to manipulate the other’s mental representation.
- The speaker intends the other to recognize that intention.
- The other’s recognition of the intention becomes, at least in part, the reason for the other’s belief.
The example is wine. I want a friend to pour me more, so I move my empty glass. The friend recognizes this as an intention, and also recognizes that I intend that recognition. That is meaning. The author says this is another name for ostensive-inferential communication (p. 52).
From here he criticizes animal-communication research for using the word carelessly (p. 53), treating “meaning” as if animals had the same intentions as humans in everyday speech.
(2) Meaning in evolutionary biology. Here the distinction between ultimate explanation and proximate explanation comes in (p. 54).
| Question | Example: an infant crying | |
|---|---|---|
| Ultimate | Why does the behaviour exist? (function) | To draw care and protection from caregivers |
| Proximate | By what mechanism does it occur? | Cold or hunger triggers it; endogenous opioids and the like are involved in stopping it |
The author says these are different kinds of explanation, complementary, and neither is more important (p. 55).
He then puts it this way: the “meaning” of an animal’s communication is an explanation at the ultimate level of signal-producing behaviour, that is, the signal’s function, while the Gricean sense of meaning is an explanation at the proximate level (a metapsychological mechanism). Even if a mating call’s function is rewritten as the command “mate with me,” it does not follow that the monkey intends this (p. 56). Conflating the two produces misunderstanding in comparisons with human language.
Note
The quoted dialogue is ToL’s paraphrase of the book’s Peter-and-Mary example. Details such as the Pirahã recursion debate and the years of road-side changes follow the book. I did not read the chapter-1 summary.