My earlier discussion of Evans and Levinson's critique of universal grammar was vague on details. Today I wanted to look at one specific argument.
Funny words
Evans and Levinson briefly touch on universal semantics (variously called "the language of thought" or "mentalese"). The basic idea is that language is a way of encoding our underlying thoughts. The basic structure of those thoughts is the same from person to person, regardless of what language they speak. Quoting Pinker, "knowing a language, then, is knowing how to translate mentalese into strings of words and vice versa. People without a language would still have mentalese, and babies and many nonhuman animals presumably have simpler dialects."
Funny words
Evans and Levinson briefly touch on universal semantics (variously called "the language of thought" or "mentalese"). The basic idea is that language is a way of encoding our underlying thoughts. The basic structure of those thoughts is the same from person to person, regardless of what language they speak. Quoting Pinker, "knowing a language, then, is knowing how to translate mentalese into strings of words and vice versa. People without a language would still have mentalese, and babies and many nonhuman animals presumably have simpler dialects."
Evans and Levinson argue that this must be wrong, since other languages have words for things that English has no word for, and similarly English has words that don't appear in other languages. This is evidence against a simplistic theory on which all languages have the same underlying vocabulary and differ only on pronunciation, but that's not the true language of thought hypothesis. Many of the authors cited by Evans and Levinson -- particularly Pinker and Gleitman -- have been very clear about the fact that languages pick and choose in terms what they happen to encode into individual words.
The Big Problems of Semantics
This oversight was doubly disappointing because the authors didn't discuss the big issues in language meaning. One classic problem, which I've discussed before on this blog, is the gavagai problem. Suppose you are visiting another country where you don't speak a word. Your host takes you on a hike, and as you are walking, a rabbit bounds across the field in front of your. Your host shouts "gavagai!" What should you think gavagai means?
There are literally an infinite number of possibilities, most of which you probably won't consider. Gavagai could mean "white thing moving," or "potential dinner," or "rabbit" on Tuesdays but "North Star" any other day of the week. Most likely, you would guess it means "rabbit" or "running rabbit" or maybe "Look!" This is a problem to solve, though -- given the infinite number of possible meanings, how do people narrow down on the right one?
Just saying, "I'll ask my host to define the word" won't work, since you don't know any words yet. This is the problem children have, since before explicit definition of words can help them learn anything, they must already have learned a good number of words.
One solution to this problem is to assume that humans are built to expect words of certain sorts and not others. We don't have learn that gavagai doesn't change it's meaning based on the day of the week because we assume that it doesn't.
More problems
That's one problem in semantics that is potentially solved by universal grammar, but not the only. Another famous one is the linking problem. Suppose you hear the sentence "the horse pilked the bear". You don't know what pilk means, but you probably think the sentence describes the horse doing something to the bear. If instead you find out it describes a situation in which the bear knocked the horse flat on its back, you'd probably be surprised.
That's for a good reason. In English, transitive verbs describe the subject doing something to the object. That's not just true of English, it's true of almost every language. However, there are some languages where this might not be true. Part of the confusion is that defining "subject" and "object" is not always straightforward from language to language. Also, languages allow things like passivization -- for instance, you can say John broke the window or The window was broken by John. When you run into a possible exception to the subject-is-the-doer rule, you want to make sure you aren't just looking at a passive verb.
Once again, this is an example where we have very good evidence of a generalization across all languages, but there are a few possible exceptions. Whether those exceptions are true exceptions or just misunderstood phenomena is an important open question.
-------
Evans, N. and Levinson, S. (2009). The myth of language universals: Language diversity and its importance for cognitive science Behavioral and Brain Sciences, 32 (05) DOI: 10.1017/S0140525X0999094X
photo credit
The Big Problems of Semantics
There are literally an infinite number of possibilities, most of which you probably won't consider. Gavagai could mean "white thing moving," or "potential dinner," or "rabbit" on Tuesdays but "North Star" any other day of the week. Most likely, you would guess it means "rabbit" or "running rabbit" or maybe "Look!" This is a problem to solve, though -- given the infinite number of possible meanings, how do people narrow down on the right one?
Just saying, "I'll ask my host to define the word" won't work, since you don't know any words yet. This is the problem children have, since before explicit definition of words can help them learn anything, they must already have learned a good number of words.
One solution to this problem is to assume that humans are built to expect words of certain sorts and not others. We don't have learn that gavagai doesn't change it's meaning based on the day of the week because we assume that it doesn't.
More problems
That's one problem in semantics that is potentially solved by universal grammar, but not the only. Another famous one is the linking problem. Suppose you hear the sentence "the horse pilked the bear". You don't know what pilk means, but you probably think the sentence describes the horse doing something to the bear. If instead you find out it describes a situation in which the bear knocked the horse flat on its back, you'd probably be surprised.
That's for a good reason. In English, transitive verbs describe the subject doing something to the object. That's not just true of English, it's true of almost every language. However, there are some languages where this might not be true. Part of the confusion is that defining "subject" and "object" is not always straightforward from language to language. Also, languages allow things like passivization -- for instance, you can say John broke the window or The window was broken by John. When you run into a possible exception to the subject-is-the-doer rule, you want to make sure you aren't just looking at a passive verb.
Once again, this is an example where we have very good evidence of a generalization across all languages, but there are a few possible exceptions. Whether those exceptions are true exceptions or just misunderstood phenomena is an important open question.
-------
Evans, N. and Levinson, S. (2009). The myth of language universals: Language diversity and its importance for cognitive science Behavioral and Brain Sciences, 32 (05) DOI: 10.1017/S0140525X0999094X
photo credit
2 comments:
I haven't read the paper yet -- I'll do so this week. But isn't the big thing they're questioning universal grammar? Universal "mentalese" is a much more radical, minority view whereas universal grammar is (to my mind) very well established.
Their target really is UG, and I'd say there are a lot of people who doubt UG as a hypothesis. I should say that I'm assuming a broad (but common) definition of UG, where UG is whatever is common to all languages and what allows humans to acquire language. The split between UG and universal mentalese might in some places be unclear.
The point of this post is that there are broad similarities in how people in all languages choose to describe things. Quine's gavagai paradox is one of the easier cases to describe, though I can see how that looks more like mentalese than grammar. The (very similar) Linking Problem is more clearly a linguistic issue.
Post a Comment