The story so far: Steven Pinker’s new book The Stuff of Thought argues that the meaning of words is found in the abstract concepts that are either inborn or assembled from inborn concepts (thesis summarized in Pinker’s Anti-Whorfian Hypothesis). If Pinker is correct, an account of the origin of speech has to include the evolution of a system for translating the abstract concepts into concrete words. But a system for expanding abstract information into true, apt information won’t work (see The Abstract/Concrete Divide).
One difference between speech and computer output is the way people revise themselves. “It was a smooth ride …well, not that smooth. In fact at one point we hit a pothole.” Or they correct themselves, “I saw Tom Jackson the other day. I mean Pete Jackson. His name is Pete.” Or they notice something unintended, “I made my fame at that poker game. Ha, fame game. I’m a poet and don’t know it.”
Because of these sorts of changes I completely agree with Pinker’s basic thesis that our thoughts govern our language. Yet these sorts of changes also make me still more doubtful about his claim that these thoughts consist of abstract concepts. The concepts he describes are too close to language to provide a second opinion, so to speak. They are some sort of computational language that generates output in whatever cultural language we happen to speak. Meanwhile, real speakers are struggling to put what they think into the right words. And I know Pinker is familiar with what I’m saying. He writes way too well for his book to be anything close to a first draft.
Most of Pinker’s book is devoted to the semantic analysis of sentences and the differences in meaning that subtle changes make. There is also extensive consideration of things we never say which, under straight syntactical analysis, might seem permissible. So we can say This glass breaks easily but not Babies kiss easily, or That dog slaps easily. Why not? We can say it is easy to break this glass, it is easy to kiss babies, or it was easy to slap that dog. What’s the difference? Most people cannot explain that in the examples given, breaking alone is the effect of an action. Kissing and slapping are just actions. To break is an action in the sentence it is easy to break this glass, while breaks easily is about the effect of an action. You can say If you throw stones at the window, the glass breaks easily, but kissing is not an effect of doing something else. You cannot say If you do A, the baby kisses easily. This kind of semantic analysis is the heart of Pinker’s book. Considerations of these distinctions run on for page after page, chapter upon chapter, so if the kind of semantic revelation offered in this paragraph does not interest you, Pinker’s book is probably not for you.
Pinker’s argument is straight forward. Our speech has these rules that most people cannot explain, and yet people obey them. The rules, therefore, must not be things we know consciously. They are unconscious rules whose enforcement must rely on unconscious concepts. What’s more, since they are unconscious, nobody can have taught them to us, so they must either be innate or have formed from innate concepts that have somehow come together into assemblies.
This argument emerges from Chomskyan thinking about syntax. Pinker describes some variations in rules that cannot be explained syntactically and concludes that the differences must refer to an even deeper level of rules, semantic concepts. Chomskyans may want to embrace the argument as an advance or reject it as heresy; however, there is a non-Chomskyan way of explaining “oddities” like the breaks easily/kisses easily conundrum. I put oddities in quotation marks because they are only noticeable as problems if you assume that sentences work according to the way Chomsky says.
Here’s a different way of thinking about it. Speech directs a listener's attention, and a sentence like The glass breaks easily directs our attention to a property of the glass. But if there is no property for a speaker to notice, we will have nothing to say. Nothing ever happens that draws attention to an effect that causes a baby to kiss easily, so we don’t say it. There is no need to complicate the story by blaming the absence on some sort of unconscious concept. If somebody does say a baby kisses easily, we will promptly discover that the words do not direct our attention anywhere, and we will declare it meaningless Thus, even without knowing Pinker’s abstract distinction, speakers who occasionally say the first sentence will never speak the second. Perception, not concepts, underlie the meaning of speech. And its the difference between perception and words, not concepts and words, that allow us to revise, correct, or comment on what we have said.
Freud’s claim to being a scientist has been generally discredited, but Pinker still takes it for granted that an explanation based on unconscious activity is more scientific than one based on conscious attention. It is surely true that we do many things unconsciously, including mental operations. A simple example comes from trying to remember a name. It’s gone, nowhere to be seen. Then some time later the name just pops into your head. As far as you can tell, the name may have been supplied by God, although a more likely explanation is that the brain has been working quietly. One reason for preferring the brain explanation over God is that the answer is not always correct. Thus, the actor’s name we couldn’t think of may suddenly pop into one’s head: Jack Gyllenhall. But that doesn’t sound quite right and after a moment’s attention to the possible answer comes the correction, Jake Gyllenhall.
The experience identifies at least one function of attention; it is a kind of governor, keeping the unconscious work on track. We see this same sort of thing in speech. I might say Did you see Brokeback Mountain? I thought Jack Gyllenhall was great. Jack? I meant Jake. We can also see that this kind of correction is also a function of speech. I say, I thought Jack Gyllenhall was excellent in Brokeback Mountain. You reply, You mean Jake Gyllenhall. Score one for you while I sheepishly agree, Yeah, Jake. Where’d I get Jack?
This explanation of meaning works better than Pinker’s. For one thing, it begins at the concrete level and thus avoids the complications discussed in my previous post of trying to place abstract concepts in the here and now. For another, attention-based meaning simply makes public what is already going on in the mind. It is not the evolution of something intellectually new, whereas Pinker’s concept-based meaning does require something new. As Pinker explained it, “The cognitive machinery that computes relations among things, places, and causes [has been] co-opted for abstract ideas.” [p. 242]
Ocam’s Razor favors the simpler explanation. Attention-based meaning adds one step to the evolution of speech, finding a way to make public the results of paying attention. Concept meaning adds two steps, co-opting the cognitive machinery and then making the results of that co-option public. So William of Ocam votes against Pinker.
So the arguments against Pinker's thesis are:
- plausibility: there seems to be no way to make truthful concrete remarks, if you depend on abstractions as your starting point;
- simplicity: it requires fewer evolutionary steps to evolve speech that keeps meaning at the perceptual level than it does to evolve a language dependent on unconscious meanings; and
- evidence: there is an alternate explanation for the evidence Pinker relies on (things we do and do not say), while Pinker has no explanation for the way we struggle to put our thoughts into words.
Oh, and there's a fourth argument, but I'll save that for tomorrow, my final post about Pinker's book.