Related texts

Pinker's discussion of movement of auxiliary verbs.

[1] I discuss here Steven Pinker's discussion, in the book The Language Instinct (Penguin book, 1994, paperback, 0-14-017529-6) of movement of auxiliary verbs in English to form yes/no questions out of statements. My main target is not to show that Pinker is wrong, but to show how Pinker uses demagogical tricks to convince the reader. I assume in this text that the reader has the book. Page numbers are from the paperback edition. The part that I discuss is approximately in the middle of chapter 2 (Chatterboxes). All quotations below are from this book unless stated otherwise, and apart from text inside square brackets are exact quotes, including Italics.

[2] I would expect many readers to think that I am being pedantic in the discussion below, and that the points I raise are minor. However, Pinker's arguments are based on using many minor distortions of logic and common sense to achieve large distortion in the final conclusion.

[3] Pinker achieves many of these distortion by using implications, i.e. making statements which do not seem to have a point unless the reader accepts some propositions that were not actually made. In the text below I make these implications explicit and show how Pinker uses them. I believe in all cases the implications are obvious, and all readers will make them.

[4] Before discussing Pinker's text, several points should be made:

[7] Linguists express the same idea in a much more formal manner. For example, in the paper that Pinker is quoting in page 42, the same idea is expressed this way (Crain and Nakayama, 1987, Language, 63, 522-543, p.526 top):

"Hypothesis II: In yes/no questions the auxiliary verb in the main clause of the declarative is inverted with the subject noun phrase."

The formal language make it seems difficult and non-intuitive, but it is actually the same obvious rule that I wrote in the previous paragraph.

[8] Crain and Nakayama also claim that a procedure similar to what Pinker describes (below) is computationally simpler, and base this on a reference to Chomsky (from 1971). I didn't bother to follow this reference, but clearly Chomsky did not bother to take into account the information that a listeners must have by the time they understood the sentence.

    1. It makes it clear to the listener/reader that the sentence is a question right from the first word.
    2. It tells the listener/reader what tense is the proposition that the question is about from the first word.
    3. It gives the listener/reader an idea what kind of proposition it is (an existent/attribute proposition with a be verb or other kind).
    4. It helps the listener/reader to find the point of the question because it is where the sentence stops conforming to normal rules.

On the other hand, it doesn't make it much more difficult to understand the sentence, as auxiliary verbs contribute little information, and most sentences with auxiliary verbs can be interpreted reasonably easily without the verb.

[10] Pinker uses this topic to convince the reader that (p.40, bottom):

In fact, we can show that they [children] know things they could not have been taught.

[11] Pinker starts the discussion this way (p.40, bottom):

Consider how you might turn the declarative sentence A unicorn is in the garden into the corresponding question, Is a unicorn in the garden? You could scan the declarative sentence, take the auxiliary is, and move it to the front of the sentence:

[12] Well, of course "You could scan the declarative sentence", but does it make sense? For a person that understands the sentence, of course it doesn't: they already know what is the verb of the sentence, so can easily pick it out and move it to the front. Scanning the sentence to find the verb is an extra work, which is completely redundant. But Pinker wants to convince the reader that it is natural to assume that this is what the children will do unless they have the logic of language wired. Instead of making this proposition explicitly, he implies it by considering it first and ignoring any other possibility.

[13] It is also worth noting that Pinker has selected a very unnatural sentence as an example. First, talking about unicorns is rare. Secondly, most of people will use the sentence "There is a unicorn in the garden" (and "Is there a unicorn in the garden?") instead. I suspect some people will actually regard the question "Is a unicorn in the garden?" as un-grammatical. So why does Pinker use such unnatural example? The main advantage of this is that it makes it more difficult for the reader to use their intuition.

[14] An important point in the maneuver that Pinker performs is that at the first time he introduces the scanning procedure, he doesn't give the impression that it is a significant concept in the discussion. The significance of the scanning procedure as the sole alternative to "wired language" emerges later, after the user hopefully already got used to it. This way he makes it easier to the reader to go along with the argument, without actually trying to figure if it really make sense to consider the scanning procedure.

[15] Pinker continues the maneuver of implicating the scanning as the natural assumption if children don't have language wired on p.41. He uses the scanning as the obvious alternative for dealing with the more complex sentence (top of p.41), and then refers to the procedure as the "simple procedure". At the bottom of p. 41 he claims it is "simpler and presumably easier to learn". At no point, though, does he make explicit the claim that is required for the argument to make sense, i.e. that the linear search is the natural assumption if language is not innate. By avoiding making the claim explicit, he makes it less likely that the reader will consider if this claim is reasonable.

[16] In the middle of p.41 Pinker explains why it is the second is that is moved. The explanation is almost reasonable, but it uses formal terms without justification. For example, what is the evidence for "a mental label, like "subject noun phrase" or "verb phrase.""? Linguists use these labels for formal analysis, but that doesn't tell us how it works in the brain. The difficulty that people have with learning such labels strongly suggests that we don't use similar labels.

[17] Pinker's explains why the second is is moved this way (p.41):

It [the real rule] looks for the auxiliary that comes after the phrase labeled as the subject.

[18] The interesting thing about this explanation is that it is not the way linguistics normally put it. In linguistics, they would refer to the verb as "the auxiliary verb of the main clause," e.g. the way Crain and Nakayama put in [7] above. So why doesn't Pinker use the standard terms? The most plausible explanation is that it is too close to the common sense explanation (see [6] above), and Pinker needs to reduce the chance that the reader will think about the common sense explanation.

[19] Next Pinker writes (p.41, last paragraph):

Chomsky reasoned that if the logic of language is wired into children, then the first time they are confronted with a sentence with two auxiliaries they should be capable of turning it into a question with the proper wording.

[20] Here Pinker uses a quite neat demagogical trick, based on the ambiguity of "the first time they are confronted with". The normal interpretation, and the way almost all of the readers would understand it, is that this means "the first time they hear these kind of sentences". This would lead them to believe that Chomsky's claims is about the ability of children to generate the question the first time they hear this kind of sentences. Since Pinker says that the experiment he describes later supports Chomsky's claim, they will be lead to believe that there is experimental support for this kind of ability in children. If this was really true, it would have been really remarkable, but it is not, and even Chomsky itself wouldn't claim it.

[21] Anybody who is familiar with linguistics literature, and certainly Pinker himself, knows that this is not Chomsky's claim. Chomsky's claim is that children can generate the question before they hear any derived question from this kind of sentences, but not necessarily before they heard any of the declarative sentences themselves. Therefore, anybody who knows the literature will interpret "confronted with" as meaning this claim. This claim, however, is much weaker, and as I wrote above in [6], has a simple and obvious explanation.

[22] By using an ambiguous expression, Pinker succeeds to mislead most of the readers, and to do it in a way which is "hidden" from other linguists. The latter know the literature, and hence automatically interpret the sentence towards the correct claim, and don't realize that the bulk of the readers interpret it differently.

[23] Very alert readers may be able to figure out that something is odd here if they read the rest of the paragraph very carefully, but the bulk of Pinker's readers are not that alert. Even those that are will still find it very difficult to work out the real claim, because the text does not make it explicit anywhere.

[24] Next Pinker writes:

This should be true even though the wrong rule, the one that scans the sentence as a linear string of words, is simpler and presumably easier to learn.

[23] As already mentioned above [12], Pinker uses here again an argument by implication. For the experimental evidence to actually support that "language is wired into children", he must also claim that if it doesn't children will tend to form the wrong question. If he made this claim explicitly, some readers may actually think about it, so Pinker does not make it explicitly. Instead he implies it by stating that the scanning rule is simpler and presumably easier to learn, and let the reader infer the claim.

[25] Next sentence is:

And it should be true even though the sentences that would teach children that the linear rule is wrong and the structure-sensitive rule is right - questions with a second auxiliary embedded inside the subject phrase - are so rare as to be nonexistent in Motherese.

[26] Two minor demagogical tricks here: First, Pinker continues to push the scanning rule by writing as if it is the only alternative to wired language. Second, that is the sentence where very alert readers may figure out that they misunderstood Chomsky's claim (see [19] above). Pinker makes it more difficult to figure this out by referring first to the questions as "sentences". While questions are formally sentences, in normal usage they are not referred to as "sentences". Therefore the readers will think on the declarative sentences through the first half of the sentence, and probably will fail to re-interpret the first half properly. Note that it would much more natural to use a term like "information" or "data" instead of "sentences".

[27] Next sentence (last on p.41) is:

Surely not every child learning English has heard Mother say Is the doggie that is eating the flower in the garden

[28] First let us note that this is not relevant to the argument, because the information in the declarative sentence is enough to form the question correctly (see [6] above). However, it also worth noting that Pinker, like in [13] above, is using again the trick of an example that is highly unlikely semantically. This is both because doggie normally does not eat flowers, and because for the speaker to know that doggie is eating the flower, they must see it and hence know that it is in the garden. Reversing the clauses gives a much more natural question: "Is the doggie that is in the garden eating a flower?", which can be made even more natural by replacing "flower" with "bone". So why does Pinker use such bizarre examples, both here and in [13]? The most plausible explanation is that he tries to confuse the reader. It should be noted that it is not only the plausibility of the question that the reader gets confused about by such bizarre examples. Because it is difficult to use intuition with these examples, it causes the readers to "shut-off" their intuition (to some extent), and hence makes it easier to get them to accept counter-intuitive ideas.

[29] After Pinker hopefully convinced the reader that if children can generate the right questions it shows "wired language", either because they accepted that the scanning procedure is the only alternative, or because they think about the wrong claim (see [19] above), he presents the experiment in the top of page 42. Then he pretends to discuss possible objections, by discussing one possible objection. He presents this objection thus:

Now, you may object that this does not show that children's brains register the subject of the sentence.

[30] That is clearly nonsense, because to understand a sentence the listener must "register the subject" in some sense (though not necessarily the way linguistics think they do it). Thus the only objection that Pinker discusses is clearly nonsensical, and hence easy to dispatch of. Pinker somewhat expands the objection in the following two sentences, but not in a way that makes sense, i.e. it is not obvious how, according the objection, the children are forming the question.

[31] Pinker's counter-argument to the objection (whatever it is supposed to be) is based on examples like "It is raining" and "There is a snake", and this maybe also regarded as a possible refutation the common sense explanation in [6] above. However, constructs of "it is" and "there is" are special constructs, which are very common in English, and therefore children learn them as a special case. You cannot understand sentences like "It is raining" or "There is a dog in the garden" using the general rules of English, so to understand these sentences children must learn to treat them in a special way. There is no reason not to assume that they also learn to convert them to question as a special case, and they certainly don't lack the evidence for it. The special rules they need to learn are very simple (e.g. "It is X." => "Is it X?"), and in fact they can learn them even before they realize that it, is and There are separate words. Therefore, that children can get these questions right tells us nothing.

[32] Pinker is implicitly doing here a typical error of linguistics, which is the assumption that a feature that can be derived from a general rule is necessarily performed by using the general rule. This assumption is clearly not logically valid, and the only justification for it is that when creating a formal description of the language it makes the description simpler. However, children don't create a description of the language, they learn to communicate by interaction with older people. In this situation the obvious route that any learning mechanism will take is first to learn simple rules (because it is easier to learn simple rules, provided there is enough evidence for them), even though these have limited applicability, and only later learn the more complex rules that are generally applicable.

[33] To make it more difficult for the reader to realize that his examples are invalid, Pinker interrupts the discussion by a completely irrelevant piece from Alice's Adventures in Wonderland (bottom of p.42). This can be justified because it makes the discussion livelier, but it also has the useful (for Pinker's purpose) attribute of making it more difficult to follow the logic of the argument.

[34] At the top of page 43 Pinker also gives as an example sentences with abstract subjects ("subjects that are not things") like "Ask Jabba if running is fun". These examples prove that children can cope with abstract subjects, but nothing else.

[35] Pinker deviates from typical linguists texts by using the movement of auxiliary verbs as an evidence for the arbitrariness of language. First, it should be noted that for somebody that believe that language ability evolved by natural selection (as Pinker does), showing that language is arbitrary is not a supportive evidence for its innateness. Only innatists that believe that language arose by chance can use this kind of evidence. However, Pinker discusses evolution much later in the book, and he relies on his readers to forget the details of the current argument by then.

[36] Second, as discussed in [9] above, the rule is clearly functional. Pinker's way of arguing against the functionality of the rule is to avoid even mentioning the possibility, to prevent the reader from thinking about it. He starts his argument this way (p.42, second paragraph):

The universal constraints on grammatical rules also show that the basic form of language cannot be explained away as the inevitable outcome of a drive for usefulness.

Pinker here uses the standard demagogy of extremifying the opposition position. Nobody claims that the basic form of language can be explained as "the inevitable outcome of a drive for usefulness." Note that as usual, Pinker doesn't explicitly blame the opposition for making such claim, he just implies it. In addition, Pinker writes here about "universals", even though the following sentence makes it clear (to an alert reader) that the auxiliary movement rule is not universal.

[37] Few sentences later he write (p.42, middle):

One could just as effectively move the leftmost auxiliary in the string to the front, or flip the first and last words, or utter the entire sentence in mirror-reversed order (a trick that the human mind capable of; some people lean to talk backwards to amuse themselves and amaze their friends).

[38] The treatment of reversing the sentence is especially interesting. The text in the brackets is pretending to support it as a possibility, but it clearly doesn't, because the claim is that it could be used "just as effectively" as the existing rule, and the fact that some people can do it doesn't show it is easy to use. So why does Pinker find it useful to give such a blatantly nonsense idea? The blatant nonsense effect is an obvious possibility, but in this context it seems to work more by serving as a lightning rod. By giving such a nonsense example, and "trying" to argue for it, Pinker uses up any critical tendency that the reader has, and thus protects the other ideas, particularly moving the first auxiliary verb, from critical thinking. In addition, compared to the other two nonsensical ideas, the first idea seems reasonable. It actually requires some thinking to realize that normally people don't note the location of words in sentences, so this rule will require extra effort, and that it will cause a break in an unuseful position. Thus by building the sentence this way, Pinker increases the chance that the reader will accept that moving the leftmost auxiliary is a reasonable suggestion,



 Yehouda Harpaz

10 Sep 2001