Why Do Illiterates Do So Badly in Logic?

My title is a complex question. Any attempt at answering it presumes a 'yes' answer to the question, Do illiterates do badly in logic? Illiterates certainly do not give the answers to very simple exercises in formal logic that you or I would give, and I am going to explain why. However, my explanation does not account for why illiterates do badly because I do not think that there is anything wrong with their answers. My title intentionally makes use of a misleading question in order to focus attention on the presumption behind the question.

The ethnographers whose work is the concern of this essay are cognitive psychologists interested in how literacy affects the ability to do logic.1 The illiterates they interview belong to a culture that has just had a written language introduced, or to a social class or tribe who customarily never receive any schooling. These subjects, together with a schooled and literate control group, are given very simple logic exercises to solve. The schooled and literate subjects answer them as they are supposed to be answered, but the illiterate subjects do not, and the psychologists speculate on why or how they go wrong.

What interests me is why the answers of illiterate subjects are counted as wrong and why the answers of educated subjects like ourselves are considered to be correct. As it happens, much more experimentation has been done, not with the kinds of exercises that only illiterate subjects seem to find difficult, but with deceptively simple exercises, such as the Wason Selection Task, that even logic students and teachers, myself included, get wrong. My reaction to the fact that I did badly on these exercises was to be defensive and to question whether my answers really were wrong. I distanced myself from the illiterate subjects when I reacted in this way because I would not have given the answers to the simple logic exercises that they did. In this essay I try not to distance myself from them.

8.1 INTERPRETING THE RESPONSES OF ILLITERATE EXPERIMENTAL SUBJECTS

I want to begin by remarking on how strange, if not surrealistic, the problems given to these illiterate subjects are. I marvel at the good will of the subjects and find it touching that they even agreed to participate in the experiments. Despite the apparent mindlessness of the questions, many of their answers have a considerable charm, as is evident in the following exchange with an illiterate Vai tribesman, as cited by Sylvia Scribner (1997, p. 134).2

This reply, although it stands out for its imaginativeness, is typical of the responses of the illiterates because of the subject's reliance on outside information to support his answer, a mistake that Scribner refers to as 'empirical bias.' The interviewer's question really calls for an answer that is based only on the information given, as one of the subjects seems to have understood:

A subject also reveals a form of empirical bias when, like Abdurakhm, an illiterate Uzbek whose interview is quoted in Luria (1976), he refuses to answer because he says he knows nothing about it.

Abdurakhm persists in not giving an answer of the kind the interviewer is seeking because he thinks that the answer calls for personal knowledge which he does not possess.

I have quoted this long exchange because it illustrates how stubborn an interviewer can be in persisting with his questioning despite ample evidence that the subject does not understand what the interviewer is asking. Abdurakhm tries to explain his unwillingness to answer by saying that they have different tsars, by which he seems to mean that he can't be expected to answer questions about a world different from his own. There is an insight in this reference to different tsars because a struggle is going on over whose turf they are to occupy. Even though he does not realize it, Abdurakhm is resisting being driven from his own familiar turf where questions are answered on the basis of his experience, to one where different rules seem to apply, where, as we will see, a specialized discourse is employed.

Empirical bias also accounts for why the subject relies on personal knowledge to confirm (or reject) a premise or the conclusion, as is evident in the following exchange from Scribner (1997):

What the subject says about where schools are built is obviously true, but his reliance upon this truth is wrong (biased) because he is not supposed to be using anything outside what is given.

Luria argues that the responses by illiterates show that they could not think logically.

Luria's belief that illiterates are incapable of abstract thought is based on interviews where subjects are shown cards depicting an object and asked which belong together. Many of the illiterates do not make the selections that Luria thinks they should make, such as picking the hammer, hatchet and saw because they are all tools, and omitting the log because it is not a tool. They do not do so even when asked which of the things would be referred to by the same word, or told that other subjects left out the log, or that the hammer, saw and hatchet are all tools. Luria links their failure to classify correctly to the absence of the cognitive apparatus that makes it possible to derive conclusions from given premises.

Luria's thinking is suspect because it violates what James Hamill (1990) calls a 'fundamental principle of ethnography' (p. 31), namely, that responses not be counted as wrong because they do not conform to the ethnographer's expectations. It does not occur to Luria that he is violating this principle because he believes that objects have essential features on the basis of which they belong to the categories that they do, and that our common nouns are names for these categories. He rejects the reasons illiterate subjects give for grouping things together, such as that the log has to be there in order for the others to be used, on the grounds that they are thinking in terms of concrete and familiar activities that utilize the objects, rather than the abstract terms that Luria thinks should be used in grouping the objects. Luria's assumption that things just are similar or different independently of any interest we have in grouping them is unwarranted; it is that assumption which seems to he in back of his belief in the existence of essential features of objects that make it possible for them to be named (classified) as they are.

All that really interests me is Luria's claim that the responses of illiterate subjects to simple logic exercises show that they are unable to reason logically because they are unable to derive conclusions that follow logically. The claim is suspect when its support is the fact that many illiterates give answers based not on what can be logically derived from the given data, but on what seems true in their experience. For the claim to follow Luria would have to show that these subjects are incapable of deriving a conclusion that follows logically, not from the premises they are given, but from those they actually are using.

What complicates matters is that experiments have shown how difficult it is for illiterates to reproduce the problem accurately. The following interview, quoted by Scribner (1997), with an illiterate member of the Kpelle tribe in Liberia, illustrates the difficulties a subject can have when asked to repeat the problem.

The second repeat, which is not as obvious an attempt at story telling, is closer to the original because of its inclusion of a version of the first premise about all stores in Kpelleland being in town. It still is different from the original because of the omission of 'Kpelleland' and the substitution of land for town in the first premise, and the change from the store being in Kpelleland to its being the one there.

These difficulties that illiterate subjects experience when trying to repeat the problem are a complication because unless we know how they understand the problem any speculation on whether their reasoning is logical is ill founded. Mind you, it is easy to explain why their illiteracy makes it difficult for them to repeat the problem accurately--the ability to read and write helps to make it easier to appreciate the differences between different wordings. But this explanation is based on the assumption that the subjects understand that they are being asked to repeat an exercise in determining what conclusion follows from certain given premises. This assumption seems not to be true not only when applied to the responses of the subjects asked to repeat a problem, but also to the responses subjects give when asked to solve an exercise rather than to repeat it.

Consider the case of Abdurakhm. He seems to think that the interviewer really wants to know about the color of the bears in the far north. And, of course. Abdurakhm knows nothing about it. That is why he claims ignorance, and keeps on insisting that he won't talk about things he knows nothing about. Cora Diamond (1989) reads into his behavior something which the facts do not seem to support, that it is his practice to talk only from personal experience (pp. 24-5). She supplies no evidence that he would not accept what other Uzbeks tell him about his long dead ancestors, or what they saw on their travels or what they heard from outsiders. Moreover, she fails to realize that there is a much simpler explanation for his behavior, namely, that he refuses to answer because he thinks that the questioner really wants to know about the color of the bears in the far North and he doesn't know anything about these bears.3

Or consider the answer of the subject who answers that Boima is a tax collector. If I were in his position, I would be suspicious. 'Why do you want to know? Is Boima in trouble because of not paying taxes?' Instead, the subject is cooperative, perhaps because he thinks he is being asked to do something familiar, namely, explain something that seems hard to understand: 'People are taxed on their homes and yet Boima doesn't pay house taxes. How do you explain it?' This is how he seems to understand the problem, and he responds with a possible explanation--Boima is a tax collector and is exempt from taxation. (Another subject suggests that Boima is too poor to pay the taxes).

The only mistake the subject makes is to suppose that the interviewer really is puzzled about how Boima could own a house and not pay house tax. What Scribner is calling 'empirical bias' seems to be the understandable presumption that the interviewer really has a question or problem. Abdurakhm makes the same presumption, but he does not see himself as being given a puzzling situation to explain, and so he refuses to answer. Neither of these subjects understands that the questioner is only posing a story logic problem designed to determine what effect the subject's illiteracy has on his ability to reason logically.

Scribner's diagnosis is that because of their lack of schooling, illiterate subjects are unfamiliar with story problems, like those used in teaching arithmetic or algebra, that are to be solved by relying only on the information given in the statement of the problem. She refers to these story problems as constituting a 'specialized language genre that stands apart from other genres in ways that may be difficult to define but are readily recognizable' (p. 141). I agree with her that it is our familiarity with this specialized genre that explains why the rest of us give different responses to the logic exercises than do the illiterate subjects. But we disagree about whether the answers of the illiterate subjects are wrong.

8.2 WHAT IS THE RIGHT ANSWER TO THE LOGIC PROBLEM?

I wonder whether a logic exercise like the one about Boima really has the solution that Scribner is convinced it has. Her conviction about its solution is based on the validity of the following:

Scribner thinks what every cognitive psychologist and logician thinks, that the conclusion follows logically from the premises.

Before discussing some of the problems with her thinking it, something needs to be said about the vagueness and generality of the first premise. The exercise is being given to a Vai tribesman, so the reference to all house owners is confined to Vailand, and the present tense applies to the time period when the interview is being conducted. The subject would understand it that way, and that is how it should be understood.

To see that there is a problem with the claim that the conclusion follows from the premises, imagine trying to explain to the illiterate subject who thinks that Boima is a tax collector that the right conclusion is that Boima cannot own a house. Presumably, he would understand the first premise as a statement about what the law requires, and he would find it naive to suppose that the law is always followed. That Boima does not own a house just does not follow when the premises are understood as the subject would understand them.

But, as any logician can tell you, the reasoning seems incorrect only because it has been misinterpreted. If the subject understands the first premise as a statement about what the law requires, then the second premise must be given the same interpretation:

This formulation ensures that the premises are given the reading that is required in order for them to he the premises of a logically correct inference.

The reasoning that Boima must own a house still seems naive. If the law excuses him from paying a house tax then there must be loopholes or exceptions to the law, or so the subject would insist who assumes that the experimenter really has a question about Boima.

However, the logician would object to a reading of the second premise which makes it seem that Boima is being singled out either by the law itself or by someone who is being consulted about his tax liability. The law itself says nothing about Boima; otherwise it would be explicitly excusing him from having to pay what house owners are supposed to pay. Nor is a tax consultant or government official saying that the law on house taxes does not apply to Boima, something there would be a point in saying only if Boima was a house owner. When the premise is interpreted either as making an explicit reference to Boima or as the response of someone who has been told about Boima's situation, the wrong impression is given that what is being requested is an explanation as to why Boima does not pay taxes. To ensure the reading it needs to be given in order for the conclusion to follow logically, we must interpret the second premise as an instruction:

This phrasing is designed as a preemptive strike against the impulse to supply a context in actual discourse for the second premise. A comparable reading should also be given, for the same reason, to the first premise:

Otherwise, there is a possibility of a loophole or exception. As with the second premise, here too the reading makes explicit the fact that it is the logician who is talking, and that the discourse is that of the specialized logic story problem genre.

The logician would complain that if the 'all' in the first premise is read as it needs to be in order for the conclusion to follow, then it must be read as not allowing for any loopholes or exceptions. C.R. Hallpike (1979) seems to think that there is no other reading for 'all' when he complains that illiterate subjects have 'no conception of logical quantifiers,' and so do not appreciate the 'logical implication' of 'some', 'none' and 'all' (p. 121). He is right when he is talking about uses of 'all' in the specialized discourse of formal logic.

But, he is not right if he is thinking about its use in actual discourse about taxation, where 'all' does not seem to function as a universal quantiñer. When you wonder whether you have to pay the house tax, someone operating in an official capacity is not going to say something like, 'The law requires that almost all house owners pay a house tax.' No doubt there are exceptions, but the statement of what the law requires is not going to make a reference to that fact, even though an expert on the law may want to do so. Because a house tax is presumed to be universal, there is little point in the use of 'all'--it is sufficient for the official say, 'House owners pay a house tax.' Moreover, even if the authorities have tried to eliminate any exceptions, and want to emphasize that fact by using 'all' or even 'each and every', nevertheless loopholes are always possible; someone may not be considered a house owner as far as tax laws are concerned, even though he or she really owns one.

Everyday discourse does seem to be the source for a different reading, where instead of referring to what the law requires, the premises refer to taxpaying practices. The first premise, on this reading, becomes something like a census report on everyone's taxpaying practices. Of course, any such report has a margin of error. Perhaps, from the perspective of the person supplying the information, the perspective of an omniscient narrator, the possibility of any errors in the compiling of the census may be considered to be ruled out.

Even so, it is possible for Boima to own a house without paying house taxes if he became the owner of it after the census was done. A rewording of the premise is needed:

However awkward the wording, the intent is to make the time frame be the same for all of the statements, premises and conclusion. Nevertheless, it is hard to see how it is something someone could say outside of the story problem genre. The problem is the reflexive nature of the discourse, the fact that the audience is being told how to interpret, for instructional purposes, what it is being told. No doubt that is permissible when the story is merely the setting for an exercise, but not when it is some other kind of story.

Another problem is that the conclusion that Boima does not own a house does not follow. To see this you have to take the claim that he cannot own a house as a challenge to show how he can. Maybe someone else pays Boima's house taxes because Boima does not want it known that he is the owner or because he cannot afford to do so himself. Then the records would show that the taxes are paid for on his house, but that Boima, the legal owner, does not pay them. The omniscience of the narrator is no guarantee of what does or does not follow logically. Unless the narrator is understood to be introducing what is merely a logic story problem--rather than a challenge to the audience to imagine circumstances where the conclusion does not follow--when he says that everyone without exception pays a house tax he cannot be understood to be ruling out the kind of possibility we have been considering.

Of course, the possibility that the taxes on his house are paid by someone other than Boima can be explicitly ruled out. And we can do the same thing for any other possibility that we think up, such as that Boima is dead, and no taxes on his house have been paid because ownership of his house is still in his name but no one is in charge of his estate. This possibility can also be ruled out by a suitable reformulation, only to leave room for still another possibility.

To thwart my logical perversity, we can treat the question of what follows from the premises as a matter of what can be concluded on paper from the census information:

The conclusion is about what the census data reveals, about what can be concluded from it. This leaves little or no opportunity for challenging the conclusion because on paper nothing changes, and so nothing can happen that can go wrong.

This last reformulation of the problem once again makes explicit that we are operating within the specialized logic story problem genre. The proper reading for each statement in the sequence is that it is a substitution instance of a schema in the following inference schema:

Any question about how to read the schema 'All A (homeowner) is B (taxpayer)' can be answered by reference to what is supposed to be done with the schema--every element that is referred to as an 'A' is also to be referred to as a 'B'. Moreover, the use of schematic letters is designed to ensure that every substitution instance of the same letter refers to the same collection (category; class) or individual, within the space of the argument. The context is Platonic, where nothing happens, and so nothing happens that can go wrong. The tense of each statement reflects that fact by being the timeless 'is' of mathematics. What makes the discourse specialized is that each of the statements of the problem is to be understood in this mathematical sense.

The logic problem should not be understood as the experimenter understands it, especially when it is not being given to logic students. This is the key claim I want to make, that there can be no basis for the presumption that the reading the interviewer wants to force on the problem is the right reading. Perhaps we have to play by the logician's rules when we are on his turf, when we are learning how to solve problems or do exercises in the specialized discourse of formal logic. But subjects in an isolated village who are not learning that discourse should not be expected to do so. So, there is no reason for counting the answers they give as incorrect.

There is also no reason to count as correct the answers that we would give. Like the unschooled subjects of the experiments, we are not students when asked to solve the problems. But, unlike them, we have something to unlearn, something that our schooling makes it difficult for us to appreciate--that we are not always in the classroom. Abdurakhm says that we have different tsars. I am saying that we all have the same one, actual discourse, by which I mean that the specialized logic discourse is being given authority when it has none. So concerned with seizing power are the logicians and their allies, the cognitive scientists, that they are not even aware that there is any power struggle going on when they count as correct the answers they or we would give and as incorrect the answers that the illiterates give.

8.3 THE ARGUMENT THAT THE ILLITHRATES DO REASON DEDUCTIVELY

Abdurakhm really does have the same tsar as we do. This is what Terry Moore (1986) argues. He agrees with Scribner that the interviews with illiterate subjects like Abdurakhm reveal that, Luria to the contrary, at least some of them they are quite capable of 'reasoning deductively.'

Moore also restates Abdurakhm's 'last word' as the valid argument:

Moore's evidence that Abdurakhm can argue or reason logically is that his refusal to answer is 'extractable' as a deductively valid argument.

Moore's reasoning does not seem logical. The main problem with it is that it both generalizes and abstracts what Abdurakhm is saying to an extent which seems unwarranted. Instead of being a response to a specific question about the bears in Novaya Zemlya, the extracted paraphrase attributes to Abdurakhm a very general, albeit awkwardly expressed, statement about when he can and cannot tell. And even if the paraphrase is made less general, it is abstracted from the context in order to satisfy what I referred to in the sixth essay of this collection as the PC Requirement that the premises and conclusions be all there is to the argument.

To see why this generalizing and abstracting is a problem, consider that the paraphrase attributes to Abdurakhm a position that seems indefensible, that he cannot say or tell anything about things he has not seen. He is replying to a question by a stranger about something quite remote from his experience. Surely, he is not to be understood as saying that he would not report to other villagers what he has been told by a scout, that he will never tell even the members of his family anything he has heard from a fellow villager about an incident or adventure that befell that other person.

The generalizing and abstracting is a consequence of the assumption that Abdurakhm is giving an argument when he explains his refusal to answer, and that that argument is to be paraphrased as a proof or demonstration. As we indicated earlier, Abdurakhm seems to think that the questioner really has a question, and so he is bound to be frustrated by his questioner's persistence in getting him to give an answer. Under the circumstances, his refusal to answer is easy to understand as defensiveness at being badgered to speculate about things he knows nothing about, especially when he is in the dark about why he is being asked about them.

What seems to confuse Moore is the fact that Abdurakhm expresses himself so forcefully and in such general terms ('We always speak only of what we see; we don't talk about what we haven't seen.' 'Those who saw can tell, and those who didn't see can't say anything.') By confining himself only to the way Abdurakhm expresses himself, Moore mistakenly attributes a proof or demonstration to Abdurakhm, and so misreads him as reasoning deductively.

Scribner also responds to Luria's claim that the answers of the illiterate subjects reveal limitations in their ability to reason by citing evidence to the contrary. Her response is based on the insight of Mary Henle (1962) that if the subject's responses seem illogical it is because the statements he actually is using as premises and conclusion are different from the ones he is supposed to be using.4 The case Scribner (1997) cites is the following:

Scribner supports her claim that the subject is giving an 'elegant piece of logical reasoning from new evidence,' by restating his response as a sequence of premises and conclusion:

The restatement is confusing for a number of reasons. For one, there seems to be no need for the first premise about Sumo's drinking. For another, the conclusion does not follow unless it is assumed that people like the Chief are vexed only by someone's drinking. For still another, the answer the subject gave was that the Chief is not vexed at anyone, not, as Scribners restatement has it, that the Chief is not vexed at Saki.

Even if we adopt her version of the conclusion, Scribner still is taking liberties with her restatement. The subject says that sometimes when Saki drinks palm wine he will not give people a hard time; as he tells the story, on this occasion Saki was sleeping it off and not giving anyone a hard time. So, the subject's reasoning might be restated as follows:

This formulation misrepresents the subject's thinking. The second premise is much too strong and questionable a claim to attribute to the subject. Surely people get vexed with Saki when he is sleeping it off if he is supposed to be doing something else.

A better reading is that the subject is contrasting the behavior of Sumo, who is a nasty drunk, with the behavior of Saki, who sleeps it off. The second premise should be restated:

Even so, the conclusion that he was not vexed with Saki on that occasion hardly follows. This is because a claim about what vexes the Chief is not to be read as the hard generalization that on each and every occasion when Saki sleeps off his drinking the Chief is not vexed with him. The premise might be reworded by introducing a qualifier such as 'probably. But, the subject is not saying that the Chief is probably not vexed with Saki, and so this rewording won't work.

The problem with all these restatements is that they confuse a narrative with a proof (or demonstration). The challenge to the subject is to find some narrative in terms of which the datum that Sumo is not drinking and Saki is drinking can be incorporated with the datum that the Chief gets vexed when either one of them is drinking. Scribner (and Moore who agrees with her) drops the datum that the Chief gets vexed if either Sumo or Saki is drinking in her restatement of the subject's reasoning because she mistakenly supposes that the story teller is giving a proof, and the premise that the Chief gets vexed if either of them is drinking is incompatible with the premises that Saki is drinking and Sumo is not drinking. Behind this mistake is a more general misconception, that a story constitutes a proof of anything.

Another problem is even more significant, especially when considering cases where something more proof-like does seem to be involved, such as an argument for a position on a controversial issue such as capital punishment. The problem is that the translation into the sequence of statements that are to constitute the argument inevitably distorts or misrepresents it. The problem arises because the translation of what the subject (or anyone else) is saying is based on the assumption that what makes the argument logically correct is a matter of form and so is independent of the content of what is being argued. This assumption leads to the development of a specialized discourse in terms of which the argument is restated. A level of generalization is required when making the translation into that discourse that distorts or misrepresents the reasoning, as we saw with the Boima, or Sumo and Saki examples. Even more disturbing, the discourse eliminates the rhetorical context--what is said in each premise or the conclusion of the restatement is to be understood without considering whom the subject is addressing or why the subject says what he does. The unwarranted assumption behind the reformulation is one that we have already discussed, especially in the sixth essay of this collection, that everything about the rhetorical context that needs to be known can be incorporated into the translation--what is produced or devised for the logic workshop can effectively take the place of actual speech or writing.

8.4 IS THERE A LOGIC LANGUAGE FUNCTION IN EVERYDAY REASONING?

To see why the conflict between us is significant consider how Scribner and I disagree even when we agree. We agree that the subject does not understand what he is being asked. But we disagree about its significance. She thinks that the subject shows that he has reasoning ability, even though he is lost when trying to perform the logical task he is given; whereas, I believe that the subject only seems lost from the point of view of the cognitive psychologist, in terms of which the problem is mistakenly supposed to have a certain solution.

Although we agree that the subjects show that they have reasoning ability, we disagree over how they show it because we disagree over what constitutes reasoning. She thinks that he shows that he can arrive at a conclusion that follows from the premises he actually utilizes, whereas I think that he shows that he can tell a plausible story that includes certain required elements or that explains away something puzzling. I would say that the subject who offered an imaginative resolution to what he understood to be genuine question about whether the Town Chief was vexed that day may be said to have succeeded in thinking up a narrative when others might not have been able to do so. By contrast, Scribner would say the subject is reasoning because his response can be restated as an inference or argument with a valid form. At issue between us is where, outside of the artificial confines of the logic genre, the kind of reasoning she has in mind is to be found.

Her answer is that it is to be found in everyday discourse. She concedes that people do not speak in syllogisms or other argument forms. She maintains that a specialized discourse is needed, the discourse of formal logic, to 'talk about a language function that has hitherto not been isolated from the other functions in which it is ordinarily embedded' (p. 141). This logic language function makes it possible to distinguish the 'topic-neutral' logical relations between statements from their 'topic-bound' truth or falsity.

The key point of conflict between us is her claim that this language function, which is present in the logic problem, also is present in everyday reasoning. Scribner seems to minimize a feature of the logic exercise that I believe is significant, the fact that its solution requires that we do what Hallpike (1979) says that the illiterates fail to do, 'consider statements purely as statements, dissociated from the context of utterance and the status of the interlocutors' (p. 121). She fails to realize that we are making this dissociation when we assume we know what the correct solutions are to the various exercises we have been discussing in this essay. We are reminded of it only when we realize that the answers of illiterate subjects are based on assuming that the information they are given is taken from actual discourse. So, I want to question whether the language functions which are present in the specialized discourse of the logic story problem genre where statements are considered 'purely as statements,' also may be found in everyday discourse, where statements cannot be understood as dissociated from the 'context of utterance.'

At issue between Scribner and me is whether what follows in the specialized discourse of logic really follows. The question is not whether it follows when understood as the logician wants it understood, but whether it really follows when understood as something someone might actually say or think. Scribner thinks it follows because she assumes that there is reasoning in actual discourse that follows because of 'topic-neutral relations rather than topic-bound content' (p. 141). I think that this assumption is unwarranted.

Many logicians will want to argue that if the replies of the illiterate are not correctly restated as deductively valid arguments along the lines that Scribner or Moore proposes, it is because a different logic needs to employed, one that is fuzzier or more dynamic and dialogic. Perhaps they are right, although I doubt it, because none of these logicians has tried to utilize such a logic in connection with the kinds of simple logic exercises we have been discussing, and because all of them, even those who think of argument and inference in dialogical terms, want to offer paraphrases of an argument that end up dissociating it from the context of utterance.

The conflict between Scribner and me has implications for cognitive psychology. Many in the field will want to question her belief in the existence of a context-independent mental logic. Some argue that we reason by implementing mental models (Johnson-Laird and Byrne 1991). Others maintain that certain cognitive biases or inform our reasoning, and that they are responsible for the mistakes we make on such seemingly simple exercises as the Wason Selection Task (Evans 1990)5 Still others attribute our successes and failures to context-dependent pragmatic reasoning schemas (Cheng and Holyoak 1985) or darwinian algorithms (Cosmides 1985). However, all of them assume that the answers that logicians count as correct really are correct; the differences between them are over how to explain why subjects do not give the correct answers.6

What is more to the point of this essay, all of them assume that there is nothing to explain about why everyone other than illiterates give the answers that logic counts as correct to the very simple exercises that we have been discussing in this essay. To see why this assumption is a problem consider that to theorize about how subjects reason we must be clear on what there is to explain. Scribner, for example, is trying to explain why the subjects give the answers that they do, rather than the answers they should be giving. Her question is based on the mistaken idea that the subjects are trying to determine what follows logically from certain stated information. What we should ask instead is why the subjects tell the stories and offer the resolutions that they do. Why does the subject come up with that particular story about Boima being a tax collector or Sumo being a bad drunk, when so many other good answers could have been given? That, as Scribner supposes, a 'schema for handling the type of discourse in the logic problem' (p. 142) is responsible for the subject's story seems implausible, and most difficult to test. It is far easier to theorize about why the subject went wrong, than about why he made the imaginative choices that he did.

An easier question to answer can be asked about the psychology of Scribner and the other cognitive psychologists who have assumed that they know the right solutions to exercises on what follows from certain artificial and contrived premises. Instead of trying to imagine circumstances in which someone would actually say or think such things, they think of the information given as instantiations of schematic letters in quasi-mathematical formulas. Why do they think this way?

The answer is that they are suffering from a logicentric bias. They assume that the contexts in which they are operating when they devise what they think of as samples of reasoning are like contexts of actual discourse. It is this bias that explains why Scribner would attribute an empirical bias to her illiterate subjects when the only thing wrong with them is that they do not give answers she would count as correct. It is the logicentric bias that explains why answers that would be counted as correct in a logic classroom are presumed to apply outside of it. Scribner is right when she thinks that their schooling explains why literate subjects will see that a logic problem has a solution that is based solely on the data that is given. What she fails to consider is whether this schooling is necessarily a good thing.

8.5 CONCLUSION

Let me conclude with some reflections on my psychology in writing this essay. I know that I risk seeming patronizing when I speculate on why the experimenters went wrong. My intention is ironic--I want to return the charge that the illiterate subjects are operating with a bias back to its senders. Also ironic is the implication that I, too, am doing cognitive psychology, when what I really am doing is trying to make a case against the key assumption behind their theorizing about why illiterates do badly in logic. My psychologizing really is part of that case; I do not imagine, as they do, that I can ignore the question of why I am doing it when I speculate on why they think as they do.

If my attempts at irony are not successful, that may be because I am more like the targets of my irony than I would like to believe. If I try to distance myself from them, it is because I want to compensate for the fact that I am like them in thinking that the questions the illiterate subjects are asked really have the answers that logic says that they do. This is not surprising; I am as influenced by schooling as they are. To overcome this influence, I find it helpful to try to see things from the point of view of the unschooled subjects, because they don't have to unlearn what they never learned in the first place. That I do explains my psychology in writing this essay: I am fighting against my impulse to believe what my educated self tells me must be right.7

Notes

1. Verbal logic problems were first used to investigate the influence of literacy on reasoning in 1931-2 by Luria (1976). The experiments he conducted were with peasants in remote regions in Uzbekistan. Some subjects were asked to solve a simple exercise in deduction; some were asked only to repeat it. Cole, Gay, Glick, Sharp (1971) conducted similar studies in Liberia with the Kpelle and Vai, as did Sharp and Cole (1975) with Mayan and Spanish-speaking villagers in the Yucatan. Scribner (1997) conducted recall studies with Kpelle and Vai subjects, which she first reported in 1975. Hamill (1990) conducted similar studies with Navajos (and others). He was more interested in the interpretation of cross-cultural experimental results than the effects of literacy on logical reasoning. Consequently only a few of his subjects happened to be illiterate.

2. Although the reference to bias implies a criticism of the subject, Scribner does not conclude that the subject is being illogical. Her willingness to see things from the subject's perspective may he due to her history. As Cole (1992) explains in her obituary, she had a long career in the labor movement before becoming a graduate student. Although her postdoctoral research was on the effects of literacy on social development, her last years were spent on research that challenges the distinction between manual and intellectual labor.

3. After this essay was published in Philosophical Investigations, Arthur Cody convinced me that my confidence about what Abdurakhm was doing in responding to Luria was misplaced, and that I was as guilty of reading into Abdurakhmn's responses what I wanted to find there as those whom I criticize. This same criticism applies to my readings of the behavior of the other experimental subjects that I discuss. Rather than presume to know what these subjects are doing when they answer as they do, which is what I did, what I should have done instead is suggested that there are other explanations of their behavior than are given by the cognitive psychologists. My argument does not depend on my knowing why Abdurakhm (or any of the others) answered as he did. My interest is in how the psychologists know what is supposed to be the right answer, because their theorizing is directed to explaining why Abdurakhm does not give the right answer.

4. Henle would concede not that Abdurakhm was illogical, but only that he failed to 'accept the logical task,' and so his refusal to answer does not provide any basis for concluding anything about his ability to reason.

5. Wason (1968) in one of the first of his many discussions of the Selection Task. For a more detailed discussion of the significance of the results researchers have gotten on the failures of subjects to perform the Wason Selection Task, see the first section of the next essay in this collection.

6. Cohen (1981) also is sceptical about conclusions about how people reason that are based on experiments where subjects are given problems to solve. Perhaps the most significant difference between us is revealed in a later essay (1986), where he makes clear that he agrees with a diagnosis like Scribner's diagnosis (p. 153), and so accepts as normative what formal logic has to say about the solutions to the problems the illiterate subjects are asked to solve, even though he rejects as normative what Bayesian probability has to say about the solutions to certain statistical problems. Also significant is that he argues that formal logic must be wrong about the argument forms--most notably those involving the conditional--it counts as valid when our intuition tells us otherwise. By contrast, I am not prepared to accept what formal logic has to say about these argument forms, because I believe that anything that instantiates them must be understood in terms of what I have been calling the 'specialized discourse' of formal logic, and because I distrust intuitions such as his which are not based on trying to imagine when someone would actually say or think what he is citing as a counterexample.

7. As I remarked in an earlier footnote, when this essay was published I did not realize how presumptuous I was in supposing that I could see things from the point of view of the unschooled subjects of these experiments. I would like to believe that that presumption does not affect the argument of this essay because my real concern is with showing that the answers the experimenters count as correct are only correct when the questions are understood as exercises in classroom logic. However, I do regret that I did not do enough to resist the impulse to think that I knew what the cognitive psychologists did not know, what the experimental subjects were doing when they answered as they did.

I am indebted to Moira Gutteridge for her helpful criticisms both as a commentator when I read an earlier version of this essay at the 1993 Northwest Conference on Philosophy, and as a critic when I sent her a later version of it.