Human Ways of Thinking:
Common Sense
Logic is doubtless unshakeable, but it cannot withstand a man who wants to go on living.
Franz Kafka (cited in Méro 1990, p. 4)
Most philosophers and logicians are convinced that truths of logic are `analytic' and a priori; they do not like to think that such basic ideas are grounded in mundane, arbitrary things like survival. They might admit that natural selection tends to favor good logic -- but they would certainly hate the suggestion that natural selection defines good logic.
Douglas Hofstadter (cited in Méro, 1990, p. 211)
I don't know if most philosophers today would agree with Hofstadter's characterization. Many have participated actively in the explosion of studies on cognition in the past decades. These studies have modified many of our notions of human thought -- its development, its relationships to language, the importance of social scripts, deductive reasoning, and abduction. This chapter reviews some of this research with the aim of pulling together elements that can help us understand the possibilities and limits of our thought.
Cognitive Development
The continuuity between animal and human cognition is especially evident in the first years of life. Various researchers have examined the acquisition of Piaget's developmental stages in different primates. In Piaget's theory, children pass through a series of stages, each characterized by the first appearance of some new cognitive skill. Piaget divided the first level of development -- the sensorimotor -- into six stages: In the first stage reflexive capacities are developed. The sucking reflex, for example, is perfected when the child recognizes the contact of its cheek with the nipple, and is able to turn its head to be able to suckle. The child also learns to follow with its eyes, a slow-moving object. According to Piaget, the child at this stage does not recognize "objects" with substance or permanence, but lives strictly in a world of phenomena, some of which the child can provoke (like moving its mouth toward the nipple).
In a second stage, the child learns to coordinate the schemata acquired in the first. At first the child only sucks its thumb if it is by chance somewhere near the mouth. Little by little it learns to bring its thumb to the mouth. The child also learns to link the visual and the tactile, looking at an object put in its hand, for example. With time, the simple fact of looking at an object is enough for the child to attempt to grasp it.
In the third stage (which lasts from 4 to 9 months), the child begins to repeat actions to produce different effects. For example, it hits a toy repeatedly to see it move. In this stage the child seems to have goals -- it wants to produce different effects--, and seems to recognize "objects" and not just phenomena, since it can return to hit an object after it has moved. This suggests the child has acquired a notion of the permanence of objects. Still, the child recognizes this "object" only in terms of the schemata used to invoke it. For example, if an object is removed, the child repeats its actions (for example, extends its hand) in an attempt to recover it. But the child will not look for the object in space. It also does not recognize the object if turned in another position, and loses interest if an object is hidden behind a cloth. The child acts as if the object simply disappeared.
In the fourth stage the child learns to perceive an object in different positions and to explore these different positions, looking at the object from different angles, for example. Also, the child looks for an object that has moved. And when a desired object is partially hidden, the child learns to move the obstacle. But the child has difficulties with higher abstractions. For example, although it looks for an object hidden under a pillow, it does not look for the same object if it is visibly then hidden in another place. The child does not seem to understand the notion of "hide," but continues to activate the schemata that recovered the object the first time -- it looks under the same pillow.
In the fifth stage the child begins to carry out small experiments to see what objects will do. According to Piaget, this requires another level of abstraction, since the child begins to perceive the object as having properties of its own, and not simply as something that appeared when the child activated some schemata. The child may let an object drop several times to observe it, or may put it in a position to roll, etc. It is in this stage that the child also learns the relationship between objects. For example, it learns that an object may not pass between the bars of its playpen because the object (and not just the hand) is obstructed by the bars. It is also in this stage that the child learns to use one object to acquire another -- for example, it may pull a cloth to obtain a toy that was on top of the cloth.
It is in the sixth stage that the child learns to simulate events, that is, imagine sequences of actions without having to see them first. The child can represent the direction of a house without seeing it. This is the period when language begins.
The passage to the other levels of development (pre-operational, concrete operations, and formal operations) that follows this first phase involve the same principles: With experience a child learns certain regularities, abstracts on these regularities, abstracts on these abstractions, etc. One example is the sequence in recognizing objects, then recognizing number of objects, then relations between numbers (adding, for example), etc. This sequence of cognitive skills has been examined in many different societies. As Ember (1977) and Dasen and Heron (1981) have noted, these studies have found the same sequence in the appearance of these different skills in the most diverse cultures, although the ages at which they appear may vary a little. This is the case of the sensorimotor stages, pre-operational and concrete operational stages. But it is not true of the formal operational stage which does not appear until adolescence, and which has not been verified to occur everywhere. There are also variations in what Piaget called décalages horizontales within the same cognitive level. For example, comparative studies have showed that in the age of concrete operations Australian Aboriginal children were better earlier at spatial tasks while European children where better earlier at conservation of liquid tasks. Dasen and Heron (1981) argue that the Aborigines give more importance to spatial relations (due to the need to always find one's way in the desert), while European children did not need to learn this.
More impressively is that the same sequence of sensorimotor stages also occurs in several non-human primates, such as macaques, cebus, gorilas (Antinucci 1989) and chimpanzees (Spinozi 1993). And the last levels of cognitive development achieved by families of primates (with some exceptions) generally correspond to phylogenetic relations (Parker 1985). Lorises and lemurs show the tendency to grasp reflexively characteristic of the first two stages of sensorimotor development, but never demonstrate the skills characteristic of the fourth and fifth stages (Box 1984, p. 195). Gorillas and and cebus achieve stages 5 and 6 in some areas of sensorimotor development. Another contrast: the macaques and cebus can organize sets of objects based on a single criterion (which may vary), but they cannot classify objects simultaneously according to two criteria (ex. by color and by shape). Chimpanzees can (Spinozi 1993).
The few differences found between primates and humans are explainable not in terms of differences in cognitive development, but rather by differences in physical development. Human babies cannot grasp objects until after being able to coordinate hand movements with vision, but gorillas (Spinozi and Natale 1989) and cebus (Spinozi 1989) grasp objects before being able to coordinate vision and hand movements. The reason is simple. These animals grasp objects with their mouths. Human babies at this stage do not possess the physical capacity for this.
Piaget was able to find tasks that form a sequence in terms of difficulty for children the world over, and even for other animals. This, in itself, is a great achievement. But there is disagreement over his interpretation of these sequences. Anderson (1990) questions the Piagetian explanation of sensorimotor development during the fourth stage with regard to the child's inability to find objects hidden and then transferred . He suggests that this has nothing to do with the child's inability to recognize the substantiality or permanence of an object as Piaget theorizes. Rather it has to do simply with memory limitations. This is seen by the fact that a child looks with equal likelihood behind any cloth (or pillow), even a third cloth that was never associated with the object. In the same way, the incapacity for recognizing conservation of material in the pre-operational level may be interpreted as resulting from a confusion due to strange stimuli (glasses in different shapes and sizes, for example). Besides the studies in our society, research in other cultures also support this position. People in different cultures pass the test of conservation of material when known materials are used (like clay among potters), but do not pass the tests if other materials are used (Cole and Scribner 1974). Possibly the use of familiar materials implies less distraction from the task at hand. Thus, for Andersen, some of these cognitive developments may consist simply in the development of better memory, or in a greater capacity to process more information at the same time. They may not imply conceptual changes as Piaget theorizes.
Another questioning of Piaget's theory comes from McKenzie (1990). He cites research that compares different levels of interest (measured by the time an infant looks at something) of newborns. This research shows that newborns recognize the differences between trapezoids and squares, even when shown in different angles.This suggests that they are born with a certain capacity for construcing "objects." Thus objects are not constructed out of experience with phenomena. Pinker (1994: 424) cites similar studies showing that neo-nates not only recognize objects, but also understand the interaction between two objects that collide, and they distinguish between objects and living things (that do not collide).
Also questioning Piaget, Kagan (1972) shows that children at nine years of age can elaborate simple hypotheses, in the sense of recognizing patterns (in this case musical patterns), and creating expectations with respect to the next element in a sequence. In Piaget's scheme, this ability should appear only much later.
To sum up, cognitive development in children is very similar the world over, and has many points in common with cognitive development in animals, especially in the great apes. The developmental stages elaborated by Piaget have been well studied and have received a good deal of support. Nevertheless, there are reasons to doubt whether the construction of thought proceeds according to the Piagetian theory. Some skills considered by Piaget to be constructed via experience may be innate (the ability to recognize shapes, for example), and other capacities (such as the ability to formulate hypotheses) may appear too early. Other stages (such as the décalages horizontales) show clear environmental influences. Perhaps it would be better to see cognitive development as being a bricolage (tinkering) that arbitrarily puts together some more specific preprogrammed abilities and other more general capacities that permit greater diversity in learning. Unlike the systematic and consistent theory of Piaget, this bricolage would be more typical of the opportunism of natural selection.
Language Development
We have already looked at ideas about the phylogenetic origins and linguistic capacities of humans. Here we need to look at its ontogeny. As Bowerman (1981) noted, the academic interest in how children learn to talk grew out of Chomsky's ideas with respect to the existence of a language acquisition device in the human brain. Different researchers want to know if this device exists, and, if so, what its relationships to other aspects of cognition might be. Some ties with general cognition seem clear. First, in all cultures, and for all languages, the appearance of some linguisitic skills occurs correlated with sensorimotor capacities. For example, the first words appear when the child begins to walk. Linguistic fluency occurs when children can walk tip toe or walk backwards three meters. Besides this, the "critical period" for learning a language is between birth and puberty. A language learned after this period remains defective (you speak with an accent, for example). This is true even for deaf mutes who learn to speak (using sign language) for the first time after puberty: they speak the gestural langauge with an accent (Andersen 1990). Compared to small children, adults and older children learn new languages more rapidly, but only those who learn younger learn the fine points (Andersen 1990).
The universality of these stages in language learning suggests that there must be a correspondence between general cognitive development and linguistic skill. But it is still not clear if this correspondence is due simply to independent parallel development of the language acquisition device, and other cognitive skills, or if there is interaction between these. The idea that these skills may be relatively independent is supported by several types of evidence. First, as we saw in the last chapter, animals already think with many concepts, but do not have language. Also suggestive are the separate brain areas dedicated to concepts and language, especially as seen in brain-damaged people. To these cases we might also add the cases of the "chatterboxes" (Pinker 1994) who, because of brain damage, cannot think well at all, but whose linguistic capacity is quite good. Research on children's language learning is also revealing. Lorenz (1973) cites the case of the deaf and blind Helen Keller, who first began to "speak" when she was almost seven years old. She seemed to have already developed many important linguistic capacities. In less than one month after receiving the first signs from Anne Sullivan, Helen Keller could use 18 nouns and 3 verbs, and had begun to ask about the names for things. Within three months she was already reading and writing in braille. More recent research with languageless deaf-mutes confirm the reports of Helen Keller (Pinker 1994). Obviously, these children already possessed many concepts (and linguistic capacities) long before they could talk.
Research with normal children also show the relative independence of concepts and language. Perhaps the most revealing research deals with bilingual children. Slobin (cited in Bowerman 1981) studied children brought up speaking Hungarian and Serbo-Croatian. He noted that before two years of age these children correctly used inflections to indicate locatives such as "inside of " "outside of" and "on top of" when they spoke Hungarian. But these same children could not use the equivalent locative forms in Serbo-Croatian. For other linguistic forms the children were better in Serbo-Croatian. These observations suggested that children already understand the concepts they want to express, but cannot always express them. The difference in the linguistic capacities in the two languages could be explained by differences in the difficulties of the languages themselves.
These examples clarify the differences between language and thought. To this distinction I would also add the distinction between language and non-linguistic communication (see Burling 1993). This includes body language (like laughing), or vocal sounds as in cries, shouts or singing. The independence between language and these forms of communication becomes clear in brain-damaged individuals who cannot speak, but who can sing, as well as produce and recognize other non-linguistic communication signals (Pinker 1994). Universals can be found in language, concepts and non-linguistic communication.
Language
As Pinker (1994) points out, hundreds of linguistic universals have been identified -- some absolute, others more a question of preferences. An example of an absolute universal is that apparently no language uses the word order: object-subject-verb. Also, all languages place adjectives near the nouns they modify (Andersen 1990). Many languages take the form of implications. For example, if a language has nasal phonemes, it will also have non-nasal phonemes. Other universals take the form of universal preferences. For example, if the basic word order of a language is subject-verb-object (as in English), then that language will have prepositions and will generally put question words at the beginning of the sentence. If the language has a word order subject-object-verb (e.g. Japanese), then that language will have post-positions and will generally put question words at the ends of the sentence. Also, generally topics (the already known contexts) precede comentaries and old information generally precedes new infomation (Givon 1989: 222-223).
Slobin was one of the first to argue that the "language acquisition device" consists in learning procedures that give preference to some linguistic structures over others. Basing himself on analysis of language learning in 40 different tongues, Slobin suggested the existence of seven universal "principles": 1) pay special attention to the ends of words. Only if you cannot find the meaning at the end should you look at the beginning. For this reason the Hungarian locatives (which occur afterwards) were easier than the Serbo-Croatian locatives (which occur before, 2) pay special attention to the order of words and morphemes, 3) avoid interruption or rearrangements of linguistic units, 4) mark semantic relationships clearly, 5) avoid exceptions, 6) look for semantic meaning in the grammatical markers, and 7) feel free to systematically modify phonological forms of words.
The existence of universal preferences for expressing oneself is also apparent when studying the errors made by children in learning to speak. For example, everyone has observed the use of principle 5 (avoid exceptions) in children's speech -- `he bited me!' instead of `he bit me'. The universal principles also become apparent in studies of creole languages in different parts of the world. These languages originate when people of different language groups are thrown together (as with slavery). To communicate with each other, during the first generation the adults develop a type of language known as a pidgin, which is limited in the types of messages that can be transmitted. Generally pidgins are based on the ruling class language. But the children in these communities must speak with each other, and they develop a complete language based on elements from all of the languages in their environment. As Bickerton (1985) has noted, all creole languages in the world possess some characteristics in common, and these characteristics correspond to the "errors" children typically show in learning a language. For example, conforming to principle 5 of Slobin, all of these languages avoid exceptions. Also, in agreement with rule 3, these langagues do not modify the order of words to change declarations into questions. These questions are expressed with intonation (In English, "he went to the store." becomes "he went to the store?" instead of "did he go to the store?"). Besides this, all use double negative ('no want nothing').
These principles are only preferences (or default options to use computer jargon). This leaves a lot of margin for cultural differences. Only when the default option cannot account for what the child hears, are other options considered. Still, the simpler forms of language may continue to be easier to understand, even in adults. Perhaps this helps to explain something discovered by Bichakjian (1988) in his comparisons of changes that have occurred in Indo-European languages during the past 5000 years. Amazingly, all of these languages seem to be changing in the same direction! They have all dropped morphological and phonological forms that children have difficulty learning. For example, Russian children do not sort out all of the proper declensional suffixes until they are 7 or 8 years old. This is very late in development. Proto Indo-European was characterized by declensions, but modern languages have been dropping these grammatical forms, and no modern Indo-European language has ever added a declension. Similiar phenomena have been occurring with complicated phonemes. Bichakjian cites evidence that a similar process has been occurring in Sino-Tibetan languages. Perhaps the increased communication needs of more complex societies have made it important to eliminate all unnecessary linguistic complications so that the really important information can get passed on clearly.
Non-Linguistic Communication
Universals in non-linguistic communication are well known. For example, in all human societies people recognize the same facial expressions for anger, fear and happiness (Ekman 1982), and use the same signs in similar circumstances (for example, raising eyebrows on greeting someone) (Eibl-Eibesfeldt 1967). Some of these forms of communication may also mix with language. For example, Fónagy (1988) found what seem to be universal "errors" in adult speech. We all apparently round our lips (as in a kiss) when speaking of something with affection, and make our voices tense when we are angry. Many of these forms of communication show up very early in life -- even in neo-nates (Alexander 1987; Pinker 1994; Eibl-Eibesfeldt 1967).
Concepts
We should not equate words with concepts. Nevertheless, the appearance of the same words in the same order in the children's speech of different languages, suggests the universality of these concepts. As Macnamara (1978) points out, compared to adults, children use nouns that refer more often to concrete objects, and less often to abstract notions, and their verbs more often refer to actions. In addition, the first nouns learned by children refer to stable objects, never to unstable objects (like smoke). Macnamara argues further that children are very sensitive to other characteristics of objects. For example, they pay more attention to whether objects extend in one, two or three dimensions (such as string, paper, or ball, respectively). These distinctions are especially marked in many languages. Also, children are very sensitive to the difference between objects that are quantified by mass or by number (a little milk, but a few cookies). We still need to verify in cognitive tests if children intuitively use these categories (or at least learn them more easily) in tests of classification.
Givón (1989, p. 125) also noted the universal tendency to construct vocabulary from more concrete words, and adds that, compared to other languages, pidgin languages use a higher percentage of concrete words. This tendency to construct abstract concepts through the use of concrete concepts is seen in the etymology of many words. Givón (1989, p. 150) describes etymology as "tracking" how analogies are used to construct abstract concepts. As Lorenz (1973) suggests, many of these analogies deal with spatial orientation (We are "deep" or "superficial" "superior," or "inferior," "ahead of" or "behind" our colleagues, etc.). Even as adults we continue to find concrete words more salient. People evaluate more easily and more accurately sentences spoken with concrete than with abstract words. Concrete words evoke more images. Perhaps this explains their positive effects in publicity (Percy and Rossiter 1983). It's worth noting here that Lévi-Strauss's (1962) view of the "savage mind" also applies to the "civilized mind" -- we both construct more abstract concepts, like bricoleurs (or tinkerers) do, putting together rather opportunistically different concrete concepts. If some intellectuals (like "engineers", according to Lévi-Strauss) attempt to define concepts in more abstract terms, this is a question to be dealt with in the next chapter.
Besides concreteness, there are also other conceptual characteristics of a child's first words. For example, the first utterings usually aim to achieve some objective ("gimme," "mommy," etc.). Only afterwards does the child begin to use phrases that show it understands the nature of objects that are not present, or the perspective of other people. Comparative studies also suggest that there is a universal developmental sequence in the use of forms to join phrases. First, phrases are joined by "and" then by "because" then by "before" then by "after." In the same way, locatives also follow a sequence of increasing difficulty when they enter the child's vocabulary. ("in" before "between" for example) (Bowerman 1981).
Cultural Variation and Universal Cognition
Many years ago Benjamin Whorf, an authority on native American Indian languages put forth the idea that languages not only express our thoughts, but also mold them. For example, Whorf argued that Hopi did not possess direct references to Time, to the past, present or future. He argued that this reflected a profound difference in the ways the Indians saw time, since they did not think in terms of a continuous process. Givón (1989) criticized as very "naive" Whorf's grammatical analysis of Hopi, and the linguist Pinker (1994) saw it as one of the great anthropological hoaxes.Whorf's analysis is unacceptable today. But the question of the relationship between concepts and language continues.
Cultural differences in the terms used for colors illustrate well some of the questions involved. As Cole and Scribner (1974) point out, upon discovering that Homeric Greek had no terms for many colors, some authors imagined that they must have been colorblind. Research in other cultures such as the Dani of New Guinea showed the error of this idea. The Dani have only two terms for colors: "light" and "dark." Nevertheless, they perceive equally well the whole gamit of colors. Also, the Dani, like Europeans, recognize the special status of certain colors called focal colors. These are "primary" colors -- "real" red (which is more red than other more-or-less red colors)., "real" yellow, blue, and green (Cole and Scribner 1974). In memory tests, even the Dani remember focal color chips better than other colors. The emotions associated with these colors also seem to be universal. For example, more saturated colors are seen as good, or happy, while more washed-out colors are seen as weak or sad (Johnson 1986).
With American subjects, Boster (1986) showed that there is a sequence in the salience of different colors. When asked to successively divide colors into two categories, Americans first separate lights from darks. With three categories, reds are separated out. The four category consists of lighter yellows-greens and blues. With five categories the yellows get separated from the greens and blues. Then the greens get separated from the blues. Research in other societies shows the same sequence. Politically and technologically simpler cultures tend to have terms only for dark and light. With slightly more complexity a term for red enters the vocabulary. Then a single term to refer to blues, greens and yellows. Then the yellows are distinguished from the blue/greens. Only in the most complex cultures are there terms to distinguish blues from greens, and to refer to other tones, like purple, pink or brown.
How might this cultural variation be explained? One possibility is that having a term for a non-focal color helps memory. Where a language has a word for a color, or can at least express it easily, people find it easier to identify this color at a later time (Cole and Scribner 1974). Perhaps in more complex societies where people work with paints, or other colorings, we simply need more terms in order to communicate and remember them.
These studies of color show how ideas about cognitive universals might be used to explain cultural variation. If we think that a concept, or the association of concepts with emotions is universal, we can make predictions about cultural preferences. For example, if red is universally associated with action and excitement, we can try to see if cultures where action is highly valued also tended to prefer this color (warring societies for example). If a tense voice and heavy accentuaded speech are universally associated with aggression (and remember that Fónagy found this in the speech "errors" of the French and Hungarian), then we can try to see if people in societies with more aggression sing with more tense tones and heavily marked accents (see Lomax 1968 for verification of this idea). If certain facial expressions are universally associated with aggressivity, we can verify if these facial expressions are especially prominent in the masks or artwork of more warring societies (see Aronoff 1975 for an examination of this idea).
There is good evidence that concepts precede words, so there is good reason to believe that thought affects language. But the effects of language on thought are more difficult to verify. Outside of its effects on memory, is there any other evidence that language can affect thought? The Russian psychologist, Vygotsky (1991) thought he had clear evidence for this in the "private speech" of children. Children between four and eight years of age talk a good deal to themselves, and Vygotsky saw this talk as a bridge between culture and thought. In his view this speech has social origins (and so varies from culture to culture and from person to person), and affects our later thought, because with time this private speech is internalized and becomes an internal speech that guides our thinking as adults (Diaz and Berk 1992). In an extreme version of Vygotsky's position, Bakhtin considers this speech as the way our experience is simultaneously organized and expressed. For G.H. Mead, the consciousness of an act does not exist before the communicative act (Ramirez 1992).
The data on linguistic learning, and on aphasics, together with data on thinking in other animals without language, suggest that this extreme form of Vygotsky's theory is not correct. Concepts exist before language. To consider language as the only way to organize experience doesn't even explain some things about private speech, such as its limited use in non-verbal tasks (Goudena 1992, p. 219). Also, various researchers (Diaz and Berk 1992) have shown that the use of private speech helps with problems of intermediate difficulty, but not with problems that are too easy or too difficult. Finally, not all private speech is aimed at solving problems. Sometimes children sing, recite rhymes, or play other linguistic games.
But if we consider a less radical version of Vygotsky's theory -- that private speech is a way (among others) to organize and guide experience, and to solve problems, the theory seems more valid. I think we can better understand how private speech works if we consider it as one among other ways of facilitating thought or access to memory. We have already seen how being able to describe a color helps us remember it later. It is in the learning of algorithms for action that private speech most seems to help children: "first a block here, then this block..." The explanation for this help might include several elements. In part this internal conversation may make the child remember earlier conversations, helping it reconstruct a series of instructions. One problem with this idea is revealed in a study by Azmitia (1992). She showed that conversations with "specialists" (more skilled children) helps children solve a problem when they first encounter it, but they do not reproduce this conversation in later attempts when alone, and are no better able to solve the problems, although they continue to use private speech. Another possibility is that the private speech simply leaves the child more motivated to solve a problem, or that private speech helps a child keep its attention on the task at hand. A third possibility is that private speech, especially when it consists mostly of concrete words, helps evoke images that guide thought. Future research will need to solve these questions.
Cultural and Personal Scripts
All animals with a minimally complex nervous system have routines that consist in a generalized "representation of classes of events with recurrent patterns" (Smith 1991, p. 224). In the case of simpler animals these routines are more stereotyped. Ants that smell the "death smell" painted on a live ant, will carry the dead-smelling ant out of the nest, time after time (Wilson 1980). These routines can be more or less fixed or flexible, both in terms of what is learned, as well as when it is learned. Honey bees, for example, seem to have an "agenda" in which slots must be filled with information on time, smell, color, size, form, landmarks, and location. If one of these slots is not filled in, the bee will not record anything, and if one of the items is modified, the bee must relearn the whole set (Gould 1979). The moment of this learning is also fixed. The bee learns the location and landmarks only when leaving, for example.
In human language learning we have already come across this interaction between fixed and flexible elements. For example, there is a critical age for language learning, and there are universal preferences as to phonemes, the characteristics of objects that serve for categorization, word order, etc. But even so, human languages vary tremendously. Now it is time to look at these same questions with regard to human "scripts." Michel (1991) defines these scripts as consisting of "concrete episodes of action and reaction" and argues that they serve as prototypes for classifying events and for guiding our action. We select a particular script to represent the situation, and assume one of the roles in the script. This lowers the cognitive cost of always having to evaluate and judge a situation. In the words of Fayol (1985, p. 69) we are dealing with "prêt-a-pensers", which we might also call "canned thoughts" that we can easily mobilize when needed.
Some more repetitive scripts may become almost reflexive. On learning to drive or play the piano we need to concentrate a good deal on what we are doing. But after repeatedly coming across similar problems, we soon learn to turn on the "automatic pilot." We recognize the stereotyped situations and execute the necessary actions without much reflection. Other times we are more attuned to possible changes in the script. In social interaction it is common to provisionally accept a script , but we remain alert to possible changes. Goffmann (1961) and his disciples have dedicated a good deal of attention to the ways people manipulate these scripts. We can be more or less "conscious" of these scripts. Perhaps it is easier to recognize when someone else is following a script than ourselves. It is common, for example, to hear friends comment "I've seen this film before" when someone begins a fight, a love affair, or some other typical undertaking.
Scripts may be learned alone or via contact with others. For example, the sequence of actions in taking a shower may be based on individual experiences. People may never have seen others take a bath or listened to details. Some may let the water run before entering the shower stall. Others may enter first. Some may use a washcloth or a sponge, others just soap. Some may shave under the shower, others only at the sink, etc. Individually learned scripts are not necessarily idiosyncratic. Because people are confronted with similar problems, they may often come up with similar solutions. Those who have sensitive skin, for example, may be more likely to have discovered more or less by chance that shaving under a hot shower causes less irritation than shaving at the sink. We might then expect a correlation between shaving under the shower, and senstive skin. By the same token, peoples' ways of washing dishes may also have to do with common problems. People who live alone and have fewer dishes to wash may wash under running water, while those with large families may wash the dishes in a basin. In both of these cases we would be dealing with shared and learned scripts, that are not based on communication with others, symbolic or otherwise. This point is important when we consider the concept of "culture" in anthropology. Should we consider as "cultural" only that which is transmitted symbolically, or should we consider any learned and shared behavior of a group as cultural?
Whether idiosyncratic or shared, how do we learn these routines? In the case of the ant's reaction to the "death smell" there seems to have been no learning at all. But in the case of the bee's "agenda" there are specific "slots" that must be filled in with learned information. In more complex animals, the role of learning is clearer. In the previous chapter I distinguished between "imitation" and "social facilitation". We saw that many animals learn by social facilitation. That is, they observe another animal doing something and become "inspired" to experiment with the object or technique that they see. (Remember the case of the bottle-opening tits in England). But these animals do not seem to imitate in the sense of following and repeating what they see. For some researchers (e.g. Visalberghi e Fragaszy 1990; Tomasello 1990) imitation, not just social facilitation is necessary in order to have culture. But do humans really imitate?
The idea that we learn by imitation is very dubious. First, we find it very difficult to imitate. For example, few people can really imitate the accents or gestures of others, or follow the tricks of a magician. More systematic research also suggests that our minds do not work by imitation. In the previous section I cited research on "private speech". These studies show that children do not repeat what they learned from another child or adult, and cannot imitate their solutions to problems (Azmitia 1992). Also revealing are studies in which people are asked to remember a sequence of words or objects presented to them. People do not recall these in the order of presentation. Rather they reorganize the items, taking advantage of new classificatory schemes (Cole et alii 1971).
Doubtless, there are some differences in the ways humans learn. Our scripts are more flexible, complex and varied than those used by other animals, and, more than with other animals, may be constructed via "mental experiments" which go beyond conditioned learning. Some cognitive ethologists argue that the capacity to see things from the point of view of different social roles was selected to allow for social manipulation and the detection of deceit. This capacity would be responsible for the "interpenetration of phenomenal fields" that permits "comprehension" and "empathy" (Michel 1991, p. 260). It is perhaps this capacity for "comprehension" that has helped us to elaborate or apply many of our "canned thoughts" to physical phenomena.
In any case, we should recognize that our scripts also have their limits. Probably, the more "basic" these scripts are in terms of survival, the more "instinctual" they are. The ways we chew food are relatively fixed, for example. But the ways we sing or tell a story allow for more variation. Yet even with these more "abstract" scripts, that are divorced from immediate necessities, there is evidence of biases and limits.
Fayol (1985) has reviewed research on universals in narratives. These studies show that people generally consider as narratives only texts that oppose an initial state with a final state (something happens to change the life of one of the characters, for example). Generally, on recalling a narrative people re-organize the text, sometimes introducing logical connections that were not in the original, and eliminating details that are logically superfluous. There is also a tendency to expect certain elements at specific points in the narrative. If an element appears at other points, it takes longer to understand the narrative, and more errors are made in its reconstruction. Generally we expect a summary or a contextualization at the beginning, then the statement of a goal, then perhaps, sub-goals. The later events should be linked to these goals. Changes in perspective need more treatment to be understood. For example, if a narrative is written from the point of view of a child who goes out to by an ice cream cone, it is easy to understand that the child bought an ice cream cone. But it is harder to understand that an adult sold an ice cream cone. It is also easier to understand narratives whose events follow a real-time sequence. For example, it is easier to understand "John ate dinner before taking a bath," than "Before taking a bath, John ate dinner." For Fayol these expectations are not limited to narratives, but also apply to other phenomena such as films, instructions, and solving daily problems.
In childhood, the development of a sense of narrative also follows a uniform sequence. Children from four to seven years old consider as a story, any report, as long as there are characters. Later, children reject "narratives" composed of semantically unrelated statements. It is only between nine and 10 years of age that the child begins to reject stories without all of the necessary elements for a complete narrative.
These studies on narrative show that we are similar to honey bees and their "agendas," in the sense that we have a predisposition to expect specific elements at specific times. I suspect this applies to other less abstract scripts as well. These studies also illustrate our tendency to change and to reinterpret narratives we hear (and other scripts as well). These reinterpretations have been the object of study in cultural anthropology. For example, Leal (1985) documented the different interpretations given to Brazilian soap operas by lower and middle class viewers. The "why" of these reinterpretations is the object of greater debate. In part these cultural scripts may be modified to better fit already existent symbolic schemata. But whether scripts need to fit together is not clear. Strauss (1990) analyzed the beliefs of three working class men in the United States, and noted that at the superficial level of discourse, the three seemed to share the same beliefs with respect to why some people get ahead economically, and others do not, including apparent contradictions between individualistic beliefs (It's the courageous or working individual who gets ahead) and the belief in exploitation (people get ahead because they exploit others). But more detailed analysis showed that the three had organized differently these "contradictory" beliefs. One simply separated the two discourses and applied them in different contexts. Another attempted to integrate the two discourses in a more consistent whole. But in spite of the different levels of integration of their discourses, the three workers held similar beliefs. This study, then, shows that the "need" for symbolic integration is not universal.
Cognitive psychologists and advertisers (who have a very practical interest in this matter) have suggested other ideas about when and why people accept "canned thoughts" and when they don't. Several researchersthat most of the time people simply do not stop to reason about what they are doing. They simply enter a routine (script) and continue in it. What, then, would make people stop to think? Hewstone (1989) reviewed psychological studies on this theme. These studies have taken various forms: 1) Content analyses of newspaper articles were carried out, in which the moments were coded when the authors felt obligated to provide an explanation or not. Correlations were then sought between types of events, and types of explanations. 2) Observations of individuals (like prison officials) at moments when they must make decisions, and 3) experiments in which different situations were created (either verbally or through videos), and subjects are asked to analyze the situation. In these cases, attempts to seek more information, or to attribute causes were correlated with the elements being analyzed. Three main factors have arisen from these studies.
First, people tend to reason more when a situation is unexpected. Normally, in ordinarly situations, people simply describe the events with the use of a known script, without explanation. This makes sense in terms of economy of cognitive effort. Givón (1989, p. 293) emphasizes that questions like "why" are appropriate only if the questioned items are surprising, and the degree of surprise depends on our pre-existing schemata. It seems that the cognitive algorithm adopted is: first look for a known script. If it works, stay with it. If the situation does not follow expectations, then attempt an explanation. This also makes sense in terms of information processing. Automatic processing is faster and more efficient when information is predictable, certain and consistent (Givón 1989, p. 251). The way people view unexpected events is revealed in an experiment in which individuals, on seeing a video, were instructed to press a button whenever there was a new significant action. People pressed the button few times on seeing ordinary seens, but in seeing less familiar scenes they pressed the button more often. In this case, the subjects could not sum up these sequences with "canned thoughts."
The second factor that inspired explanation attempts was failure. People are more likely to explain why they failed at some task, than why they succeeded. In part this might be explained as an attempt to rescue self-esteem, or justify failure. But additional research suggests another perhaps more important factor: the need (innate?) to feel "in control" of a situation. As O'Connor (1991) notes, research shows that the loss of a sense of control leaves people more stressed and more depressed. On the other hand, a feeling of control over a situation leaves people happier. O'Connor suggests that this helps to explain the pleasure we feel in solving puzzles or other practical problems. Usually, playing is pleasant mostly because in play we learn to control more and more situations. All of this makes sense in terms of natural selection. An animal that felt pleasure in solving problems would have a greater chance of passing offspring on to the future. An animal unhappy about failure would be more motivated to stop and try to solve the problem. In line with this view is the finding that bad moods are associated with more attempts to reason. People in good moods use more heuristics (canned thoughts), they reason less systematically, and they reduce the complexity of their judgments and decisions, which they make rapidly and simply (Hewstone 1989; O'Connor 1991). Supporting the same argument, Percy and Rossiter (1983) discovered that people in good moods evaluate more positively the products they see in advertisements, and make more optimistic judgments about events, using more heuristics and less reasoning.
The third factor that seems to stimulate reasoning is the importance of a situation for the person doing the analysis. In seeing a sequence of actions in a video, people give explanations for the acts 1) if they see these actions as having serious consequences, 2) if they anticipate future contacts with the people in the video, and 3) if they feel empathy for the actors (Hewstone 1989, p. 44). In experiments with different types of publicity, Petty and Cacioppo (1983) showed that people pay more attention to the quality of arguments when the advertised material is seen (via experimental manipulation) as affecting them personally and in a more important way. But when the advertised matter is seen as less important personally, then the "peripheral" characteristics of the advertisement (like the authority status or the physical appearance of the actors) become more important, and are more likely to affect buying decisions.
There is still a fourth factor that affects, not the initial tendency to search for an explanation, but the tendency to continue searching for an explanation when earlier attempts have failed. When do people give up trying? Psychologists refer to the phenomen of fixing on the last belief one had as "freezing." In a series of experiments by Kruglanski (cited in Hewstone 1989) people "froze" or "unfroze" their explanations depending on their capacities and their motivations. People's capacity to come up with an explanation depends a good deal on the amount of specific information they have about the matter, and the number of their ideas, The higher the capacity, the greater the tendency to keep on trying. Motivation depends on 1) a psychological trait, "need for structure," 2) the desire to reach a specific conclusion (which can make one stop searching for explanations when this desired conclusion is reached!), and 3) the need for validity (that is the cost of a wrong explanation is high, so there is more incentive to find an explanation that works).
Anyone familiar with the anthropological literature will recognize in these psychological studies an old anthropological theory -- Malinowski's hypothesis about magic. Elsewhere (Werner 1992) I reviewed some of the studies based on this theory. To sum up, Malinowski argued that people use magic in situations that are unpredictable, uncontrollable, and important -- exactly the situations that evoke attempts to explain. Later anthropologists have confirmed this theory in many different societies, and for such diverse phenomena as illness (it is the most unpredictable illnesses that receive magical treatment), sports (it is the players in the most unpredictable positions -- like the baseball pitcher -- who use the most magic), the use of water witching (used mostly where geology makes finding underground water especially unpredictable), and even for phenomena like "cargo cults" (which arise in areas where the receipt of important Western goods becomes unpredictable and uncontrollable). The reflections of Mary Douglas on the special magical or sacred status of things that go against our expectations (such as bizarre animals) may also be related to the natural tendency to seek explanations for the unexpected. That these phenomena have a basis in natural selection is also suggested by the "superstitious pigeon" experiments described in the last chapter, since the pigeons "froze" on arbitrary behaviors when the receipt of their food became arbitrary. It seems that humans, like pigeons, attempt to seek explanations for important, unpredictable and uncontrollable events, but when they can find no valid explanation, they "freeze" their beliefs. Even the "freezing" may be adaptive, in that the search for explanations also has its cost, and may not always be worth it.
Verbal Reasoning
In the previous chapter we looked at some forms of animal reasoning, and at the adaptive value of this reasoning. We saw, for example, that monkeys can reason in terms of analogies -- photograph A is to photograph B (a mother to a daughter) as photograph C (a female) is to ? (choice of photographs of sisters, mothers, daughters, etc.) We also saw how different animals understand the notion of transitivity -- if A>B (two animals compared in terms of dominance level) and B>C, then A>C. Finally we saw how animals solve problems involving inference (such as using a stone to cut a cord to open a box to reach a banana), and involving planning (the case of the ravens obtaining meat hanging from a string). In all of these cases it was possible to understand the adaptive value of this reasoning -- either in terms of social living, or in terms of food-getting and predator avoidance.
There is no reason to believe that humans do not also have these reasoning capacities. But human thought goes farther, probably at least in part because humans have a symbolic language. For many scholars, language has allowed humans to free themselves from the more concrete ways of thinking of animals. We can reason about abstractions, not just about more concrete phenomena. Piaget saw formal reasoning, achieved through the use of arbitrary, abstract symbols as the last stage in human cognitive development. But how much has language really freed us from more concrete reasoning? We can examine this question both at the level of deductive formal logic, and at the level of abduction.
Formal Logic
Recently many researchers have become interested in a little test of logical reasoning developed in 1972 by Wason and Johnson-Laird (cited in Méro 1990): The test consists in showing four cards, as drawn below:
The aim of this game is to decide if the following statement applies to these cards: If a card has a vowel printed on one side, then it has an even number printed on the opposite side. The trick is to know which cards have to be turned over in order to verify this statement. You should turn over the minimum number of cards possible.
A long series of studies has compared the results of this test using slightly different contents. Here's one example:
You work in a bank, in the check cashing sector. Your job is verify that checks above $500.00 have a signature on the back. Of course you're very busy like everyone else, and so do not have time to look at all of the checks. Here are four checks on top of your desk, two face up, and two with reverse side up. Which checks do you need to turn over to verify?
You managed to give logicians what they want if you answered cards A and 7 on the first test, and the $600.00 check and the check with no signature on the back in the second test. Logicians see these two questions as variations on the same inference: p --> q (read as: p implies q). This inference means that if p exists, then q also exists (known in logic as modus ponens). Of course q could also exist without p, so p's absence says nothing about q. A check could have a signature and still be less than $500. That doesn't matter. A card could have an even number on its back, without having a vowel on the front. However, if the inference p --> q is valid, then whenever we fail to find q, we will also fail to find p. In logic this is written as: ~q --> ~p (not q implies not p and is called modus tollens by logicians). Everytime we have an inference in the form p --> q we can also deduce its modus tollens:
~q-->~p. So the correct response to any logical problem of this sort is always p and ~q. The cards with ~p and with q do not imply anything. We don't need to turn these over since we can't infer anything from them anyway.
If you are like most people, you probably found the first task more difficult than the second. In the original research of Wason and Johnson-Laird carried out among university students, only 4% gave the right response (p and ~q) in the first task (cited in Méro). Many more gave correct responses in the second task.
Where do these differences in difficulty come from? One explanation is that the second task involves a better known, more concrete situation. Cole et al (1971) suggested this argument after analyzing the situations in which the Kpelle of West Africa solve or fail to solve syllogisms. However, more recent research suggests that this is not the problem. Lawson (1991) used different variations of Wason and Johnson-Laird's test to compare the effects of different types of known contents. For example: "If a car is working, then there is gas in the tank," "If a person has a driver's license, then that person can legally drive a car." These two propositions are both of the form p -->q, and the two represent well known situations. However, almost 60% gave the correct answers about gas in the car, while only 5% correctly solved the driver's license proposition. Lawson suggested that people reinterpreted inferences based on their previous knowledge, and converted the second proposition into a biconditional in which p -->q and q -->p, that is they restated the proposition as "a person can legally drive, if, and only if, that person has a driver's license." With this interpretation it is necessary to turn over all of the cards, and this is what almost 80% of the researched people did.
The idea that it is familiarity with a problem that is important is also questioned in a study by Gigerenzer and Hug (1992). These researchers found familiar contents that resulted in many more errors than imaginary contents. Looking only at inferences that could not be interepreted as biconditionals, they found a good deal of support for another argument -- the "social contract" argument. They suggest that humans have an algorithm to detect cheaters in their social contracts. One of the first things that we do in evaluating a contract is to verify that we are not being deceived. An example taken from another similar project (Politzer and Nguyen-Xuan 1992) illustrates this phenomenon. The researchers began with the inference: "if a sale is for more than 10,000 francs, then the salesperson must put a voucher for gift (a gold watch) on the back of the receipt." But in this case, the people being studied were presented with two different introductory explanations. Some were advised that they represented a consumer protection agency, and must check receipts to verify if the rule is being followed. Others were told that they were store managers, who also needed to verify the rule. From the consumer's point of view, what is important is that everyone who bought more than 10,000 francs worth of goods in fact received a gift voucher. As expected, people who received these instructions were more likely to choose the correct reponses: p (receipts for more than 10.000 francs) and ~q (receipts without vouchers) when asked which receipts needed to be turned over. But from the store manager's point of view, what is important is that salespeople do not give away more watches than they must. For the manager it is important that only those who bought more than 10.000 francs worth of goods receive vouchers. That is, the proposition becomes "to receive a gift voucher, it is necessary to have spent at least 10.000 francs," which is logically q -->p. For the manager it is most important to check q (receipts with vouchers) and ~p (receipts for less than 10.000 francs). The people who were told they were managers, in fact, had a greater probability to choose this answer than the logically more "correct" answer, p and ~q.
Gigerenzer and Hug analyzed this question even further, by choosing different situations. For example, in one inference only one member of the "contract" had something to gain by the rule: "If someone wants to drink alcohol, then that person must be over 18 years old." The waiter who serves the table, has nothing to gain by this rule. The results of these inference tests also fit the idea of a "Darwinian algorithm" which specifies that people tend to interpret inferences so as to avoid cheaters. The conflict between this (natural?) algorithm and formal logic became evident in the answers of some students who responded "correctly" to these inference tests. Most were from mathematical and natural science areas, and expressed their sense of awkwardness at their responses. One commented that "sometimes, in reality, another answer would have made more sense."
These results support the theory of "Machiavellian intelligence" proposed by cognitive ethologists (Whiten and Byrne 1988a, 1988b). They also ressuscitate the Durkheimian theory about the social origins behind logic. But while Durkheim (1915) emphasized how the natural "need" for social solidarity created bad feelings when people contradicted their peers, these data suggest a somewhat more cynical view of human nature.
The importance of emotions in reasoning is especially clear in the case of people who, because of brain damage, have lost their capacity for feeling. As Damasio (1994) points out, these people are good at solving abstract logical problems, but are totally incapable of making decisions. They don't know where they want to go. They cannot hold on to jobs or maintain their social ties. In fact their reasoning is totally incapacitated.
But there are also other factors behind people's logical reasoning. Markovits and Savary (1992) suggest that the fact of having "multiple models" in one's head may be important for logical inference. When experimental subjects receive instructions that suggest the existence of different scenarios, they do better on Wason's and Johnson-Laird's test. This may also explain the findings of Lawson described above. If subjects give a correct answer to the inference "if a car runs, then it must have gas," this is because they remember immediately that there are many reasons, besides lack of gas, that may keep a car from running. But for the inference "if a person has a driver's license, then that person can legally drive" it is more difficult to imagine other reasons (besides lack of a driver's license) that would prevent a person from legally driving. The problem of multiple models is especially evident in syllogisms based on quantifiers like "some a are b; no b is c." From these premisses we could conclude that "some a are not c." Or the syllogism "no a is b; some c are b." From this we might conclude that "some c are not a." To solve these syllogisms we need to construct diagrams. Boolean diagrams are very useful, but can admit more than one design for a set of premisses. For example: "no a is b; some c are b." The different diagrams below are all compatible with these premisses.
From these premisses we can conclude that
some c are not a. But there are many possible models for this. In
one series of experiments with this type of syllogism, Newstead
et al (1992) showed that people attempt to construct multiple
models. However, if they find a believable solution in the
beginning, they often neglect to examine other possibilities. For
example, if the above syllogism is stated as "No pigeon is a
sparrow; some birds are sparrows," many people build the
mental model in the middle above, and conclude erroneously that
"some birds are pigeons." This tendency to stop when we
find an answer that satisfies, may result from the
"algorithm" that explains the tendency to
"freeze" a search when we have found an explication we
find successful.
It is in the light of these experiments that we need to analyze the use of formal logic in non-Western societies. For example, the avoidance of logical conclusions from syllogisms among the Kpelle was attributed by Cole et al. (1970) to their lack of familiarity with the psychological instruments. But we have wonderful examples of the spontaneous use of logical reasoning among so-called "primitive" societies that belie this idea. Malinowski (1929) showed that the Trobrianders can justify, very logically, their own theories about why males are unnecessary for conception. And Goodenough (1990) showed how a Micronesian navigator justified with impeccable logic his belief with regard to how the sun revolves around the earth. In both cases we are dealing with moments when anthropologists and the people they study are asked to contrast alternative ideas (a Western belief vs. a native belief). Note that we are dealing here with with new arguments specifically elaborated because of a disagreement. What may have been missing in the studies of Cole et al, is exactly the stimulation of multiple mental models.
On analyzing their results about different cognitive phenomena, Gigerenzer and Hug (1992) suggest that there is no reason to consider formal logical solutions as more "correct." Natural selection gave us a capacity to use appropriate algorithms for different tasks, and the separation of a problem's content from its form or structure is artificial. To discard the content and analyze only the structure implies losing information, and drawing inappropriate conclusions for a given context (for example, the manager who evaluates an inference from the point of view of a consumer). This does not mean that formal logic is false. It means only that the "correct" ways of thinking depend on the context. This context may be more or less abstract, or concrete. The results of Cummins (1992) studies show how people in fact reason this way. Cummins asked students to analyze different logical-mathematical problems and to group them in "similar" categories. She advised one group to pay special attention to the structure of the questions. She said nothing to the other group. The group receiving the advice, organized problems according to the mathematical formulae needed to solve them. The other group organized them according to the concrete matters they presented. The "concrete" students had more difficulty in solving the problems, but should we conclude from this that they analyzed poorly the questions?
To sum up, we can once again recall the joke about the woman who won in the lottery and who looked down on the reporter who "corrected" her logic. Recent research on logical reasoning suggests that humans adopt different algorithms or heuristics for their reasoning depending on the situation at hand. In this light, it is not unreasonable to conceive of these logical algorithms as similar to other "prêt-a-penser" (canned thoughts), used to organize our behavior. Perhaps we are dealing with a continuum in levels of abstraction, and not with a rigid division between "canned" relatively concrete thought, and formal, relatively abstract logical thinking. In the end these ways of thinking are all forms of abduction.
Abduction
We saw that even apparently purely deductive logic involves the analysis of contents, and of contexts, not just of structure. It seems that even when we solve "purely" logical problems, we act, in fact, as if we were analyzing information. Abduction consists in conjectures (throwing out hypotheses) with regard to a phenomenon. We arrive at these conjectures, neither through deduction, nor through induction. Instead, we use analagoies and metaphors as a basis for "blind guesses" about something. Abductions are ways to explain or account for phenomena that interest us. We saw above how the search for an explanation occurs most often in situations that are unpredictable, uncontrollable and important, and that it makes adaptive sense that an organism would try to control or at least predict important situations. Givón (1989) sees "explanation" as consisting in recognizing that a problem is part of a larger context. This context may consist in 1) some pre-established script (in some canned thought), 2) in a formal functional or semantic structure in which the problem is embedded, or 3) a more direct relationship between a few variables, in which one "causes" the other. In all cases we are comparing a model with something "out there."
We have already considered the use of canned thoughts, and, in analyzing language learning and narrative structure, we have also considered limits as to how people structure their thought. We now need to consider the third type of "explanation" or "contextualization": For centuries philosophers have been debating the meaning of causality, without being able to agree on what they mean. I think the problem stems exactly from the possibility that our notion of "cause" is a "primordial" concept, so basic and so concrete in our minds, that we cannot define it in terms of anything else. In all human languages children learn to ask "why?" very early in life. To better understand the concept of cause, we need to leave philosophy and look a little more at psychology.
What, in ordinary common sense, do we consider a cause? We still need more research to distinguish the roles of notions like "necessary/sufficient," "time lag," "change" "covariation" or "contiguity" that seem to have something to do with how we think about "cause," but some psychological research has already helped clarify a few questions about "causality." Hewstone (1989, p. 45) argues that causal attributions are generally seen as "conditions that 'make a difference' between a normal event and some such event." The problem is to define what determines that an event is normal or not. Cheng and Novick (1991) carried out experiments to distinguish five different arguments about cause. All of the arguments assume that causes refer to "what makes the difference." Two arguments suggest that it is the need for communication that affects what "is normal" and what requires explanation. The first emphasizes the need to supply "missing" information (for example, to explain why my plants grow so well, I tell my mother that I acquired an interest in plants after reading a book; I inform my neighbor that I bought seeds in an agriculture store). The second argument (also "informative") suggests that it is the relevance that defines cause (I explain that watering plants every day is essential, even if my neighbor has seen me water the plants every day). Two other arguments assume that people consider as "cause" whatever is "not normal" defining normality either in terms of norms (I water my plants, in spite of the drought, because I don't believe I'm wasting that much water), or in terms of statistical tendencies (The plants do well because I'm the only one here who was brought up on a farm). The last argument suggests that it is covariation within a given context that defines what we mean by cause (my plants do better than other people's because they don't water their plants every day, and do not buy good seeds, while I do). The experiments best supported the covariation argument about how people conceive of causality. But the covariation was always seen within a context. Factors that do not vary within a given context are seen as "necessary conditions," but not as "causes." However when other contexts make these factors vary, then they also are considered "causes." Cheng and Novick attribute to their lack of knowledge about contexts many of the seemingly "profound" or "scientific" questions that children ask -- "why do things fall," "why is the sky blue?"
But if covariation helps "define" causality, this does not mean that people are especially good at recognizing covariation (see Hewstone 1989). Instead of looking for evidence that might distinguish between different hypotheses with regard to variation, people tend to look for information that confirms their preferred hypotheses. Pressupositions about causes make people see covariation where it does not exist, and ignore covariation that does exist. Perhaps this can be explained in terms of information theory. It seems that disconfirmations require more processing time than that required for confirmation (Hewstone 1989, p. 23-29, p. 86-90).
But there are differences between people with regard to causal attributions. FIrst, there are age differences. Small children show relatively little interest in cause and effect. Older children restrict themselves to one cause for each effect. Adolescents are more likely to invent possible multiple causes (Lawson 1992). Also there may be differences with regard to the tendency to explain human acts in terms of "dispositions" versus "causes." Hewstone (1989, p. 20) described a dispositional attribution as consisting in proposing a label to characterize someone -- for example, "explain" that someone robbed because "he's a thief." These attributions are more spontaneous, and require less thought and effort than "causal" attributions.
Some people are also more creative than others. There is an enormous literature on the reasons for creativity, but here I cite just one factor that may have cross-cultural implications. Houtz et al. (1989, 1979) noted differences between psychologically "inner directed" and "outer directed" people. The "outer directed" people depend more on others' evaluations for self-evaluation, while "inner directed" people evaluate themselves on the basis of personal progress or regression. These two types of people display different types of creativity. When presented with a typical problem (like "Imagine different things that might be done for a drug addict), the "outer-directed" people are able to imagine more items. But the items imagined by the "inner-directed" people are more original. In addition, the "inner-directed" people are more creative when presented with pure fantasy problems ("Imagine what you could do if you had six fingers on each hand"). The authors suggest that "outer directed" people are better in administrative positions, since they can better anticipate arguments, objections, problems, etc. that the others will come up with. But the "inner-directed" people are more creative in other spheres. Perhaps this phenomenon also partly explains the case of shamans in many societies. Shweder (1972) tested the creativity of shamans and lay people among Zinacanteco Indians of Mexico. He found that the shamans were more likely to offer explanations and descriptions for out-of-focus photographs. In addition, they offered more original explanations. Perhaps what is sometimes seen as "deviant" in shamans' behavior in some societies, is simply a reflection of the shaman's more "inner-directed" orientation, in which social pressures toward conformity have less effect.
Independently of our creativity or our age, it seems we also have some general baises about "causes." First, we generally assume we need a confluence of various causes to explain rare phenomena (Only when A and B occur together do we get the rare phenomenon C). But when a phenomenon is common, we assume that the multiple causes are independent and sufficient to explain it (Either A or B is enough to cause the common phenomenon D). For example, success on an easy task, or failure on a difficult task are generally attributed to a single cause, but success on a difficult task, and failure on an easy task require us to join different causes (Hewstone 1989, p. 25-26). Perhaps because of our "Machiavellian intelligence" we also have a tendency to attribute causes for a phenomenon to human intentions instead of to random, physical or situational factors (Hewstone 1989; Humphrey 1988, p. 24). This tendency is so strong that even when we known that it is a situation that influenced an action (for example, when we know that a researcher asked someone to defend a given position), we still have a tendency to attribute the action to personal characteristics, and not to the situation (Hewstone 1989, p. 18). We also have a tendency to see two events as related if both are more distinctive than other events, or if we already begin with the idea that they ought to be linked. We ignore abstract and statistical information and depend more on concrete and vivid examples to inspire ideas. We also have biases with regard to how much time is given to an argument, or its order of presentation (Hewstone 1989, p. 94-96; Michel 1991). Another bias is the tendency to seek causes with characteristics similar to the phenomenon one wants to explain (Hewstone 1989, p. 95). Anthropologists recognize this phenomenon in "sympathetic" magic -- for example, the belief that a penis-shaped plant can serve as an aphrodisiac.
The acceptance or verification of causes also involves general biases. The general biases affecting the elaboration of hypotheses also affect their testing. But there is more. We tend to adopt stronger positions with regard to matters that affect us more directly (Hewstone 1989, p. 18). Perhaps this is explained by our greater care in arriving at conclusions that affect us more directly. In experiments in which students believed they would be directly affected by a matter, they paid more attention to the quality of the arguments given. But when the matter was seen as having little personal relevance, students tended to value more the number of arguments rather than their quality (Sherman et al. 1989). Another very important factor in the acceptance or rejection of an argument is social pressure. For example, experiments with students show that as many as a third end up agreeing that something is other than what they clearly saw (e.g. a green pen instead of a red pen), simply because their peers report having seen something else (Worchel and Shebilske 1989). The tendency to believe in authorities is also well documented, both in cognitive psychology and in advertising (Milgram 1971; Hewstone 1989; Petty 1983). Often simple body posture, facial expressions, or tone of voice is enough to create confidence in or suspicion of a person's arguments (DePaulo 1992).
The existence of these general biases does not, however, imply the absence of a "scientific" attitude toward testing. Probably in all human societies there are individuals who pay more attention to the agreement between conjectures and inputs from the outside world, and there are moments when everyone pays more or less attention to these inputs. For example, Werner (1984) observed a Kayapó Indian carry out a controlled experiment similar to those carried out by agronomists in our society. Not knowing how to plant beans, the Indian planted some in the center of a garden (where it would receive a lot of light), and some on the border (where outlying trees would shade it somewhat). He repeated this experiment in different seasons.
Dreams and Altered States of Consciousness
We spend around a third of our lives sleeping. How can we explain this tremendous investment in terms of time? Evans and Evans (1984) used the analogy with computers to suggest that sleep, and especially dreaming, serve to "bring programs up to date." They note that large computers periodically must remain "off line" in the sense of shutting off all external contact (inputs and outputs) to bring routines up to date. This, they suggest, corresponds to the sleep of animals. This theory explains sleep, not as having a direct adaptive value (after all, the animal remains very vulnerable during this period), but rather as a side effect of an increased cognitive capacity. In support of this argument, Evans and Evans note that more complex animals generally sleep more than less complex animals -- at least when "sleep" is understood as involving brain activity different from the brain activity typical of wakefulness. Primates sleep a great deal, carnivores sleep more than herbivores, mammals generally sleep more than birds, which sleep more than reptiles, which sleep more than fish. In the case of insects, their periods of inactivity seem to be related directly to environmental factors (like temperature) rather than to cognitive factors. Electrodes placed in insects' brains also fail to show the characteristic signs of sleep in more complex animals, and there are no special implications of their being kept "awake."
Besides sleeping more, more complex animals also seem to dream more. Humans spend around 20% or 25% of their sleep in REM ("rapid eye movement") periods. If awakened during these periods, they remember what they were dreaming. Animals like cats and dogs also show REM movements, and make small gestures with their paws, or vocalizations suggesting sleep. If awakened during these periods, they show surprise reactions. Normally our muscles are especially flacid during dreaming. A chemical inhibits the transmission of brain signals to the muscles. The elimination of this substance in cats makes the animals act out their dreams. Judging by their movements, it seems that cats dream about capturing animals, even if they have never needed to do this. Animals like cows and sheep also show REM and EEG's indicative of dreaming, although less than with cats or dogs. Studies of EEG's in reptiles like alligators have found no indications of dreaming.
The independent (?) evolution of longer periods of sleep in cetaceans and primates seems to support the cognitive argument. Since cetaceans need to surface for air, it is not easy for them to sleep. The adaptive solutions are bizarre. For example, the bottlenose dolphin sleeps with only half of its body at a given time. In these periods one eye stays shut. The Indus dolphin which inhabits waters with rapid currents, runs the danger of being thrown against rocks. It sleeps in short spurts of 90 seconds each. But it seems that dolphins do not dream! This finding goes against practically everyone's view of the functions of dreaming.
Evans and Evans (1984) suggest that dreaming is important to "run through" our "files," order them in memory, and formulate and analyze routines. In this latter function dreams have characteristics like play -- they both serve to simulate different mental models. In support of this argument, the authors note that "imprinting" (like the goose's acceptance of the first moving object it sees as mother) which would normally occur in animals, fails to occur if these animals have been deprived of sleep. On being able to dream again, the animals recover their lost dreaming (they dream more), and can imprint. Jouvet (1994) would explain this as due to the working out of inborn iterative programs during sleep. That younger animals sleep and dream more than older animals would be explained by the greater learning of routines that occurs in this period of life. A relationship between play and sleep is also indicated by the fact that most children's dreams have more to do with play activities than with school. Although children spend more time in school, these experiences do not enter dreams, perhaps because the problems confronted in school are more easily solvable, or perhaps because school is not so natural as play, and so our brains have not naturally been programmed to deal with school problems. Possibly dreams help solve problems in that they evoke different capacities of human thought, beyond verbal reasoning. Ornstein (1977) and others have dedicated a good deal of attention to these other forms of thinking. In any case, the impact of dreaming on problem solving is suggested by more systematic research showing that we learn best shortly before sleeping, and by anecdotal reports of scientists and artists who solved great problems on the basis of a dream.
The importance of sleep is demonstrated in epidemiological research (reported in Newseek) showing that sleep deprivation (due to work, etc.) is responsible for more illness than excessive use of tobacco or alcohol. Also illustrative, are the cases of people who tried to beat records in terms of the number of hours they managed to stay awake. These people went through a sequence of phases that began with light hallucinations that became stronger and more vivid as time went on, and eventually led to paranoid reactions until the sleep was recovered. Evans and Evans explain these phenomena as an overloading of the cognitive system, in which the brain's need for organization ends up invading the awake brain. Possibly, sleep deprivation causes health problems because it does not permit the mental simulation of possible solutions to stressing questions. As Jouvet (1994) points out, stressed mice generally have more paradoxal sleep (presumably dream more), than do non-stressed mice. The depression that results from non-resolved problems leads to sleep and to dreaming, which may help solve problems and thus stop the depression and stress. To sum up diagramatically we might hypothesize that:
Still, there is one problem
with this argument. As Jouvet (1994) points out, individuals
taking anti-depressive drugs that inhibit dreaming do not appear
to be harmed in their memory or other cognitive functions. Jouvet
argues that they simply fail to develop their own individuality.
Several authors (Langdon 1992; Kracke 1992; Winkelman 1990) have emphasized the exceptional ability of shamans to enter into trance, and to control their dreams. Since there are many relationships between dreaming and trance, and especially hallucinations, we are perhaps in a better condition to understand the shaman's role in society. The shaman, besides dealing with specific illnesses, also deals with general community problems. The use of dreams and trances perhaps helps reorganize everyone's thought -- especially those forms of thought most closely tied to the social "scripts" or to canned thoughts in general. In our society there are people (called "lucid" dreamers) who manage to control their dreams, and there is laboratory evidence that this ability can be trained. Ethnographic reports suggest that shamans can, to a certain extent, control their trances and their dreams. It would perhaps be interesting to check these abilities using the techniques of sleep laboratories, and observing better the techniques used by shamans to control their trances. In any case, it is possible that the greater originality of shamans in coming up with explanations for ambiguous phenomena (Shweder 1972) has something to do with this capacity. The fact that shamanism is practically universal and doubtlessly the first "specialization" both intellectual and non-intellectual to appear in human cultural evolution, underlines the importance of the development of this side of human cognition.
Conclusions
In this chapter we have seen many continuities between human and animal cognition, and have observed many limits on human thought. Human psychological development has its parallels in different primates. The development of concepts among humans seems to precede the development of language, clarifying that even in humans, language does not determine thought. Language learning itself is partly preprogrammed in that we have definite linguistic preferences in terms of phonology, syntax and semantics, and can learn to speak like natives only during certain ages. There is also good evidence that many non-linguistic forms of communication, and that many concepts are also universal.
Studies of human cognition suggest the existence of different modules of thought. More stereotyped and repetitive routines may be thought of in terms of scripts that demand little reflection. These scripts, even the ones most divorced from immediate practical matters (such as narratives) also show apparently universal preferences. When things are more important, unpredictable and uncontrollable humans tend to use more reasoning. Logical reasoning itself seems to depend on the context and content of the problems to be solved, and not just on their structure. Biases in the analysis of logical questions (like the tendency to look for "cheaters") may have their adaptive value, just as do biases that appear when elaborating or testing abductions.
The existence of natural biases in human cognition may have its adaptive value in many situations, but it may also be harmful. This for two reasons. First, these biases may blind us, making problems that do not fit into these biases especially difficult to solve. Second, these biases may be used by more capable individuals to manipulate others. Bousefield and Davis (1980) suggest that the use of false arguments for social manipulation is universal. In their words, "it is inconceivable that a society that has rationality, would not also have sophistry." In the next chapter we will look at some of the implications of these limitations on the thought of those humans who are specially trained to make their livelihood by thinking.