related texts
2 May 1997 [Last modified 7 Sep 2010]

Evolutionary psychology (EP): innate vs. learned

1.abstract

Evolutionary psychology (EP) is an emerging branch of anthropology and psychology, which have been gaining ground lately. A fundamental tenet of EP is that large part of psychology is innate, as opposed to learned, to the point of rejecting the concept of "learning" altogether (e.g. 3.2.7 below). Here I discuss the evidence and arguments that are used in EP for the innateness of psychological traits.

Layout of the text

In section [2] I outline the main arguments for and against innateness and learning.

In section [3] I analyse general discussions of the EP approach. These come from 'The Adapted Mind' by Barkow, cosmides and Tooby (1992), which at the time of writing was the most prominent book in EP, and an earlier chapter by the same authors. Here I discuss only those articles and chapters that touch on the point of learning.

In section [4] I discuss specific chapters from 'The Adapted Mind'. In section [5] I discuss other examples.

The reader should note that in the text in sections [3-5] I am not intending to bring evidence against innateness. The evidence against innateness is outlined in [2.2] below. What I am trying to show in sections [3-5] is that the evidence for innateness is not real.

In the text below indented quotes are always exact quotes, except words in square brackets, which are my addition for clarification. Numbers in square brackets (e.g [3.2.1]) indicate a section in this text.

2.The innateness assumption in EP

2.1 Evolutionary theoretical arguments: what we should expect

The base for most of EP literature is the belief that significant parts of human mental abilities are directly coded in the genome, in other words innate. This is based on the logic that the mind, like the body, would evolve useful adaptations to the world.

This logic is wrong, because humans can learn, and can pass behavioural patterns to following generations as `culture' (which causes learning). Learning and cultural changes are far faster processes than genetic evolution, so they are much more efficient in generating useful adaptations. Therefore they will tend to eclipse any innate mental abilities, hence removing the selection pressure on them. This will cause innate mental abilities to degrade, or more likely not to evolve at all. Since learning and cultural adaptation and functional degradation of unselected abilities are all fast processes in evolutionary scale, we shouldn't expect innate mental abilities that can be learned (with or without cultural effects) to exist at all.

There are two major points that EPs seem to miss here:

  1. At the basic level, learning is understanding regularities in the world. That means that facts of the world do not need culture or anything like that to be learned. Culture is one of the factors that affect the world, and hence determine what is learned. EPs tend to confuse learning with culture, hence to fail to see culture-independent learning.

  2. Culture is subject to `mutation' and selection{1}. Cultures which do not teach their members the right things (either explicitly or implicitly) will either die, by physical death of their people or by being overwhelmed by other cultures, or adopt the appropriate traits from more successful cultures that get it right. Thus cultural practices which cause members of the culture to learn some skill or behaviour can evolve even if no member of the society is aware of either of the skill or of the relation between the practices and learning the skill {2}. This evolution is still much faster than genetic evolution.

This argument still does not tell us that there are no innate mental abilities, because:

  1. some abilities are required before the learning process actually succeed in doing anything (e.g. crying).
  2. Mental abilities that are essential for learning will necessarily survive.
  3. There may be abilities that are too difficult to learn for humans.
  4. some innate mental abilities may still survive anyway, because of reasons that we don't understand.
However, it does tell us that, in general, we should not expect mental abilities (except learning) to be innate, and there is a need for evidence for or against innateness. Note that the argument is applicable `as is' to the physiology of humans, but in this case nothing can be learned, so everything must be innate (coded by the genome).

2.2 Evidence from neurosciences

The conclusive evidence against innate mental abilities comes from brain damage studies, and from neurobiology of the brain.

2.3 The EP argument for innateness

Evolutionary psychologists cope with the arguments above mainly by ignoring them. The stochastic connectivity and brain damage studies are totally ignored. The learning question, while almost always ignored, is sometimes tackled with some nonsense arguments, which are discussed below in sections [3],[4],[5].

The evidence that EPs bring for innateness is of two types:

  1. Universality of a psychological trait. This is an evidence only the trait can be shown not to be learned. Mostly EPs ignore this, and sometimes they use some nonsense argument
  2. Showing that young babies know things `before the could learn them'. This is always rely on strong overinterpretation of the behaviour of the babies (see [3.2.2] and [5.1] below

2.4 The `learning is unexplained' argument

Additional argument against learning is that it is unexplained phenomenon, so cannot be used to explain other phenomenon (see below in [3.2.7], [3.4]). This is nonsense, because there is no reason not to assume the existence, and use for explanation, a phenomenon that we don't understand, provided we have good evidence for it. The evidence for general purpose learning mechanism is that human can acquire skills like horse riding and training, growing plants, building houses, using fire, making clothes, reading, driving cars, brushing teeth, building spaceships, playing chess, using bow and arrow and many others, for which there cannot be an innate code. A general purpose learning mechanism is currently the only explanation for these. EPs simply ignore all these skills in their literature.

2.5 Arguing against the wrong alternative

The major alternative to innateness of mechanisms is a general purpose learning mechanism, that learn domain-specific mechanisms to actually perform actions. This is mainly done during infancy and childhood, when the basic domain-specific mechanisms (perception, movement, reasoning, language, social behaviour, moral views etc.) are learned. This alternative is rarely tackled directly (the only place I could find is discussed in [3.2.1] below). Instead, most of the argument is against the general purpose mechanism that solve everything, or against `capability-for-culture'. While it is true that the latter two ideas are sometime promoted, debunking them does not allow EPs to say that they can prove innateness. They must consider the general-purpose learning of domain-specific mechanisms as well, and they don't.

2.6 Evolutionary analysis

Underlying some of EP thinking is the assumption that it is possible to analyze the environment that humans have evolved in and hence predict from this anlysis what psychological traits humans will evolve. For example, see [3.4.3] below, and the reference to EEA in [3.3.3].

This is simply wrong, because currently we cannot (and are very far from being able to) predict evolution of complex traits. There are two fundamental problems in predicting evolution:

  1. The number of possible mutation is large, and the number of possible combinations of mutations is much much larger. Out of this huge number of possible combinations of mutations, relatively small number actually occur. Since natural selection can only select from those combination of mutations that did occur, evolution is restricted to these combinations. As far as we know, the occurence of mutations is essentially random.

  2. Even if know which combinations of mutations do occur, to predict the evolution we have to know the phenotype of each combination of mutations. We are very far from knowing how to do that for complex traits. For example, given the DNA of an unknown animal, it is not possible to deduce the attributes of the animal, except by comparing it to DNA sequences of other animals.
Problem (1) may be unsolvable in principle, while problem (2) may be solvable, but at the moment we haven't got a clue of how to do it. Thus we cannot predict emergence of complex traits.

The illusion that we can do effective evolutionary analysis is based on the ability to predict of the evolution of existing traits, under very strong selection, inside a short time period (in evolutionary terms). In this case, there will be many mutations which cause variability in the trait, and strong selection is likely to change the trait in a specific, predictable direction. Outside this restricted range, however, it is not possible to predict evolution.

2.7 Claiming that learning and innate mechanisms are not contradictory

Sometimes EPs claim that there is no contradiction between innateness and learning, and the question is what innate mechanisms underly learning. The problem is that they go on to take for granted that the learning mechanisms are specific, rather than general. For example, in their 'primer', Cosmides and Tooby say, in the subsection "Innate" is not the opposite of "learned" :

The interesting question is what are these unlearned programs? Are they specialized for learning a particular kind of thing, or are they designed to solve more general problems?
(compare with [3.2.7] below).

They simply ignore the main (and obvious) alternative, that these unlearned programs are specialized for learning, not a particular kind of thing, but anything (i.e. any regularity in the input).

Later they say:

For any given behavior you observe, there are three possibilities:

1.It is the product of general purpose programs (if such exist);

2.It is the product of cognitive programs that are specialized for producing that behavior; or

3.It is a by-product of specialized cognitive programs that evolved to solve a different problem. (Writing, which is a recent cultural invention, is an example of the latter.)

Here, the possibility that the cognitive programs (in possibility 2) are learned is not actually excluded explicitly. However, this is not discussed further . The authors clarify what they think three paragraphs later:
Most EPs acknowledge the multipurpose flexibility of human thought and action, but believe this is caused by a cognitive achitecture that contains a large number of evolved "expert systems".
Learning, by now, have completely disappeared. Thus the assertion that innate and learning mechnaisms are not contradictory is just a trick to avoid arguing against learning.

2.8 Ignoring learning altogether

The main strategy that evolutionary psychologists use for against the possibility of learning is ignoring it completely. For example Cosmides and Tooby (1994) are calling for `evolutionary rigorous cognitive science', but ignore learning altogether.

Buss, in "Evolutionary pschology, new science of the mind" (1999) also does not discuss learning at all. Buss shows his contempt for any non-genetic factors by selecting the formation of calluses on the feet as an example of interaction between genes and environment.

3 General discussions

3.0 Evidence from neurosciences is ignored

In all the texts that I discuss here, the way the supposed genetical adaptation is actually implemented in the real system (i.e. the brain) is not discussed at all, except in the discussion of pregnancy sickness (chapter 8 in `The adaptive Mind'), which stands out because it is nothing to do with psychology. When models are offered in other chapters, they are all in abstract terms which cannot be mapped to neural structures.

3.1 The introduction of `The Adaptive Mind'

In the introduction (Cosmides, Tooby and Barkow 1992) the counter-arguments that I brought in [2.1 and 2.2] above are not mentioned at all. This is particular evident in the questions that the contributors to the book were asked (p.6), which does not include the question whether the adaptation is learned or innate, and how it fits with neuroscience evidence. These subjects are simply not of interest for EP, even though they claim that EP is the basis for conceptual integration.

3.2 Chapter 1 of `The adaptive mind': The basic EP approach

In the first chapter of `The Adaptive Mind', Tooby and Cosmides (1992) introduce the EP approach and argue for it. In introducing the approach, the authors rely heavily on research in vision (mainly Marr) and linguistics. I show the nonsense of these two in separate texts (Critique of Vision by Marr and discussion of psycholinguistics).

3.2.1 preliminaries

The authors first introduce the Standard Social Science Model (SSSM), and their arguments are all against this model. Most importantly, they attribute to it the assumption that all mental abilities are based on general-purpose mechanisms. The more sensible option is that mental abilities are based on specific mechanism which are learned by general purpose learning mechanism (as discussed in [2.5] above), but this possibility is not discussed in this chapter (but see in [3.2.5] below).

Before presenting the central elements of the approach, the authors express an amazing optimism in the approach, which reaches a climax on pp. 68-69, where they say:

Just as one can now flip open Gray's Anantomy to any page and find an intricately detailed depiction of some part of our evolved species-typical morphology, we anticipate that in 50 or 100 years one will be able to pick up an equivalent reference work for psychology and find in it details information-processing descriptions of the multitude of evolved species-typical adaptation of the human mind, including how they are mapped onto the corresponding neuroanatomy and they are constructed by developmental.

Most of this can be regarded as naive optimism, but the last part is plain ignorance. Brain damage research clearly demonstrated that there is not such mapping, except for input/output.

In this paragraph the authors also show the typical analogy between body organs and `mental organs'. This analogy is broken, because there is no way in which a human can acquire a new organ, or modify existing organ to different function, but humans can acquire new skills. Thus organs are not analogous to mental skills.

3.2.2 What babies can do

One of the arguments for innateness is that young babies have `surprising' abilities, which they `couldn't learn'. The authors discuss this on p. 70:
For example, contrary to the piagetian notion that infants must "learn" the object concept, recent research has shown that (at least) as early as 10 weeks - an age at which the visual system has only just matured - infants already have a sensorily-integrated concepts of objects as entities that are continuous in space and time solid (two objects cannot occupy the same place at the same time), rigid, bounded, cohesive, and move as a unit (e.g., Spelke, 1988, 1990, 1991). Indeed, when infants of this age are shown trick displays that violate any of these assumptions, they indicate surprise -- one could almost say in such cases that the object concept embodied in their evolved mechanism causes them to "disbelieve" the evidence of their senses (Leslie, 1988).
This is quite impressive collection of misleading hints, some of which may be called straightforward lies:
  1. At most, the experiments can show that learning is faster than Piaget thought. That is not an evidence against (or for) learning, because we have no idea how fast a system with the complexity of the brain can learn the basics of behaviour of physical objects.

  2. The baby's vision does not mature at 10 weeks. It is already working when the baby is born. If there is a learning mechanism which requires visual input, it is in action from birth.

  3. Most of the experiments that are reported in the references that the authors give were done with older babies. The authors unashamedly confuse results of older babies with younger, and even have the effrontery to hint that babies younger than 10 weeks show the understanding they claim, even though the research clearly show a very little ability at 10 weeks, which gradually improves with age. The only experiments with `10 weeks' infants that is referred to in Spelke (1991, but published as a research paper in Spelke et al(1992) (experiment 3)), and even this one is with a range of babies of 10-13 weeks.

    The way the point is put in the last sentence (`when infants of this age') is particularly blunt, and this is clearly an intentional `inaccuracy' to mislead the reader, in other words a lie. The primary research that this statements based on is from infants of 3.5-4.5 months for simple cases, and 6 months or older for the more complex tasks (see [5.1] below for a discussion of the actual experiments).

  4. None of the babies ever `indicate surprise'. In all the experiments, the sole dependent variable was looking time, i.e. the time a baby spent looking at different events. That longer looking indicates surprise, based on understanding of the physics of the event, is an interpretation which is not based on any evidence at all. I think this can also reasonably be called a lie, because there was no reason why the authors would not say `look longer' instead of `indicate surprise'. It would be fair to say that the original research papers made this assumption as well, see below in [5.1].

  5. Babies don't `indicate surprise' (look longer) at `any trick display'. Rather, they look longer at `some trick displays'. Again, there was no reason why not say `some', or maybe `many', instead of `any'.
That Tooby and Cosmides (1992) felt the need for such blunt `inaccuracies' shows that they themselves don't think that the real data support their arguments, and hence need to `improve' it. Since there was no reason for the authors not to use data that supports their argument without `improvement', we can deduce that the authors are not aware of any data that support their arguments without `improvement'.

In the rest of the page, the author present other abilities of young children, but do not give any reason why these could not be acquire by learning. As the authors themselves emphasize, all the abilities are based on recognizing real attributes of the world, so any learning mechanism that learns facts about the world will learn them.

3.2.3 `the central elements of evolutionary functional analysis'

The basic approach of EP is presented (Tooby and Cosmides, 1992) in the section called `the central elements of evolutionary functional analysis' (pp. 73-77). This describes the various steps in this analysis, but does not include considering if the skill under consideration is learned. In fact, the only mention of learning is in this form (p.74):

We want to emphasize that we are looking here for a mechanistic or causal description of how the system generates output given input. Statements like, "the human child learns its culture through imitation and generalization" are not models of how input generates output. They are too unspecific to qualify as hypotheses or explanations;

Since there is no other reference to learning in this section, it seems reasonable to assume that the range of statements which are `like' the statement in double quotes include all learning possibilities. Thus Tooby and Cosmides think that the possibility of learning should not be considered at all, because it is `too unspecific'. However, how specific a statement is does not tell us if it is true or not. The statement in double quotes is specific enough to be refuted (and hence also be supported) by experimental evidence, and it is therefore a hypothesis in the scientific sense of the word. Tooby and Cosmides, however, already decided what kind of explanations they want, and since this statement contradicts them, they reject it without bothering about evidence.

3.2.4 `Learnability'

Later in the chapter Tooby and Cosmides do try to tackle learning. First they claim that there are proofs that some things are not learnable. This relies mainly on evidence from psycholinguistics (p. 95, p.111). I show that this is nonsense in linguistics blatant nonsense (specially [2.11-13]).

3.2.5 The "guidance system"

The discussion on pages 100-102 is the closest that the authors get to actually tackling learning. Their analysis leads to the recognition of the need for a "guidance system", which I wholly agree with. In fact, one of the main problems that I try to solve with my theory is the question of what drives the behaviour of humans. However, once they concluded the argument for the need for a "guidance system", the authors drop this line of argument, and do not return to it. It is probable that they believe that this "guidance system" is impossible, but they don't actually say this or support it in any way.

Something like a discussion of the "guidance system" is found in P.111. The authors say:

Because a domain-general architecture by definition has no built-in content-specific rules for judging what counts as error and success on different tasks, it must have a general rule.

That is nonsense. There is no contradiction between general learning mechanism and domain-specific targets. For example, an `eating' model that is compatible with the data is that humans have some mechanism that tells them that `something' is missing, and stops after they eat, but the actual actions of getting food, processing it and then eating it are learned by a general-purpose learning mechanism (sucking is an exception). All that is required is that the mechanism that detect hunger is connected to the general purpose learning mechanism.

In the next sentence, the authors say:

Unfortunately, there is no usable general cue or criterion for success or failure that can apply across domains.

That is true when applied to basic needs, which need specific "target setting" mechanism, but does not mean the actual actions have to be innate as well, as outlined above about eating. When applied to domains which are about understanding the world, the last quote is simply false. When an individual understands the world, there is a better fit between the way shhe thinks the world is going to behave, and the way the world realy does. Thus a mechanism that gives positive response when there is a good fit between internal "predictions" and perceptual input is a usable general cue.

3.2.6 `The weakness of content-independent Architecture'

The rest of the discussion in this section (p.101-114) is based on a failure to distinguish between special, domain-specific, content-dependent mechanisms and innate mechanisms (see in [2.5] above). It tries to argue for innateness by arguing for domain-specific mechanisms, ignoring the possibility that special, domain-specific mechanisms can be learned by general-purpose learning mechanism.

The point is seen most clearly in the summary in p.111 (point 2). In arguing against a general purpose learning mechanism, the authors say:

Such mechanism [specialized] will be far more efficient than general-purpose mechanisms, which must expand time, energy, and risk learning these relationships through "trial and possibly fatal error" (Shepard, 1987a).

The risk and effort are only part of the learning, not of the actual performance, which is done by a learned specialized process. This is in good agreement with reality, because it is clear that humans spend a lot of effort in learning tasks to achieve good performance. The quote about "trial and possibly fatal error" is specially amusing, as this is exactly what babies and young children do, so this argument actually supports general-mechanism. This is an example of theory-driven blindness to the facts.

The discussion contains additional errors. For example, the `combinatorial explosion', which is introduced in p.102, is a problem only for systems that use `search and compare' algorithms. The brain doesn't use these algorithms (it can't, because neurons cannot implement pointers, so cannot perform comparison in the sense that these systems require. See in brain-symbols), so does not suffer from this problem. For example, the authors bring this example (p.103):

For example, if cognitive mechanisms are attempting to infer the meaning of an unknown word, there is an infinite set of potential possibilities.
This is problem only if the mechanisms actually check all the possibilities. However, children (and adults) simply increase the association of the word with those concepts that are active in their brain when the word was perceived (adults also use various other (learned) algorithms as well). In neural systems, this is a simple operation which is implemented by Hebbian mechanisms. Thus the infinite possibilities are not a problem for neural systems.

The next sentence involves an additional erroneous assumption:

If perceptual mechanisms are trying to construct a three dimensional model of the local world from a visual array, there is an infinite number of different ways to do it that are all consistent with the array.

Logically, this sentence is correct, but it is interesting (relevant to human brain) only if we assume that the brain:

  1. Enumerate the possible models to see which one fits.
  2. Use only visual input.
The last assumption is clearly wrong, because infants use tactile information to learn about the structure of the world, and adult use the knowledge they have acquired about it. The first assumption is unsupported by any evidence, is counterintuitive, and is incompatible with neural systems. Thus the sentence is not relevant to humans.

Point 5 on P.111 is another example of theory-driven blindness. The authors object to general-purpose mechanisms because they cannot learn courses of actions that are based on data that is not available to individual. That ignores the fact the learning is based, among other things, on the culture. As explains in [2.1], the culture can adopt practices that direct individuals to learn the appropriate course of action even if it depends on invisible factors, because cultures are subject to the same selection forces that are applicable to the genome.

3.2.7 Conclusion: total rejection of learning

In the conclusion of this chapter Tooby and Cosmides (1992) say (p.122):
Of course, as most cognitive scientists know (and all should), "learning"- like "culture," "rationality," and "intelligence" - is not an explanation for anything, but is rather a phenomenon that itself requires explanation (cosmides & Tooby,1987; Tooby & Cosmides, 1990b).
As was discussed in section [2.4] above, learning is the only plausible explanation for a large range of skills that human can acquire. The authors does not bother to consider alternative explanation for the acquisition of all these skills. They simply ignore them.

That such a nonsense can actually be written by these authors show how much they are driven by their own beliefs, and that facts of life that do not fit with their beliefs simply do not count for them as relevant facts.

In case the reader may think this was just a slip of the pen, the authors clarify their position in the following paragraphs, and say on p. 123:

We expect that the concept of learning will eventually disappear as cognitive psychologists and other researchers make progress in determining the actual causal sequences by which the functional business of the mind is transacted.
Again, the authors do not bother to tell what other explanations they think are possible to the range of mental skills that human can acquire.

3.3 The `correct Darwinism'

In the second chapter of `The Adapted Mind', Symons (1992) tries to establish the correct use of darwinism. Most of the discussion is rejections of other's people claims for adaptations. I will discuss only the Symons view of adaptation as it comes out in this chapter.

3.3.1 What is `adaptation'

On P.140 Symons (1992) presents `The Adaptionist Program'. Here he explains when can a trait be regarded as an adaptation. In this discussion, he regards as an adaptation traits that arise by differential reproduction.

This is a possible definition of adaptation, but it means that to show that something is an adaptation you have to show that it has arisen by differential reproduction. Currently, there is no way to prove this directly. Something of a proof can be achieve by showing that other explanations are unlikely. Symons, however, does not discuss this point at all. It seems that, like most of EPs, he assumes that any trait that cannot be proved to not arise by differential reproduction did arise by differential reproduction. This is simply baseless assumption.

Later, After defining adaptation as a trait that result from differential reproduction as a result of a design, he says

Finally, given modern understanding of the genetical basis of reproduction, to claim that a trait is an adaptation is to make certain kind of claim about genes.

That is simply false, because differential reproduction may be a result of a better culture vs. worse culture (e.g a culture that uses iron vs. culture that doesn't).

3.3.2 Innateness of tendencies in sexual attraction

Symons (1992) demonstrates his point by discussing sexual attraction. the argument for the innateness of tendencies in sexual attraction starts like this (p. 143):
If the perception of sexual attractiveness were the product of some sort of generalized "capacity-to-culture" mechanisms, then the standards of sexual attractiveness would vary capriciously cross-culturally and would be impossible to predict in advance for a heretofore unknown people.

This is blatant nonsense. For a group of people to survive, it must be the case that members of the group are attracted to healthy humans (as opposed to animals) of the opposite sex in the appropriate age for reproduction (this is only partial list). Therefore each member of the group must be attracted to those attributes that distinguish humans from animals, the opposite sex from his own, associated with good health and are typical of an individual of the right age. This is necessarily true in any group of people that survive for evolutionary length of time, independently of the means by which sexual attractiveness is determined. Even if we find a human society in which sexual attractiveness is determined by a silicon chip that is inserted into the head of members of this society, but this practice is quite old, and the reproduction mechanisms are the same as the rest of humans, we can predict with absolute certainty that the attributes above will be programmed, in some way, on this chip. Otherwise, this society would have died after few generations.

This is the same theory-driven blindness that Tooby and Cosmides (1992) do in point 5 on p.111 of the same book (see [3.2.6]). above), but here is a concrete example, and it is easier to see its nonsenseness.

The confusion between `culture' and `learning' is also demonstrated here, when Symons talks about "capacity-to-culture" mechanisms. What humans has is capacity-to-learn mechanisms. What they learn is determine by several factors, one of which is culture.

3.3.3 Where human nature come from

On p.144 Symons (1992) says, without any hint of supporting evidence or argument:

Since the brain/mind mechanisms that collectively constitute human nature were designed by natural selection in the EEA [Environment of Evolutionary Adaptation], They must be described solely in terms of phenomena that existed in the EEA;

Thus Symons simply ignores the possibility that learning, based on the facts of the world and culture, affect human nature, without feeling any need for an argument against this possibility.

3.4 `Evolutionary Psychology as a missing link' (earlier chapter)

Cosmides and Tooby (1987) is an earlier discussion of the approach. From the beginning to p.291, the discussion is dependent on the assumption that mental mechanisms are innate. On p. 291, the authors start to discuss learning, and end in p. 299. From this point onward the authors assume that they can ignore learning. I will discuss here only the sections about learning (pp. 291-299).

3.4.1 confusion of the meaning of the term "learning"

In their section 8, the meaning of the term `learning' is quite confused. Normally, when `learning' applies to humans (and animals), it means something like `a robust internal change that causes an improved performance in some task, which is not coded in the genome'. The authors sometimes seem to understand it differently. For example, pp. 291-292:
analytically, however, the only meaning to the word "learned" is "environmentally influenced." As a hypothesis to account for mental or behavioural phenomena, it is nearly devoid of meaning.

This skips both that "learned" is based on some internal changes, which may be figured out and analysed, and that this change was not coded in the genome. These are refutable statements, so they have meaning, but the authors ignore them in this quote.

In the second paragraph of p.292 the authors say that the argument between chomsky and behaviourists was not if language is "learned", but how it is learned. This can be said only if the term "learned" is neutral about innateness, because Chomsky is a strong believer in innateness. On the bottom on p.292 the authors say:

Suggesting that "learning" is an alternative hypothesis is comparable to claiming that an alternative hypothesis to a well-specified theory of vision, such as Marr's (1982), is, "light hits the retina and this causes the organism to see three-dimensional objects."

That, of course, is nonsense. The correct alternative is something like: "Internal changes occur inside the organisms, which are directed by some learning mechanism and its interaction with the visual input, and result in mechanisms that analyse the visual input into three-dimensional object." The failure of the authors to come with the correct alternative hypothesis is probably because they don't feel they have a good argument against it. In the top of page 293 the authors come with this bizarre statement:

A complex, domain specific cognitive program is a learning mechanism;
I don't think the that this sentence can be interpreted without giving the term "learning" a radically different definition than the usual one.

3.4.2 Evidence for learning

On page 292, the authors say:
The belief that the human mind contains only one, simple, domain general cognitive process that results in "learning" -- be it "induction" or "hypothesis testing" or "conditioning" -- is nothing but conjecture. It has no basis in fact, and can only be explained as metatheoretical holdover from the heyday of behaviourism.
The authors here confuse several issues. The alternative to innate mechanisms is learning by general purpose learning mechanism. This mechanism is unlikely to be simple, is not necessarily a single unit, and it is not one of "induction" or "hypothesis testing" or "conditioning". It is some neural structure in the brain (probably most of it), which allows humans to learn new understanding and skills.

The second sentence, about `basis in fact', is again an example of ignoring the ability of human to acquire new skills, as was discussed in [2.4] above.

3.4.3 What is the issue

During the discussion in section 8, the authors say (p.292):
The issue is, What kind of learning mechanisms would natural selection have produced?
That is an inaccuracy. The real issue is `what kind of learning mechanisms did natural selection produce?'. The way Cosmides and Tooby put the question implies that we can find the answer to the real question (what was produced) by answering their question (what would have been produced), so we can ignore other evidence. They do it because they are ignoring other evidence (ability of humans to learn novel tasks, neurophysiology evidence), and need justification for it.

As discussed in [2.6], we cannot deduce evolution of complex traits.

3.4.4 What can be used as explanation

During the discussion in section 8, the authors say (p.293):
It is improper to invoke an undefined process as an explanation. "learning" should not be invoked to explain other phenomena at this point in the history of science, because it is itself a phenomenon that requires explanation.
The first sentence of this quote is correct, but it is irrelevant, because "learning" is a defined process. The authors try to confuse the reader between `undefined process' and `unexplained process'. The second sentence was discussed in section [2.4] above.

3.4.5 blindness to cultural learning

In the next section (section 9, p. 294) the authors discuss the question of how humans know if to eat fruit or feces. Here they show their blindness to cultural evolution, as was discussed [3.2.6].
Does learning from others constitute a solution to the problem? Imitation is useless unless those imitated have themselves solved the problem of the adaptive regulation behaviour. If the blind lead the blind, there is no advantage in imitation.
This is nonsense, because the `solution' could have been found by cultural evolution, i.e. groups of people in which the appropriate manners where followed did better, and these groups pass it on, both to their children and to other groups.

The authors also use here the trick of restricting learning to imitation, which is specially nonsense in this discussion about preferences in eating. These are (obviously) passed on mainly by direct instruction, and also by habit forming.

3.4.6 Do human try new things?

On page 296 the authors say:
Moreover, why would others choose to learn through trial and error while the observer does not? The population of self-experimenters would be selecting themselves out, compared to the observers who parasitize their risky experiments.
That, of course, is plain false, as humans do try things that they haven't try before, so the self-experimenters have not been selected out. However, that is only a fact, so the authors do not fill obliged to take it into account.

3.4.7 Domain-specific criteria vs. innateness

Next, the authors confuse Domain-specific criteria with domain-specific and innate mechanisms. They say (p.296):
Admitting smell or taste preferences is admitting domain specific innate knowledge. Admitting the inference that foul-smelling or foul-tasting entities should not be ingested is admitting a domain specific innate inference.
That is intentional misleading argument, because what is `admitted' when the effects of taste and smell are acknowledged is not an `inference' about what should be digested. All that is `admitted' is that smell and taste can give positive/negative signals to the learning mechanism.

Next, authors ask(p.296):

Without domain specific knowledge such as this, what kind of mechanism could result in learning to avoid feces and ingest fruit?
This is quite amazing, because:
  1. The specific knowledge of taste obviously exist, and nobody would argue against this.
  2. Direct instruction by parents is obviously the answer, and it is clear that this is how infants know how to avoid feces altogether.
Both of these are obvious facts, but not to Cosmides and Tooby.

3.4.8 Evidence from incest research

On page 297 the authors use `evidence' from research about incest. This research is discussed in sections [5.4,5.3] below, where I show how much nonsense it is.

3.4.9 Domain-specific criteria

On p. 298 the authors conclude their argument in this way
Simply put, there is no domain general criterion of fitness that could guide an equipotential learning process toward the correct set of reponses.
(italics in the source).

If we assume that the italicized sentence is the key conclusion, then it is quite reasonable, though this is true only for question of fitness, and not question of facts of the world (e.g. learning to see 3d objects). As discussed in section [3.2.5] above, for these there is a single criterion, and that this is the ability to match internal `predictions' about the behaviour of the world with the real behaviour of the world (more accurately, matching predictions of the sensory input with the real sensory input).

With this caveat, the conclusion that authors reached in the quote above is reasonable, and probably correct. However, the authors continue (p. 298):

Consequently, there are only two ways the human mind can be built. either
  1. All innate psychological mechanisms are domain general, and therefore do not track fitness at all, or
  2. Some innate psychological mechanism are domains specific darwinian algorithms with procedural knowledge specialized for tracking fitness in the concrete situations hominids would have encountered as Pleistocene hunter-gatherers.
This simply does not follow. The criteria have to be domain specific in some cases, but that does not entail that they have to be innate, or that they have to include `algorithms with procedural knowledge'. For example, in the cases that the authors discuss (eating fruit vs. eating feces, p.295), there are two significant criteria:
  1. The test and the smell of fruit feel better.
  2. Input from parents and other adults.
The first is innate, but is not `algorithm with procedural knowledge'. The second is not innate.

The rest of the discussion in this section is using the arguments that are discussed in [2] above.

4 Specific chapters in `The Adaptive Mind'

Here I will discuss the various evidence and arguments that are brought in the rest of 'The Adaptive Mind' for the innateness of adaptation. I will skip chapters 4,7,9,10,11,14,17 and 18, because these do not try to argue for innateness. Most of these chapters are written with an implicit or explicit assumption of innateness, but there is no discussion of innateness vs. learning, and the evidence cannot be used to support or reject either of them.

I will also skip chapter 8, because it is not about psychology. As mentioned above, this is the only chapter where there is discussion the implementation of the adaptation in the real brain.

4.1 Chapter 3: Cognitive adaptations for social exchange

Cosmides and Tooby (1992) spends quite a lot of space (pp. 165-203) on showing that people are specially good in analysing social exchange, and in particular in identifying cheaters. Then they go on to discuss the possible explanation, and actually tackle the possibility of a general learning mechanism learning a specific mechanism, which they skip in chapter 1 ([3.2] above, see in [2.5] above) (p.207-209). The rest of the chapter is a diffused discussion about variations between cultures, which does not contribute anything to the question of innateness vs learning. I will discuss only the section that argues against learning.

4.1.1 Learning specific mechanisms

The first mistake that the authors do is that they give a weight to the fact that people particularly good at detecting cheaters (p.207). Since cheaters will tend to hide their cheating, and there is less chance that they will correct their cheating on their own, it is more important to detect cheating than to detect (for example) other people mistakes. Thus being particularly good at detecting cheating is useful for all people, and therefore will be learned.

Then the authors say (p.207):

Procedures specialized for solving ancient adaptive problems, such as social exchange, would be no more likely to develop than procedures specialized for solving the many evolutionary novel problems posed by life in a modern, postindustrial culture. Evaluating a scientific hypothesis about the effects of dietary cholesterol, detecting someone has misfiled a document, and following the strip of "if-then" directions on one's tax returns would be as effortless -- and require as little explicit instruction -- as detecting cheaters in a situation of social exchange among members of an unfamiliar tribal culture who value cassava roots.
That is a blatant nonsense, because:
  1. A general-purpose learning mechanism would learn the things that the person has experience with. People are engaged in social exchange from birth onwards, so will be good at it.

    To say that the social exchange is in `unfamiliar tribal culture' is ridiculous, because in the experiments the authors use stories about unfamiliar tribal culture, not real situations. The subjects in the experiment simply interpret the stories as analogies of their own familiar experience.

  2. as mentioned above, detecting cheaters is more important than the other tasks for all people.

The authors ignore both of these points, in another theory-driven blindness.

4.1.2 Where would be expect social exchange

Cosmides and Tooby (1992) then explain what they think makes people good in social exchange, and then say (p.208):
In contrast, A Standard Social Science Model approach would, if thoroughgoing, maintain that social exchange is culture-specific and historically contingent, existing in some places and not in others.
This is another example of blatant nonsense. The advantage of groups over individuals is so large for humans, that humans always will live in groups, and whenever there is a group there is social exchange.

4.1.3 What is needed for learning

Next the authors say (p.208):
Moreover, the Standard Model would have to predict that wherever social exchange is found to exist, it would have to be taught or communicated from the ground up.
This is an inaccuracy, because the prediction is that social exchange will have to be learned from the ground up. Humans can learn things even if these are not explicitly communicated to them. This is an example of confusion between culture and learning.

In the rest of the paragraph the authors claim that the fact that people cannot articulate their culture argues against learning. This a baseless argument, because learning is not dependent on being able to articulate what is learned, though it is helped by it.

4.1.3 The need for shared assumptions

The authors continue by discussing the need for shared assumptions for communication. In the bottom of p.208, they say:
Consequently, for someone ignorant of a culture to learn it in the first place, it is necessary that she already share many assumptions with those from whom she is learning the culture. If human minds truly were initially tabula rassa, with no prior contentful structure, then no anthropologist or immigrant to a culture could ever learn about it (quine,1969).

The first sentence is correct. The second sentence, however, is blatant nonsense. That is because any human, even if started as tabula rassa, will share knowledge (assumptions) about the physical structure of the world, the significance of animal and plants, the physiology of humans, the basic rules that are required for a society to survive, and the basic rules of communication. These are learnt by all humans, because they are facts of the world, so the anthropologist/immigrant would share them with a novel culture.

4.1.4 Children learning

On p.209 the authors reach child learning:

If the evolved architecture of the child's mind contained no meaningful content at all -- that is, if it were truly content-free -- then children could never learn the culture they are born into: They would have no reliable means of interpreting anything they saw and hence of placing the same construction on it that members of their culture do.
The argument is based on the assumption that children must have 'reliable means of interpreting anything they saw' to be able to learn the culture. That is a ridiculous argument, because it is obviously true that children indeed 'have no reliable means of interpreting anything they saw', and develop their ability to interpret the world as they grow up. That is another theory-driven blindness to the facts of the world.

4.1.5 Effectiveness of learning vs. genetic evolution. (from earlier paper)

The chapter that was discussed above in [4.1.1-4.1.4] is based on an earlier research paper (Cosmides and Tooby, 1989). In the earlier paper, the arguments above do not appear. Instead, the authors say (Cosmides and Tooby, 1980, p.93):
It is implausible to expect that natural selection would leave learning in such a domain to the vagaries of personal experience processed through some kind of general-purpose learning mechanism.
That ignores the fact that evolution based on cultural changes is far faster than genetic evolution, and therefore much more effective in getting it right in social exchanges.

It is amusing that the authors here, contrary to their habit of discussing culture when they should be discussing learning, refer here only to learning instead of culture.

4.3 Chapter 5: Mate Preference Mechanisms

In this chapter, Buss(1992) discusses mate preference mechanisms, and discusses their origins (p. 261-263). This discussion stands out in being sensible. The author seems to be inclined towards innateness ("hard-wired") mechanisms, but discusses the evidence in an almost objective fashion.

The only evidence that the author brings for innateness is the article by Langois et al (1987), which shows that infant looks a little longer (10-20%) at more attractive faces. However, Langois et al themselves say that their data cannot distinguish between a preference to attractive faces and preference for more general attributes, like mirror symmetry, smoothness etc.

Apart from this, Buss (1992) does not bring any evidence against learning, and instead say (p. 263):

Domain-general learning mechanisms, however, are unlikely to evolve in contexts where an important and specific adaptive problem must be solved (Cosmides & Tooby, 1987, 1989; Tooby & Cosmides, 1989).
(The two 1989 references are the precursors of chapters 1 and 3, which are discussed in [3.2] and [4.1] above, The 1987 references is discussed above in [3.4]). So it is not the evidence that show innateness, but 'theoretical arguments'.

Buss (1992) finishes the discussion in this way:

Our current state of knowledge does not allow us to differentiate among these alternative causal accounts [learning and hard-wired] of the origins of the closed calibration between mate preferences expressed by one sex and tactics of intrasexual competition displayed by members of the opposite sex.

This is a reasonable evaluation of the behavioural data, but it ignores any evidence from brain damage and neuroanatomy, which clearly rule out any innate mate preferences mechanisms in the brain. If there are any such innate mechanism, it is in the physiology of the body, rather than in the brain/mind.

4.4 Chapter 6: The evolution of sexual attraction

Ellis (1992) discusses if the sexual attraction can be learned, and does exactly the same mistake that Symons (1992)([3.3.2] above) makes, i.e. ignoring the fact that cultures that get sexual attraction wrong cannot survive in evolutionary time scale.

Ellis goes even further, and ignores the human physiology. He claims that mainstream social scientist would predict that sexual attraction cannot be predicted, and then says (p.284):

By this reasoning, men and women could just as easily be attracted to tree trunks as other people if they were culturally conditioned to do so, and distribution of sex differences across cultures should be random.
By making a concrete example, Ellis make it easier to see that he writes nonsense: any human which is mature enough to have sexual attraction has enough experience with his/her body and other objects to know that having sex with a tree is not going to be enjoyable. If any human does fail to make this judgement, shhe is going to make it the first time shhe tries. Note that this is because of the physiology of the human body and the physical properties of trees, and the ability of humans to learn and think. It is not dependent on any special mechanism.

4.5 Chapter 12: natural language and natural selection

Most of the discussion in this chapter simply ignores the possibility of learning of language specifics by general-purpose learning mechanism. The authors do address this point directly on p.474, where they discuss a general-purpose learning mechanism that learns from examples. This is obviously nonsense, because there is nothing to stop the learning mechanism to use any other techniques.

I discuss the ideas of psycholinguistics separately.

4.6 Chapter 13: Perceptual organization of colours

Shepard (1992) spends a considerable length (pp. 495-516) on arguing that the trichromaticity of visual perception is adaptation for allowing colour constancy, and that other features of colour perception are adaptation to other features the terrestrial scene. The problem with this theory is that it does not make any predictions at all.

On p. 517, however, Shepard claims that trichromaticity is not only coded by having three types of cones in the retina, but that there is also an innate specialized structure for it. In theory, this is possible, because the argument from the stochastic connectivity of neurons is not applicable to colour, which is already coded in the receptor, so does not requires any further processing, and therefore does not require specific connections below the level of neuronal populations. However, considerable amount of research about the visual system of humans and other vertebrates did not find any single hint for specialization for trichromaticity anywhere (there are identifiable cells sensitive to colour in the cortex, but not identifiable cells which are sensitive to a specific colour), which is a strong evidence against it. ([21Jan2003] There is now a a a paper in Nature 421, 535 - 539 (2003) that finds areas sensitive to specific colours in Macaques. These areas are variable across individuals (Figure 2), so they are not genetic specialization, and there is no reason in the paper to think that areas have any significance).

Shepard (1992) simply ignores this evidence completely. Instead, he brings his own research, which shows that colour-blind people give the same estimate of distance between named (as opposed to presented) colours. The author claims that this shows that internally they also have the three-dimensional structure like colour-normal people.

That is nonsense, because it ignores the possibility that colour-blind people learn the attributes of different colours from the people around them. They have a strong motivation to do this, because it helps them to understand how other people see the world. In fact they must have learned these attributes: even if they have the proper internal structure, the only way they could associate the names of the colours (which are arbitrary) with these structures is by learning the attributes associated with the colour names and matching them with the internal attributes. Thus learning must be invoked, and once it is, there is no reason to assume internal structure. This is an impressive example, because Shepard ignores learning even though it is necessary.

Shepard (1992) also offers us this pearl (p. 517):

One particularly articulate protan insisted that although he could not distinguish the (highly saturated) red and green we showed him, neither of these papers came anything near matching up to the vivid red and green he could imagine!
This, of course, shows nothing, because there is nothing to stop this person to use the label red and green for any arbitrary 'vivid' (whatever that means for him) image he has. There is no way, either for him or anybody else, to check the correctness of these labels, and it is probably motivated by giving the person the illusion that he does not miss as much as he actually does. That is an example of the `relying on introspection' error (see in Reasoning errors, [3.20]).

In the rest of the chapter (pp. 514-525), Shepard (1992) discusses colour `categorization'{4}, i.e. the fact that human see distinct colours even in continuous spectrum. Since colour is detected by cones in the retina, and that is the only known specific-colour specialized part of the visual system, it makes sense to start to look for the proximate reasons for this distinction in them. For example, a border between two colour may be where the response of one (or more) of the cone types has a large gradient, or where the response of one of the cone types emerges out of the background noise. Amazingly, Shepard (1992) completely ignores the cones in the retina in his discussion of colour distinction.

4.7 Chapters 15-16: Environmental preference

All the data that is presented in these chapters can be explained by the following three facts:

  1. Humans like interesting environment, because they like interesting things in general (in other words, they are curious). Why humans are curious is a different question, but they are curious.

  2. Most of the time, humans like to feel safe. Note that this is not only from a real physical danger. For example, humans also try to avoid being seen doing embarrassing things (where the definition of embarrassing is variable), so humans feel somewhat unsafe when they cannot see everybody that can see them.

  3. Humans like plants, because they like their smell, they find them very interesting and beautiful, and they are useful in various ways.
Orians and heerwagen (1992) seem to be unaware of this kind of explanation, and consider only additional, more specific, evolved adaptations. Kaplan (1992) is aware of these kind of explanations, but still insists on looking for evolutionary explanation, without any justification. While less blatant than some of the examples above, this is another example where theory drives the interpretation of the evidence, even though the theory itself is not supported by any evidence.

5. Other examples

In this section I will discuss various examples, showing how EPs tend to use nonsense arguments to prove their points. This is not supposed to be a comprehensive survey of the literature, but to give the reader a taste of the kind of mistakes he should expect to (and will) find when reading EP texts.

5.1 infants Object concept: the primary experimental evidence

Baillargeon's research is a typical example of the kind of research that EPs relies on to prove that babies have innate knowledge on the world. In the discussion here I show how the data is being overinterpreted as we go from primary research paper to secondary sources. The papers which I discuss here are those that were quoted by Leslie (1988), which is quoted in Tooby and Cosmides (1992) above ([3.2.2]). It should be noted that the latter is the reason for selecting Baillargeon's papers, rather than the papers themselves being specially bad. In fact, if we ignore the unscientific inferences which are listed below, the raw data is quite interesting.

5.1.1 first experiment

In the first paper, Baillargeon (1986) describes an experiment with 6 and 8 month old babies, in which in the `impossible' event it looks like a toy car moves through a block. After describing the experiment, in which the infants looked longer at the event that for adults would look impossible, Baillargeon (1986) says (P.34):
What is clear, however, is that the infants did show a preference for the impossible over the possible event. This preference suggests that the infants were surprised to see the car reappear from behind the screen when the box stood in its way.

The first sentence here is already slightly problematic, because it implies (at least weakly) that the infants looked longer (showed a preference) at the impossible event because it is impossible, which the data does not tell us. The suggestion in the second sentence is completely spurious, as there is nothing in the infants behaviour that indicates surprise. The most that can be said is that they found it more interesting for some reason. There may be several reasons why the infants found the impossible event more interesting. The simplest is that their experience tell them that when two objects come to the same place in space interesting things happen.

This spurious suggestion may seem pretty harmless, until we read the next sentence:

Such results implies that the infants understood (1) the box continued to exist, in its same location, after it was occluded by the screen; (2) The car continued to exist, and pursued its trajectory, after it disappeared behind the screen; and (3) the car could not roll through the space occupied by the box.

The first two `understanding' can be regarded as reasonable, provided the term `understand' is taken in its weakest meaning (so, for example, `understand that the box continue to exist in the same location' means the `keeps some association between the box and the location'). If the term `understand' is taken in the way the term is usually used, these assertions are already a strong overinterpretation of the data. The third assertion, that the infant understand that the car could not roll through the box, is completely spurious. It is based only on the spurious assertion that the baby was surprised to see the car appears in the `impossible' case.

In the conclusion of the paper, Baillargeon (1986) goes even further (p. 40):

Further, they are able to use representation to reason about simple collision events involving hidden objects.
Obviously, there is nothing in her data that shows anything about reasoning, and this is evidence-free theorizing.

In this paper, Baillargeon does not actually discuss whether the understanding of the infants is learned or innate, but in the last sentence she implies (weakly) she thinks it is innate (P.40):

From this perspective, the general problem for research becomes that of examining infants' intuitive "theories" about the rules that underlie the displacement of objects, and of specifying the nature and range of their causal reasoning ability.

5.1.2 second experiment

In the next paper, Baillargeon (1987a) do a simpler experiment than the previous experiment, because only one object is hidden, and it is the hiding screen that moves and `collide' with it in the `impossible' case. The subjects are also younger, 3.5 and 4.5 months. Baillargeon does not actually tell us why she did not repeat the previous experiment with younger infants, though the reason is clear: it was too complex for them. After describing the experiment, Baillargeon (1987a) says (P. 662):
The 4.5 month old in experiment 1 and the 3.5 month olds in experiment 2 and 3 who were fast habituators all looked reliably longer at impossible than at the possible event, suggesting that they understand (a) the object behind the screen (i.e. box or Mr. potato head) continued to exist after the screen rotated upward and occluded it and (b) the screen could not move through the space occupied by the object.
The first `understanding' (a) may be reasonable interpretation, provided we understand the term `understand' weakly, as above. The second `understanding' (b) is spurious. Like in the previous experiment, all that could be said is that the babies found the `impossible' event more interesting. Apart for the same explanation as in Baillargeon (1986) above, they could have been just waiting for the box to appear once the screen went through it.

In this paper, Baillargeon explicitly discuss whether the babies abilities are learned or innate, and she says (p. 663):

The results of the present experiment are not sufficient to determined which (if either) of the hypotheses mentioned above better explained the presence of object permanence in 3.5-month old infants.

5.1.3 third experiment

The third experiment that Leslie (1988) cites was done with 7 months old infants (Baillargeon, 1987b). In this case a screen moved toward a hidden object, which was in variable orientation or distance from the screen. In the conclusion of the first part of the experiment, Baillargeon (1987b) says (p.190):
The infants in the two screen conditions had reliably distinct looking pattern: the infants in the sliding-screen condition looked longer at the horizontal- than at the vertical-box event, whereas those in the rotating-screen condition looked longer at the vertical- than at the horizontal-box event. These results suggest that the infants (a) represented the location (sliding screen) the height (rotating screen) of the box behind the screen, (b) used this information to estimate at what point the screen should reach the box and stop, and therefore (c) were surprised when the screen continue to move after it reached this point.
Again, conclusions (a) and (b) are reasonable, provided `represent' and `estimate' are used in a weak sense. The third conclusion is, again, a spurious overinterpretation.

The paper also contains an experiment with compressible/incompressible objects, which is analysed in the same way.

In the conclusion of this paper, Baillargeon seems to want to claim that the infants have symbolic representation, but realizes that her data does not actually support it. On p.198 she says:

According to Piaget (1952) and, indeed, most developmental psychologists (see mandler, 1983, 1984 for a critique), infants less than 18 months of age do not have a symbolic representation system.
She goes on to claim that her experiments refute this, thus implicitly saying that the infants do have symbolic representation, though she does not say it explicitly.

5.1.4 Important points to notice in Baillargeon papers

The following points should be noted in Baillargeon papers when they are used in discussing the question of innateness:
  1. In all the experiments the sole dependent variable was looking time.
  2. There is a graduation in the complexity of experiments, and only the simpler experiments are reported for the younger infants (3.5-4.5 months). Baillargeon does not actually tell us why is this, and leaves it to the reader to figure out that the younger infants could not cope with the more complex experiments. Note that even in the simple experiment, only part of the 3.5 months old infants gave a significant response.
  3. Baillargeon always interprets longer looking time as indicating surprise, without any evidence.
  4. Based on the attribution of surprise, Baillargeon attributes to the infants understanding and reasoning, even though the data is only behavioural (looking time).
  5. Baillargeon seems to prefer innate explanation, but she does not think that her data proves it.
With the exception of the last point, all these points are applicable to the research of Spelke (1988, 1990, 1991; Spelke et al 1992). In contrast, Spelke strongly believes that the skills that the infants have are innate, though her data does not contain any evidence for this either.

5.2 "triangular awareness" in chimpanzees

Brown (1991) writes in P.111:
For example, Franz De Waal's Chimpanzee Politics (1982) provides evidence among chimpanzees for what the author calls "triangular awareness," the ability of an individual A to calculate the interdependence of the three separate relationships composed by his relationships with individual B and C and the relationship between B and C themselves. This sort of ability among our nearest relatives in the animal world suggests that the same ability in us has phylogenetically deep roots and is therefore innate.

The last sentence is a clear nonsense, because chimpanzees are intelligent enough to learn this kind of things. Therefore, the fact that both chimpanzees and human has "triangular awareness" is not an evidence either for or against its innateness. It makes sense only to people that ignore the learning ability of humans and chimpanzees.

5.3 Incest avoidance as a universal

In his book about Human Universals, the only universal that Brown (1991) discusses in depth is incest avoidance (chapter 5, pp.118-129), and this chapter "looks in greater detail at the means of explaining universals" (P.6). The main hypothesis he is discussing is the observation by Westermark that there is "a remarkable lack of erotic feeling between persons who have been living closely together from childhood." (p. 119), and the assumption that this is a result of an innate imprinting mechanism.

5.3.1 evidence from kibbutzim

The first `evidence' is from groups of people that were grown together, mainly in kibbutzim in Israel (P.120-121,123). Brown quotes research that shows a very small numbers of marriages (and allegedly also small number of shorter sexual relations) inside groups of people that were grown together. Brown emphasize that there are no taboos that restrict this kind if relations. There are at least two major faults in the discussion:
  1. There is a very simple explanation for the lack of intragroup relations. When a member of a close-relations group develops special relations with another person, the rest of the people in the group feel that something has been taken away from them. They will rarely admit it, but it irritate them significantly. They show this irritation by subtle changes in their behaviour, to some extent towards the person of which they `lost something', but mainly towards the person that `took' it. If the latter is an outsider, this does not have a large effect, but if shhe is part of the group as well, shhe feels this very strongly. As a result, members of close-relations group learn to look for special relations with outsiders.

  2. In most the cases, pairs are of different age, with the male almost always older. Marriages of the same age are rare. The data that Brown quotes does not take this very significant factor into account, and is therefore meaningless.

    Brown himself is actually aware of this factor, and he quotes Spiro about this (P. 128). However, this quote is only half a paragraph, separated by several pages from the main discussion of the data. In a serious discussion, this point would be raised when the data is presented, or even better, the data would not be presented at all, as it is meaningless.

5.3.2 Evidence from chinese `minor-marriages'

The second evidence is from `minor-marriages' in china (p.121-122). In this kind of the marriage, a young girl is moved to live with her future husband (also a child). Brown bring evidence that this marriages are less successful than normal marriages, for example they are 30% less fertile, and regards this as an evidence that living together as children cause the future husband-and-wife to be imprinted against each other. Brown discusses and dismisses several possible objections. He also brings a similar evidence from an arab village (p. 122).

However, Brown ignores the most obvious objection, that minor-marriages are effectively random pairing. Normal marriages, even when they are arranged by relatives rather than by the groom-and-bride themselves, are done with at least some considerations of the personalities of the bride and groom. In contrast, in a `minor-marriage' this information is not available, because you cannot predict what kind of person a child is going to be.

Considering this randomness, it is actually very surprising that `minor-marriages' actually succeed to reach 70% fertility of normal marriages. Most likely, this number ignores `minor-marriages' that have been broken before the bride reached maturity.

5.3.3 Incest in animals

The next kind of evidence that Brown brings is that incest avoidance exist in non-human animals, and then he says (P. 124)

Consequently, the assumption that human incest avoidance is fundamentally is cultural phenomenon now rests on the inelegant assumption of a double discontinuity with the animal kingdom; unlike other species we lack innate avoidance mechanisms; unlike other species we therefore avoid incest via cultural prohibition (Arens 1986:94).

The only reasonable interpretation of `inelegant' here is something like `nonparsymonious'. This is bad argument, because the two `discontinuities' are strongly dependent on each other. If a learned mechanism to avoid incest evolves (culturally), and it covers all the cases that the innate mechanism does and more, the innate mechanism will be degraded by random drift. Alternatively, if the innate mechanism become ineffective (e.g. because it worked only in small groups, or only in short-childhood species, or only in a specific environment), it will give a strong drive to develop another mechanism, which in a learning species is more likely to be learned than innate, because learning mechanisms evolve much faster than innate mechanisms. Thus the double discontinuity is a consequence of a single process.

5.3.4 Evidence from actual incest

Brown also tries to support his ideas by bringing evidence from actual incest. For example, he tries to suggest that imprinting can explain why father-daughter incest, and specially stepfather-stepdaughter incest, is more common than other kinds of incest. However, as Brown admits, there are many other possible explanations, so this is not and evidence for or against imprinting. Brown, however, presents it as if it is a supporting evidence for imprinting.

5.3.5 Summary

In short, all the evidence that Brown brings has straightforward explanations without any assumption about innateness. On the other hand, nobody as yet came out with any direct evidence of the imprinting, but Brown does not bother to discuss this point at all.

5.4 Incest: pseudo-science

Shepher (1983) develops model of incest avoidance, and `supports' it by evidence. The discussion in this book, however, is full of overinterpretations. For example, after bringing evidence that there is more in incest in dyads (pairs of people) of a father and daughter, Shepher (1982) says (p.127):
Thus my prediction has been verified: differentiated parental investment results in different psychological mechanisms in the different incestuous dyads, which are reflected in the differential occurrence of committed incest in these dyads.
This is clearly nonsense, because only the differential occurrence of committed incest was verified. The data tells us nothing about the underlying reasons, so the first part of the `prediction' has not been verified.

Next, he says (p.128):

My second prediction was that, because females lose more than do males, females will resist incest more than males. Evidence for the second prediction seems to be overwhelming. In Maisch's (1968) German courts study females constituted 91% of the victims and 9% of the aggressors.
Again, the data does not tell us anything about the underlying reasons.

This book is an example of pseudo-science, but is still regarded highly and quoted many times. For example, it is quoted 6 times in barkow, Cosmides and Tooby (1992), and is quoted by brown (1991) above.

5.5 What kind of explanations

As mentioned above in [4.3], Cosmides and Tooby decide what kind of explanations they want before considering the evidence. The strongest expression of this is in Tooby and Cosmides (1987), p.32:
Models of psychological phenomena need to be expressed in an algorithmic, procedural form, or at least as structured and well-defined "cause and effect" models, instead of a vague qualitative descriptions, or as patterns found in behaviour.
That approach is correct only if the way the psychological phenomena are implemented is always `structured and well-defined'. Even a-priori this is unlikely, and because of the stochasticity of low-level connectivity in the cortex, psychological phenomena are unlikely to ever be well-defined, so this approach is almost guaranteed to go wrong.

5.6 Totally ignoring other possibilities.

In a new book, which applies Evolutionary Psychology to mental disoreder (Baron-Cohen, 1997), the editor says in the preface ( P. x (the roman numeral )):
Having identified universal aspect of behaviour or cognition, evolutionary psychology then encourages the researcher to consider its adaptive significance: What advantages does this behaviour or cognitive process confer on survival and reproduction? How might this behaviour or cognitive process have shaped by natural selection in our ancestral landscape ?

Such universals also need to be explained in terms of neural mechanisms that control them --here, evidence from neuropsychology, neuroimaging and neurophisiology methods need to be considered. Ultimately, the genes that code for specific universal neurocognitive mechanisms will need to be identified. Both behavioural and moleculer genetic techniques need to feature in such accounts

Thus the author takes for granted that any universal neccesarily:
  1. Confer advantage (Ignoring the possibility that it a side effect).
  2. Was evolved by natural selection.
  3. Has a specific neural mechanisms to carry it.
  4. Is specifically coded by genes.
It worth noting that the author does not ignore neurobiology as most EPs do, but does not give it any role in confirming (and hence maybe refuting) the psychological theories. He takes it for granted that the universals are implemented by specific mechanisms, and neurobiology is just about finding these mechanisms.

5.7 A typical research paper

I selected
Price, M.E., Cosmides, L. & Tooby, J. (In press). Punitive sentiment as an anti-free rider psychological device. Evolution and Human Behavior 23 (2002) 203-231
for commenting on because it is available online, the first author published it in this message, and the senior authors (Cosmides and Tooby) are central researchers in Evolutionary Psychology. [10 May 2002] It also appears in the front page of the Center for Evolutionary Psychology in UCSB.

The most outstanding feature this paper is what it does not contain: any mention of the possibility that what they observe, i.e. the punitive sentiment, is a result of learning (most simple hypothesis: people grows in a society where almost everybody thinks free-riders should be punished, so they learn to believe it too). In fact, the word "learn" does not appear in this paper at all.

At one point the authors seem to get close to actually discuss this point. They actually write (p.205, bottom):

Therefore, we wish to raise the following questions: Is there evolved neurocognitive circuitry that causes people to punish free riders in collective action contexts?
Which suggests that they may consider the possibility that punishing free reiders can be a result of mechanism that is not "evolved neurocognitive circuitry". However, that is not what the authors mean, and they clarify it in the next question:
More importantly, does our motivational system show sufficient evidence of special design for this outcome that we should conclude it is a solution to the problem of punishing free riders, rather than a byproduct or maladaptive misfiring of a mechanism designed for solving some other kind of problem?
I.e. they completely ignore the possibility of learning mechanisms. They can only think about designed mechanisms. They are going even further on p.206, in the beginning of the section titled: "Punitive sentiment as an anti-free rider psychological device":
By a punitive sentiment, we mean the designed expression of evolved, reliably developing circuitry in the motivational system: specifically, a desire that the target of the sentiment be harmed.
I.e. they actually define "punitive sentiment" to be an expression of evolved circuitry (by which they mean "genetically evolved circuitry"). This is clearly different from the way "punitive sentiment" is understood, which is any expression of wish to punish, independently of the underlying causes and mechanisms. In the rest of the paper, they take it for granted that what they see is a "punitive sentiment", which would be justified with the normal understanding, but not with their definition (because they don't show any evidence that it is evolved). Thus they use this bogus definition to "justify" the assumption of evolved circuitry, which they also take for granted in the rest of the paper.

The total avoidance of any mention of learning and taking evolved cicuitry for granted are a typical approach in evolutionary psychology. It is quite effective, because readers of this kind of paper get the impression that existence of the evolved circuitry have actually been established. In effect, it is an extended example of the lying by implication maneuver. This is the most significant point of this paper.

In addition to the major maneuver, the paper also contain several logical mistakes.

  1. On p209 :
    (a) If encouraging participation by otherwise reluctant individuals is one adaptive function of a system that causes punitive sentiments toward nonparticipants, then those who are most likely to benefit from the achievement of a group goal should differentially act to induce others to achieve this goal (e.g. Alexander, 1987, pp. 191-192; Irons, 1991; Cronk, 1994).
    This is true in principle, but evolution does not necessarily does what it should, and in many cases it achieves some effect without doing it in what looks to us as the optimal way. This can be either because evolution didn't happen to have the right mutations, or because achieving the optimal state on behavioural grounds involves costs that make it non-optimal overall. Large portion of the following discussion is based on this logical error.

  2. on p. 209 they write :
    (b) The predicted relationship between own expected benefit and sentiment for punishing others should remain significant, even when one controls for the individual's own willingness to participate. After all, encouraging self-sacrificial participation by others provides the largest net benefit, even for a free rider.

    (c) If encouraging participation by others is this adaptation’s only function, then after controlling for perceived benefit, any relationship between the individual's willingness to participate and punitive sentiment toward others should disappear.

    These are both based on the assumption of optimal behaviour in (a) above.

  3. On p.209(d):
    There is nothing inherent in the problem of labor recruitment that privileges punishment over reward as an incentive - each might serve to motivate recruitment.
    That is simply false, because giving a reward must involve a cost (the reward must be some resource that is made unavialble to the rest of the group). Punishment does not necessarily require high cost. (Note that the high cost of the modern penal system is mostly for identifying correctly those who deserve punishment, rather than the actually administrating the punishment, and that the penal system is so costly only in modern states).

  4. On p.220, when discussing the question of (lack of) punishing individuals that contribute less than the optimum but more or equal than the average, the authors say :
    The optimal recruitment function predicts that subjects will feel punitive even toward individuals who are contributing their fair share, as long as they would benefit by these individuals contributing even more.
    This is another false statement. The underlying mechanism are assumed to be shared among inidviduals, so if an individual has a mechanism that makes him punish individuals that contribute their fair share, Other individuals will have this mechanisms too. Hence the other individual will punish the individual, reducing his benefits. Clearly, there is an optimal level for requiring cooperation from other individuals, and that is the "fair share" level (more or less, we cannot assume that the optimization reaches the actual peak). Individuals with higher requirements can have an advantage only when they form only small fraction of the population. If their number become large, punishment become very common event, and punishment become ineffective in getting cooperation.

    The authors seem not to realize this point. They seem to think that there is some optimum level which is above the "fair share" level (p.220) :

    If motivating optimal participation were the selection pressure responsible for designing punitive sentiments, then one should feel punitive toward anyone contributing below the optimum. In Fehr & Gächter's games, an individual benefits the most when others contribute everything. Thus each individual should view the optimum level of participation by others as the maximum contribution.
    It is not obvious what the "optimum" in the first sentence mean, but it clearly meant to be different from the "fair share" level. The last sentence suggests it means maximum contribution. As discussed above, that is not the optimum level for the population. Maybe the authors mean the optimum for the individual. That makes sense only if it is possible to decouple the behavior of each individual from the behaviour of other individuals, but this is not possible if the underlying mechanisms are evolved.

    It is also quite ludicrous the way the authors use the experimental results. Whatever are the underlying mechanisms of the "punitive sentiment", they did not develop (genetically or otherwise) to optimize behaviour in Fehr & Gächter's games. Thus the fact that they did not achive optimal behaviour in these games cannot tell us anything.

It should be noted again that these logical errors are secondary. The primary problem with the paper is that it completely ignores learning, and "argues" for evolved circuitry by assuming it.

5.8 A more review-like paper

Here is another example of contemporary writing in EP (Marc D. Hauser and Elizabeth Spelke, Evolutionary and developmental foundations of human knowledge, Chapter for "The Cognitive Neurosciences, III" (Ed. M. Gazzaniga) (to be published 2004)). In contrast to the research paper above, the word "learn" and derivations appear many times in this chapter. However, the idea that the "core knowledge systems" themselves are learned is not considered at all. The authors clearly believe strongly that these systems are innate, and express this by the language they use (e.g. these systems "evolved before humanity" , and infant are "endowed" with capacities or "have" capacities, rather than acquire or develop them), but there is no explicit discussion.

It is possible that the authors consider some of the points that they make as a relevant argument to the question of innateness. For example, they may consider the fact that some capapcity is common to several species as an evidence that it is innate (clearly false argument, because it can be learned by all the species, or learned by only some of the species). Since they don't make the point explicitly, it is not possible to argue against it.

The uthors make repeated a serious error of assuming much more power for a behavioral characterization than is reasonably warranted. They say (p.6, middle):

If the same signature characteristics and limits are observed at different ages, then those common signatures provide evidence for a common system of core knowledge at work at the different ages.
That is clearly invalid inference, because "signature characteristics and limits" maybe characteristics of the task itself, or characteristics of the whole cognitive system rather than of a "core knowledge" system. The error is made clear by the discussion of stereopsis in the bottom of p.6, most clearly when they regard horizontal selectivity as such characteristics, because the horizontal selectivity is clearly a result of the horizontal pairing of the eyes. The other two characteristics (hyperacuity and critical range) are also a plausible a result of the characteristics of the visual system as a whole, rather some "core knowledge", and since there isn't any anatomical evidence for "core knowledge" of stereopsis, this former is the more plausible explanation.

5.9 Another example

[18 Sep 2003]

In this article(Monkeys reject unequal pay, Nature 425, 297 - 299 ), the authors report that some monkeys behave in a pattern that can be explained by a notion of fairness. They conclude (end of first bolded paragraph):

These reactions support an early evolutionary origin of inequity aversion.
This conclusion, ofcourse, is sensible only if you already assume "evolutionary origin" (i.e. that it is innate). If the behaviour is learned in humans, then the observatios on monkey tell you nothing. Thus the authors "argue" for the innateness of the fairness notion by implication, rather than make the argument explicit.

The word "learn" appears once in this article, in the context of the subjects learning what they are going to get.

This article is also quite an ovrinterpretation of the data. Monkeys are too complex subjects to assume that you can understand their behaviour from comparison between small number of conditions.

Interestingly, in this case the commentary in Naure Science Update is quite even-handed. They let one of the authors say that it is innate in the second paragraph, but then quote other researcher that suggests the possibility that it is learned, and highlight this quote. It is interesting that the other researcher make the point that it is learning is captivity, giving the impression that he also doesn't believe that they can learn it in the wild.

-------------------------------------------------- --------------------------------------------------
-------------------------------------------------- --------------------------------------------------
-------------------------------------------------- --------------------------------------------------

Comments:

{1} Dawkins (1976,1989) (chap 11, p. 203 in the 1976 edition, p. 189 in the 1989 edition) discussed the idea of cultural evolution, but he makes two significant errors:

  1. He discussed it in terms of `memes', which are well specified separated units. This is in parallel to genes, but has no logical basis. For evolution, there is no need for such units, and the `replicators' can be flowing like liquids. The only important attribute of the replicator is that mixing two or more `allels' does not give the average of them (otherwise mutations are diluted away).

    Using 'memes' as parallel to genes (and 'memetics' parallel to genetics) suggests that results from genetics can be applied to 'memetics'. This is simply false, because all the results of genetics are applicable only when inheritance is from parent to child. In cultual evolution, evolution can go in all directions, and therefore results from genetics are inapplicable.

  2. More importantly, Dawkins limits the transmission of `memes' to imitation. Cultural traits are transmitted in many ways, and the most important are:

    1. Giving the child a specific experience. For example, children take part in social interaction from their childhood, so they get used to it, which increases their tendency to do it as adults.

    2. Implicit instruction, by making what the individual has to learn a fact of life. For example, that `having a proper job is the best way for earning your living' is actually true for most of people, so they learn it like any other fact of the world.

    3. explicit instruction.

    4. weak and continuous coercion. For example, showing sexual attraction to members of the same sex is in general discouraged, so people learn (normally as children) to direct their sexual attention to the other sex.

    A cultural trait can be, and normally is, transmitted by more than one way.

Imitation is only one way of transmission, and not a good one, because it is unreliable.

To most of people, both transmission by imitation and well-definedness of ideas seem suspicious, and rightly so. By limiting transmission to imitation and requiring it to be by a well-separated units (memes), Dawkins makes cultural transmission, and hence cultural evolution, seems like an unlikely mechanism.

Dawkins also makes the mistake of assuming that `memes' have to be associated with specific structures in the brain, which is obviously nonsense.

{2} Sometimes it is claimed that cultural trait cannot develop by evolution, because this is group selection, and `group selection have been shown to be wrong'. However, both the notion of 'group selection' and the argument against group selection are based on the assumption that the 'replicators' cannot move horizontally from group to group and inside the group from individual to individual. This is true for genes, but not for cultural traits. In other words, there is no well defined 'vehicle' for cultural traits, only replicators.

{4} It is a common error in cognitive sciences to regard the distinction between colours as `categorization', and use is for research about human categorizing. Because the information about colours is available right from the receptors, it is a very different operation from categorizing, which has to rely on information that is found only after considerable processing of the input (e.g. animal kinds, tools etc.). This difference means that we cannot use colour distinction as analogy to categorizing in general, because the former is several order of magnitudes simpler.

------------------------------------------------------

references:

Baillargeon, Renee (1986). Representing the existence and the location of hidden objects: object performance in 6- and 8-month old infants. Cognition 23, 21-41.

Baillargeon, Renee (1987a). Object performance in 3.5 and 4.5 month old infants. Developmental Psychology, 23, 655-664.

Baillargeon, Renee (1987b). Young infants reasoning about the physical and spatial properties of hidden object. Cognitive Development 2, 179-200.

Baillargeon and Graber (1987). Where's the rabbit? 5.5 month-old infants representation of the height of a hidden object. . Cognitive Development 2, 375-392.

Barkow, Cosmides and Tooby (Eds)(1992). The adapted mind: Evolutionary psychology and the generation of culture. Oxford University Press.

Baron-Cohen, Simon (Ed.)(1997). The Maladapted Mind: Classic Reading in Evolutionary Psychology. Psycholofy Press, Hove, UK. Brown, Donald E.(1991). Human Universals. Mcgraw-Hill, NY.

Buss(1992). mate Preference mechanisms: Consequences for partner choice and intrasexual competition. In Barkow, Cosmides, and Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. New York: Oxford University Press.

Cosmides and Tooby (1987). From evolution to behaviour: evolutionary psychology as the missing link in Dupre (Ed.) The latest on the best . Cambridge, MA: MIT press.

Cosmides and Tooby (1989) Evolutionary psychology and the generation of culture, part II. Case study: a computational theory pf social exchange. Ethology and sociobiology 10, 51-97.

Cosmides and Tooby (1992). Cognitive adaptations for social exchange. In Barkow, Cosmides, and Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. New York: Oxford University Press.

Cosmides and Tooby (1994). Beyond intuition and instinct blindness: toward an evolutionary rigorous cognitive science. Cognition 50, 41-77.

Cosmides, Tooby and Barkow (1992). Introduction: Evolutionary Psychology and Conceptual Integration. In Barkow, Cosmides, and Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. New York: Oxford University Press.

Dawkins, Richard (1976,1989). The selfish gene. Oxford University Press.

Dupre (ed) (1987). The latest on the best . Cambridge MA: MIT press.

Ellis(1992). The evolution of sexual attraction: evaluative mechanisms in women. In Barkow, Cosmides, and Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. New York: Oxford University Press.

Kaplan (1992). Environmental preference in a knowledge-seeking, knowledge-using organism. In Barkow, Cosmides, and Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. New York: Oxford University Press.

Langois et al (1987). Infant preference for attractive features: rudiments of stereotype? Developmental Psychology 23, 363-369.

Leslie (1988). The necessity of illusion: perception and thought in infancy.in Weiskrantz, (ed), Thought without language. Oxford: Clarendon Press, pp.185-210.

Nesse and lloyd (1992). The evolution of psychodynamic mechanisms. In Barkow, Cosmides, and Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. New York: Oxford University Press.

Orians and heerwagen (1992). Evolved responses to landscapes. In Barkow, Cosmides, and Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. New York: Oxford University Press.

Pinker and Bloom(1992). Natural Language and natural Selection. In Barkow, Cosmides, and Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. New York: Oxford University Press.

Shepard(1992). The perceptual organization of colours: an adaptation to regularities of the terrestrial world?. In Barkow, Cosmides, and Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. New York: Oxford University Press.

Spelke(1988). The original of physical knowledge.in Weiskrantz, (ed), Thought without language. Oxford: Clarendon Press, pp.185-210.

Spelke (1990). principle of object perception. Cognitive science 14, 29-56.

Spelke (1991). Physical knowledge in infancy: Reflections on piaget's theory. In Carey and Gelman(1991). The epigenesis of mind. Hillsdale.

Spelke et al (1992).Origin of knowledge. Psychological reviews 99, 605-632.

Joseph Shepher (1983). Incest: Biosocial View. Academic Press.

Tooby and Cosmides (1989) Evolutionary psychology and the generation of culture, part I. Theoretical considerations Ethology and sociobiology 10 29-49.

Tooby, J. and Cosmides, L. (1992). The psychological foundations of culture. In Barkow, Cosmides, and Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. New York: Oxford University Press.

------------------------------------------------------------------

Yehouda Harpaz
yh@maldoo.com
2May97
http://human-brain.org