Monday, February 28, 2005

On Anonymity

As you've probably noticed, I blog anonymously. Well, that's not completely true. My name really is Chris, but my last name remains one of those great mysteries, like how they get the filling into twinkies, or how many licks it really takes to get to the Tootsie Roll center of a Tootsie Pop. Not all bloggers think anonymous blogging is a good thing. Some even think it's cowardly. Well, here's what I say to those of you who think anonymous blogging is wrong: Go [Twinkie] yourself with a [Tootsie Pop].

I suspect that there are all sorts of reasons for blogging anonymously. For one, the internet is full of crazies (ask PZ Myers, who has received angry phone calls from Powerline fanatics), and I have a child. The last thing I want is one of those insane pro-Evolutionary Psychology freaks to know my name and where I live! But the main reason that I chose to be anonymous has to do with my career. I personally find my career more important than a stupid blog, but I also see no reason to not blog simply because I don't want to step on the toes of individuals who might be responsible for determining whether a particular institution hires or fires me. So, I blog, but do so anonymously. Perhaps someday I'll put my last name at the end of each post, but I can't promise that I will. I certainly don't see any reason why I should have to.

One more thing. In the above-linked post, the always slow-witted Keith Burgess-Jackson writes:
Imagine a world in which there were no anonymous utterances. It would force people to be civil, fair, and charitable; to be responsive to the facts; and to be logically consistent—for the absence of any of these things would constitute a black mark on one’s record.
Clearly, he's never actually observed a political debate between onymous pundits. Hell, he's apparently unaware of the bulk of the political discourse in this country, period. Only a fool would fail to notice that the absense of anonymity does not force people to be civil, fair, and charitable, responsive to the facts, or logically consistent. But since KBJ blogs onymously, and exhibits none of these features, it's not surprising that he's failed to notice this.

UPDATE: The Anal"Philosopher" has more to say on the topic, here. Once again, he demonstrates the civility, fairness, and charity that makes him so representitive of the onymous bloggers of whom he speaks.

Sunday, February 27, 2005

Philosophers' Carnival X:

The tenth Philosophers' Carnival is now up at E.G. I haven't had time to read all of the posts, but the ones I have read, by Richard on personal identity, Andrew on the Chinese Room (my least favorite non-thought experiment thought experiment ever, by the way), Clark on "critical common sense," Brandon on Darwin's logic, Hugo on theory-ladenness, and Jonah on thought experiments, are all worth reading. I wish I had more time to say something about them, particularly the one on thought experiments, because I've been thinking a lot about thought experiments in conjunction with my own work on counterfactuals (I see thought experiments and scientific experiments as being instances of counterfactuals, and I'm working on a cognitive account of counterfactuals).

The Evidence for Recovered Memories: A Short Critique

In my previous post on recovered memories, I focused mainly on the evidence we have from real-world (naturalistic) and laboratory studies of traumatic memory in general. I focused on this research because the false memory research is already well known, at least generally, even by non-experts, and because one of the best reasons for disbelieving the claims of the repression hypothesis (the hypothesis that traumatic memories can be repressed, and later recovered) is that it does not accord with anything we know about the nature of memories of trauma, or memory in general. With the false memory and traumatic memory research, the case against the repression hypothesis looks something like this:
  1. Because traumatic memories are actually more memorable, more easily recalled, and more difficult to forget, the burden is on the proponents of the repression hypothesis to demonstrate how some traumatic memories, in some individuals, are different, and what as of yet undiscovered mechanisms are involved in causing these memories, in these individuals, to behave differently.
  2. There is ample reason to believe that many memories can be explained as instances of false memories.
Given 1 and 2, it is reasonable to be highly skeptical about claims of recovered memories. However, there are types of evidence that can overcome 1 and 2 and provide support for the repression hypothesis. Such evidence would have to involve cases in which it is demonstrable that there exist individuals who a.) were abuse during childhood, b.) had no recollection of the abuse for an extended period of time after the abuse had occurred, and c.) now have accurate memories of the abuse that were not created by (i.e., implanted) external sources, such as therapists, family members, or friends who knew the details of the abuse. Research purporting to demonstrate the existence of individuals displaying a-c does in fact exist. The most prominent studies reporting such evidence are presented in two papers, one by Linda M. Williams, and the other by Alan W. Scheflin and Daniel Brown. Each of the studies presented in these two papers contains some serious flaws (in fact, the Scheflin and Brown paper, bolstering my claim that the proponents of recovered memories are the cognitive scientific analog to ID theorists, is downright deceptive). Detailing these will help to demonstrate further evidence against the repression hypothesis, and thus further strengthen the case against it.

Williams, L.M. (1995). Recovered memories of abuse in women with documented child sexual victimization histories. Journal of Traumatic Stress, 8(4), 649-673.

In this paper, Williams presents the results of a study in which 129 women with documented cases of abuse occurring in the early 1970s were interviewed almost 20 years later (between 1990-1992) about the particular case of abuse that was documented. They found that 67% of these women remembered the specific instance of abuse that was documented. This is not evidence that the other 33% have repressed the memory of the abuse. Most of the women were abused over an extended period of time, and not recalling a particular instance when asked open-ended questions about abuse is thus not surprising. However, of the 67% who recalled the abuse (80 total, 5 of whose data had to be discarded due to errors on the part of the researchers), 16% (12) of the women answered yes to the following question:
Was there ever remember a time when you did not remember that this happened to you?
Williams believes that this data provides evidence for the existence of repressed and recovered memories of trauma in some individuals. However, there are two major flaws in her data that cause it to fail to address criterion b. The first is reported by Williams herself. The average age at which the documented instance of abuse occurred, for the 12 women who reported a period of forgetting, was 6.5 years, while the average age for the women who did not report such a period was 9.5 years. This is important, because it hints that her data may be explained by the existence of childhood amnesia, a well-documented phenomena in which, up until about age 6, people have a great deal of difficulty in forming and retaining autobiographical memories, even traumatic ones. Some of the 12 "forgetters" were 3 years old during the time of the documented abuse! It is unreasonable to expect them to remember the events, even today, without having been given information from external sources. In other words, any memory of abuse at age 3 is likely to have been implanted, even if accurate, rather than to have been caused by the event itself. Memories of events before age 6 that were caused by the remembered events are likely to be very sketchy, a fact that explains another aspect of Williams' data (her finding that the "forgetters" were less confident in their memories, and felt that they did not remember everything about the events).

The second problem, which may be even more important, is the question that Williams asked. It is widely known that memory for memory (i.e., memory for whether we remembered something at a particular time in our past) is extremely poor1. In fact, researchers have shown that people will report that there have been periods during which they have forgotten events, even though there is evidence of them talking about those events during the "forgetting period!"2. The question is further problematic, as some have noted3, because it implies that the individual remembers not remembering. This would require that the person remember abuse that he or she does not remember! Thus, the sort of self-report data elicited by Williams' question about periods of forgetting is essentially worthless.

What we've learned from the Williams study, then, is that self-report data about periods of forgetting is insufficient to provide evidence that satisfies criterion b above, and that another potential explanation for instances of recovered memories is childhood amnesia. Williams' data, if it shows us anything, actually supports the case against the repression hypothesis, by demonstrating that those who report forgetting tended to be abused at an age for which we would not expect them to have good, if any, autobiographical memories that were not obtained indirectly through outside sources.

Scheflin, A.W., & Brown, D. (1996). Repressed memory or dissociative amnesia: what the science says. Journal of Psychiatry and Law, 24, 143-188.4

This is a deceptively titled review of the pro-recovered memory literature. It is the most thorough to date, and any anti-recovered memory position will have to address it. Four years after its publication, an article was published in the same journal by Piper, Pope, and Borowiecki, titled "Custer's last stand: Brown, Scheflin, and Whitfield's latest attempt to salvage 'dissociative amnesia.'" Here I'll present most of their criticisms, along with a few of my own.
  1. In arguing that the burden of disproof is on those who do not believe that repressed memories exist, Scheflin and Brown write:
    Opponents of repressed memory... cannot meet the Daubert standards of science. Other than a few pro-false memory reviews, they can cite no data-based scientific studies in support of their view.
    Piper et al. note that Brown and Scheflin are not following one of the basic tenets of the scientific method, in which we do not posit explanations that are not needed to explain the data. In other words, the burden of proof is on the pro-repressed memory crowd to provide data that requires the repressed memory hypothesis for explanation. Otherwise, the repressed memory hypothesis is merely a hypothesis without an empirical grounding.

    Furthermore, as I have already noted, there are data-based reasons for not believing that repressed and recovered memories exists. While these reasons do not amount to a thorough disproof of the existence of repressed memories, from a scientific standpoint, they serve to place the burden of proof squarely on the shoulders of the proponents of the repression hypothesis.
  2. Scheflin and Brown focus exclusively on a few studies (25), claiming that they are the only studies on the subject. They write in the abstract, "A total of 25 studies on amnesia for CSA [childhood sexual abuse] now exist, all of which demonstrate amnesia in a subspopulation; no study failed to find it." I can honestly say that this is one of the most deceptive statements I have ever seen in a peer-reviewed publication. The fact is, as Piper et al. note, there are many more studies than this, and the majority of them do not find any evidence for amnesia for CSA. Piper et al. cite their own meta-analysis of 63 studies, containing analyses of data from more than 10,000 individuals who had suffered some sort of trauma, in which they find not a single instance of amnesia.
  3. Scheflin and Brown blatantly misrepresent many of the studies they present. On pages 160-171 of the paper linked above, Piper et al. provide a lengthy list of mistatements," false statements," and quote mining, and "distorted quotes" in Brown, Scheflin, and Whitfield, many of which are also present in Scheflin and Brown.
  4. The bulk of the studies cited by Scheflin and Brown, as well as Brown et al., are retrospective studies like Williams'. In these studies, individuals are asked about episodes of forgetting in the past. For the reasons discussed above in regard to Williams' study, these studies are essentially worthless.
  5. The prospective studies that both the Scheflin and Brown papers cite fail to include important data, such as the age of the individuals at the time of abuse, making them impossible to evaluate. Leaving out the age, for instance, makes it impossible to rule out the childhood amnesia explanation.
  6. One of the arguments against the childhood amnesia explanation, offered by Brown et al., is that sexual abuse prior to age 6 is so traumatic that it is likely to be remembered. In other words, it is different from other traumatic events that children are likely to forget at that young age. However, this claim is contrary to the evidence. Piper et al. note that the research on sexual abuse shows that, by and large, most children, particularly young children, are not disturbed or adversely affected by most instances of sexual contact. In other words, these events are likely not that different from other autobiographical events which children regularly forget. Piper et al. argue that, since most children do not immediately experience sexual contact as negative, the repression hypothesis would require that those children have a highly sophisticated knowledge of the social conventions that make such contact wrong, and thus repress the memories based on that knowledge, rather than the qualitative character of the experience itself.
In short, the Scheflin and Brown article, along with their follow-up in Brown et al., not only fail to address the criticisms of repressed memory research, and the repression hypothesis itself, but in the process of avoiding these criticisms, they misrepresent the evidence and deceive the readers with misquotes, quote mining, and the presentation of flawed (and in one case, perhaps non-existent) studies.

To wrap up then, there is to date no evidence, nor the hope for any eivdence, that means criteria a-c, and thus counters the evidence against the repression hypothesis. Furthermore, since studies like the one by Williams actually give us more reasons to doubt the existence of repression and recovery (e.g., the potential for childhood amnesia to explain many cases of "repression"), it is difficult, as a scientist, to believe in them. Now the evidence against the repression hypothesis looks like this:
1. The nature of traumatic memory & 2. False memory research, from above.
3. The fact that, as Williams' data suggest, childhood amnesia may explain many actual instances of forgetting. Add to this the fact that for young children, it is unlikely that sexual abuse, the most commonly cited elicitor of repression, is actually traumatic enough to be remembered by young children when most autobiographical information is not.
4. The huge methodological flaws in all retrospective and prospective studies purporting to demonstrate the existence of repressed and recovered memories.
Based on this evidence, no court should ever admit recovered memory testimony as evidence, and no jury should ever give it any weight. Yet, as the case linked in the previous post demonstrates, the courts and juries aren't always interested in "what the science says."

1Arnold, M.M., & Lindsay, D.S. (2002). Rembering remembering. Journal of Experimental Psychology: Learning, Memory, & Cognition, 28(3), 521-529.
2 Schooler, J. W., Ambadar, Z., & Bendiksen, M. (1997). A cognitive corroborative case study approach for investigating discovered memories of sexual abuse. In J. D. Read & D. S. Lindsay (eds.), Recollections of trauma: Scientific evidence and clinical practice, (pp. 379-387). New York: Plenum Press.
3 Mcnally, R.J. (1999). Review of Memory, Trauma Treatment, and the Law. International Journal of Clinical and Experimental Hynosis, 47, 374-382.
4 See also Brown, D., Whitfield, C.L., & Scheflin, A.W. (1999). Recovered memories : the current weight of the evidence in science and in the courts. The Journal of Psychiatry and Law, 27, 5-156, which was the actual target of the Piper et al. paper.

Diversity in Academia: A Proposed Study

It is very difficult to prove discrimination, especially when one's goal in proving it is to develop and/or justify group-specific legislative or other institutional remedies for discrimination. Trust me on this one, because I spend a lot of time (too much) trying to do so, for extra dough. Since The City of Richmond v. J. A. Croson Co., any group-specific (in most cases, non race-neutral) program designed to counteract discrimination must meet two major requirements. First, the program must be designed to serve a "compelling government interest," which means the government must have a damn good reason for any group-specific action that attempts to alleviate the effects of discrimination which is over and above, "an effort to alleviate the effects of societal discrimination is not a compelling interest," because the court held that this alone is not a "compelling government interest." This requirement thus demands two things:
First, it must identify the past or present discrimination 'with some specificity.' Second, it must also demonstrate that a 'strong basis in evidence' supports its conclusion that remedial action is necessary.
After compelling government interest is proved, the program must meet another requirement. It has to be "narrowly tailored" to meet the government interest specified in meeting the first requirement. In other words, it has to focus on the areas in which the government has a compelling interest, and on the specific groups that are subject to discrimination in those areas, along with the ways in which they are discriminated against. These two requirements, which are the requirements that all programs must meet if they are to stand up to scrutiny by the courts, make proving discrimination exists, and that we should act on it, very difficult.

Given all this, it surprises me that there has been so much loose talk of "discrimination" against conservatives in academia, based, almost exclusively, on studies showing a disparity exists in the number of "liberals" (operationally defined, almost laughably, as those who vote for or donate to Democratic candidates) and "conservatives" (those who vote for or donate to Republicans). I am perfectly willing to accept that universities are, by and large, populated by liberal professors. My own experience in universities, spanning now more than a decade, is consistent with this. Yet, does this disparity amount to evidence of discrimination, as so many conservatives allege? And furthermore, does it amount to discrimination that is accompanied by a compelling (governmental, or university) interest to use affirmative action-like programs, like the ones proposed and lobbied by David Horowitz and his epigones? As far as I can tell, no one has actually attempted to produce the sort of evidence for discrimination against conservatives in academia, along with compelling interest, to justify the claims of discrimination of Horowitz call for affirmative action programs to aid conservatives and "promote diversity" in universities*. So, I want to propose a study to do so.

Let's consider compelling interest. What, exactly, is the interest of the government and universities in alleviating discrimination of conservative intellectuals in terms of hiring, firing, and promotion, on university faculties? The answer that is always given is to "promote diversity." Yet, as Aaron Swartz (link via Preposterous Universe), an undergraduate at Stanford, so aptly (though sarcastically) notes in response to claims of discrimination at his university, "diversity," at least in the realm of ideas, is not an inherent goal of universities. Universities are in the business of educating and scholarship, both of which require that ideas be held to some standard, of truth for instance. Swartz writes:
I have found that only 1% of Stanford professors believe in telepathy (defined as "communication between minds without using the traditional five senses"), compared with 36% of the general population. And less than half a percent believe "people on this earth are sometimes possessed by the devil", compared with 49% of those outside the ivory tower. And while 25% of Americans believe in astrology ("the position of the stars and planets can affect people's lives"), I could only find one Stanford professor who would agree. (All numbers are from mainstream polls, as reported by Sokal.)
It would be reasonable to say that universities like Stanford discriminate, albeit indirectly, against believers in astrology and telepathy, as promoters of these ideas will have a very hard time meeting the standards of scholarship that such universities demand. Few of us would argue that universities have a compelling interest to remedy the effects of such discrimination. While it may not be fair to compare political conservativism to astronomy and telepathy, Swartz is making a point that supporters of programs designed to increase the representation of conservatives in academia must address, namely, if there is in fact discrimination against conservative intellectuals in American universities, are there explanations for this discrimination that effectively eliminate the universities' compelling interest in remedying such discrimination?

Yet, even to answer that question, we must first ask, is the disparity between liberals and conservatives in academia evidence of discrimination? The answer to this question isn't obvious, either. As affirmative action cases have shown, when attempting to justify remedial action, it is not sufficient to merely show a disparity. We must also show that this disparity cannot be attribute to factors other than race, or in this case, political ideology. Since, to date, there is no study showing anything more than a disparity, I propose that we actually conduct a new study. In this study, we will do more than simply collect voting and political donation records, or hiring, firing, and promotion decisions. This information alone can only provide evidence of disparity, not discrimination. To do this, we need to rule out alternative explanations for the disparity. Thus, we will need to collect information about job applicants and faculty that is relevant to hiring and promotion. Thus, we should collect information related to publication, citation, teaching evaluations, ongoing research, etc., that hiring and promotion committees consider when making their decisions. Using this information, along with the relative number of liberal and conservative applicants, we can apply fairly simple statistical tools (e.g., regression analysis) to determine whether the ideological disparity that exists in American universities is a result of discrimination.

So here is my challenge to anyone who supports programs like the Academic Bill of Rights that are designed to get more conservatives onto university faculties through legislative action: get me this data, and let me analyze it. We can discuss exactly what data we will need. Publication data, for instance, may be problematic, because I can imagine that some conservative intellectuals believe that they are discriminated against by journal editors and reviewers as well. We should probably get the data from a large public university system, with enough departments to give us a big enough sample size. We can also discuss other forms of evidence (e.g., anecdotal evidence), though I'm not sure exactly how you would collect that.

Until we have this data, and can answer both questions related to the "compelling interest" requirement, we will have no legally-sufficient evidence for discrimination or justification for remedial programs. After we have such evidence, if it exists, we can then go about producing narrowly tailored programs designed to remedy the discrimination that we ourselves have demonstrated empirically. These programs will, of course, have to be different for every university, as the "narrowly tailored" requirement demands that they be specific to the area in which the programs are enacted. In fact, they may need to be tailored to specific departments. It is reasonable to believe, until it has been demonstrated otherwise, that discrimination based on political ideology is less prevalent in most physics departments than in most cultural studies departments. If this is the case, then "narrow tailoring" requires that we develop programs aimed only at those departments in which discrimination exists.

One more thing. If our study shows that factors other than discrimination explain the disparities, then I reserve the right to use it to counter any attempts to enact legislative or university programs designed to get more conservatives onto university faculty.

* Note also that Horowitz' "Academic Bill of Rights" could hardly be considered narrowly tailored, but he does try to get around that by giving it the appearance of being ideologically-neutral).

UPDATE: Apparently unfamiliar with the last 16 years of judicial decisions on discrimination and affirmative action, the legal scholars at The Volokh Conspiracy claim that disparity alone constitutes prima facie evidence of discrimination. Shame on them, for making claims that anyone who's read any court case on the topic would know are false. Maybe they'll be willing to get me the data.

Freud on Repression

Repressed memory proponents often reference Freud as the origin of the repression hypothesis. Since most experimental psychologists don't really like Freud, and since they've probably never read him, they're probably more than willing to believe that it was he who came up with such a bad idea. However, I'm an experimental psychologist who has read Freud (because studying the history of psychology is a bit of a hobby of mine), and I can say with certainty that the current conception of repression held by some psychologists is not Freud's. However, you don't have to take my word for it. Read what the man wrote, in "The Aetiology of Hysteria."

More on repression soon.

Friday, February 25, 2005

Traumatic Memories: The Case Against Memory Recovery

Every time I read or hear about the outcome of a criminal case resting on testimony from a "victim" who has recently recovered memories of the crime, of which he or she had no recollection for many years between the incident(s) and the memory recovery, I cringe. I mean I really cringe. To cognitive psychologists, such stories are worse than fingernails on a chalkboard. Recovered memory theories are our Intelligent Design theories, only instead of having only a little success in affecting textbook and curriculum decisions, our equivalent of ID proponents are actually getting people convicted of serious crimes, and sent to prison for the rest of their lives. As is the case with ID in biology, much of the general public is ignorant of the relevant research, and of the nature of the issues involved, but since members of the public serve as jurors in criminal cases, their ignorance can have dire consequences. A juror in a recent child molestation case in which the prosecution's primary witness was a man who had recovered memories of the abuse in his 20s, told reporters
We agreed after discussion that you can experience something up to a point, and then not think about it and have plenty of other things in your life that are more important.
She said this after she and the other jurors had convicted a man who was subsequently sentenced to 12 to 15 years in prison. Yet, her defense of her and her fellow jourors' acceptance of the recovered memory testimony displays a complete ignorance of memory in general, and even what recovered memory proponents believe recovered memories are. They are not memories that have simply not been reflected on for a period of time, or that have been overshadowed by more pressing concerns and other memories (perhaps even from the same time period). Recovered memories are memories that had been mysteriously repressed, and thus completely unavailable for retrieval, until some triggering event (usually psychotherapy) made them available again. The jury members couldn't distinguish between going a year or two without thinking about someone they dated a couple times in college and not being able to remember incredibly traumatic childhood sexual abuse for decades. Yet, they convicted a man anyway. I am disturbed by the fact that religious fanatics who are ignorant of biology can influence education policy, but I am terrified by the idea of people who are so blatantly ignorant of the science of memory could convict me, or anyone else, of a crime that could result in the loss of freedom for years. Almost as disturbed as I am by the fact that recovered memories, for which there is almost no non-anecdotal empirical evidence, while there is a wealth of evidence calling their existence and accuracy into a question, are even admissable in criminal cases.

Don't misinterpret that angry paragraph as implying that I do not believe that it is possible to repress and later recover memories for traumatic events. As is the case with an intelligent designer, I am agnostic about the possibility of the existence of repressed and recovered memories. Since there is no empirical evidence for the existence of such memories that does not come from either case studies or self-report surveys (and asking people if they did not remember their trauma at some point in their life invites misunderstandings like those displayed by the juror in the quote above, making this data even more worthless), and to this point no one has any viable suggestion for the mechanisms involved in repressing and recovering memories. Add to this the fact that the existence of repressed memories goes against everything we do know, from experimental investigation, about memory and its mechanisms, and it seems irresponsible to claim that they do exist, to say nothing of convicting someone of a crime based on them, but cognitive science is a young discipline, and there is much that we do not know, so it's not very responsible, scientifically, to claim that they definitely do not exist, either.

So, now you know where I stand. Since I'm an intellectual (or at least, someone who holds himself to intellectual standards), I should probably say why that's where I stand. I don't want to run through the evidence for false memories. It's easy to find and read about, and most people who have read this far have probably heard a lot about that already. False memories, and the manipulability of memory in general, are reason enough to question any accusations and beliefs based on recovered memories, without any independence evidence to back them up. Yet, our propensity for the production and belief of false memories doesn't say much about the existence of accurate recovered memories. To be able to speak to that, we have to look at what we have learned from actual studies on memory for traumatic events. Since that research is less-often discussed in public debates on recovered memories, I will talk about it instead.

The first, and perhaps most important thing to note about research on traumatic memory is that it suffers from serious methodological hurdles. It is impossible, for ethical reasons, to draw research participants randomly from the population, and then randomly select some of them to suffer from extremely traumatic events, and the other to experience emotionally stimluating or mundane events of other sorts. Short of that, experimental research into the nature of traumatic memory will be incomplete. Still, there are methods with which we can gain some knowledge of memory for traumatic events, and then compare what we've learned to the claims of recovered memory proponents. Here are some of the things that we know:
  • There are some extreme psychological syndromes, most notably psychogenic amnesia, Dissociative Identity Disorder, and fugue, which may result from traumatic events. Psychogenic amnesia is notoriously difficult to diagnose, because ruling out all physical causes is next to impossible. Furthermore, psychogenic amnesia for traumatic events rarely lasts for more than a few hours, in documented cases, and begins during the traumatic event itself. Repressed memories, on the other hand, are said to be repressed for years, and the repression itself is generally said to begin at some point after the traumatic event. Disocciative Identity Disorder, if it in fact exists (a question that is also hotly debated), is extremely rare (with a prevalency of about 1%, over the last decade or so), and because of the other symptoms involved, can not be used to explain the vast majority of recovered memory cases. Fugue is even rarer (.02%), and again, involves symptoms that distinguish it from most recovered memory cases.
  • Laboratory studies overwhelmingly show that higher arousal levels increase memory accuracy and the ease of retrieval, while at the same time decreasing suggestibility, at least for the arousal-inducing events themselves2. In fact, unlike non-traumatic memories, memory for traumatic events actually improves over time. This is particularly striking in cases of Post Traumatic Stress Disorder, in which the repeated retrieval of traumatic memories makes them more difficult to forget, and in many cases, more detailed. This is likely due to the fact that traumatic memories tend to be repeatedly retrieved (and, particularly in the case of PTSD, involuntarily retrieved). Repeated recall has been shown to improve recall ability and accuracy, particularly for traumatic memories3. There are plausible neurological explanations for this. For example, it is likely that the brain regions activated and hormones released during emotionally-stimulating events are the same ones that are involved in memory storage4. It's not surprising, then, that increased arousal produces better memory.
  • In PTSD, memories are almost impossible to forget, or to repress. However, the difficulty of forgetting, and increased accuracy and memorability for traumatic memory is not exclusive to people who suffer from PTSD. Two famous studies illustrate this well. In the first5,26 children were interviwed immediately after being abducted in their school bus, and buried for more than a day, and interviewed again four years later. During both interviews, their recall of the event was very good, and they showed no real decrease in recall over the four year period. In the second study6, more than 70 Holocaust survivors who had written about their experience soon after the end of World War II were asked to recall their experiences. Despite the study being conducted in the late 1980s, more than 40 years after the traumatic experiences, the survivors' memories for the events were exteremely accurate when measured against their written accounts from after the event. This second study speaks directly to one of the more concrete claims of repressed memory proponents, made by Lenore Terr7, the author of the school-bus kidnapping studies. She has argued that there are two types of traumatic experience, which she calls Type I traumas and Type II traumas. Type I traumas, like that experienced by the school-bus kidnapping victims, occur only once. Type II traumas, such as repeated sexual or physical abuse during childhood, occur repeatedly over extended periods of time. Terr argues that the latter can produce repressed memories, while the former produce enhanced memories. However, the Holocaust survivors suffered years of repeated traumatic events, and their memory accuracy was excellent over time. In another study, young children who had been subjected to painful and embarassing medical procedures also experienced no decrease in memory recall8, a result that is also at odds with Terr's theory. Since the only evidence for Terr's Type II trauma and its effects on memory comes from a study of children who experienced trauma prior to age 5, when children's memories are notoriously inaccurate and short-lived, and there is a great deal of evidence that traumatic memories are better remembered, over time, than ordinary memories, Terr's theory, the only serious attempt to explain the experimental data, is difficult to support.
In sum, the evidence from laboratory and real-world studies of traumatic memory doesn't provide us with much of a basis for believing in the existence of repressed memories. Combine this with the well-demonstrated suggestibility of human memory and the frequency of false memories, and, as scientists, we have to come to the conclusion that the repression and recovery of traumatic memories are both extremely unlikely. There is certainly not enough evidence to admit recovered memory testimony in court cases. Lie detectors, which aren't admissible as evidence of guilt in 49 states, have accuracy rates of around 70%. There is no reason to think that repressed memories have an accuracy rate anywhere near that, and plenty of reason to think their accuracy rate is near 0%. Yet, judged continue to allow them, and juries continue to convict solely on the basis of them.

UPDATE: After careful consideration (and a strong suggestion from a friend), I'm pretty sure that I stole the recovered memory-Intelligent Design analogy from PZ Myers, who compared the proponents of recovered memory to creationists in this post. Given the importance of source-monitoring in the debate about false memories, I find my inability to remember the source of the analogy (I thought I was clever enough to come up with it on my own, but it turns out I'm not) kind of funny.

1 Ross, C.A. (1991). Epidemiology of multiple personality disorder and dissociation. Psychiatric Clinics of North America, 14(3), 503-517.
2 Christianson, S.A. (Ed.) (1992). The Handbook of Emotion and Nemory: Research and Theory. Hillsdale, NJ: Erlbaum.
3 See for example Bornstein, B.H., Liebel, L.M., & Scarberry, N.C. (1999). Repeated testing in eyewitness memory: a means to improve recall of a negative emotional event. Applied Cognitive Psychology, 12(2), 119-131.
4 LeDoux, J. (1996). The Emotinoal Brain. New York: Simon & Schuster; McGaugh, J. L. (1992). Affect, neuromodulatory systems, and memory storage. In S.A. Christianson (Ed.), The Handbook of Emotion and Memory: Research and Theory, pp. 245-268. Hillsdale, NJ: Erlbaum.
5 Terr, L. (1981). Psychic trauma in children. Observations following the Chowchilla school-bus kidnapping. American Journal of Psychiatry, 138, 14-19; Terr, L. (1983). Chowchilla revisited: The effects of psychic trauma four years after a school-bus kidnapping. American Journal of Psychiatry, 140, 1543-1550.
6 Wagenaar, W.A., & Groeneweg, J. (1990). The memory of concentration camp survivors. Applied Cognitive Psychology, 4, 77-87.
7 Terr, L. (1994). Unchained Memories. New York: Basic Books.
8 Goodman, G. S., Quas, J.A., Batterman-Faunce, J.M., Riddlesberger, M.M., & Kuhn, J. (1994). Predictors of accurate and inaccurate memories of traumatic events experienced in childhood. Consciousness and Cognition, 3, 269-294.

Sunday, February 20, 2005

What Intelligent Design Could Be... But Isn't

The cognitive revolution arose out of the Turing (and then Chomsky, and others)-inspired notion that the mind is a computing machine. Since computers generally don't have sensory modalities, as we traditionally conceive them at least, and they don't really need them, their representations tended to be digital, and thus amodal. Since cognitive scientists were treating minds as computers, it was natural to make their representations digital, and amodal, as well. And since the 1950s, that has been the orthodox view of cognitive scientists: representations are discrete information states that can be subjected to syntactical processes, just like the machine language of a computer. Sure, there have been some dissenters, like the cognitive linguists and dynamic systems theorists, but they have existed on the fringe of cognitive science for the most part, being either silly (cognitive linguists) or near incoherent (system theorists)1. A few years ago, though, Larry Barsalou and some of his colleagues (see e.g., this book) began to argue that it was time to leave our amodal, disembodied representations behind, and return to 17th century empiricism. To do this, we have to return to the belief that cognition is "inherently perceptual," i.e., that representations are inherently tied to sensory modalities. Thus, perceptual symbol systems theory was born.

Now, there are lots of problems with perceptual symbol systems theory (PSS), though I won't get into most of them (see here for a thorough critique). The most general criticism is that, ultimately, PPS doesn't make any predictions that theories utilizing amodal representations couldn't handle. This makes it difficult, empirically, to decide between the two alternatives. More specific criticisms include the fact that PPS theorists always end up positing a whole bunch of amodal (sometimes they call them multi-modal, just to avoid admitting that they're amodal) mediating representations to get their theories to work. They also have a hell of a time with some of empiricism's old archenemies, like abstract concepts (though not for lack of effort; see here) and arbitrary signs (like words). All-in-all, though, PPS has been good for cognitive science. It has forced us, especially those of us who study concepts, to take perception seriously again.

The way PSS has managed to get concept researchers to take perception seriously is by inspiring research that those of us who are firmly entrenched in the amodalist world-view would never have thought of, producing findings that would likely have otherwise gone undiscovered. For example, in one experiment2, participants were given the names of items like "watermelon" and "lawn," and asked to list their properties. Some of these participants received further instructions to picture the items in their heads before listing properties. Furthermore, some of the participants received the normal names, while others received modified names (e.g., "sliced watermelon" or "rolled-up lawn). The predictions Barsalou derived from PSS were twofold: people who pictured the items would produce different lists of features, depending on whether they received the normal or modified names, because the perceptual features would be different (e.g., lawn has green grass, while a rolled-up lawn has roots and grubs and all sorts of other creepy crawlies), a prediction that Barsalou believed amodal systems theories would also make, and also that the people who hadn't received the picture instructions would produce feature listings that were not significantly different from those of the people who did receive them. This second prediction, Barsalou argued, could not be made by amodal systems theories, because it required that these people spontaneously use perceptual simulations of the items to represent them. And it turned out that the two groups did produce the same listings, with watermelons having features like "green," "striped," and "round," and sliced watermelons having features like "red" and "seeds." In an even more fascinating study, Simmons and Barsalou3 asked participants to categorize objects presented in photos. Some participants were told to make hand motions that were compatible with some of the items (e.g., the motion used to turn on a faucet). These participants categorized the items faster than participants who did not perform compatible movements. Simmons and Barsalou argue that this is because the perceptuo-motor actions facilitated the perceptual simulation of the items, resulting in faster categorization.

The results of these PSS-inspired experiments, and others, whether the support PSS over amodal theories or not (again, while they were derived from PSS, they appear to be consistent with amodal theories), have sometimes wide-ranging implications for theories of concepts in general, such as the types of information we represent, how that information is retrieved, and the role of context and imagination. Thus, while the the prospects for PSS as a theory of mental representation, in its current form at least, may not be very good, its contributions to cognitive science in the form of novel hypotheses that lead to theoretically important empirical findings is undeniable.

At this point, you're probably wondering what all of this has to do with intelligent design theories (ID). After all, the title of this post is about ID, not cognitive science or PSS. Well, I think that PSS and ID have some very interesting commonalities, which can allow us to use PSS as an analogy when reasoning about ID. For instance, they're both 17th century ideas (which, in more rudimentary forms, go back even further) that are being revived in order to challenge more recently developed scientific orthodoxies; the adherents of both claim to be inspired by empirical data that they believe their more entrenched alternatives cannot handle; and they're both ultimately likely to fail as scientific theories due, among other things, to their ultimate inability to empirically distinguish themselves from their alternatives.

Once we align the two theories so that these attributes are in correspondence, we can then begin to look at their differences in informative ways. In particular, we can try to understand one key difference: PSS has led to novel hypotheses, which have produced interesting experimental results that have advanced our knowledge of concepts and mental representation, while ID has yet to produce any hypotheses that have led to experimental results of any kind. It might be objected that ID concepts like Behe's "irreducible complexity" have led to research demonstrating that certain biological mechanisms are, in fact, producible through natural selection, but since biologists were already aware that the evolution of these mechanisms was, as of yet, unexplained, it's hard to argue that ID had anything to do with this research, other than adding another sentence or two to the discussion sections in the papers describing it.

What other difference might explain this big one? The answer is obvious. PSS theorizes that concept representations are produced and processed as simulations within the same perceptual systems of the brain that represent and process non-conceptual information from the senses. Thus, we can use our knowledge of those perceptual systems, theorized properties of the simulation process, to make predictions about concepts. ID has nothing analogous to this feature of PSS that would allow it to make predictions. In place of such a feature, ID has an intelligent being of a "mysterious and incomprehensible nature," about which the natural phenomena that we are trying to explain can tell us nothing. Because we know nothing about this being's nature, we have nothing to work with when it comes to making novel predictions. Thus, we're left with a theory that can add nothing to science, because it gives us no new positive knowledge. At most, it can merely trumpet the null hypotheses of its alternatives.

So, PSS serves as a good example of the sort of theory that ID could be, but isn't, and unless it makes some positive statements about the nature of the designer that go beyond attributing the vague property of intelligence to it, never will be. If the adherents of ID were actual scientists, this is the problem they would be working on. Merely attempting to show that there are cases which evolutionary theory cannot yet handle isn't going to cut it. Since ID provides nothing in the way of positive alternative explanations, the best way to deal with the fact that evolution has not yet explained everything is to note that evolution is still the only potential explanation. If the adherents of ID were real scientists who were genuinely interested in the scientific process, they would be working very hard to come up with empirically testable theories concerning the nature of the designer. Short of that, they've got nothing.

1 I purposefuly leave out connectionism, because while connectionist representations can be ostensibly analog, and even considered modal (depending on the model), when you strip away the surface-level representations of the models themselves, what you have is a multi-dimensional space with points and vectors serving as representations, and that seems pretty damn digital, or at least amodal, to me.
2 Barsalou, L.W., Solomon, K.O., & Wu, L.L. (1999). Perceptual simulation in conceptual tasks. In M.K. Hiraga, C. Sinha, & S. Wilcox (Eds.), Cultural, typological, and psychological perspectives in cognitive linguistics: The proceedings of the 4th conference of the International Cognitive Linguistics Association, Vol. 3 (209-228). Amsterdam: John Benjamins.
3 See here, p. 11.
4 It might be objected that ID concepts like Behe's "irreducible complexity" have inspired research demonstrating that complex biological mechanisms can, in fact, evolve through natural selection, but biologists were already well aware that there were some mechanisms for which they did not have an evolutionary mechanism, and thus needed to study more thoroughly. Thus, it doesn't appear that "irreducible complexity" has actually inspired any research, and it certainly hasn't produced any novel hypotheses, but has instead built a case around the null hypothesis that scientists would have worked with anyway.

How Evolutionary Psychology Can Make You Look Like an Ass

There is something about evolutionary psychology (EP) that makes it very attractive to non-psychologists (and to undergraduate psych majors -- you should see them rushing to register for EP courses). I've never been entirely sure what it is about EP that makes non-experts find it so fascinating, and more often than not, swallow it's claims without hesitation. Perhaps it's the simplicity and intuitiveness of many of the explanations. Cheating is bad, and harmful, therefore it is adaptive for us to have evolved a mechanism for detecting it. That's pretty simple and intuitive, right? Of course, this is one of the many reasons that most psychologists don't seem to find EP very attractive. The explanations generally rely on little more than intuition bolstered by sketchy, usually non-experimentally derived data. A careful review of the EP literature would give a scientist little confidence in its claims. However, there are plenty of non-psychologists who are happy to read some trade books on EP, and treat it as gospel. Doing so leads them to come up with all sorts of nonsensical arguments about human behavior. This is especially true when EP "theories" are used to make political arguments, as was the case in a recent essay by Will Wilkinson for the Cato Institute.

Now Wilkinson's formal training is in philosophy, and while he does cite three books on EP and politics, making it reasonable (or at least less charitable) to assume that he's read something on the topic, it's quite clear that his knowledge of EP is minimal. We see this, for instance, in his citing of Oliver Goodenough and Kristin Prehn, whose name he misspells, as the sources of original research on the detection of moral transgressions, when, in fact, she has not published any such research, only a review of the literature for a special issue of a philosophy journal. He furthermore cites the Tooby and Cosmides' social exchange theory, for which there is preciously little (if any) evidence, and even cites their Wason selection task experiments, which have been shown to provide no evidence for their conclusions. He also cites Robert Kurzban's research on social groups as having "shown," which I assume means demonstrated with certainty, that certain aspects of in-group, out-group dynamics have an evolutionary basis. However, if he had even read Kurzban's research, he would know that there are plenty of other alternative theories that explain the data, and that Kurzban has produced preciously little new data in support of his own theory. Given this, we can be sure that Wilkinson is not interested in the science of EP, but only in accepting its claims uncritically, and drawing his own conclusions from them. And that is in fact what he does. I'll go through his major claims, one by one.

1.) The evolutionary basis of in-group, out-group dynamics leads to the following conclusion:
We cannot, however, consistently think of ourselves as members only of that one grand coalition: the Brotherhood of Mankind. Our disposition to think in terms of "us" versus "them" is irremediable and it has unavoidable political implications.
Wilkinson's conclusion is probably true, though it's certainly not new. Social psychologists have been making similar points for decades, without the benefit (burden) of evolutionary stories. In fact, the two evolutionary stories that Wilkinson cites here, the social exchange theory, which can be thrown out due to a lack of evidence, and significant amounts of recalcitrant data, and Kurzban's theories of stigmatization and social categorization, which also has preciously little empirical support, are utterly superfluous in Wilkinson's argument. They're little more than distractions from the real point: that in-group, out-group dynamics appear across cultures, and that understanding them is important if we are to successfully transcend perceived group boundaries.

3.) Wilkinson writes:

There is evidence that greater skill and initiative could lead to higher status and bigger shares of resources for an individual in the EEA. But because of the social nature of hunting and gathering, the fact that food spoiled quickly, and the utter absence of privacy, the benefits of individual success in hunting or foraging could not be easily internalized by the individual, and were expected to be shared. The EEA was for the most part a zero-sum world, where increases in total wealth through invention, investment, and extended economic exchange were totally unknown. More for you was less for me. Therefore, if anyone managed to acquire a great deal more than anyone else, that was pretty good evidence that theirs was a stash of ill-gotten gains, acquired by cheating, stealing, raw force, or, at best, sheer luck. Envy of the disproportionately wealthy may have helped to reinforce generally adaptive norms of sharing and to help those of lower status on the dominance hierarchy guard against further predation by those able to amass power.

Our zero-sum mentality makes it hard for us to understand how trade and investment can increase the amount of total wealth. We are thus ill-equipped to easily understand our own economic system.

These features of human nature—that we are coalitional, hierarchical, and envious zero-sum thinkers—would seem to make liberal capitalism extremely unlikely. And it is. However, the benefits of a liberal market order can be seen in a few further features of the human mind and social organization in the EEA.

This argument borders on the nonsensical. Even if we ignore the merely speculative assertions about the EEA (Environment of Evolutionary Adaptedness), Wilkinson's conclusions simply do not follow from his non-EP premises: humans are coalitional, hierarchical, and envious. In fact, it is the coalitional and hierarchical nature of human groups that makes economic and power hierarchies so natural, and readily accepted by most individuals. Thousands of years of human society demonstrate this, making Wilkinson's ultimate conclusion baffling. Even if we refer to the primate data, we will still come to the conclusion that economic hierarchies are natural, easily understood, and widely accepted. In nonhuman primates, higher-status individuals have access more and better resources (including food and mates), and, in direct contrast to Wilkinson's claim, the best way to move up in the hierarchy is to invest in groups of similarly low status individuals in order to gang up on higher-status individuals. While higher status individuals, in both human and nonhuman species, often do steal and use force to gain resources and maintain their status, it is not automatically assumed that they have done so by "cheating," but instead have done so through natural social processes. It's not surprising, then, that exchange and unequal distributions of wealth have dominated human societies even during the times when group sizes were relatively small, i.e. in the EEA.

In short, then, Wilkinson's premises lead to a conclusion contradictory to his own. In fact, as economists, political scientists, and social psychologists have noted for decades, various systems of economic and political hierarchies flow quite naturally out of our seemingly innate understanding of social hierarchies. This is because they mirror the structure of hierarchical social structures, be it through tradition, the consolidation of power, or the control of production and resources (Habermas has an entire book on this stuff, with no mention of evolution).

4.) Citing a paper he clearly has not read by Goodenough and Prehn, Wilkinson argues that neuroscientific research has demonstrated that we have innate property concepts, out of which property rights naturally flow. This claim is absurd given the data. Even if we did have consistent research demonstrating that certain brain areas are active during reasoning about property rights in adult humans, this would hardly be evidence that property concepts are innate. Furthermore, we have no such evidence! It doesn't appear that the detection of property right violations are any more automatic than other forms of moral reasoning, and the lack of cross-cultural data makes it impossible to know just how natural they are. What we do know about moral cognition and its neural correlates goes against everything Wilkinson says in the section on property rights. In particular, imaging studies of moral reasoning demonstrate that its neural correlates are spread throughout the brain, utilizing brain centers that likely developed for other tasks (e.g., decision making), and that there is likely no "moral reasoning module," much less a "property rights violation module," as Wilkinson implies there is. If Wilkinson had read any of the literature (including the paper by the authors he actually cites), he would know this. In particular, I recommend three papers by William D. Casebeer (one with Patricia Churchland), which are here, here, and here (the third is a fairly extensive literature review), along with this review paper by Greene and Haidt.

5.) Citing social exchange theory again, Wilkinson writes, "The human mind is 'built' to trade." Well, social exchange theory, in all its falsifiedness, certainly doesn't demonstrate this. It is true that reciprocity appears to be universal in humans, though this universality arises largely out of cooperative behavior, not trace specifically. Trade may, then, be a natural offshoot of reciprocity, but nothing here implies that trade is innate.

From there, Wilkinson goes on to argue, taking EP speculations about the EEA as certain truths, that modern capitalism is difficult for most humans to comprehend because we evolved to exist in smaller groups, with all of the features he claims in 1-4. This makes absolutely no sense. First, even during the EEA, when humans did exist in small groups for the most part, inter-group trade was common in some areas, and for at least 10,000 years, it has been the norm (consider early modern human remains found in China that are more consistent with European lineages, indicating trade between two widely different and geographically separated groups). In fact, a more plausible evolutionary story is that this is in fact one of the main reasons why we are so good at detecting cheaters. Trading within groups is fairly safe, because we have guarantees that group members will play by the rules (if they don't, they'll be ostracized or worse), while non-group members, with whom we had to trade on many occasions, could not be automatically trusted. As with his claim that the existence of social hierarchies makes capitalism difficult to comprehend, his arguments here have no real connection to his premises.

The lesson to be learned here, aside from reading about things you cite, is that unless one is willing to critically evaluate research, one shouldn't draw conclusions from it. Wilkinson, and most other non-psychologist EP fans seem to think that it is OK to take EP at its word, without deigning to evaluate the theories, research methods, or data. And what we get from them is the sort of nonsense that Wilkinson's essay represents so well.

(Link to Wilkinson's paper via Positive Liberty.)

Saturday, February 12, 2005

Categorical Beginnings

It's never easy to begin a paper, especially a science paper. You want to start out general enough that people will understand the general importance of the paper, and you want to capture the attention of your readers, but you're bound by the principles of scientific writing, which can be pretty restricting. I myself have written entire papers before even attempting the first paragraph. There is one area of cognitive psychology, though, in which starting papers seems to be even more difficult than usual. This is the only explanation I can think of for the continual rehashing of the same beginning for at least 4 decades. The area? The study of concepts and categorization.

In order to demonstrate what I mean, here are a few examples of actual opening sentences and paragraphs from papers on categorization:
The world consists of a virtually infinite number of discriminably different stimuli. One of the most basic functions of all organisms is the cutting up of the environment into classifications by which nonidentical stimuli can be treated as equivalent.
-From Rosch, E., Mervis, C.B., Gray, W., Johnson, D., & Boyes-Braem, P. (1976). Basic objects in natural categories. Cognitive Psychology, 8, 382-439, which, by the way, would make a good addition to the list.

One of the major components of cognitive behavior concerns abstracting rules and forming concepts. Our entire system of naming objects and events, talking about them, and interacting with them presupposes the ability to group experiences into appropriate classes. Young children learn to tell the difference between dogs and cats, between clocks and fans, and between stars and street lights.
-From Medin, D.L., & Schaffer, M.M. (1978). Context theory of classification learning. Psychological Review, 85, 207-238, which is another classic.

Categorization is one of the most basic cognitive functions. Why is the ability to categorize events or objects important to an organism? An obvious answer to this question is that categories are important because they often have functional significance for the organism. Another familiar answer is that grouping objects into categories allows for efficient storage of information about these groups of objects.
- From Corter, J.E., & Gluck, M.A. (1992). Explaining basic categories: Feature predictability and information. Psychological Bulletin, 111, 291-303.

Is the plant edible or poisonous? Is the person friend or foe? Was the sound made by a predator or by the wind? All organisms assign objects and events in the environment to separate classes or categories. This allows them to respond differently, for example, to nutrients and poisons, and to predators and pray. Any species that lacked this ability would quickly become extinct. [Here they actually includes a citation! To a previous paper by the first author.]
- From Ashby, F. G., & Ell, S. W. (2001). The neurobiology of category learning. Trends in Cognitive Sciences, 5, 204-210.

Beachcombers categorize flotsam as man or fish. Players of 20 questions categorize things as animal, vegetable or mineral. Guards categorize approachers as friend or foe. Bystanders categorize flying objects: ‘Look, up in the sky! It’s a bird; it’s a plane! No, it’s Superman!’ Categorization permeates cognition in myriad protean variations.
- From Kruschke, J.K. (2005). Category Learning. In: K. Lamberts and R. L. Goldstone (eds.), The Handbook of Cognition, Ch. 7, pp. 183-201. London: Sage. This one actually begins with a quote from The Tempest, an indication that the author may have struggled even more than is usual, to come up with an opening.
So, you get the point. Categorization is one of the basic functions, major components, and adaptive necessities of cognition, and it "permeates cognition in myriad protean variations." I'm particularly impressed with the Ashby and Ell opening, which includes a citation for its view that categorization is important. It's not even a citation that begins with "e.g." I mean, given the frequency with which that same point is made in the opening of papers on categorization, you'd think a citation for that view would include a tediously long list of papers, or at least suggest that the paper cited is just one example of many that begin that way.

I wish these five examples were unusual, but they are not. Every year, at least two or three papers on categorization begin with a similar cookie-cutter opening paragraph. At this point, do we really need to be convinced that categorization is important? I'm pretty sure Aristotle taught us this more than two thousand years ago. The only explanation I can come up with for these is that it's just really hard to start a paper on categorization. With a literature spanning two millennia, it must be tough to come up with something original.

100 Most Influential Works in Cognitive Science

With all the recent lists of the best, most important, or must-read books in various sub-areas of analytic philosophy (interesting that none of the philosophy of mind lists included books like Psychology from an Empirical Standpoint), I thought some people might be interested in the recently compiled list of the "100 Most Influential Works in Cognitive Science."

Like the other lists, all of the books and papers have been influential, though the ranking seems a bit silly, and as you might expect, not everyone agrees with many of the inclusions and omissions (including me). It seems silly to put Turing's "Computing machinery and intelligence" third behind a paper by Chomsky and David Marr (both of which are excellent, though judging from the influence in my own corner of the discipline, I would have put Marr's Vision ahead of Chomsky), when both of the works ahead of it depend on it, in many ways, for their very existend. Hebb's 1949 work is also a justifiable choice for #1, since it has had an unmeasurable influence on neuroscience, and essentially created the idea of neural networks.

There are some works that are glaringly misplaced. While Fodor's The Modularity of Mind definitely merits inclusion, placing it at #7 tells me that the panel of judges had a slight philosophy bias. I mean, it's ahead of Shannon's "A mathematical theory of communication" (#12), Neisser's Cognitive Psychology (which pretty much launched the discipline, and is at #20), and Piaget's The Child's Conception of the World (which has defined the study of cognitive development for 75 years, and is ranked #46)! That's sort of like putting The God Particle ahead of "The principle of relativity."

The one work I would have left off the list is Edelman's Neural Darwinism (#87), but I'm not a neuroscientist, so maybe that convoluted work is more influential than I think (I've never read, or until I saw this, heard of Koehler's The Mental Life of Apes, #90, so I can't really speak to its inclusion). There are also works not on the list that I would have included, like Lewin's Dynamic Theory of Personality, Gentner's "Structure-mapping: A theoretical framework for analogy," Feldman and Ballard's "Connectionist models and their properties," and one of the Bransford and Johnson papers on memory from the early 1970s.

Anyway, maybe someone out there has a list of books they would have included, or excluded.

Thursday, February 10, 2005

Vaccines and Autism

Andrew of Universal Acid has a great post on the (non-existent) connection between the MMR vaccine and autism, and why some people still believe there is one. There is nothing I could possibly add. I just thought it was worth linking, so that's what I'm doing.

Ward Churchill

I know you're all sick of this whole Ward Churchill thing, but I don't care. I'm going to say something anyway. Controversies in cognitive psychology rarely create a political fervor. Sure, there is one debate that has taken on a life of its own, and nearly cost one very good researcher her career, but for the most part, "academic freedom" isn't much of an issue for those of us who are publishing the bulk of our work in empirical psychology journals (there are ethical issues, and while the debates over those may be politically charged, I think the issues themselves generally fall outside of the purview of "academic freedom"). Still, I feel strongly about academic freedom, and I think anyone who believes in critical thinking should. Sometimes good ideas are unpopular, and if we don't give academics a great deal of leeway, we will inevitably silence some of the voices that might be expressing those good but unpopular ideas.

Naturally, there are cases in which academic freedom can not be used to excuse the expression of ideas. For instance, if a university professor intentionally writes so as to intentionally incite unjustified violence against an individual or group, academic freedom should not be used in his or her defense. Even in this extreme case, though, things are more subtle than they at first appear. If a professor writes a paper detailing a case for going to war, he or she has certainly written to incite violence, but is the violence justified? It is virtually impossible to say when considering hypotheticals, and ultimately each case has to be evaluated individually.

Fortunately for all of us, that is what the University of Colorado appears to be doing, by allowing Ward Churchill to defend what he wrote about the victims of the September 11, 2001 attacks. Ultimately, I think Gerald Dworkin is right when he writes:
While some of the language is disgusting (little Eichmanns for those killed in the WTC) and some of the claims are bizarre (were the secretaries, janitors, fireman, waiters in the restaurants, stock clerks, etc. also part of the “technocratic corps at the very heart of America’s global financial empire"?) the main theses represent moral, political, and empirical claims about the cause of the attack, and its moral character. No faculty member should be dismissed because of such claims. Whether someone who has, and publishes, such views should continue to be retained in an administrative post is a more difficult question.
What Churchill is guilty of is a terrible analogy. It seems more consistent with his position to compare some (certainly not all) of the World Trade Center victims with the German people under Hitler, not with Eichmann or others who directly participated in the Genocide. If he had used a better analogy, some of his claims might not have seemed as absurd (which is not to say that I agree with them, just that they could at least be transformed in a way that would make them more rational). For instance, of the WTC victims Churchill writes:
They formed a technocratic corps at the very heart of America's global financial empire – the "mighty engine of profit" to which the military dimension of U.S. policy has always been enslaved – and they did so both willingly and knowingly.
One can't help but wonder in what sense Churchill means "willing and knowingly" here. His comparison to Eichmann suggests an answer, but it's an absurd one. Surely he doesn't believe that most of the WTC victims have ever considered themselves as members of the "technocratic corps" of which Churchill speaks, and even those who have probably did not consider their role in it. Most of them were likely completely ignorant of the political and economic dynamics of which Churchill wrote in that article, as most who still participate in Churchill's "technocratic corps" still are. However, if Churchill had used a more fitting analogy, he might have stated his case in a way that, while it wouldn't have gotten the attention it has now (I'm not sure that wasn't his aim), would have at least been more intellectually sound. He could have said that, much like many of the German people under Hitler in the 1930s and 40s, the people in the World Trade Center, and people throughout the U.S., have (mostly) unconsciously ignored the signs that the system of which they are a part is unethical. He might then have written:
They formed a technocratic corps at the very heart of America's global financial empire – the "mighty engine of profit" to which the military dimension of U.S. policy has always been enslaved – and their willful ignorance of this situation is no excuse.
That still might be wrong, but it's certainly more empirically viable, and avoids the hyperbole that turns his claims into inflammatory nonsense.

I would argue, then, that what Churchill is guilty of is sloppy scholarship, in the form of exaggeration and terrible use of analogy. He hasn't violated any scholarly ethics that I can think of. He hasn't falsified evidence, for instance. He's merely uttered what, in the form that he wrote them, are extremely unpopular ideas (the idea that corporate or capitalist interests drive the "military dimension of U.S. policy," which has, and still is, being used unethically, is certainly not new, or entirely unpopular among many academics). This is hardly a reason to fire a tenured faculty member, no matter how offensive we find his ideas. Instead of calling for him to resign or be fired, any academic who disagrees with him, liberal or conservative, should be sitting in front of a computer working on a paper with arguments that counter Churchill's. That's how academia is supposed to work.

Concepts III: Exemplars

Research on prototype theories was designed to show that they could deal with the issues listed in the last post, including typicality effects, fuzzy concepts, relationships between features, and relationships between concepts. While prototype theories handle these well, they ran into another set of problems, and we must now consider it necessary to explain these as well. They include:
  • Intra-category variance: Prototypes are just averages, but the variance within a category appears to matter too.
  • Interrelations between specific features: Prototypes capture category-wide feature correlations, but often miss local ones (e.g., the correlation between wooden and large in the category SPOONS).
  • Conceptual combinations: Why is a goldfish a typical pet fish, while it's neither a typical pet nor a typical fish?
  • Ad hoc categories: How do people produce and represent categories on the fly in particular contexts (e.g., things to take on a picnic)?
  • Linearly separable categories: for some natural kind and artifact concepts, all members have the same (or similar) values on a particular dimension.
For some, the way around these problems with prototype theories was to develop two-process theories that encorporated rules and prototoypes. This provides an easy solution to the problem of linear separability, and over the last 10 or so years, prototype models of this sort have done when modeling experiments that use linearly separable categories, but the other problems are more difficult to handle. In the late 70s, some researchers argued that the way to deal with these problems was to treat category representations as a collection (or cluster in high-dimensional space) of all of the members of that category that we have encountered. These theories have generally been called exemplar theories.

As with prototypes, the fundamentals of expemplar theories are pretty straightforward. We encounter an exemplar, and to categorize it, we compare it to all (or some subset) of the stored exemplars for categories that meet some initial similarity requirement. The comparison is generally considered to be between features, which are usually represented in a multidimensional space defined by various "psychological" dimensions (on which the values of particular features vary). Some features are more salient, or relevant, than others, and are thus given more attention and weight during the comparison. Thus, we can use an equation like the following1 to determine the similarity of an exemplar:
dist(s, m) = åiai|yistim - ymiex|
Here, the distance in the space between an instance, s, and an exemplar in memory, m, is equal to the sum of the values of the feature of m on all of dimensions (represented individually by i) subtracted from the feature value of the stimulus on the same dimensions. The sum is weighted by a, which represents the saliency of the particular features. The distance is converted into similarity by feeding it into a function in which the similarity decreases exponentially as the distance increases.

There are a couple ways to determine to which category an instance belongs using similarity calculated from the equation above. First, we could have some similarity threshold, and the first category to reach that threshold is the one into which we place the instance. This is how random-walk models of classification work2. Exemplars from different categories are retrieved from memory, roughly in the order of their similarity to the instance, and contribute incrimentally to the similarity of their category to the instance. More commonly, the instance is classified as a member of the category that has the highest similarity relative to the total similarity of all of the retrieved categories to the instance.

The classic experiment used to argue for the use of an exemplar model, the context model3, over a prototype model, used stimuli of the following form:

Category A Category B
1. 1111 4. 0000
2. 1110 5. 1100
3. 0001 6. 0011

Each series of four numbers represents an exemplar, and each 1 or 0 represents the value on one of four dimensions. If we take the averages on each dimension for the two categories, we get prototypes of 1111 for A and 0000 for B. If participants learn these two categories during a training phase, and, during the testing phase are presented with a new exemplar, 7, with the feature values 0101, then absent any information about the salience of the dimensions, prototype theory would predict that the probabilities of us classifying 7 as a member of A and B are equal. However, using an exemplar model equation like the one above, and the relative similarity (rather than a similarity threshold), the probability of us classifying 7 as a member of A would be 61%. The experiments showed that people actually classified stimuli like 7 at rates more consistent with the predictions of the exemplar model than those of prototype models.

As you might imagine, exemplar models are incredibly powerful. By storing all, or many of the exemplars of a category in memory, and comparing new instances to those in memory, we capture all of the important features of concepts, such as the interrelations between specific features (since most wooden spoon exemplars that we've stored in memory share the value "large" on the size dimension, these features are automatically associated with each other in any similarity calculation), the variation within categories (when you've got all the exemplars, you've got all the variation), the ways in which linearly separable categories are categorized (on the whole, if all members of a category share a value on a particular dimension, then the similarity of new instances with that value will be higher to members of that category than to members of other categories with different values on that dimension), and the role of context in classification (the probability of classifying an exemplar into a particular category is dependent on both its similarity to the members of that category as well as to the members of all of the other retrieved categories). It can also explain conceptual combinations, because instead of simply combing the two prototypes, we can combine specific exemplars. In fact, since exemplar models store all of the instances of a concept that we've encountered, it's hard to imagine any feature of permanently-stored concepts that it can't capture.

While that is a blessing when it comes to fitting exemplar models to data, it's not much of one when attempting to use the models to say something positive about concepts and their representations, and this turns out to be the biggest flaw of exemplar models. In many cases, they look more like statistical tools for analyzing classification data than actual theories of concepts, because they can model any possible data set (since they contain all the information!). What have we learned about categories if exemplar models can explain any data set, be it logically possible but empirically impossible, or actually obtained through empirical research? Not a whole hell of a lot. This problem doesn't make exemplar models any worse than prototype models, however. It turns out that if we tweak the parameters of prototype models, we can model pretty much any data set with them that we can with exemplar models. So, we have a problem. We have two types of models that are too powerful to tell us anything about concepts.

Exemplar models also have arough time with ad hoc concepts. By definition, these categories are produced on-line, rather than stored permanently in memory. If people actually can and do use these categories, as research suggests, then how do we account for them with a theory that requires that we compare new instances to old instances that have been previously stored in memory? Part of the problemhere is that exemplar models tell us nothing about the relationships between features, other than that the relationships exist. For many concepts, especially ad hoc concepts, the thematic and causal nature of relationships between features can be important for distinguishing members from non-members. Without accounting for this information, our theory of concepts will be incomplete.

For the most part in the concept literature, the problems that prototype and exemplar models share have been ignored. Instead, there has been an often violent battle between the prototype and exemplar camps, focusing more on what one type of model can do that the other can't (which quickly becomes what the latter type can do better, and on and on). Fortunately, over the last decade or so, some concept researchers have become fed up with both types of theories, and with similarity-based approaches in general. This has led to the construction of several alternative types of theories, some of which actually predate the similarity-based approaches (e.g., rule theories), and some of which are fairly new (e.g., theory theories, decision-bound theories, multiple-systems theories, and causal reasoning theories). In the next couple posts, I'll describe some of these alternatives, and try to wrap all of this up. It won't be easy, because as you've probably noticed already, things are pretty messy. It's hard to say, at this point, exactly what it is that we know about concepts.

1 Kruschke, J.K. (2005). Category Learning. In: K. Lamberts and R.L. Goldstone (eds.), The Handbook of Cognition, pp. 183-201. London: Sage. This equation describes most exemplar models fairly well.
2 Nosofsky, R.M., & Palmeri, T.J. (1997). An exemplar-based random walk model of speeded classification. Psychological Review, 104, 266-300.
3 Medin, D.L. & Schaffer, M.M. (1978). Context theory of classification learning. Psychological Review, 85, 207-238.

Tuesday, February 08, 2005

Causal Reasoning

Brandon, who studies David Hume, and has probably forgotten more about his work than I will ever know, wonders about the treatment of some of Hume's issues in cognitive science. Now, since I'm no expert on Hume, I can't do much to relate cognitive research to Hume's philosophy, but I can say a little about what cognitive scientists think about causal reasoning.

More than 250 years ago, David Hume wrote:
All reasonings concerning matter of fact seem to be founded on the realtion of Cause and Effect... I shall venture to affirm, as a general proposition, which admits of no exception, that the knowledge of this relation is not, in any instance, attained by reasonings a priori; but arises entirely from experience, when we find that any particular objects are constantly conjoined with each other.
For the most part, since the inception of experimental psychology in the 19th century, this is how psychologists have believed that people reason about causes and effects, as well. To this day, the most prominent models of causal reasoning argue that the primary type of information used in detecting causes and effects is co-occurence information.

Among the co-occurence based accounts of causal reasoning in cognitive science, there have been two general types: associationist and probabilistic. Associationist models, inherited from behaviorism, argue that the processes involved in causal reasoning are similar to those involved in classical conditioning. In fact, the most widely cited associationist model of causal reasoning was originally designed to model Pavlovian conditioning1. The model looks like this:
ΔVC,O = αβ(λ - ΣVS,O)
In which VC,O represents the associative strength between a cue, C, and an outcome, O. The salience of the cue and outcome are represented by the two parameters, α and β, λ represents the outcome of a single trial (if O, λ = 1; 0 otherwise) and ΣVS,Ois the set of all cues for outcome O (for all you math people, yes, there should be a little s under the sigma, but I'll be damned if I know how to do that). In other words, the associative strength between a cue and an outcome is equal to the difference between the outcome on the present trial (λ) and the expected outcome based on all of the co-occurence information from all trials, weighted by the salience of the salience of the cue and outcome.

Needless to say, there is a problem with this model when it is used to describe causal reasoning. Under this view, the repeated co-occurence between a cue and outcome is sufficient to establish a causal connection. However, as we all know, correlation is not causation, and people are generally pretty good at distinguishing mere co-occurences from causal relationships. For instance, every morning, without fail, when I leave my apartment my nextdoor neighbor is standing outside waiting for his dog to do its business. The associative strength of the cue (me leaving in the morning) and outcome (my neighbor walking his dog) would be 1, but I've never believed that my leaving causes my neighbor to be outside walking his dog. This is because I have some knowledge of likely causes (the dog has been holding it all night, and my neighbor has to take him out first thing to make sure he doesn't pee on the carpet).

In order to capture this fact, co-occurence theories have more recently begun to model causal reasoning using Bayesian techniques. The most prominent of these is the probabilisic contrast model, and its more recent supplement, the causal power theory. The probabilistic contrast model represents causal strength (ΔP) with2
ΔP = P(E|C) - P(E|~C)
in which P(E|C) represents the probability of an event (E) occurring with a cue (C) and P(E|~C) representing the probability of the event when the cue does not occur. Thus, the causal relation between an event and a cue is determined by subtracting the probability of the event occurring without the cause from the probability of it occurring with it. Because this equation alone can ultimately lead to the same problem that doomed the associationist models (co-occurence doesn't always signal causation), the causal power theory was added onto it3. This states that the causal power of a cue, C, is a function of the causal strength ΔP divided by 1 - P(E|~C), at least in the case that C causes E to occur (there are other versions for other situations, such as when C tends to cause E not to occur). This addition helps to explain situations in which two events co-occur with a high frequency, but are not treated as causally related, by making the causal power dependent on the value of P(E|~C), or on the probability that one of the events will occur without the other. While different probabilies of one event occuring without another can yield equal causal strengths, the causal power will be different if P(E|~C) is different.

Others have dealt with the problems of the associationist and probabilistic contrast model (sans causal power theory) by introducing the concept of causal mechanisms. Consider the following (true) story4:
The number of never-married persons in certain British villages is highly inversely corelated with the number of field mice in the surrounding meadows. [Marriage] was considered an established cause of field mice by the village elders until the mechanisms of transmission were finally surmised: Never-married persons bring with them a disproportionate number of cats.
According to Glymour5, this example illustrates a co-occurence (between the number of "never-married persons" and the number of field mice) that is mediated by a mechanism (the number of cats). A mechanism is thus a cue which, if removed, will cause the correlation between another cue and an outcome to disappear. This method allows us to bring in explicit background knowledge about the relationships between different potential causes and outcomes in order to explain co-occurence relationships that are not causal.

While co-occurence theories have dominated since Hume, recent empirical evidence has called their viability into question. For instance, Dennis and Ahn6 have shown that the perceived causal strength between two events differs for depending on the initial information people receive. If people first receive information suggesting that one event causes another to occur, they will perceive the causal strength as being higher than if they initially receive information suggesting that the same event causes another not to occur, despite the fact that over the course of the study, participants received the same co-occurence information. In another study7, they showed that when participants are presented with co-occurences of two events, A and B, and co-occurences between B and a third event, C, they will infer a causal relationship between A and C, despite the fact that these two events never occurred with each other.

Due to these results, and others, researchers have begun to develop theories of causal reasoning that depend on information other than co-occurence. In these theories, co-variance information is still important, but it is secondary to, and dependent on background knowledge.The earliest of these was the causal-model theory8, of which Lagnado et al. wrote9:
According to this proposal causal induction is guided by top-down assumptions about the structure of causal models. These hypothetical causal models guide the processing of the learning input. The basic idea behind this approach is that we rarely encounter a causal learning situation in which we do not have some intuitions about basic causal features, such as whether an event is a potential cause or effect. If, for example, the task is to press a button and observe a light, we may not know whether these events are causally related or not, but we assume that the button is a potential cause and the light is a potential effect. Once a hypothetical causal model is in place, we can start estimating causal strength by observing covariation information. The way covariation estimates are computed and interpreted is dependent on the assumed causal model. (p. 6-7)
Lagnado et al. provide a (non-exhaustive) list of four possible sources of information used in causal reasoning, (statistical) covaration, temporal order, intervention (by which they mean that human action on the world is a source of both specific causal knowledge and our general conception of causality), and prior knowledge. These four sources, along with others, could be used on conjunction with each other, especially in cases where there are conflicts.

A second theory in which information about co-variation comes after, or is dependent upon, other types of knowledge is the mechanism view of Ahn and Kalish10. This view arises out of the intuition that people's concept of cause has as a component the concept of force, or causal power. They write:
We believe that the core component of the idea of "cause" is a sense of force. If A causes B, then A makes B happen, or Be had to happen given A. It was no accident. It is this sense of necessity that distinguishes genuine causal relations from mere correlations. (p. 200-201)
Under their view, prior to other types of information such as co-occurrence, we have theory-like beliefs about causes in general (specifically, that causal powers are involved in causation, and that causation is a process), and specific causal relationships. They give the example of germs causing illness, writing:
Consider getting sneezed on and getting sick. If people think the sneeze is the cause, then they also believe that there must have been a basic process or mechanism by which the sneeze forced the illness to come about. In modern Western cultures, we typically understand the mechanism to be infection; getting sneezed on infects you with germs that make you sick. A relatively elaborate notion of the mechanism might include the ideas that germs posses the causal power to make a person sick, that the person's immune system has causal powers to counteract germs, and that the person's immune system can be weakened by lack of sleep. (p. 201)
One of the benefits of this approach is that in addition to providing an explanation of causal induction, for which most of the co-occurence models are designed, it also provides a straightforward explanation of causal abduction. In abduction, we have knowledge of the occurence of an effect, and we must make an inference about its cause. These inferences will, in most cases, be constructed through a sort-of hypothesis testing process, in which we compare possible explanations, seek more information based on the hypotheses we are considering, and pick decide on our best guess. This process is likely to involve a great deal of theory-like knowledge of the interconnections between events that go beyond mere co-occurence.

In support of this view, Ahn et al.11 conducted a series of experiments in which participants were able to elicit different types of information as they attempted to determine the cause of an event. In three separate experiments, participants almost always sought information about causal mechanisms, as opposed to co-occurrence information, and

To sum everything up, co-occurence theories of causal reasoning have dominated psychology for more than a century, and philosophy even longer. These theories posit that people use co-occurence information exclusively when reasoning about causal relationships. However, the day of co-occurence may be coming to an end. Recent empirical studies have produced findings that are difficult to reconcile with co-occurence theories. Instead of attempting this, some theorists have begun to develop alternative models in which co-occurence information still plays a role, but takes a back seat to other types of knowledge about causal relations. I wish I were qualified to relate these more recent views to Hume's own arguments, but I'm afraid that I would fail miserably. In closing, I will note that, considered from the perspective of philosophy, there is much in the Ahn and Kalish account that resembles the work of Wesley Salmon on causation (e.g., the focus on causal power and the treatment of causation as a process), which was developed, in large part, to overcome some of the problems raised by Hume.

1 Rescorla, R. A., & Wagner, A. R. (1972). A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement. In A. H. Black & W. F. Prokasy (Eds.), Classical conditioning II: Current theory and research (pp.64-99). New York: Appleton-Century-Crofts.
2 Cheng, P. W., & Novick, L. R. (1990). A probabilistic contrast model of causal induction. Journal of Personality and Social Psychology, 58, 545-567.
3 Cheng, P. W. (1997). From covariation to causation: A causal power theory. Psychological Review, 104, 367- 405.
4 Baumrind, D. (1983). Specious causal attributions in the social sciences: The reformulated stepping-stone theory of heroin use as exemplar. Journal of Personality and Social Psychology, 45, 1289-1298.
5 Glymour, C. (1998). Learning causes: Psychological explanations of causal explanation. Minds and Machines, 8, 39-60.
6 Dennis, M.J., & Ahn, W. (2001). Primacy in causal strength judgements: The effect of initial evidence for generative versus inhibitory relationships. Memory and Cognition, 29(1), 152-164.
7 Ahn, W., & Dennis, M.J. (2000). Induction of causal chains. Proceedings of the Twenty-second Annual Conference of the Cognitive Science Society, (pp. 19–24). Mahwah, NJ: Erlbaum.
8 Waldmann, M.R., & Holyoak, K.J. (1992). Predictive and diagnostic learning within causal models: Asymmetries in cue competition. Journal of Experimental Psychology: General, 121, 222-236.
9 Lagnado, D., Waldmann, M.R., Hagmayer, Y., & Sloman, S.A. (In Press). Beyond Covariation: Cues to Causal Structure. In A. Gopnik and L.E. Schultz (eds.). Causal Learning: Psychology, Philosophy, and Computation. In the chapter, they provide empirical justifications for the inclusion of each of these, and if you are interested in the topic, I highly recommend reading it.
10 Ahn, W., & Kalish, C.W. (2000). The role of mechanism beliefs in causal reasoning. In F.C. Keil and R.A. Wilson (eds.), Explanation and Cognition (pp. 199-226).
11 Ahn W., Kalish C.W., Medin D.L,. Gelman S.A (1995). The role of covariation versus mechanism information in causal attribution. Cognition, 54(3), 299-352.