Tag Archives: Outliers

Memory is fallible, but then again, there’s super-recognizers

It appears that super-recognizers (people with very good face recognition ability) are mentioned in the new book The Memory Illusion by Dr Julia Shaw, but I cannot find a preview of that bit of text. I’d be interested in reading what Shaw wrote about supers, because I believe that we are very good evidence against the argument that this book, and some other pop psychology books have offered, that human memory is unreliable and open to interference. I’ve noticed that writing by researchers and authors who offer arguments against the reliability of human memory (such as Elizabeth Loftus) and also those who offer arguments against the idea of natural or inborn talent (such as K. Anders Ericssson) tend to ignore or gloss over the many things that science already knows about face recognition, face memory and super-recognizers.

I’m happy to admit that people who perform amazing feats of semantic memory such as remembering huge lists of random facts or meaningless digits using new or ancient memory techniques have trained their own memories with many hours of practice, but super-recognizers are very different to those people. We do not knowingly or deliberately train ourselves and we do not consciously use tricks or techniques. Maybe we self-train and invent strategies in an implicit manner, but it is also true that super-recognition does seem to run in families, so there seems to be an important genetic contribution to the elite ability or talent, just as there is clearly a genetic component to developmental prospagnosia (very poor face recognition ability).

Face memory researchers have been investigating the phenomenon of super-recognition since it was first described in 2009, and there seems to be ample evidence that supers have very long-lasting, adaptable, and reliable memory of the faces of other humans. We can remember faces across many decades and across changes in facial appearance by forces such as ageing. I believe I am very good at spotting facial family resemblance and facial phenotypes across gender and age. Super-recognizers can also display very accurate face recognition after being briefly shown images of only faces (no hair etc) of a large group of faces of same gender and similar age, some of them very degraded images. This accuracy requires being able to avoid false positives and false negatives. There’s no denying that supers are bloody good at faces. There’s also no denying that some other people are very poor at face memory, so authors of these pop psychology books that denigrate human memory are able to state with a vague air of truth that human memory for faces is fallible. But such a statement ignores what we know about supers, and this is why I have issues with the common practice of psychology researchers of roputinely discarding data from outliers in their studies. If any of that discarded data is from outlier study participants that did incredible well in tests of face recognition or memory, then those participants could be supers and their data tells an important story about human memory and human face recognition.

I think supers are interesting examples of a type of human memory that stands out from other types of human memory as reliable, long-lasting, easily or unconsciously enmcoded and accurate, so one should wonder, why is the face memory of supers so great? My bet is that this niche example of human memory has two characteristics that give it special power: it is disributed across a broad network of neurons throughout the brain (and this is why it might be found along-side synaesthesia), and it is also a type of visual memory, which I can only assume is a very ancient and well-evolved type of human memory that predates stuff like writing and language, that happens in areas of the brain that work amazingly and unconsciously because they evloved well before there ever were humans. I cannot imagine how genuine face memory could ever be interfered with by suggestion or manipulation, because the tricks that some memory researchers have used to fool around with the memories of study participants work on a conscious level communicated by verbal means. Genuine face memory is implicit and visual. It is safe from such nonsense.

The Memory Illusion by Dr Julia Shaw:

https://books.google.com.au/books?id=OdKOCwAAQBAJ&source=gbs_navlinks_s

 

A link between autism and super-recognizer ability, or am I reading this wrong?

I was just having a look at an Australian/UK study of face recogniton ability that was published last year in the open-access science journal PLoS One. The subjects in one of the two studies reported in this paper were parents of autistic kids, and they were tested with the Cambridge Face Memory Test (CFMT). The CFMT happens to be one of the tests that were used in the 2009 paper by Russell, Duchaine and Nakayama that established the concept of the super-recognizer. There are a couple of problems with comparing scores between these different papers – the 2009 paper used both the short and long forms of the CFMT and gave raw scores, while the 2010 paper used only the 72 question short form of the CFMT and gave age-standardized z-scores based on a study of the Australian population. But having looked at the 2009 study I don’t think the long form does that much better a job of sorting the super-recognizers from the controls than the short form does.

I’m happy to stand corrected, but to my eye it looks as though there is an interesting score in the CFMT reported in Figure 2 of the 2010 paper. If the vertical axis is in standard deviations then I guess that the top score from a father of an autistic child that is nearly level with the number two is close to super-recognizer class. He almost looks like an outlier. According to the authors of the 2010 paper, none of the parents of autistic kids in this study scored in the range of prosopagnosics, who apparently typically score less than two standard deviations below the control mean.

Definitions of prosopagnosiacs and super-recognizers can be found in the 2009 paper; “Most developmental prosopagnosics we have tested in our laboratories score around 2–3 SDs below normal on the CFMT short form. In comparison, 3 super-recognizers scored around 2 SDs above the mean on the CFMT long form.” It appears that the short and long forms of the CFMT are comparable with regard to SDs and face-blindness, and also I presume with regard to super-recognizers. What would really be interesting would be to see what kinds of scores the autistic kids would get on these tests.

References

Wilson CE, Freeman P, Brock J, Burton AM, Palermo R (2010) Facial Identity Recognition in the Broader Autism Phenotype. PLoS ONE 5(9): e12876. doi:10.1371/journal.pone.0012876
http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0012876

Russell R, Duchaine B, Nakayama K Super-recognizers: people with extraordinary face recognition ability. Psychonomic Bulletin and Review. 2009 Apr;16(2):252-7. http://pbr.psychonomic-journals.org/content/16/2/252.full.pdf      http://visionlab.harvard.edu/members/ken/Ken%20papers%20for%20web%20page/157Russell_supersPBR.pdf

Are super-recognizers the type of people that researchers discard?

I guess I’m unusual among housewives in my keen interest in the sciences that are relevant to human life, to the degree that I go beyond reading popular books and articles on the subject and I read complete science journal papers. I was once a psychology student at a university, but I didn’t complete that degree. I’ve never been satisfied to accept dumbed-down, summarized and officially-approved explanations of how things work, so I’m a bit of an autodidact, and I like to find out what goes on behind the stories that we are told by science journalists, alternative medicine advocates, doctors and pop science writers. To this end I have taught myself a bit about how research studies are conducted and written-up, and how they should be conducted and written up. One thing that many people might not be aware of is that researchers routinely discard a particular type of data from their studies, and this selective exclusion of data isn’t seen as a scandal. Outliers are observations that deviate markedly from other observations in the sample. Outliers are data points that are a long way from the mean of the sample. They are extreme values. When super-recognizers take tests of face recognition, I guess their scores could look like outliers. I’d love to know whether exceptional performances of super-recognizers have been recorded in studies, but the observations of those performances excluded from data sets. But there is no way to know, because researchers often don’t give information about excluded outliers in their papers reporting studies. Apparently the exclusion of outliers is now frowned-upon and is not regarded as best practice, but it has been a common practive in years past. Today I have been looking at a study involving a test of face processing that was published in 2006. The paper’s authors explained that two outliers were excluded. They gave no information about these excluded outliers except for the gender of the relevant study subjects. It’s a pretty poor way to conduct research, don’t you think?