Imaging a Superior Mnemonist

In neuroscience, we spend most of our time trying to understand the function of the "normal" brain -- whatever that means -- hence, we are most interested in the average. Under most occasions when scientists take an interest in the abnormal neurology, it is usually someone with who has something wrong with them -- has brain damage or a disorder of some kind. In these cases, we try and understand what brain functions they have difficulty performing as a way to understand what each part of the brain does (and hopefully to someday be able to help them).

The point is that when neurologists study the abnormal, it is typically on the non-functional end of the spectrum rather than the highly functional. This is why I found a paper in the journal Neurocase quite interesting. The authors, Raz et al., placed a superior mnemonist -- an individual who can memorize long lists of arbitrary items -- into an fMRI scanner to try and get an idea how they did it.

Raz et al. imaged a single participant known throughout the study as "PI." PI is remarkable in his ability to recite the constant pi to around 2^16 decimal places with very few errors. (I have no idea whether the selection of the pseudonym is because P and I are actually this guy's initials or it is some sort of pun.) The authors subjected participant to a psychological battery and two scanning experiments in the fMRI to try and see how this individual was so gifted at memory.

PI claims to be able to remember so many digits using a variant of a memory strategy called the Method of Loci (MOL). MOL is a very old method of memorizing material -- known even to the ancient Greeks and Romans -- where the user attempts to associate each item in memory with a spatial location or a visual scene. The user then mentally "walks" between each location and visual scene and uses this to help recall the associated items in memory. Basically, if you were trying to memorize all the phone numbers of all the people in your building, you might associate each phone number with a different turn on your drive to work. By visualizing the turn, you could help retrieve each phone number more easily. This MOL technique is demonstrably effective at improving memory performance, and many superior mnemonists claim to use this technique.

In interviews with PI, the subject claims to use a variation on the classical MOL technique. Rather than visualizing a mental journey, PI claims to try and associate each item in memory with an emotionally-laden image -- sometimes funny, sometimes horrible. In this way, the subject can "co-opt" the emotional salience of the visual image to make the memory item more memorable. (More on the reliable of the subject's reports and the relevance of MOL on the findings in a second.)

When the authors tested PI on a variety of memory tests -- testing memory for different types of material --- they found that PI's abilities were around average with one exception. This finding is consistent with what we know about how memory is organized in the brain. Memory is not monolithic. Rather, it is organized into several parallel memory systems that sometimes act cooperatively and sometimes act antagonistically with one another. Each system is responsible for different classes of information. (For more information, I would read this.) Thus, it is reasonable that PI would be particularly good -- probably through training -- at certain types of memory, but not at memory as a global concept.

What was the type of memory that PI was good at? Neuroscientists refer to working memory as the type of memory that is necessary to keep information online over short intervals. This is the type of memory you use when you are trying to remember a phone number written on a piece of paper away from the phone before dialing. Measurements of PI's working memory were off the chart (99th percentile). This is consistent with other superior mnemonists, and we think that some element of rehearsal -- saying the item over and over again to yourself -- within working memory may have to do with their excellent performance.

The authors then placed PI into a scanner and asked him to perform two tasks. In the first, they asked him to recite pi. This experiment asked: what areas of his brain are activated when he is retrieving this long number string? In the second, they asked him to memorize a list of 100 arbitrary (and novel for him) numbers. This experiment asked: what areas of his brain are activated when he is encoding new information of this type? Note that both of these two experiments are necessary. Memory is a multi-part process. For something to be remembered, it must first be encoded in the brain -- analogous to writing something down. Then, after an interval, this encoded information must be successfully recalled -- or retrieved.

When the authors asked PI to recite pi, they found increased activation in areas of the frontal cortex including the medial frontal gyrus and dorsolateral prefrontal cortex. This is interesting in relation to PI's gifts at working memory. According to some current theories of how working memory and intelligence work (most notably P-FIT), frontal regions participate with regions in the back of the brain in the parietal cortex to form what is called a fronto-parietal loop. Information circulates between these two regions allowing it to stay online in the brain and continue to occupy working memory. We also believe that activity in the fronto-parietal loop may have a function in allocating attention.

What is interesting about this finding that this activation differs from structures that we traditionally associate with memory -- structures like the hippocampus. Superior mnemonists appear to co-opt other memory systems -- such as working memory -- to increase performance.

Similarly, when the authors asked PI to encode a new random string of 100 numbers (which he later recited flawlessly to confirm that he had retained them), he activated parts of his frontal cortex as well as various association cortices (parts of the brain involved in associating different classes of stimuli). Activation in assocation cortices are consistent with PI's reports of using the variant MOL technique of associating emotionally-laden images with the numbers. Again, the activity in frontal regions may be the result of co-opting working memory to assist in encoding.

All of these findings provide an interesting look into how a superior mnemonist's brain works, but there are a couple of caveats.

First, this is a study with only a single subject, and we should be at least mildly skeptical of it on that basis. While the findings presented here conform broadly with those of other superior mnemonists, we are not dealing with a large enough data set to make definitive conclusions. Further, the authors acknowledge that we may not even be imaging this subject at the correct time. Clearly, PI has a lot of training applying the variant MOL technique. What would be really interesting is to follow a naive individual applying this technique to see how activation changes over time as the new technique is applied. Studies of this nature have been performed, but it would be interesting to watch subjects reach even higher levels of performance.

Further, we must always be skeptical of first person reports in science. I have no doubt that PI is a honorable person and provided accurate information about his thought processes to the best of his ability. But our insight into our own mental processes is always limited. One aspect of this that I found interesting is PI's claim about the variant MOL procedure. PI claims that the variant MOL includes associating the digits with emotionally-laden imagery. Emotionally-laden stimuli in the brain is processed in part by a brain region called the amygdala, and individuals exposed to these stimuli often show activation in this region. This type of activation was not observed in either retrieval or encoding in this case. Whether this is because self-generated emotional imagery is processed differently or whether PI's insight into the nature of his strategy is imperfect, it is very difficult to tell. I am just pointing out that we need to be skeptical of how people describe their memory experiences.

It may be that one reason that such little research goes on about superior human functioning is because exemplars are so rare. It is difficult to corral enough superior subjects together, and each subject may be unique in their superior function, such that it makes comparisons difficult to make.

Still I think this is a very interesting paper that offers some insight in how the human brain can be co-opt to perform amazing feats of memory.