Show me your Amygdala and I’ll tell you how you feel April 14, 2007Posted by Johan in Emotion, Neuroscience, Social Neuroscience.
When cognitive neuroscientists discuss MRI, the focus is usually on the functional applications: with the help of primarily the blood-oxygen-level dependent (BOLD) method, it is possible to get an (indirect) measure of activity in the brain as the participant views a stimulus or makes a response.
However, the “vanilla” structural MRI, which produces high-resolution images of the brain in vivo, can be interesting to cognitive science in its own right. The typical paradigm in this area is to correlate the size of a given brain structure with behavioural measures, a design not entirely dissimilar to the lesion studies in neuropsychology, where deficits in brain-damaged patients is related to the site of the lesion. To look at this structural MRI application further, I looked at two recent papers that investigated how the size of the amygdala relates to behaviour.
Is emotional memory mediated by amygdala and hippocampus size? This is the basic question that Weniger, Lange and Irle (2006) set out to answer by comparing a group of women with recent-onset major depressive disorder (also known as clinical depression) with a group of controls matched on age, years of education, and height (!).
A simple comparison showed that the depressive participants had larger amygdalas and smaller hippocampi, compared to the controls. The depressive participants were also impaired at a task which assessed memory for facial expressions. Within the depressive group, a negative relationship emerged between left amygdala volume and performance on the emotional memory task, and between left amygdala volume and self-reported anxiety. Self-reported anxiety also correlated negatively with left hippocampus volume.
So to summarise, depressed people have larger amygdalas and smaller hippocampi. Weniger et al (2006) would surely like to conclude that these structural differences cause the behavioural differences in emotional memory and self-reported anxiety, but naturally this conclusion cannot be made from a study that has no experimental manipulation. It’s possible that a third variable (ie, another brain structure – note that Weniger et al didn’t compute correlations for anything apart from the amygdala and hippocampus) causes both the behavioural and the structural differences. Alternatively, as Weniger et al hint at, negative life events act on the amygdala and hippocampus. It’s easy to assume that because one variable is a brain structure, this variable is necessarily the cause, since we all know that the brain is the cause of behaviour in a more general sense… In the individual case this is not necessarily true, since the brain can be shaped by the environment just like the environment can be shaped by the brain.
Nacewicz et al (2006) used a fairly similar design to carry out their two studies, with some variations: they used a facial expression identification task rather than learning task, and instead of looking at the performance of depressed people, they compared people with high-functioning autism or Aspberger’s Syndrome to controls matched by age and sex.
As an aside, it’s worth noting that this control group is less stringently matched than the one employed by Wegener et al (2006). Indeed, IQ measures showed that the two groups differed by a fairly massive 20 points, which is somewhat problematic if you want to conclude that any behavioural differences between the groups are mediated by differing amygdala volume, rather than differing intelligence (note however that Nacewicz et al attempted to control for this by using IQ as a covariate).
Briefly, Nacewicz et al (2006) found that within the autism group, amygdala volume was negatively related to reaction times in the facial expression identification task. There was also a negative relationship between facial expression identification RT and eye region fixations while carrying out the task, which neatly suggests a possible mediator of the identification RT performance. In an attempt to include a longitudinal scope, Nacewicz et al (2006) also used the Autism Diagnostic Interview to show that amygdala volume was related to the degree of retrospectively self-reported social impairment in early childhood. In contrast to the depressed participants studied by Wegener et al (2006), the participants with autism in Nacewicz et al’s (2006) study had smaller amygdala volume than the control group.
This last finding is a little surprising. The amygdala is commonly thought to play a part in the experience of fear or anxiety, and anecdotal evidence suggests that individuals with autism experience these emotions like everyone else, or even more strongly (this documentary illustrates the point well). So while Wegener et al’s (2006) findings make sense in that you’d expect depressed people to be anxious if not afraid, it’s difficult to make sense of why people with autism should have smaller amygdalae, if they are not less afraid or less anxious than others.
Davidson (one of the co-authors) believes that this can be explained by allostasis, that is, neural adaptation to chronic stress. In this view, people with autism (and depressives) have hyperactive amygdalae. This hyperactivity causes initial growth (as would be the case in the depressive sample studied by Wegener et al, 2006), but eventually gives way to atrophy and shrinkage. This is believed to occur early in life for autistic individuals, which would explain why the sample Nacewicz et al (2006) studied had smaller amygdalae than the controls. In this view, then, the amygdalae of people with autism is atrophied but hyperactive. This may sound like a just-so story, but there is a bit of evidence to support it (see the original paper).
While structural neuroimaging is a powerful tool especially when looking at individual differences, these two studies illustrate some of the potential problems with this approach. Unlike functional studies, where a typical analysis involves subtracting the activation of the whole brain in the control condition from the activation in the experimental condition to see which areas are specifically involved in processing the experimental stimulus, these structural studies generally only look at correlates of one or two previously-defined brain structures. This is done for practical reasons – while subtracting one activation pattern from another in fMRI is mainly a matter of processing power, defining brain structures in a structural study requires manual measurement, participant by participant. Unfortunately, one consequence of this practical limitation is that there is no way of knowing what other areas correlate with the relevant behaviour, if any.
That being said, this method of correlating brain structure size with behaviour does offer an interesting source of convergent evidence for “traditional” fMRI studies, which are also fraught with limitations. Subtly different limitations, that is, so if the functional and structural results agree, you can have some confidence in concluding that the results are not a side-effect of your method.
Unfortunately, both these methods share a common limitation – they are correlational, not experimental. All the problems of interpretation that were outlined above would apply equally if the Wegener et al (2006) study had tried to correlate BOLD fMRI amygdala and hippocampus activation with behaviour instead.
Nacewicz, B.M., Dalton, K.M., Johnstone, T., Long, M.L., McAuliff, E.M., Oakes, T.R., Alexander, A.L., & Davidson, R.J. (2006). Amygdala Volume and Nonverbal Social Impairment in Adolescent and Adult Males With Autism. Archives of General Psychiatry, 63, 1417-1428.
Weniger, G., Lange, C., & Irle, E. (2006). Abnormal Size of the Amygdala Predicts Impaired Emotional Memory in Major Depressive Disorder. Journal of Affective Disorders, 94, 219-229.