On tantrums and black holes

Following on from my post on how research findings cluster together with huge expanses of blackness in between, here is a post on one area that I think of a huge expanse of blackness. It’s a massive and (I think) really important area, where there’s virtually no research at all at the moment.

I’ve been researching arousal levels in typical infants for a while now. (Arousal derived from a combination of heart rate, movement patterns and electrodermal activity - see publications here, here, here).

One of the things I noticed early on was that for heart rate, for example, if you plot a histogram of how the values for that measure are spread out across a whole testing session, your data often looks positively skewed. For example, the mode for heart rate in infants is typically c.120 BPM. The values to the left of the mode are typically quite closely clustered in, but then you get a tail, going out to the right hand side - often up to 180 BPM.

I hardly thought anything of it at the time, as I’m so used to looking at reaction time data that is positively skewed.

But then it occurred to me – why is this? We couldn't think of any artifactual explanation that would explain why the same phenomenon would be observed across different measures. Previous research has examined stochastic fluctuations in arousal states, and also how arousal is lagged (slow-moving), so you get inertia in arousal. But neither of these would predict a positive skew.

Maybe, the explanation for this skew comes from research into how anxiety disorders arise in adults. In panic disorder, for example, an initial small event can lead to an increase in physiological arousal which is misinterpreted as indicating an on-coming heart attack. This in turn triggers greater arousal, leading to a self-sustaining loop of incrementally increasing anxiety. Similarly, in tantrums in young children, a minor event can trigger an increase in arousal which can then be reinforced by behaviours such as oppositional behaviours, thrashing and crying. Again, as with a panic attack, a small initial increase in arousal can trigger a self-sustaining episode, as arousal levels become amplified over time.

Lots of research has looked at different arousal states (when I’m in a high state of arousal, my behaviour is in state X, whereas when I’m in a lower state my behaviour is in state Y). Other research has looked at arousal reactivity (reactivity to a particular, experimentally controlled stressor). But I think that this is a different way of looking at arousal completely: it looks at the transitional probabilities between arousal states. When I am in a state of mildly elevated arousal, how likely is it that this will lead to an episode of highly elevated arousal? It’s similar, in some ways, to the idea of attractor basins, that have been used to model multi-stability in conditions such as epilepsy.