Why mirror neurons isn’t the whole story April 16, 2007Posted by Johan in Developmental Psychology, Neuroscience, Social Neuroscience.
How do we understand the thoughts and actions of others? Traditionally, this is explained by positing that humans use a kind of naive theory of psychology to make sense of the often limited and contradictory input that we receive in social situations. This naive theory is far from perfect. For example, we tend to assume that the behaviour of others is far more likely to have been caused by dispositional than situational factors (the fundamental attribution error). So as naive theorists, we are inclined to jump to conclusions when it comes to explaining the behaviour of others.
More recently, this “theory” view of predicting and explaining actions has met some challenge from a group of theories that I will refer to collectively as simulation theory. Fueled by the recent discovery of mirror neurons in monkeys (e.g., Rizzolatti & Craighero, 2004; see a previous post), the basic idea of simulation theory is that others are understood by simulating their minds in one’s own mind. When I see a woman on the subway with her head in her hands and a stricken expression, I need not refer to poorly defined and unparsimonious concepts or naive theories about how this behaviour is to be interpreted. I simply simulate (with the aid of a mirroring system) how I would feel if I were in the same position, with the same facial expression. Simulation theory draws a lot of its appeal from the apparent parsimony of using our intuitive understanding of our own mental states to understand the mental states of others.
The basic idea of simulation theory is simple and parsimonious, yet it seeks to explain something as complex and seemingly-irrational as human social behaviour. Unsurprisingly, it runs into some problems. A paper by Saxe (2005) gets at many of these problems by showing how systematic errors in perception and reasoning (for example, the previously-mentioned fundamental attribution error) can’t be explained by simulation theory. Saxe (2005) concedes that a mirror system may well underpin “simple actions and some basic emotions” (p. 174), but maintains that a naive theory must be used for more complex attributions and predictions. To make this point, she draws on empirical examples of cognitive errors and biases:
- Children frequently confuse “not knowing” and “getting it wrong.” When 4-year-olds perform a task where they but not an adult observer know the colour of a bead (of a known set of two colours) that has been moved into a container, the children generally reason that the adult will believe that the colour is the wrong one. This is incorrect, as the adult is equally likely to guess that the colour is the right or the wrong one, since he or she does not know which colour it really was.
- 3-year-olds fail to realize that all properties of an object are not known automatically by a person who interacts with it. Thus, they do not understand that an object’s colour can be known if the object is seen, but not if it is merely touched, or that the firmness of an object can be known through touching, but not seeing.
- These egocentric errors do not all disappear at the same age. For instance, children are able to understand that other people have different desires and perceptions at two years, but will not understand that beliefs can also differ until a year later. This implies that different concepts underlie the different tasks, which simulation theory does not allow.
- Adults are generally convinced that their reasoning is entirely rational and free from such errors as the self-serving bias, but when this is probed experimentally, actual behaviour is a far cry from normative logic. If we use simulation to predict how we will reason, why is the actual behaviour so different from the simulation?
- In situations where the self-serving bias is known, adults engage in”naive cynicism,” that is, an overly pessimistic estimation of another person’s self-serving bias. For instance, when married couples were asked how often they were responsible for desirable and undesirable marital events, and how often they thought that the other spouse would assume responsibility for desirable and undesirable events, the spouses estimated that the other would be self-serving, even though the actual data showed that both took credit equally for desirable and undesirable events. It’s difficult to explain why this occurs without again invoking different concepts.
All the above examples probe the same inconsistency: our failure to consider others in the same way that we consider ourselves. In its simplest form, simulation theory predicts that these social biases simply should not occur.
Perhaps imperfect simulation can explain these discrepancies? It’s reasonable to suggest that any simulation is going to be inferior to first-hand experiences, and we do not lose much of the ever-alluring parsimony of simulation theory by positing that the simulated experience is weaker or somehow degraded. Saxe (2005) doesn’t buy this explanation. For example, the case of the beads outlined in point 1 above cannot be explained in this way. To re-iterate, the child sees the colour of a bead being put into a container, while an adult does not. The child is then asked what colour the adult will think that the bead is. Either the child simulates the mind of the other appropriately, in which case performance should be good, or it fails to simulate the mind of the other, and assumes that the adult knows what the child knows. But this isn’t what happens – as described above, the child insists that the adult will always pick the wrong bead, instead of guessing (proper simulation) or picking the right bead (failed simulation). Similar problems arise with the self-serving bias examples in points 4 and 5. Why is performance so different from the norm in the couple situation?
Fundamentally, the problem may go back to the tired old controlled/automatic distinction. Simulation theory posits a bottom-up, automatic process, which is unaffected by beliefs, expectations, and other aspects of a naive theory. As such, simulation theory works very well when explaining behaviour in paradigms such as facial expression identification, where people seem to have an uncanny ability to identify expressions extremely quickly, and with little effort. The “theory” approach has a better shot at explaining top-down influences in reasoning. It’s worth noting that Saxe’s paradigm, where children are essentially told stories and are then asked what different actors in the story might know or think, is one well-suited to this approach. The examples Saxe (2005) cites as evidence against simulation theory (in points 1-5 above) rely on making inferences in this controlled manner.
Perhaps it is too much to ask that one theory should explain both the fast, automatic processing required in typical simulation paradigms, and the thoughtful, controlled processing in typical “theory” paradigms.
Rizzolatti, G., & Craighero, L. (2004). The Mirror-Neuron System. Annual Review of Neuroscience, 27, 169-192.
Saxe, R. (2005). Against Simulation: The Argument from Error. Trends in Cognitive Sciences, 9, 174-179.