Thursday, August 18, 2011

Mirror Neuron Forum - some additional discussion - Part II

In my answer to Question 1 I suggested that mirror neurons can be viewed analogously to canonical neurons, that is, as a sensory-motor association system involved in action selection, not action understanding. Here is Gallese's response to this suggestion:

According to GH, both classes of neurons instantiate the action-oriented coding typical of the dorsal stream, whereas object and action semantics would be exclusively provided by the ventral stream. However, an exclu- sive action-oriented characterization of the dorsal stream falls short of explaining the functional role exerted by the ventral part of the dorsal stream (the ventro-dorsal stream) that reci- procally connects cortical areas of the inferior parietal lobe to ventral premotor areas (see Gallese, 2000, 2007a, 2007b; see also Rizzolatti & Gallese, 2006; Rizzolatti & Matelli, 2003). The meaning we attribute to objects is not exclusively the out- come of their visual description as instantiated by extra-striate visual areas within the ventral stream. That is, objects are not merely identified and recognized by virtue of their physical ‘‘appearance’’ but also in relation to the effects of the potential interaction with an agent...

Here is the problem with Gallese's counter-argument: it's not an argument at all. He states that my idea "falls short of explaining the functional role exerted by the ventral part of the dorsal stream" which I take to be spelled out in his next statement that "The meaning we attribute to objects is not exclusively the outcome of their visual description..." Gallese is describing his hypothesis, not an observation that needs explanation. Basically what Gallese is saying is that my view falls short of matching what he thinks is going on. The "argument" structure is: GH proposes Function A for the system. VG proposes Function B. Function A does not include Function B. Therefore Function A is wrong. Obviously, this is not an argument and should be disregarded.

What we need instead is to discuss how well the two hypotheses account for the available data. Is there any evidence that suggests that monkeys (remember this question is about MNs in monkeys) fail to understand objects or actions when the dorsal stream is damaged? No. All we have is evidence that cells in the dorsal stream fire both during movements directed at objects and during observation of actions or objects. Correlation does not imply causation as we all know so there is no direct evidence for Gallese's claim. Do we have direct evidence for my claim? Well, object recognition deficits can be induced by disruption of ventral visual areas (e.g., see the classic work by Ungerleider and Mishkin) and while not direct, the fact that the response properties of cells in the ventral stream have the right features for object and action recognition (specificity, invariant to size, etc.) is consistent with my proposal. In other words, there are empirical reasons why the "standard view" of dorsal and ventral stream function in the monkey visual system is what it is.

The predictable response from Gallese et al. will be that yes, the ventral stream is good at recognition but this recognition is devoid of true meaning. Again, this is a non-argument in that all it is, is a counter-hypothesis that is not supported empirically in the macaque. This is where the argument slips into human data as we'll see in Gallese's next point. (But I thought there was no reason to assume human and macaque systems would be doing the same thing Vittorio?)

Gallese, V., Gernsbacher, M., Heyes, C., Hickok, G., & Iacoboni, M. (2011). Mirror Neuron Forum Perspectives on Psychological Science, 6 (4), 369-407 DOI: 10.1177/1745691611413392

No comments: