Developing Intelligence

Andersen et al discuss both the attentional and intentional aspects to the function of the intraparietal sulcus. What’s the distinction between attention and intention?

First, let’s talk about attention. The modal view, based on the biased competition model of Desimone and Duncan, and the Miller & Cohen model presented yesterday, is probably that prefrontal regions actively bias particular spatial locations, as represented in parietal cortex, in accord with the current task or goal.

However, some evidence apparently conflicts with this modal view: Anderson et al. review two studies showing reduced responsiveness of parietal neurons representing spatial locations currently within the focus of attention. As Anderson et al. note, this is compatible with the idea that ventral parietal areas (including the temporoparietal junction; TPJ) are involved in the triggering of attentional reorienting to novel events and locations.

But the spatial locus of attention is only one aspect of parietal function. Andersen et al. also describe evidence that many cells in lateral intraparietal sulcus are reflective of the intended response, and not the spatial allocation of attention. One relevant experiment trained monkeys to saccade to a distractor item before responding; at the offset of the distractor, they were to respond. In this case, the cells were modulated more by the response than by the location of the distractor item, indicating they represented movement intentions independent of spatial allocation of attention.

In addition, many of the neurons in the posterior parietal cortex are effector-specific – some cells appear specific for saccade-related activity whereas others are selective to situations in which a manual response is required. This is taken as further evidence that parietal activity reflects the specific intention of an animal.

But perhaps the most compelling dissociation of attention and intention comes from a 2006 J Neurosci article by Quiroga et al., who showed that they could nearly perfectly predict whether a monkey would generate a saccade or a reaching motion to a target, on a trial by trial basis, by decoding recordings from lateral intraparietal area and the parietal reach region (using a leave-one-out approach). However, predictions of the location of the targets were much worse than those of the movement that would be generated, indicating that the recorded parietal neurons are indicative of movement intentions, and not solely spatial attention.

The distinction between attention and intention in this work feels somewhat arbitrary – why couldn’t we say that intraparietal sulcus independently represents both attention to movements and attention to spatial locations?

This is the argument made by Lau et al. in a 2004 issue of Science. They developed an alternative test of intentionality – resting on the assumption that regions representing “intentions” should be more coupled when those intentions must be freely formed relative to when it they are fully determined by the environment. The Lau et al results show that IPS activity does not increasingly predict that in motor areas (pre-SMA to be precise) under conditions of free-choice relative to forced-choice, whereas dlPFC activity does.

Unfortunately I think the Lau et al framework suffers from a similar problem as other work attempting to find the neural correlates of intentions. Their paradigm might be isolating areas representing “attention to intentions” or the “formation of intentions.”

This ambiguity is highlighted by a curious “note added in proof” at the bottom of the Lau paper: the authors cite this Nature Neuroscience article which show that patients with damage to the parietal cortex (the angular gyrus) actually lose the ability to report their own intentions, but have a preserved ability to report their own actions. I’ll cover this paper in a subsequent post.