Working memory and fluid g

Yesterday I sketched an argument for believing that all access-conscious thinking is sensory based. But suppose this conclusion is wrong. Suppose there is some sort of workspace in which amodal (nonsensory) thoughts – judgments, goals, decisions, intentions, and the rest – can become active and be conscious. What would one predict? One would surely predict that variance in the properties of this workspace among people would account for a large proportion of people’s variance in fluid general intelligence, or fluid g. For it is conscious thinking that is thought to underlie our capacity to solve novel problems in creative and flexible ways, which are precisely the abilities measured by tests of fluid g.

However, there are now a great many studies examining the relationship between working memory and fluid g. Generally variance in the former overlaps with the latter somewhere between 0.6 and 0.9. (That is to say, the relationship between the two seems to be somewhere between very strong and virtually identical.) Many have thus come to regard working memory as the cognitive system or mechanism that is responsible for fluid g (which is itself a purely statistical construct, of course, being the underlying common factor calculated from a range of different types of reasoning task). And to the extent that other factors have been found to correlate with fluid g independently of working memory, the only one that has received robust support is speed of processing, which seems to be a low-level phenomenon (perhaps related to the extent of neural myelination).

It may be, of course, that standard tests of working memory tap into both the sensory-based system and the supposed amodal attitude-involving system. But in that case one would predict that as tests of working memory become more and more sensory in character (requiring one to keep in mind or manipulate un-namable shapes or shades of color, for example), the extent of the overlap with fluid g should go down. For these tests of purely-sensory working memory would fail to include any measure of the variance in amodal thinking abilities which would (by hypothesis) account for a large proportion of our flexible general intelligence. But this seems not to be the case. Low-level sensory tasks overlap with fluid g just as strongly (if not more strongly) than do concept-involving ones. Moreover (and just as the sensory-based account would predict) measures of sensory attentional control (using such tests as the anti-saccade task or the flankers task) themselves predict capacities for general intelligence quite strongly.

In addition, there is a separate body of evidence that pushes toward the same conclusion. This derives from studies that have presented people with a range of different sensory-discrimination tasks. Participants might be asked to order a series of color-chips by shade, arrange a series of lines by length, arrange a set of tones by pitch, order a set of identical-looking objects by manually feeling their weight, and so on. From these measures one can compute an underlying common factor (just as one does when computing fluid g from a range of reasoning tasks). While it is unclear exactly what this common factor represents, it seems likely that it has to do with capacities for sensory attention and purely-sensory working memory. Across studies, it has been found that this underlying factor overlaps with fluid g between 0.6 and 0.9 (which is the same as the extent of overlap between working memory and fluid g, note). Since for sure there will be executive and memory-search components of working memory that make no contribution to these sensory-discrimination tasks, we can conclude pretty confidently that there is no variance in general intelligence remaining to be explained by the hypothesized workspace for conscious amodal reasoning.

Given that all conscious reasoning takes place in the sensory-based working memory system, tomorrow I will say a little about how this framework can be used to explain some of the findings from the psychological literature on human reasoning and decision making.

Comments are closed.

Back to Top