Wednesday, August 22, 2007

Self-Reported Vividness of Imagery and the Cortex

In 2002, I published an article critical of the generally weak-to-nonexistent relationships between self-reported vividness of imagery and performance on tasks psychologists have often thought to involve imagery, such as mental rotation tasks and tests of visual memory and visual creativity. Differences in subjective report about imagery, I suggested, may relate only poorly to real differences in imagery experience. This fits with my general skepticism about the trustworthiness of our reports about our own conscious experience.

Yesterday, on a tip from Anibal in a comment on Monday's post, I read two articles on the relationship between self-reported vividness of visual imagery and activation in the cortex during visual imagery tasks.

The self-report measure was the widely-used Vividness of Visual Imagery Questionnaire (VVIQ). The VVIQ asks respondents to imagine various scenes (e.g., some relative or friend -- the exact contours of face, head, shoulders, and body; characteristic poses, etc.) and then rate the "vividness" of the resulting image on a scale from 1 (perfectly clear and vivid as normal vision) to 5 (no image at all, you only "know" you are thinking of the object).

Amedi et al. 2005 looked at nine subjects. Those who rated their images as more vivid on the VVIQ showed a trend (not statistically significant, though) to have more activity in their visual cortex during visual imagery tasks and, perhaps more interestingly, a substantial (and statistically significant) tendency to show less activation in their auditory cortex -- which Amedi et al. interpreted as showing a narrow focus of concentration on visual matters.

Cui et al. 2007 retested the issue of visual cortex activation and were able to confirm Amedi et al.'s trend: In their eight subjects, there was a strong and significant tendency for those claiming more vivid imagery on the VVIQ to show more activation in the visual cortex during visual imagery tasks. Although Cui et al. don't comment on this, a striking trend appears in their time course data: The self-rated poor visualizers start out with as much visual cortex activation as their vividly-visualizing peers, but that activation rapidly declines relative to other brain activity (over 10 seconds), while the good visualizers keep their level of visual cortex activation constant or increase it. The possibility occurs to me, then, that the difference between them may be in maintaining focus on the task -- which would also harmonize with the Amedi et al. results that more vivid self-rated visualizers show more selective cortical activation.

It still puzzles and troubles me that VVIQ scores should relate so poorly to behavioral performance. If only there were more research like this, showing consistent relationships between self-report of conscious experience and third-person measures!


Pete Mandik said...


Here's a hunch concerning high-vividness self-report and behavioral tasks:

If you buy into the two-streams-of-visual-processing stuff a la Milner and Goodale, then you can sort tasks into those that are more "ventral" and those that are more "parietal" (I am assuming that you are sufficienntly familiar with M&G that my shorthand isn't coming across as gobledegook.) My hunch is that the vividness goes along more with the ventral tasks.

This hunch is motivated by readings of the two-streams stuff whereby consciousness is more strongly correlated with ventral stuff than parietal stuff.

Eric Schwitzgebel said...

Interesting thought, Pete. My recollection is that the dorsal stream is usually taken to be more involved with reaching and the parietal/ventral to be more involved with consciousness and verbal labeling. But then the question is which visual *imagery* tasks would correspond to which stream (rotation, memory, etc.); that I'm not sure.

Cui et al. looked only at V1 and V2. Amedi et al., however, looked at the whole brain. It looks to me from their figures that, if anything, the ventral stream was relatively *de*-activated during visual imagery -- but the auditory cortex (where they hypothesized and found deactivation) is pretty close to the ventral stream, so I don't know.

It would make an interesting follow-up study!