Friday, April 4, 2008
TB Journal Club: Effective and Structural Connectivity of Human Auditory Cortex
Ok, so I have had a chance to read through Upadhyay et al.'s recent paper in J. Neuroscience (2008: 3341-9). It's a good one. They asked subjects to listen to short sentences during BOLD fMRI, and also collected scans for DTI imaging. Using Heschl's gyrus (HG) as a seed, they performed Granger causality mapping to identify regions that are functionally coupled with activity in HG. Two regions showed up, one anterior to HG and the other posterior to HG. The locations are pretty dorsal, involving the lateral portions of the supratemporal plane and wrapping out toward the crown of the STG in both locations (see figure). DTI analysis showed that these two sites are connected to different regions of HG: the anterior site appears to get its input from rostral HG, whereas the posterior site appears to get its input from caudal HG.
This finding is consistent with dual pathway models of primate auditory cortex which distinguish between ventral/rostral and dorsal/caudal streams and suggest that the distinction between these pathways in humans is present at the level of A1.
Mapping connectivity patterns is critically important to understanding the functional anatomy of audition including speech/language processes, and this study is a big step in that direction. But still we don't seem to be any closer to resolving the functional roles of these projections. For example, there remains debate over the extent to which anterior vs. posterior STS regions (which seem to be beyond the scope of the Upadhyay et al. analysis) support speech processing. Regions of the STS. both anterior and posterior, seem to behave very ventral stream ("what") like in their response to speech stimulation. Also, there is debate over whether the dorsal/caudal stream supports spatial functions, sensory-motor integration functions, spectro-temporal analysis, or some combination of these. And then there's the Computation Hub idea. The present study doesn't really resolve any of these questions (not that it was intended to though).
Still lots to work out. I've got plenty of ideas on the topic, and hope to put together a paper soon that lays them out. Here's a couple of preview tidbits:
1. The computational hug idea (typo intended) is wrong, at least in its broadest conceptualization. It's an interesting hypothesis, but the fact that lots of different types of stimuli are getting cozy in the planum T. doesn't mean that there is a single mechanism devoted to sorting them out into different processing streams. Hickok & Poeppel 2007 touched on this issue.
2. Evidence for pure spatial functions in the dorsal/caudal stream -- whether we are talking auditory motion or spatial localization -- is weak at best, and probably non-existent, as Robert Zatorre has suggested (Google "Where is 'where' in auditory cortex'). Here's another post on the topic.
3. Both anterior and posterior projections are involved in "what" processes (although I don't know what kind of what).
4. Sensory-motor integration is an important function of the posterior planum region, although it is probably a mistake to refer to it as part of an "auditory" stream. See this previous post.