Tuesday, January 27, 2009

Where is 'where' in the human auditory cortex?

This is the title of an interesting and highly cited paper by Zatorre and colleagues (2002) which questions the existence of a dedicated spatial processing system (or stream) in human auditory cortex. I have highlighted this issue previously here on Talking Brains in the context of my (now) former graduate student Kevin Smith's fMRI study on the topic. Having spent the last year+ trying to get the paper published, I'd like to return to the issue, in part to vent a little frustration and in part to hopefully clarify a bit of confusion -- maybe some of my own.

First the story on the paper, which now has the proud distinction of being rejected from the Journal of Neuroscience, not once, but twice! Our basic result: we failed to find a region within the planum temporale that showed a selective response to auditory spatial manipulations (more details in the previous post; see above link). As with previous studies, we easily identified a region that showed a greater response to a stimulus (we used human speech) that spatially varied than to a stationary stimulus. However, we found that simply by adding additional sound objects (more talkers) at the same stationary location we generated just as much of an activation increase in the "spatial" ROI. That is, "spatial" regions are equally sensitive to non-spatial manipulations, specifically the number of auditory objects in the signal; they are not selective for spatial manipulations. This surprised me and we thought J.Neuroscience would be interested.

First submission: a reviewer criticized our use of passive listening, stating that spatial activations in the PT were only found in studies using active tasks, and that is why we found no selective spatial activation. Editors rejected the paper. I appealed the decision (there is a formal process) citing several different papers that showed spatial activations in the PT using passive stimulation. The editors invited a resubmission of the paper along with the opportunity (actually, requirement) to pay another $100 submission fee.

Second submission: round one went fairly well (they do the reviews from scratch after an appeal) with the usual comments and criticisms, and we were invited to revise and resubmit. We addressed the concerns pretty well we thought and satisfied one of the reviewers, but another (apparently new!) reviewer criticized the paper harshly. Quoting from a 2000 paper by Rauschecker and Tian (a macaque study) the reviewer stated that our "core prediction" that of "signal increase in PT as a function of sound sources" is predicted by proponents of a dorsal, spatial processing stream. Said reviewer went on to quote from Talking Brains (!) implying that we don't understand our own data! Wasn't that the review comment David got once? :-) Editors rejected the paper again.

So this is a bit irritating (please someone correct me if I'm missing something!) because (i) our core prediction was NOT that the PT should be modulated by the number of sound sources, but rather that the PT should NOT have a spatially selective region in it, and (ii) since Rauschecker and Tian's statements, there have been several claims that there is a spatially selective region within the human PT. Here's a few quotes and pictures:

Warren & Griffiths, 2003
The present study ... suggests anatomically distinct spatial (posteromedial) and object (anterolateral) processing mechanisms within PT... (p. 5803)

Barrett & Hall, 2006:
Manipulating the temporal parameters that inform changes in spatial location on one hand and pitch perception on the other produced regions of cortical activation that were distinct apart from a small region of overlap in anterior PT. Responses to pitch occurred within and anterior to HG, while responses to changes in spatial location were restricted to posterior non-primary auditory cortex in PT (p. 976)

(Pink=spatial, yellow=non-spatial, blue=overlap)
Altmann et al., 2007:
Both pattern and location changes implicated partly overlapping areas with the pattern changes additionally activating more anterior areas and the location changes implicating posterior temporal areas... This was confirmed by a contrast analysis ... that revealed significantly stronger fMRI responses ... for pattern compared to location changes within the bilateral anterior aspects of the STG and STS. ...we observed significantly stronger fMRI responses for the location compared to pattern changes within the bilateral inferior parietal lobule (IPL), in the right temporo-parietal junction (TPJ) and right anterior insula (p. 1196-1197)

(green=spatial, red=non-spatial, yellow=overlap)

I take this to mean that more than one highly competent and well-respected groups believe that there is a region within the human PT that is selective for spatial processing. (There may also be a region that is sensitive to both spatial and non-spatial signals according to these authors, but still the claim seems to be that at least some part of the PT is indeed space dedicated.) Using a different kind of non-spatial manipulation than any of these studies, i.e., adding simultaneously presented auditory objects, we showed that the "spatial" activations are not purely spatial. Either I'm seriously misreading this literature or somebody owes me $100. (Dear reviewer: I know you read this blog, so either correct me, or pay up! ;-) )

In any case, I still think Zatorre is right. There is no "spatial" area. Instead, and consistent with Zatorre et al.'s suggestion, these spatial activations may reflect auditory stream segregation, using spatial cues, rather than spatial processing per se.


C ALTMANN, C BLEDOWSKI, M WIBRAL, J KAISER (2007). Processing of location and pattern changes of natural sounds in the human auditory cortex NeuroImage, 35 (3), 1192-1200 DOI: 10.1016/j.neuroimage.2007.01.007

D BARRETT, D HALL (2006). Response preferences for “what” and “where” in human non-primary auditory cortex NeuroImage, 32 (2), 968-977 DOI: 10.1016/j.neuroimage.2006.03.050

Warren JD, Griffiths TD (2003) Distinct mechanisms for processing spatial sequences and pitch sequences in the human auditory brain. The Journal of Neuroscience 23:5799-5804.

Robert J. Zatorre, Marc Bouffard, Pierre Ahad, Pascal Belin (2002). Where is 'where' in the human auditory cortex? Nature Neuroscience, 5 (9), 905-909 DOI: 10.1038/nn904


Anonymous said...

It is not totally surprising--though still very annoying--that you received this rejection, since you are going against the orthodoxy here. There seems to be a conventional wisdom which has solidified into a kind of doctrinaire view. Once you conceptualize these pathways as "what vs where," anything that doesn't fit with that view is automatically regarded as incorrect. Same thing happened with the visual domain in earlier days, although now the view seems to be more nuanced thanks to people like Goodale and others who pushed for an alternative view, and also thanks to anatomical papers showing considerable cross talk between the two pathways.

In the auditory domain, there will possibly be a similar eventual transition towards a view that the dorsal stream can be better conceptualized as performing certain operations, and that spatial processing happens to be one that calls on these operations. At the moment, a lot of people seem to be fixated on the handy heuristic of what/where, but eventually, the proponents of that idea will have to specify how this works out in terms of the computations that these systems engage in. I think your findings in this paper speak precisely to this issue, showing that source separation may have a lot in common with spatial separation, and that is why they recruit the same system. Unless I've misunderstood your claim, this is what your data seem to show, that the "spatial" pathway is responsible for certain types of analyses, and that spatial tasks require such analyses, as does source separation. But the converse, that such analyses are necessarily spatial, does not hold.

Hang in there!

Greg Hickok said...

The funny thing was, though, that the rejection was based primarily on the idea that no one was arguing for a spatial-selective system/region. This is incorrect as the quotes and data in this post indicate. So in a way, the reviewer's claim was that no one believes the conventional what vs. where dichotomy. Oh well. Time to move on and try another journal.

I had a nice chat yesterday with John Middlebrooks who is now a faculty member here at Irvine. His view on auditory spatial processing (at least my interpretation of his view) is that it is pretty much accomplished at lower levels, like in the brainstem, and that this information can be used at the cortical level for various functions as appropriate. E.g., if it is useful for segregating objects in the acoustic signal it will be used for object-identification functions, if it is useful for guiding action such as orienting then it will be used as such in those systems. I fully agree with this kind of position.