Faces ARE Really Special, After All

OR: Bentin, Taylor, Rousselet, Itier, Caldara, Schyns, Jacques, & Rossion respond to Thierry et al.


from Bentin et al. (2007): Supplementary Figure 4. B. These histograms derived from Thierry et al.'s (2007) data provides the best illustration that ISPV cannot account for N170 amplitude, showing an inverse relationship between within-category picture similarity (highest for faces) and N170 amplitude (highest for cars) in the low ISPV condition and similar N170 amplitudes for faces and cars in the high ISPV condition despite lower picture similarity for faces than cars (Experiment 1).

The Phineas Gage Fan Club has noted the publication of a letter to the editor in Nature Neuroscience that's critical of the ISPV paper published by Thierry et al. (2007a) in April. What's ISPV? That's interstimulus perceptual variance, and it was a factor purportedly uncontrolled in prior studies of the N170, an electrophysiological signal thought to be selective for faces (see Are Faces Special?). The specificity of the N170 for faces (vs. other objects) is part of a larger debate on whether faces have a privileged processing status or not (the latter view being that faces are just one example of a stimulus class that requires substantial expertise in order to distinguish between similar exemplars). Then the recent NN paper added more fuel to the fire.

Briefly, in their original article, Thierry et al. argued that previous studies of the N170 component did not adequately control for variability across stimulus classes, i.e., face stimuli were much more similar to each other than the non-face stimuli. When ISPV was more tightly controlled, the selectivity of N170 for faces (vs. cars) went away.

Then, Bruno Rossion, from the Face Categorisation Lab in Belgium, commented that these experiments were kind of... bad and (ironically) failed to adequately control for interstimulus variance, among other things (see also C'mon, Aren't Faces REALLY Special?). The correspondence from Bentin, Rossion, and six more authors (along with a reply by Thierry et al.) appears in the July issue of NN. Johan has done a nice job summarizing the debate in N170 face controversy continues, so I'll only add a few choice quotes:

Here we [Bentin et al.] demonstrate that ISPV was actually controlled in many studies, yet the N170 effect remained conspicuous (Fig. 1 and Supplementary Figures 1,2,3 online). Evidently, Thierry and colleagues' claim is wrong and misleading.

Continuing merrily along:

In addition to their factual error, they failed to note the striking contradiction between their hypothesis and the existing literature. Most notable are the larger N170 for inverted than for upright faces, the larger N170 for upright than for inverted Mooney faces..., [etc.] All these modulations of N170 are robust despite identical stimuli in different conditions, hence identical ISPV.

My favorite, however, is the title for Supplementary Figure 4: "Self-contradictions in Thierry et al.":

Supplementary Figure 4. A. While Thierry et al. claimed to have controlled for inter-stimulus similarity between pictures of faces and objects (exact values not reported), their own data suggest otherwise.

And the debate continues...

References

Bentin S, Taylor MJ, Rousselet GA, Itier RJ, Caldara R, Schyns P, Jacques C, Rossion B. (2007). Controlling interstimulus perceptual variance does not abolish N170 face sensitivity. Nature Neurosci. 10:801-802.

[apparently, the original title for this submission was "Much ado about nothing: controlling interstimulus perceptual variance does not abolish N170 face sensitivity," as listed in the publications of the Face Categorisation Lab. I would imagine the editors at NN did not take too kindly to the "much ado" part...]

Thierry G, Martin CD, Downing P, Pegna AJ. (2007a). Controlling for interstimulus perceptual variance abolishes N170 face selectivity. Nature Neurosci. 10:505-511.

Thierry G, Martin CD, Downing P, Pegna AJ. (2007b). Is the N170 sensitive to the human face or to several intertwined perceptual and conceptual factors? Nature Neurosci. 10:802-803.