Comments on: All smoke and mirror neurons?

By: gob

gob — Tue, 07 Jul 2009 01:42:33 +0000

I am referring to Iacoboni’s criticism that “Caramazza managed to publish with an old trick that only PNAS allows: he handed it personally to a friend of his.”
Yes, the two tracks are different, but it is the “Communicated By” track which actually describes what Iacaboni is criticizing, and that is the one that he, not Caramazza, has used (at least for the papers in question). See Tom’s post above, and the PNAS website. If nothing else, this is a nice clinic on PNAS submission tracks, and on how not to criticize a paper.

By: Sai Emrys

Sai Emrys — Mon, 08 Jun 2009 04:12:15 +0000

gob – The paper you link to does not have the same submission method as Caramazza’s.
Caramazza’s: “Edited by Edward E. Smith, Columbia University, New York, NY, and approved April 24, 2009 (received for review February 28, 2009)”; “This article is a PNAS Direct Submission.”
Your link: “Communicated by James M. Sprague, University of Pennsylvania School of Medicine, Philadelphia, PA (received for review August 9, 2001)”; no equivalent of the ‘direct submission’
I know nothing of PNAS’ submission tracks and thus make no claim as to what these mean; I am just pointing out that they’re clearly different.

By: gob

gob — Wed, 03 Jun 2009 19:12:33 +0000

It is interesting that Dr. Iacoboni is critical of a PNAS submission track that he himself appears to have used.
http://www.pnas.org/content/98/24/13995.abstract

By: ah

ah — Sat, 30 May 2009 14:55:03 +0000

It is not clear to me why Marco is making such an important distinction between measuring action potentials and measuring local field potentials. Different neuroimaging techniques (fMRI, MEG, EEG) measure slightly different things but we are able to deal with that. All these techniques measure the correlates of neural processing (of slightly different forms), and I think we do not yet have the luxury of being fussy. None of the methods are perfect (see Friston et al, NeuroImage 1996 for limits of fMRI subtraction), so I see no reason to dismiss the repetition suppression or multivariate methods out of hand.
As for the question of whether mirror neurons really exist in humans, well, probably. I think the question of what they might mean for social cognition is more important.
Antonia Hamilton
(disclosure – I have several repetition suppression papers published).

By: Ross Cunnington

Ross Cunnington — Fri, 29 May 2009 01:29:54 +0000

This study of Caramazza, finding adaptation effects when observing and then execution actions, sounds exciting. I will be very interested to see the full paper.
As mentioned in the earlier post, we also found fMRI adaptation in the IPL when participants first executed actions then observed the same actions (Chong, et al, 2008, Current Biology) – replicated in two separate studies. Interestingly, Caramazza’s result, as reported here, seems to be the other way around.
Of course, for true mirror neurons we would expect neural adaption no matter what order actions are observed/executed. Across the two studies, it now seems that there may be evidence for this (although without seeing Caramazza’s full paper, we should be cautious about passing judgement).

By: marco iacoboni

marco iacoboni — Thu, 28 May 2009 14:59:01 +0000

In the past, there used to be three ways of submitting papers to PNAS. Direct submission did not mean ‘submitted as any other journal.’ Track II meant submitted ‘as any other journal.’ Caramazza’s paper is not defined as Track II. In the past, direct submission meant that Caramazza contacted Smith and Smith accepted to function as an Editor. ‘Communicated by’ meant that a member submit his or her own paper. However, I just checked the website http://www.pnas.org/ and the description of the submission procedure is changed a bit. It’s not clear to me if Tom is correct, but definitely there is no mention of Track II anymore, so he may be right. In this case, my apologies to Caramazza and the readers of this blog. However, the scientific problems of the paper remain. See also Sawamura et al, Neuron 2006; 49: 307-318.

By: Tom

Tom — Thu, 28 May 2009 14:33:40 +0000

iacoboni says “Caramazza managed to publish with an old trick that only PNAS allows: he handed it personally to a friend of his. The paper is basically unrefereed (this is what it means ‚ÄòEdited by…‚Äô under its title).”
This completely wrong. ‘Edited by…’ after the title just tells you the name of the member editor. All PNAS papers are peer reviewed, but there are two submission routes. If it says ‘Communicated by…’ then it means it was submitted by a member (either on their own or someone else’s behalf) who has obtained reviews before submission. This pretty much guarantees acceptance, I imagine. Alternatively, if the paper is a ‘Direct submission’ (which most are these days, including Caramazza’s last PNAS paper) this means it will have been considered by the editorial board, then fully and anonymously peer reviewed, just like with any other journal. Wait until you know what route it was submitted under (and maybe even read the thing) before you start trashing it.

By: marco iacoboni

marco iacoboni — Thu, 28 May 2009 06:12:56 +0000

What phineasgage says is wrong. Both multivariate pattern analysis and fMRI adaptation have serious intepretational limitations, see Bartels, Logothetis and Moutoussis, “fMRI and its interpretation: an illustration on directional selectivity in area V5/MT” Trends in Neurosciences 2008; 31: 444-453. In short: the fMRI signal is influenced by three factors: the action potentials of neurons (what Caramazza was after in his study), the local field potential (an index of synaptic input), and global modulatory factors. The only reasonable way for fMRI to make inferences on action potentials is to have all these parameters correlating with each other. Traditional subtraction designs most likely do that (I don’t have space here to explain why), adaptation paradigms by definition don’t do that because they change synaptic efficacy, which is invariably associated with a decoupling between action potentials and local field potential. When action potentials and local field potential do not correlate, the fMRI signal correlates with the local field potential, not the action potential. This means that Caramazza is not imaging action potentials. And guess what? Mirror neurons are defined by patterns of action potential activity. I hope this helps understanding the issues.

By: phineasgage

phineasgage — Thu, 28 May 2009 00:20:00 +0000

Of course, most of the ‘direct’ evidence for mirror neurons in humans comes from studies showing that the same fMRI voxels are activated by observed and executed actions. How interpretable is that, given that a single voxel may contain several hundred thousand neurons? Fundamentally, you need to get beyond such primitive designs if you really want to directly test the existence of mirror neurons in humans. fMRI adaptation isn’t perfect, but the technique is certainly far from ‘uninterpretable’. However, we really need adaptation designs in the single unit studies in order to avoid comparing across species, designs (adaptation vs direct response) and methods (neurophys vs fMRI) all at once.
Note that this is not the first study to use fMRI adaptation to study mirror neurons – see Chong et al in Current Biology 2008 – http://dx.doi.org/10.1016/j.cub.2008.08.068
Amusingly, these authors find the opposite pattern as Caramazza et al appear to have found – adaptation crosses over from executed to observed actions in IPL, but not the other way around. This makes little sense for a canonical mirror neuron representation. It’s a little unclear so far which area Caramazza et al studied, however.
Dinstein et al, also failed to find execution-observation cross-over using multivariate pattern analysis, so the effect is not an artefact of fMRI adaptation as such – http://www.jneurosci.org/cgi/content/abstract/28/44/11231
Basically, if you want to compare the similarity of representations in humans (and this is indeed the only way to address the mirror neuron hypothesis), your only current options are fMRI adaptation and multivariate pattern analysis. They both come with their limitations, but if neither method succeeds in producing evidence for a mirror neuron representation, one has to begin to wonder how such an elusive representation can play such a key role in nigh on all human social cognitive processing, as Iacoboni and colleagues like to suggest.
See also a recent review by Hikock (2008): http://www.mitpressjournals.org/doi/abs/10.1162/jocn.2009.21189

By: marco iacoboni

marco iacoboni — Wed, 27 May 2009 15:06:59 +0000

Caramazza‚Äôs paper is seriously flawed. The technique of fMRI adaptation seemed very promising ten years ago, but careful studies on its neurophysiological correlates have demonstrated that its findings are uninterpretable. Indeed, Caramazza‚Äôs manuscript has been around for many years and nobody wanted to publish it. Caramazza managed to publish with an old trick that only PNAS allows: he handed it personally to a friend of his. The paper is basically unrefereed (this is what it means ‚ÄòEdited by…‚Äô under its title).