While the authors present compelling arguments to support that iGBR is an artifact of microsaccades, the role of microsaccades in visual information processing, especially their possible functional relationship with intrinsic neuronal oscillations remain to be elusive. Although the main result is extremely important and clears the mist around the "cognitive-gamma", the most interesting aspect is whether microsaccades contribute to information sampling or not, and whether this sampling is aligned to any neuronal oscillations or not. The study is eye opening but cries for a systematic investigation of microsaccades. For example, an analysis of microsaccades in response to a diverse set of visual stimuli, applying spectral analysis and c ... read more

This paper quite convincingly shows that the induced gamma band response recorded from scalp-EEG is an artifact of microscaades. I highly recommend this paper to everybody who has read, cited, or published (!) any papers reporting gamma band response using scalp-EEG. This paper is likely to be a must read in this research domain in the near future. The paper is very clearly written. I also recommend to read Pascal Fries' commentary in the same issue ("Finding Gamma" Pascal Fries,1,2,∗ René Scheeringa,1 and Robert Oostenveld1) and the replies from the people who published many papers reporting gamma band response with scalp-EEG (http://www.neuron.org/content/article/comments?uid=PIIS0896627308003012).

using (express) saccades as a response modality, this paper shows that humans can categorize previously unseen natural scene pictures as those that contain an animal or not, starting at 120 msec.

detailed image statistics are given in the paper. this is very compelling.

given 25 msec for preparing a saccade, how the visual system can do such a feat?

It seems the computation in IT areas is not necessary for this task (!).

Arguably, this is the strongest evidence that supports a power of feedforward-only processing in vision.

This article provides further support for fast dorsal stream / slow ventral stream ("framing" or "active black-boards") feedback theories, using CSD analysis.
It also adds important 'cortical layer' dimension to the analysis of latencies.

Thanks for the helpful comments. So, is it correct to say that, if we inject two independent (uncorrelated) Poissonian spike trains into the two neurons, then it does not matter how many spikes are we sampling, the cross-correlation will be near zero? Consequently, neither the input nor the output firing rate does matter. However, if the two Poissonians share correlated spikes, then, because of the non-linearity kicks in, the output cross-correlation will scale not only with the injected correlation, but also with the firing rate.

According to elementary statistics, the estimated correlation coefficient of a population increases with the sample size. When applying this principle to cross-correlograms, we can consider firing rate as sample size, thus, for a cell pair with a fixed effective connectivity higher firing rates will give us higher cross-correlations. Unless something is special about cross-correlograms, I don't see the need for an experimental verification on this. If someone could explain it to me why this is not obvious, I would appreciate.