On 2017 Mar 09, Lydia Maniatis commented:

First, I apologize for the error vis a vis the open-loop experiment.

With respect to "very standard procedures:" Vision science is riddled with references to "standard," "popular," "traditional" "common" "widely used" procedures that have no theoretical rationale. "It is considered safe" is also not a rationale.

With respect to fitting, you calculate r-squared by fitting data in the context of very specific conditions whose selection is without a clear rationale, and thus it is very likely that conditions similar in principle but different in detail would yield different results. For example, you use Gabors, which are widely used but seem to be based on the idea that the visual process performs a Fourier analysis - and a local one at that - for which there are no arguments in favor.

Your findings don't warrant any substantial conclusion. You claim in the paper that: “we have shown that eye and hand movements made toward identical targets can end up in different locations, revealing that they are based on different spatial information.” Only the former claim is true.

Your prior discussion reveals that the conclusions couldn't be more speculative and go far beyond the data : “This difference between hand movements and saccades might reflect the different functional specificity of the two systems… One interpretation of the current results is that there are two distinct spatial maps, or spatial representations, of the visual world.”

Your arguments in favor of this explanation are peppered with casual assumptions: "...the priority for the saccade system might be to shift the visual axis toward the target as fast as possible with little cost for small foveating errors. If integrating past sensory signals with the current input increases processing time (Greenwald, Knill, & Saunders, 2005), the saccadic system might prefer to use current input and maximize the speed of the eye movement. For hand movements instead, a small error might make the hand miss its target with potentially large behavioral costs."

Might + might + might + might (eleven in all in the discussion) means that the effect that you report is far too limited in its implications to warrant the claim you make, quite unequivocally, in your title. Most of your "mights" are not currently testable, and almost certainly represent an overly simplistic view of a system of which we have only the crudest understanding at the neural level. The meaning of the term "sensory signals" also needs clarification, as it also implies a misunderstanding of the nature of perceptual processes.

On 2017 Mar 07, Matteo Lisi commented:

Thanks for the interest in our study.

I invite you to read more carefully the article, the experiment that you indicate as "post-hoc"is actually the open-loop pointing experiment: the details of the method are described at page 4-5, and the results are reported at page 7.

The truncation of response latencies is a very standard procedure, to prevent extreme spurious RTs (e.g. due to anticipation or attention lapses) from being included in the analysis. While there is no universal agreement on the ideal procedure for selecting cut-offs criteria, it is considered safe to use extreme cut-offs that result in the exclusion of only a tiny fraction of trials (<0.5%); see for example the recommendations by Ulrich R, 1994, page 69.

I am confused by your comment regarding the "fitting". The statistical model used in the analysis is a standard multivariate linear regression, with the x, y location of the response as dependent variables. It doesn't require any particular assumption, other than the usual assumptions of all linear models (such as independence of errors, homoscedasticity, normally-distributed residuals), which were not violated in our dataset. These are the same assumptions required also by other common linear models such as simple linear regression and ANOVA.

On 2017 Mar 07, Lydia Maniatis commented:

The discussion of this paper refers to a post hoc experiment whose results apparently used in the analysis of the reported results, but gives details of neither methods nor results, nor any citation. This seems oddly casual:

"To test this hypothesis, we repeated the pointing task in a condition in which vision was blocked during the execution of the movement by means of shutter glasses (open-loop hand pointing), making it impossible to use visual feedback for online correction of the hand movement. The results of this experiment replicated those of the experiment with normal pointing; the only difference was a moderate increase in the variability of finger landing positions, which is reflected in the decreased r2 values of the model used to analyze pointing locations in the open-loop pointing condition with respect to the “normal” pointing condition (see also Figures 1D, E and 2)."

As is usual but hard to understand, an author made up a large proportion of the subjects (1/6), confusing the issue of whether naivete is or is not an important condition to control: "all [subjects] except the author were naïve to the specific purpose of the experiments."

And this:

"In the experiment involving saccades, we excluded trials with latency less than 100 ms or longer than 600 ms (0.36% of total trials); the average latency of the remaining trials was 279.89 ms (SD = 45.88 ms). In the experiment involving pointing, we excluded trials in which the total response time (i.e., the interval between the presentation of the target and the recording of a touch response on the tactile screen) was longer than 3 s (normal pointing: 0.45% of total trials; open-loop pointing: 0.26% of total trials). The average response time in the remaining trials was 1213.53 ms (SD = 351.61 ms) for the experiment with normal pointing and 1004.70 ms (SD = 209.83 ms) for the experiment with open-loop pointing."

It's not clear if this was a post hoc decision, or, whether planned or not, what was the rationale.

As usual, there was a lot of fitting, using assumptions whose rationale is also not clear.

On 2017 Mar 07, Lydia Maniatis commented:

The discussion of this paper refers to a post hoc experiment whose results apparently used in the analysis of the reported results, but gives details of neither methods nor results, nor any citation. This seems oddly casual:

"To test this hypothesis, we repeated the pointing task in a condition in which vision was blocked during the execution of the movement by means of shutter glasses (open-loop hand pointing), making it impossible to use visual feedback for online correction of the hand movement. The results of this experiment replicated those of the experiment with normal pointing; the only difference was a moderate increase in the variability of finger landing positions, which is reflected in the decreased r2 values of the model used to analyze pointing locations in the open-loop pointing condition with respect to the “normal” pointing condition (see also Figures 1D, E and 2)."

As is usual but hard to understand, an author made up a large proportion of the subjects (1/6), confusing the issue of whether naivete is or is not an important condition to control: "all [subjects] except the author were naïve to the specific purpose of the experiments."

And this:

"In the experiment involving saccades, we excluded trials with latency less than 100 ms or longer than 600 ms (0.36% of total trials); the average latency of the remaining trials was 279.89 ms (SD = 45.88 ms). In the experiment involving pointing, we excluded trials in which the total response time (i.e., the interval between the presentation of the target and the recording of a touch response on the tactile screen) was longer than 3 s (normal pointing: 0.45% of total trials; open-loop pointing: 0.26% of total trials). The average response time in the remaining trials was 1213.53 ms (SD = 351.61 ms) for the experiment with normal pointing and 1004.70 ms (SD = 209.83 ms) for the experiment with open-loop pointing."

It's not clear if this was a post hoc decision, or, whether planned or not, what was the rationale.

As usual, there was a lot of fitting, using assumptions whose rationale is also not clear.

On 2017 Mar 09, Lydia Maniatis commented:

First, I apologize for the error vis a vis the open-loop experiment.

With respect to "very standard procedures:" Vision science is riddled with references to "standard," "popular," "traditional" "common" "widely used" procedures that have no theoretical rationale. "It is considered safe" is also not a rationale.

With respect to fitting, you calculate r-squared by fitting data in the context of very specific conditions whose selection is without a clear rationale, and thus it is very likely that conditions similar in principle but different in detail would yield different results. For example, you use Gabors, which are widely used but seem to be based on the idea that the visual process performs a Fourier analysis - and a local one at that - for which there are no arguments in favor.

Your findings don't warrant any substantial conclusion. You claim in the paper that: “we have shown that eye and hand movements made toward identical targets can end up in different locations, revealing that they are based on different spatial information.” Only the former claim is true.

Your prior discussion reveals that the conclusions couldn't be more speculative and go far beyond the data : “This difference between hand movements and saccades might reflect the different functional specificity of the two systems… One interpretation of the current results is that there are two distinct spatial maps, or spatial representations, of the visual world.”

Your arguments in favor of this explanation are peppered with casual assumptions: "...the priority for the saccade system might be to shift the visual axis toward the target as fast as possible with little cost for small foveating errors. If integrating past sensory signals with the current input increases processing time (Greenwald, Knill, & Saunders, 2005), the saccadic system might prefer to use current input and maximize the speed of the eye movement. For hand movements instead, a small error might make the hand miss its target with potentially large behavioral costs."

Might + might + might + might (eleven in all in the discussion) means that the effect that you report is far too limited in its implications to warrant the claim you make, quite unequivocally, in your title. Most of your "mights" are not currently testable, and almost certainly represent an overly simplistic view of a system of which we have only the crudest understanding at the neural level. The meaning of the term "sensory signals" also needs clarification, as it also implies a misunderstanding of the nature of perceptual processes.