On 2016 Apr 09, Lydia Maniatis commented:

For comments (and author reply) on the craziness of the claim of "orientation averaging" go here: https://pubpeer.com/publications/62E7CB814BC0299FBD4726BE07EA69

Additional craziness (based on conventional visual psychophysics wisdom):

"The voluntary averaging paradigm assumes that we perceive identical samples as slightly different due to noisy sample estimates (i.e. internal noise) and therefore have to average them to find a group estimate." And the contradiction: "However, voluntarily averaging different samples is a demanding task, and we can easily imagine that observers might see no reason to average samples that appear to be identical. Indeed, we do not often perceive identically oriented Gabors as having slightly different orientations except when their signal strength is weak (with brief or noisy presentation. [No citation for this last]. Indeed, we even perceive slightly different orientations as being identical (Morgan et al 2008). This could be due to a thresholding mechanism preventing us from perceiving our internal noise...."

A few things are worth noting. First, the authors seem to be claiming both that the samples don't look identical ("we perceive identical samples as slightly different") and that they do look identical "we do not often perceive identically oriented Gabors as having slightly different orientations." Which is it? Second, the fact that identical samples look identical in no way interferes with the authors "noise" belief system. They just explain it away on the basis of "a thresholding mechanism." But the noise paradigm, aside from being arbitrary, falls at the mere hint of a logical breeze.

Some things that should be taken into consideration: Some orientations are perceived with more precision than others (vertical, horizontal); we don't perceive orientation averages (see link above for discussion); differences in orientation in a field of oriented objects are subject to pop-out effects, not averaging effects.

The casual attitude towards assumptions that I've commented on frequently is of course on display here too: "This function is typically used [so it must be right] to quantify the averaging efficiency...and, based on the averaging model, this efficiency is assumed [in psychophysics, we can assume anything we want, whether Teller (1984) https://pubpeer.com/publications/70EEEA9EF5D6A4AE003C4559D2832C likes it or not] to be the same in low noise."

How many psychophysicists can dance on the head of a pin? I'm sure there's a model for that.

Also: It seems rather strange at first that there was a condition in which people's (supposed) estimation of “average orientation” was better for the case of four patches versus one patch. We can make sense of this if we consider the authors' methods and some of the conclusions of Solomon, May and Tyler (2015) (commented on also in the link above).

First, the location of the single patch in the periphery was both brief and unpredictable. This unpredictability was designed to avoid saccades to the object.

Second, Solomon, May and Tyler (2015) concluded that the observers were “averaging” one or two patches (because there's really no such thing as an orientation averaging percept). It's a little bit of a stretch to refer to an average of a single one out of a group of patches.

Now, in the single patch condition, it takes a little time to locate and focus attention on the single relevant object. In the four patch condition, the observer could already be focussed on any of the four locations, and base their response on that, perhaps they'd also have time to attend to a second one. Since they don't really consider more than one or two anyway, knowing where the patches are going to appear is an advantage, and explains the paradox.

The proper control (assuming the experiment had been worth doing in the first place) would have been to make the four-patch condition locations unpredictable as well. But instead, the authors contrive a strained interpretation in terms of orientation averaging in one “process” vs no orientation averaging in another.

On 2016 Apr 09, Lydia Maniatis commented:

For comments (and author reply) on the craziness of the claim of "orientation averaging" go here: https://pubpeer.com/publications/62E7CB814BC0299FBD4726BE07EA69

Additional craziness (based on conventional visual psychophysics wisdom):

"The voluntary averaging paradigm assumes that we perceive identical samples as slightly different due to noisy sample estimates (i.e. internal noise) and therefore have to average them to find a group estimate." And the contradiction: "However, voluntarily averaging different samples is a demanding task, and we can easily imagine that observers might see no reason to average samples that appear to be identical. Indeed, we do not often perceive identically oriented Gabors as having slightly different orientations except when their signal strength is weak (with brief or noisy presentation. [No citation for this last]. Indeed, we even perceive slightly different orientations as being identical (Morgan et al 2008). This could be due to a thresholding mechanism preventing us from perceiving our internal noise...."

A few things are worth noting. First, the authors seem to be claiming both that the samples don't look identical ("we perceive identical samples as slightly different") and that they do look identical "we do not often perceive identically oriented Gabors as having slightly different orientations." Which is it? Second, the fact that identical samples look identical in no way interferes with the authors "noise" belief system. They just explain it away on the basis of "a thresholding mechanism." But the noise paradigm, aside from being arbitrary, falls at the mere hint of a logical breeze.

Some things that should be taken into consideration: Some orientations are perceived with more precision than others (vertical, horizontal); we don't perceive orientation averages (see link above for discussion); differences in orientation in a field of oriented objects are subject to pop-out effects, not averaging effects.

The casual attitude towards assumptions that I've commented on frequently is of course on display here too: "This function is typically used [so it must be right] to quantify the averaging efficiency...and, based on the averaging model, this efficiency is assumed [in psychophysics, we can assume anything we want, whether Teller (1984) https://pubpeer.com/publications/70EEEA9EF5D6A4AE003C4559D2832C likes it or not] to be the same in low noise."

How many psychophysicists can dance on the head of a pin? I'm sure there's a model for that.

Also: It seems rather strange at first that there was a condition in which people's (supposed) estimation of “average orientation” was better for the case of four patches versus one patch. We can make sense of this if we consider the authors' methods and some of the conclusions of Solomon, May and Tyler (2015) (commented on also in the link above).

First, the location of the single patch in the periphery was both brief and unpredictable. This unpredictability was designed to avoid saccades to the object.

Second, Solomon, May and Tyler (2015) concluded that the observers were “averaging” one or two patches (because there's really no such thing as an orientation averaging percept). It's a little bit of a stretch to refer to an average of a single one out of a group of patches.

Now, in the single patch condition, it takes a little time to locate and focus attention on the single relevant object. In the four patch condition, the observer could already be focussed on any of the four locations, and base their response on that, perhaps they'd also have time to attend to a second one. Since they don't really consider more than one or two anyway, knowing where the patches are going to appear is an advantage, and explains the paradox.

The proper control (assuming the experiment had been worth doing in the first place) would have been to make the four-patch condition locations unpredictable as well. But instead, the authors contrive a strained interpretation in terms of orientation averaging in one “process” vs no orientation averaging in another.