The Irreducibility of Vision: Gestalt, Crowding and the Fundamentals of Vision

Round 1

Reviewer 1 Report

In this review paper, Dr. Herzog discusses the fundamentals of vision. It proposes that the perception of an object is subjective in the sense that objects are interpretations of the physical world. Based on a brunch of empirical experiments on crowding summarized, time-consuming recurrent processing is needed for these processes, which need to be unconscious and followed by short, discrete conscious percepts. 

I enjoyed reading this paper as it is clearly written and provides a great summary of the recent research on crowding with insightful views on the fundamentals of vision. It well explains the topic by combining philosophy, Gestalt theory, computational science, and empirical experiments. Figures are effective at illustrating complex concepts. I think this is a timely and relevant review, which will be very influential in the field.

Minor concerns: 

1. Figure 4. The position of the upper left and right faces should switch. The upper right face is normal but the upper left is odd. 
2. Isn’t it better to also provide the exact time of the presentation procedure on Panel A? In addition, the experiment only shows a time window of 330 ms. I am wondering if there is any experiment that manipulates the time window to provide evidence of how long is needed for unconscious processing. 
3. Figure 12, line 433, typo “indiividual”.
4. Unrelated to the review, our lab has also done empirical experiments on this topic using Kanizsa figure with an fMRI approach. In the two papers by Chen et al. (2019) “Tracking the completion of parts into whole objects: Retinotopic activation in response to illusory figures in the lateral occipital complex. NeuroImage”, and  Chen et al. (2021) “Feedback from lateral occipital cortex to V1/V2 triggers object completion: Evidence from functional magnetic resonance imaging and dynamic causal modeling. Human Brain Mapping”, we show direct evidence that various grouping properties within a visual scene are integrated automatically in lateral occipital complex (LOC) and support recurrent processing:  LOC first integrates inputs from multiple neurons in lower-level cortices, generating a global shape representation while more fine-graded object details are then determined via feedback to early visual areas, independently of the current task demands. 

Finally, as a junior researcher, I feel very inspired and enlighted by this paper and also honoured to give comments on it. Look forward to seeing the publication of this paper. 

Siyi Chen

Author Response

I like to thank Siyi Chen for the flowers, as we say it. Thank you for the minors points. 1. It depends on how your rotate the figure: piece-wise or the two together. This is a very good point but it would make the layout rather complicated. Yes, we did experiments to determine the window duration. They are not published yet. Indeed, this is one of the crucial questions. 3. Thank you very much. 4. Thank you very much for the references. I tried to sneak them into the ms but could find the right place because it breaks the flow of argumentation. Let me try in the proofs.

Reviewer 2 Report

1. Speak to a specific audience.  State your goal.

Herzog’s review of the “problem of vision” and its “irreducibility” may be of interest to some, but a problem I experience from the outset is that I’m not sure who the readership is intended to be.  What is the intended audience?

Is it an introduction to visual perception?  There are many of these already and this one doesn’t get us much further than previous summaries:  (1) perception is related to physical objects, (2) physical objects are not always determinative of experience, (3) even more problematic, individual results will vary, (4) physiology doesn’t help much, (5) let’s look to physics for help –oops, they are still a work in progress too, as tiny particles don’t map readily onto big masses, (6) let’s do some psychophysics to show that physics isn’t tracked well by conscious experience, (7) let’s think of the brain as an inference machine that does its best to produce momentary glimpses that get smoothed by other inferences into the grand illusion of perception.

Is it a showcase for new results on backward masking of vernier offsets (Figures 10 and 11)?  That would really deserve its own paper(s) where we could get into the details.  There is not enough here for anyone to really get serious about the implications of the data.

Is it a safe venue for an author who has contributed much empirical research to the topic of perceptual organization to tell us about his latent assumptions, historical influences of thought, and secret biases?

Any of these 3 goals would be sufficient to kick off the article and signal to readers what the intent is.  Trying to do all 3 was not satisfying to me.  There were simply too many times when I was asking myself “what is the purpose of this journey down perception road?”

 

2. Is the question posed a good one?

Why is asking what is “fundamental” about perception a good idea?  The author assumes some form of essentialism in the framing of this question. It’s not clear to me that is a good place to start.  Perhaps this starting point could be defended to readers.  Or abandoned.  Essentialism hasn’t worked out too well so far, even for those with physics envy.  Rather, we seem to be at a place in most sciences (including physics) where we accept that the world will look differently (obey different rules) at difference levels of analysis.  We have forced to accept that not all levels of analysis can be related directly to one another.  Why does the author have higher hopes for visual perception.

 

3.  What's new?

If the summary is that “…the fundamental processing units are long-lasting discrete windows, during which the ill-posed problems of vision are resolved in a recurrent and long-lasting manner [that]…is unconscious, and only its output is rendered conscious.” are we really any further along than von Helmholtz, Richard Gregory, Irving Rock?  Julian Hochberg defeated the Gestaltists at their own game in the 1970s.  What is being offered here that is more than good old-fashioned constructivism?  Unconscious inferences that are somewhat magical.  Perception is subjective but not arbitrary or random.  What’s new?

 

Author Response

1. I aimed for a broader readership but thought the article may also been interesting to vision researchers. Obviously, it did not work out for all vision researchers as reviewer's 2 report shows. Maybe the starting point is a bit bold but I believe that the position and arguments proposed are new to many in the field and outside. The argument in the last ontology section was, to the best of my knowledge never proposed. 
2. 2&3 I do not think that there are too many who agree with reviewer's 2 and my position. Most colleagues I know do not.
3. I would be happy to improve the ms but I cannot split the article in 3 as suggested.

Reviewer 3 Report

This is a valuable and interesting paper describing deep and critical scientific argumentations on a still relevant debate started from philosophical realists and the physiological approach, on one side, and continental philosophy and Gestalt approach, on the other side. The author demonstrated through phenomenal and experimental results previously published the unsolved problems deriving from both approaches and suggested that the fundamental units of perception are unconscious processing periods, followed by a conscious percept. Therefore, only the output of the unconscious processing is rendered conscious.

The manuscript presents lot of data previously collected by the author and his collaborators in distinguished papers. The general reflection, here presented, summarize  very clearly the general theory implied in most of his scientific papers. The discussion has been quite compactly presented, mainly clear, and certainly worthwhile. I occasionally lost track in all the sections.

The general theory presented by the author is certainly interesting and susceptible to agreements, critics and further developments from different approaches to visual perception. Personally I agree with the general idea proposed, but I would like that it can become a starting point for a more exiting and involving discussion within the entire Vision Science community.

MINOR SUGGESTION

In Figs. 3b-c, I do not see a square but a rectangle.

Author Response

Thank you very much for the very positive comments and for spotting that, indeed, a rectangle and not a square is seen on fig 3b,c. I corrected main text and caption.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

As far as I can tell the author has done nothing to address the three points I raised in the review:

1. The revision does not clearly state who his intended audience is, thereby providing motivation and guidance to the reader. In fact, he misconstrues the point to say “I cannot split the article in 3 as suggested” when that was never suggested.  It should not be difficult to pick a clear intended audience, so that readers can know what the intent is here.  Life is too short to be expected to read an author’s interleaved thoughts on multiple topics at the same time.
2. The revision has not defended the author’s choice to focus on what is “fundamental” in perception. In fact, it appears to be a straw argument, since he moves on to make the case that one level of analysis is not entirely reducible to another level.  This suggests to me he could begin by posing the question whether looking for “fundamentals” (strong reductionism) is a good idea or whether it might be better to look for relations between irreconcilably different levels of analysis (experience, behavior, neurons).
3. The revision fails to acknowledge the giant shoulders that this author is standing on, by referring to the views of Helmholtz, Richard Gregory, Irving Rock, and Julian Hochberg. All of them laid out views, stretching now over 150 years, that bear a striking resemblance to what the author is claiming as news.

 

Author Response

As far as I can tell the author has done nothing to address the three points I raised in the review:
1. The revision does not clearly state who his intended audience is, thereby providing motivation and guidance to the reader. In fact, he misconstrues the point to say “I cannot split the article in 3 as suggested” when that was never suggested. It should not be difficult to pick a clear intended audience, so that readers can know what the intent is here. Life is too short to be expected to read an author’s interleaved thoughts on multiple topics at the same time.

MHH: I am really not sure what to do here. The audience is whoever is interested in the topic. I think the paper should accessible to all vision community.

2. The revision has not defended the author’s choice to focus on what is “fundamental” in perception. In fact, it appears to be a straw argument, since he moves on to make the case that one level of analysis is not entirely reducible to another level. This suggests to me he could begin by posing the question whether looking for “fundamentals” (strong reductionism) is a good idea or whether it might be better to look for relations between irreconcilably different levels of analysis (experience, behavior, neurons).

MHH: I have a degree in philosophy but cannot make too much out of this. I think the paper is explicit what is fundamental in vision. Particle physics on the external side and fundamental processing periods on the brain side. Reduction is not a topic here at all.

3. The revision fails to acknowledge the giant shoulders that this author is standing on, by referring to the views of Helmholtz, Richard Gregory, Irving Rock, and Julian Hochberg. All of them laid out views, stretching now over 150 years, that bear a striking resemblance to what the author is claiming as news.

MHH: Yes, these are giants in the field. I do not think that they are claiming similar things than what I am claiming. Von Helmholtz as Gregory and the physiological approach assume that the world of objects is fundamental and perception is inference. We propose that these objects do not exist mind-independently. Rock and Hochberg assume that Gestalts are fundamental. We think they are mind-dependent. Thus, I do not see the point

Author Response File: Author Response.pdf