Does Vergence Affect Perceived Size?

Round 1

Reviewer 1 Report

In this psychophysical study, Linton introduced a new apparatus that allows to control for vergence, accommodation, retinal image size and pictorial depth cues to investigate size constancy mechanisms under conditions in which participants are unaware of changes in absolute distance. Specifically, the present study examined the role of vergence in size constancy, as there is plenty of evidence in the literature supporting the idea that vergence eye movements are important for size perception at near viewing distances. The author found that when the above-mentioned confounding variables were controlled for, there was no evidence whatsoever of any contribution of vergence to perceived size and concluded that size constancy must rely exclusively on cognitive knowledge about changes in viewing distance.    

This is an interesting and cleverly designed study. It’s well-written in general. Although the conclusions are intriguing and novel, I have several major concerns on a theoretical point of view which the author should consider carefully. Importantly, some of the claims throughout the paper should be toned down in the light of existing evidence and the fact that the experimental set-up used here was highly artificial.

1. The author underestimates the contribution of pictorial cues to size constancy. He states that ‘…pictorial cues are neither necessary nor sufficient for size constancy’. Whilst I agree that binocular cues and other types of monocular cues are more relevant for the operation of size constancy in general, there are conditions in which pictorial cues become more prominent than other sources of depth cues. In particular, when subjects estimate objects at great distances their perceptual judgments are determined by pictorial cues as at such distances binocular cues are less effective.
2. A hypothesis exclusively based on cognitive factors cannot explain why size constancy has been reported in early infancy (e.g. Bower 1965; Granrud, 1987; Slater et al, 1990), unless top-down modulations are already operative at birth which is highly improbable. Interestingly, vergence and accommodation are sufficiently developed by the age of four months (e.g. Banks, 1983; Haynes et al, 1965). Hence, it is possible that these oculomotor signals are contributing to the basic mechanisms of size constancy observed in infants.
3. Whilst the new apparatus proposed here has the virtue to control for different sources of information (retinal size, vergence, accommodation), it is rather artificial and does not allow to investigate size constancy as it operates in the real world. This should be taken into account when making claims and drawing conclusions, as these are limited to the experimental conditions tested here. As such, the generalizability of the findings can be questioned. In a similar manner, conclusions about size constancy drawn from virtual reality studies should be limited to the virtual environments as, for instance, differences in depth perception exists between real and virtual environments (e.g. Knapp and Loomis, 2004; Willemsen, et al 2009; Naceri, et al 2010).
4. Related somehow to the previous point; the apparatus used here required a decoupling between vergence and accommodative responses, something that never happens under natural viewing conditions. Contrary to the author’s position (see discussion), I strongly believe based on previous evidence, including my work (Holoway and Boring, 1941; Chen et al, 2018; Chen et al, 2019; Sperandio et al, 2009), that accommodation is one of the most powerful sources of distance information for size constancy at close distances. In fact, when I used a 1-mm pinhole to remove accommodation in the past and tried it on myself, I was surprised to see how this affected my perceptual experience of size. I was perceiving mainly my retinal image size, despite the fact that I knew that the object was changing in distance. Therefore, I wonder if the lack of modulation in perceived size reported here was due to the fact that the only information available to the visual system was actually the retinal image size as oculomotor cues were completely ‘knocked out’ by the set-up and as such could not contribute at all to perceived size.
5. I would like to know more about the perceptual experience of the participants during the task. It has been shown that stereoscopic displays, in which the vergence-accommodation coupling does not operate normally, may result in perceived visual discomfort, such as eyestrain, blurred vision, double vision and even motion sickness (e.g. Yang & Sheedy, 2011). Did your subjects report any of these problems?
6. In methods, it is mentioned that ‘participants were asked to confirm that they saw a single fused target…’ (page 4 line 158). What did you do when participants reported difficulties with focusing on the target or double vision?
7. Relevant literature related to motion parallax has been overlooked (Point 1 - Discussion). Not everybody will agree with the statement that motion parallax is ineffective as an absolute indicator of distance information (e.g. Ferris, 1972; Ono et al., 1986). There is also evidence that motion parallax is sufficient to drive vergence eye movements under both monocular and binocular conditions (Frey and Ringach, 2011).

Minor points:

• Page 6, Line 205: question mark should be removed.
• Writing style: I would avoid the use of verb contractions.

REFERENCES:

Banks, M.S. (1983). The development of ocular accommodation during early infancy. Child Development, 51, 646-666.

Bower, T. G. R. (1965). Stimulus variables determining space perception in infants. Science, 149, 88 – 89.

Chen J, Sperandio I, & Goodale M. (2018). Proprioceptive distance cues restore perfect size constancy in grasping but not perception when vision is limited. Current Biology, 28(6), 927-932.

Chen J, Sperandio I, Henry MJ, & Goodale M. (2019). Changing the real viewing distance reveals the temporal evolution of size constancy in visual cortex. Current Biology, 23(13):2237-2243.

Ferris SH (1972) Motion parallax and absolute distance. J Exp Psychol 95:258–263.

Frey, J., & Ringach, D. L. (2011). Binocular eye movements evoked by self-induced motion parallax. Journal of Neuroscience, 31(47), 17069-17073.

Granrud, C. E. (1987). Size constancy in newborn human infants. Investigative Ophthalmology and Visual Science, 28, (Supplement), 5.

Haynes, H., White, B.L. & Held, R. (1965). Visual accommodation in human infants. Science, 148, 528-530

Holoway, A. H. and Boring, E. G. (1941). Determinants of apparent visual size with distance variant, Am. J. Psychol. 54, 21–37.

KNAPP, J. M., & LOOMIS, J. M. (2004). Limited field of view of headmounted displays is not the cause of distance underestimation in virtual environments. Presence: Teleoperators & Virtual Environments,13, 572–577.

Naceri, A., Chellali, R., Dionnet, F., & Toma, S. (2010). Depth perception within virtual environments: comparison between two display technologies. International Journ. on Advances in Intelligent Systems, 3.

Ono ME, Rivest J, Ono H (1986) Depth perception as a function of motion parallax and absolute-distance information. J Exp Psychol Hum Percept Perform 12:331–337.

Slater, A., Mattock, A., & Brown, E. (1990). Size constancy at birth: newborn infants’ responses to retinal and real size. Journal of Experimental Child Psychology, 49(2), 314–22.  

Sperandio I, Savazzi S, Gregory RL & Marzi CA (2009). Visual reaction time and size constancy. Perception, 38, 1601-1609.

Willemsen, P., Colton, M. B., Creem-Regehr, S. H., & Thompson, W. B. (2009). The effects of head-mounted display mechanical properties and field of view on distance judgments in virtual environments. ACM Transactions on Applied Perception (TAP), 6(2), 1-14.

Yang, S., & Sheedy, J. E. (2011, February). Effects of vergence and accommodative responses on viewer's comfort in viewing 3D stimuli. In Stereoscopic Displays and Applications XXII (Vol. 7863, p. 78630Q). International Society for Optics and Photonics.

Author Response

Dear Reviewer 1,

Thank you for your valuable and extensive comments. I’m attaching the following pdf that documents all of the revisions to the manuscript, that take into account the 2 editors’ and 3 reviewers’ comments on the manuscript.

Thanks ever so much,

Paul

Author Response File: Author Response.pdf

Reviewer 2 Report

I like this paper and it’s certainly a very interesting result. I’d suggest rewording the title to “Vergence Does Not Affect Perceived Size”, reflecting the fact that vergence is the only type of eye movement which has been postulated to affect perceived (distance and thus) size, and the only one investigated here.

Before I accept that conclusion fully, I’d like the author to go through the existing evidence adduced in favour of vergence as a size/distance cue, and explain his account of it (see below for some comments on that).

The data is very compelling that vergence is not a useful cue in this context (50cm to 25cm (2D) in 5s), but we should perhaps leave open the possibility that vergence might be useful in other situations, e.g. a larger or more rapid change. And I guess given the idiosyncratic asymmetry between convergence and divergence eye movements, it’s possible that divergence movements might have elicited an effect even though convergence did not.

However, I do buy the overall conclusion, which is a significant further weakening of the evidence supporting vergence as a distance cue, to add to this author’s paper from earlier in the year.

__Detailed comments__

I wasn’t clear how the conclusion that vergence does not influence perceived size fitted with the other evidence for vergence micropsia, reviewed in the introduction.  I understand that for  (4), Taylor illusion, the author’s account is that we perceive the afterimage as being  on our hand, know we have brought our hand closer, and thus perceive the afterimage as closer/smaller. But for examples like the wallpaper illusion, what is the author’s proposed explanation? We integrate the disparity cue which drives the vergence change, and it is this that informs us that the object is nearer and thus smaller, or… ?

I wonder if in this, along with telestereoscopic viewing, vertical disparity may play a key role. Vertical disparity in the periphery (>20, 30 deg eccentricity) varies strongly with vergence, and is thus a strong potential retinal cue to vergence in natural viewing. It is of course not present in the experiment of this paper (because the stimulus is small & central, and the rendering is carefully designed to remove VD cues). I’d like the author to discuss this possibility. It’s obliquely alluded to and dismissed: “binocular disparity is typically thought of as being a merely relative depth cue outside very limited circumstances (objects taking up at least 20° of the visual field; Rogers & Bradshaw, 1995)” but this is a misunderstanding. You need to be able to extract vertical disparity over at least 20deg of the visual field, ie there must be some kind of texture, but there is no need for any single object to be any particular size. We also read “one exception might be vertical disparities, but Trotter et al., 1992 proceed on the basis that they are ineffective, and the subsequent evidence in favour is equivocal at best: see Cumming et al., 1991; Sobel & Collett, 1991; Rogers & Bradshaw, 1995”. My reading of it is that studies which used small stimuli did not find an effect; studies which used large stimuli mainly have (Rogers &Bradshaw 1995, Bradshaw et al 1996 Vis Res 36, Vienne et al 2019 iPerception  ). One issue is that studies which used small stimuli may have had a conflict between the manipulated vertical disparity in their stimuli, and veridical vertical disparity due to other surfaces visible in the periphery.

So, I’d like to suggest that the author adds peripheral vertical disparity to his list of possible distance cues. If we accept that vergence is ruled out, it seems to me that vertical disparity could explain evidence (1)-(3). I also like the suggestion in Linton (2018), that vergence, and thus distance scaling, could be derived from the distribution of horizontal disparities in a scene. I think that’s a very interesting suggestion, and could account for the telestereoscopy illusion (3). I think that’s worth mentioning here, not  just referring the reader to the biorxiv paper. This could, presumably, be tested in a VR display, in which one could manipulate (a) physical vergence, (b) distribution of horizontal disparities, (c) pattern of vertical disparities across the scene, and see which if any produced a change in apparent scale.

In terms of the neurophysiology for vertical disparity, this is nicely reviewed by Sprague et al 2015 (Science Advances). The brain does contain neurons with preferences for non-zero vertical disparity (their Fig 8). Of course this doesn’t prove these are used for perception; they could be to control vertical vergence, for example.

There are some strange citations in the discussion of the evidence that vergence is represented in V1. Cumming & Parker 1999 failed to replicate this effect, possibly because their range of vergence was too small (discussed in their paper and in Trotter et al 2004)), or (as CP99 suggest) because of distance-dependent fixation disparities in the Trotter papers (which did not measure vergence).  Citing Masson et al 1997 as “further evidence for the vergence modulation of V1” is presumably just a mistake – this study didn’t look at V1, it measured eye movements. The Cottereau reference seems to be an ECVP abstract; the link given didn’t work and I couldn’t find it. Dobbins et al is very  poorly controlled: their results could be explained by a change in disparity (i.e. the animal did not fixate equally accurately at all distances, resulting in a change in retinal disparity which modulates the size tuning). The Richards 1968 paper postulating spatial remapping in LGN is very early and speculative – has it been supported by subsequent work?

__Minor points__

I couldn’t find a  clear description of the stimulus. Was it the target shown in Fig 1? Was 3deg its width and height? (before any size change).

In the author’s review of the history of vergence, it might be worth mentioning that Smith (1738) seems to be unusual in rejecting a role for vergence: “138: Nor is distance suggested by feeling the turn of the eyes in widening or contracting the interval between pupils, when we direct them to different places", although his reasoning – that vergence depends on azimuth as well as distance – is dodgy.

Can we have a clearer diagram showing how the rotating metal plates occlude the other eye’s stimulus? A photo of the set-up would also be nice!

“2. Triangulation Cues”: you could add defocus to this list.

“If there was no vergence size constancy, then participants would be at chance in determining whether there was a size change when the target’s physical size didn’t change.” – better, “the direction of the size change” rather than “whether” there was one (since that was the task).

“participants were asked to confirm that they saw a single fused target when they first entered the apparatus, and to report if the target ever went double”. If participants were naïve, is it possible any of them closed an eye? As I understand it, there were no Nonius lines or anything.

As this is a single-author paper, I wondered whether the first person singular would be preferable to the plural? Or is there someone other than the author encompassed in the “we”, “our” etc?

Author Response

Dear Reviewer 2,

Thank you for your valuable and extensive comments. I’m attaching the following pdf that documents all of the revisions to the manuscript, that take into account the 2 editors’ and 3 reviewers’ comments on the manuscript.

Thanks ever so much,

Paul

Author Response File: Author Response.pdf

Reviewer 3 Report

Linton investigated the influence of vergence on the perceived size of targets. By independently manipulating the size and the binocular disparity of a target, the author tried to disentangle vergence from changes in the retinal image. The results suggest that changing vergence does not affect the perceived size of targets.

The study addresses an interesting research question and the experimental setup is clearly innovative, but unfortunately lacks a necessary control of the manipulation of vergence. The conclusions from the results (if they are accurate) could also be defined more precisely.

My main concern is that vergence has not been measured. This is especially problematic, because the author tried to manipulate vergence by “sub-threshold binocular disparities” that are invisible to the subject. In the worst case, the absence of an effect of vergence on the size judgments could reflect a failure to manipulate vergence. I think without an appropriate measure of vergence, the experiment remains inconclusive.

My second major concern is about the interpretation of the results (if they are accurate). Based on the literature review and the current results it would seem that visual and oculomotor cues do not provide information about absolute depth. The author suggests that cognitive signals are responsible for the Taylor illusion instead. It wasn’t clear to me if the author wants to imply that size constancy in general is a cognitive rather than a perceptual phenomenon. If this is the case, how would one explain the perceptual effects in the Ames room, where two persons are being perceived as different in size, even when one knows them to have the same size? 

Minor comments:

In the introduction, the author describes four phenomena, indicating that vergence plays a role for size constancy. I think it might be useful to come back to those phenomena and reevaluate them under the light of the new empirical evidence.

Line 132: “changes in retinal image” is a bit unprecise. What do you mean by that specifically? Disparity? Size?

Line 150: From the description of the methods, it wasn’t clear to me how the targets generated subthreshold binocular disparities and why this is different to LEDs.

Figure 4A: The legend seems to be incorrect.

Author Response

Dear Reviewer 3,

Thank you for your valuable and extensive comments. I’m attaching the following pdf that documents all of the revisions to the manuscript, that take into account the 2 editors’ and 3 reviewers’ comments on the manuscript.

Thanks ever so much,

Paul

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Thank you Paul for the extensive changes you have made to the manuscript. I believe it has improved greatly since its first submission. I really like the change of tone in general. You have properly addressed all of my comments and I do not have any remaining concerns. 

Author Response

That is absolutely wonderful. Thank you so much for all your help on the manuscript, it has greatly improved it, and things are a lot clearer now. I really appreciate all your help, and the time and effort you have put into my work. Thanks ever so much, Paul

Reviewer 3 Report

The revised version of the manuscript is improved a lot. The setup and the experimental paradigm are described in more detail and the arguments in the introduction and the discussion are much easier to follow.

I think the discussion about eye tracking accuracy (line 410ff) is a bit misleading. Hooge et al., 2019 and Drewes et al. 2014 specifically investigated the artifact in measured gaze position that is induced by changes in pupil size. Both papers actually discuss ways to eliminate (constant lighting conditions) or compensate (separate calibrations for different pupil sizes) this artifact. Hence, I don't think that the eye tracking accuracy is too low in principle.

Apart from that, I spotted only two typos:

Line 559: "Equally the left eye..." => "Equally the left coin..."

Line 668: Superfluous it.

Author Response

Thanks so much for drawing that aspect of Drewes et al. (2014) to my attention. You’re quite right, I took away the headline figure (error of 2.5°) without considering how this could be ameliorated by calibrating under different luminences. I’ve now added the following passage:

“Admittedly, there have been attempts to improve eye tracking accuracy. First, since eye tracking errors are caused by changes in luminance affecting pupil size, one approach is to keep luminance fixed. But it’s not clear that controlling luminance is the right approach for our experiment. If we controlled for luminance in our experiment, as the target got larger it would have to get darker, and as the target got smaller it would have to get brighter. Second, Drewes et al. (2014) found that even if luminance is not controlled, the error can be reduced (down from 2.5° to 0.5°) by calibrating the eye tracker at different luminances. However, this still fails to control for non-luminance effects on pupil size, including changes in pupil size with vergence (the ‘near triad’: Balaban et al., 2018), and cognitive processes (Naber & Nakayama, 2013), although these are liable to be smaller than the errors induced by changes in luminance.”

I hope this addresses the concern, and thank you so much for all your help with my manuscript. It has greatly improved it, and things are a lot clearer now. I really appreciate all your help, and the time and effort you have put into my work. Thanks ever so much, Paul