A Thickness Illusion: Horizontal Is Perceived as Thicker than Vertical

Round  1

Reviewer 1 Report

This paper describes a well known, but not well documented, illusion, namely that horizontal lines appear thicker than vertical lines. In experiment 1 this illusion is established. In experiment 2 data is presented that links this illusion to the horizontal/vertical illusion, in which vertical lines look longer than horizontal ones. I think this is a good paper, that makes a valuable contribution. I only have three rather minor comments.

1) The set-up and results of Experiment 2 are not extremely clear to me, this could be described in a more accessible way. I would also like to see an "average" plot in which the results of the vertical and horizontal frame are averaged to be able to compare the results to those of E1. As it stands now, am I correct in interpreting that the frame determines whether there is a horizontal/vertical illusion and the illusion that horizontal lines look thicker than vertical lines?

2) The introduction is heavily reliant on the printing world, where letters are adjusted to compensate for the illusion that horizontal lines are thicker. By how much do they compensate? And do they compensate to the same extent as the size of the illusion found here?

3) I would like to see some discussion of other visual illusions and how they may relate to the current illusion. I've worked a lot on visual illusions presumably reliant on expectations about the visual field based on the higher-fidelity central visual signal, which then leads to filling in of peripheral perception (e.g. The uniformity Illusion, by Otten, Pinto ... Kanai, 2017; Ramachandran's scotoma; Motion Induced Blindness; Troxler fading. 

Is it possible that some similar mechanisms play a role here? For instance, our visual perception of vertical or horizontal orientations are extremely accurate, whereas our thickness perception is poorer. Moreover, perhaps vertically oriented objects may generally (in the real world) be somewhat thinner. Could it be that the visual system takes this knowledge into account and therefore adjusts our visual percept a bit? Would this lead to the prediction that under conditions of poor visibility (e.g. lines further out into the periphery), this illusion would be a lot stronger?

It would be great if the authors can speculate a bit on this.

Author Response

This paper describes a well known, but not well documented, illusion, namely that horizontal lines appear thicker than vertical lines. In experiment 1 this illusion is established. In experiment 2 data is presented that links this illusion to the horizontal/vertical illusion, in which vertical lines look longer than horizontal ones. I think this is a good paper, that makes a valuable contribution. I only have three rather minor comments.

Thank you very much for your kind comments. Please see our response below.

1) The set-up and results of Experiment 2 are not extremely clear to me, this could be described in a more accessible way. I would also like to see an "average" plot in which the results of the vertical and horizontal frame are averaged to be able to compare the results to those of E1. As it stands now, am I correct in interpreting that the frame determines whether there is a horizontal/vertical illusion and the illusion that horizontal lines look thicker than vertical lines?

We revised the Method section of Experiment 2 and changed Figure 4 for clarification. We decided not to plot the average, as we believe this would increase the complexity of the graph (or add a not so informative panel), plus we already mention the mean PSE on page 9 for comparison with Experiment 1. We have also added a few lines to the ‘results’ section which hopefully clarifies that both illusions do exist in both the vertical and the horizontal frame condition. The frame only influenced the magnitude of the illusions.

 2) The introduction is heavily reliant on the printing world, where letters are adjusted to compensate for the illusion that horizontal lines are thicker. By how much do they compensate? And do they compensate to the same extent as the size of the illusion found here?

This is an interesting question. We now briefly discuss the impact of the thickness illusion on type design on page 10 of the Discussion section. Please note that type designers are generally aware of the illusion, and take it into account in their design. However, they are often not striving for equal perceived thickness in their design, so any bias in a particular font type is not necessarily informative.

 3) I would like to see some discussion of other visual illusions and how they may relate to the current illusion. I've worked a lot on visual illusions presumably reliant on expectations about the visual field based on the higher-fidelity central visual signal, which then leads to filling in of peripheral perception (e.g. The uniformity Illusion, by Otten, Pinto ... Kanai, 2017; Ramachandran's scotoma; Motion Induced Blindness; Troxler fading. 

Is it possible that some similar mechanisms play a role here? For instance, our visual perception of vertical or horizontal orientations are extremely accurate, whereas our thickness perception is poorer. Moreover, perhaps vertically oriented objects may generally (in the real world) be somewhat thinner. Could it be that the visual system takes this knowledge into account and therefore adjusts our visual percept a bit? Would this lead to the prediction that under conditions of poor visibility (e.g. lines further out into the periphery), this illusion would be a lot stronger?

It would be great if the authors can speculate a bit on this.

While we understand the urge for relating our results to different illusions, we believe it would currently be beyond the scope of our study. Intuitively, we believe that the peripheral filling in based on central signals does not play a direct role here, but more research would be needed to say anything that is more than a wild guess. The same would go for other well-known illusions. The point about the visual system and the real world seems similar to the one made by Howe and Purves, which we discuss extensively on page 10, but whether their account also applies to thickness perception is currently unknown.

Reviewer 2 Report

Reading this paper was a pleasure.  The finding is interesting and novel, and it is theoretically and methodologically sound.  I strongly recommend that it be accepted, and I invite the authors to consider the following comments:

Lines 29-31:  It took me a moment to realize that the sentence on these lines refers to Figure 1A (to the unrotated version).  Perhaps you can make this clearer?

Figure 2, panel 2:  I found the OR to be confusing (it seemed to suggest that one of two pairs of stimuli was presented).  It might be better if you put text at the top of the panel that reads “One of the following:”

Line 84:  It’s a little strange that the paper starts out  referring to ‘participants' and switches to ‘subjects’.  This is obviously just an aesthetic comment.

Figure 4, panel 1:  I understand the impulse to depict exactly what subjects saw, but this text is too small to read.  Can you make it larger in the figure, and maybe write “‘Length' or 'Thickness'".  I don’t think this would confuse anybody if you report the actual size of the text in the methods section.

Lines 317-318:  Do they really provide no evidence on their own?  I’m not sure the humility is necessary.

Line 393:  It’s a little strange to see you endorse this explanation as the most likely, when your paper reports a phenomenon (the frame effect) which seems to support the Künnapas (1959) account.  Does the Howe and Purves (2002)’s account provide an explanation for this effect? 

I don’t think that you need to give this as extensive discussion as you give to Howe and Purves, but you might be interested in reading (and perhaps citing) this other paper offering yet another account of the vertical-horizontal illusion:  https://imi.pmf.kg.ac.rs/kjm/pub/12776261507591_kjom3301.pdf

Author Response

Reading this paper was a pleasure.  The finding is interesting and novel, and it is theoretically and methodologically sound.  I strongly recommend that it be accepted, and I invite the authors to consider the following comments:

Thank you very much for your recommendation, and your comments. Please see our response below.

 Lines 29-31:  It took me a moment to realize that the sentence on these lines refers to Figure 1A (to the unrotated version).  Perhaps you can make this clearer?

We now explicitly refer to Figure 1A for clarity.

 Figure 2, panel 2:  I found the OR to be confusing (it seemed to suggest that one of two pairs of stimuli was presented).  It might be better if you put text at the top of the panel that reads “One of the following:”

Thank you for this suggestion. We have added your idea to Figure 2.

 Line 84:  It’s a little strange that the paper starts out  referring to ‘participants' and switches to ‘subjects’.  This is obviously just an aesthetic comment.

Good point. It’s all ‘subjects’ now in line with what appears to be journal policy (looking at other papers).

 Figure 4, panel 1:  I understand the impulse to depict exactly what subjects saw, but this text is too small to read.  Can you make it larger in the figure, and maybe write “‘Length' or 'Thickness'".  I don’t think this would confuse anybody if you report the actual size of the text in the methods section.

We have come up with a creative solution that we believe makes it more clear.

 Lines 317-318:  Do they really provide no evidence on their own?  I’m not sure the humility is necessary.

We gladly agree. We removed the qualification.

 Line 393:  It’s a little strange to see you endorse this explanation as the most likely, when your paper reports a phenomenon (the frame effect) which seems to support the Künnapas (1959) account.  Does the Howe and Purves (2002)’s account provide an explanation for this effect?

We understand the confusion, our framing was not ideal. The Howe and Purves account does not provide an explanation for Künnapas’ findings. It is possible that the effect Kunnapas describes is real (i.e. the reference frame does influence perception), but is not the explanation for the v-h illusion under ‘normal’ circumstances. We re-phrased our discussion, and the abstract.

 I don’t think that you need to give this as extensive discussion as you give to Howe and Purves, but you might be interested in reading (and perhaps citing) this other paper offering yet another account of the vertical-horizontal illusion:  https://imi.pmf.kg.ac.rs/kjm/pub/12776261507591_kjom3301.pdf

Thank you for pointing us to this paper. We briefly mention this paper on p 9.