Long-Range Interocular Suppression in Adults with Strabismic Amblyopia: A Pilot fMRI Study
, Goro Maehara 4, Erin Goddard 2, Reza Farivar 2, Behzad Mansouri 5 and Robert F. Hess 2Round 1
Reviewer 1 Report
This paper reports on the absence of interocular suppression from beyond the classical receptive field in humans with strabismic amblyopia, as detectable by fMRI.
There is plenty of evidence, based on both psychophysical and electrophysiological studies, of interocular suppression resulting from dichoptic stimuli that overlap in the visual field, and the authors have previously presented fMRI data in support as well. In contrast, much less is known about interocular suppression by non-overlapping (such as centre-surround) stimuli.
In this study comprising 5 strabismic amblyopes and 5 control subjects Thompson et al find that while there are small but consistent differences between BOLD signals recorded in dichoptic vs monocular stimulus conditions there are no significant differences in either signal amplitude or latency between strabismic amblyopes and normal observers.
The data are well presented and analysed, and I have no substantial criticism other than that with a larger sample size some marginal effects might have come out as significant.
Specific comments:
1) line 159: “The was” needs correcting.
2) Figure 4: Group C (members of the amblyopia group who completely suppressed the amblyopic eye) appears to show reduced % BOLD change in dichoptic and monocular conditions. This should be commented on (unless I overlooked it).
3) line 267: “combined the lack”, delete “the”
4) When discussing the reasons for the lack of long-range interocular suppression the authors could reconsider the recent monkey data from the Kiorpes lab (ref 20, Hallum et al. 2017) which suggests that is the excitatory influence of the two eyes that is imbalanced to a degree that can be predicted from the severity of amblyopia, whereas suppression from both eyes was prevalent in all animals. This altered balance of excitation and suppression may contribute to the interocular perceptual suppression in amblyopes.
5) As further evidence for abnormal horizontal processing (spatial summation in this case) in the visual cortex the authors could also cite a recent paper by Je, S., et al. (2018). "Spatial summation across the visual field in strabismic and anisometropic amblyopia." Scientific Reports 8(1): 3858.
Author Response
Response: Corrected.
1) Figure 4: Group C (members of the amblyopia group who completely suppressed the amblyopic eye) appears to show reduced % BOLD change in dichoptic and monocular conditions. This should be commented on (unless I overlooked it).
Response: Thank you for this comment. We have now addressed this difference on lines 291-294 in the revised manuscript; “The members of the amblyopia group with complete suppression exhibited slightly attenuated BOLD responses for both conditions compared to the other groups (Figure 5C). This effect was due to lower overall BOLD responses for participants RD (inverted triangle) and GN (circle) (Figure 4, top left panel).”
2) line 267: “combined the lack”, delete “the”
Response: This sentence has now been deleted.
3) When discussing the reasons for the lack of long-range interocular suppression the authors could reconsider the recent monkey data from the Kiorpes lab (ref 20, Hallum et al. 2017) which suggests that is the excitatory influence of the two eyes that is imbalanced to a degree that can be predicted from the severity of amblyopia, whereas suppression from both eyes was prevalent in all animals. This altered balance of excitation and suppression may contribute to the interocular perceptual suppression in amblyopes.
Response: Excellent point. The following text has been added (lines 337-341): “In addition, recent primate neurophysiology data [21] indicate that interocular suppression in amblyopia is associated with a weak excitatory drive from the amblyopic eye rather than abnormally strong suppression of the amblyopic eye inputs to primary visual cortex. Our fMRI protocol was not designed to detect an overall attenuation of activity in response to the amblyopic eye because the data were expressed as a percent change from baseline. Therefore, is it possible that weak excitatory signals from the amblyopic eye in early visual cortex underlie long range suppression..”
4) As further evidence for abnormal horizontal processing (spatial summation in this case) in the visual cortex the authors could also cite a recent paper by Je, S., et al. (2018). "Spatial summation across the visual field in strabismic and anisometropic amblyopia." Scientific Reports 8(1): 3858.
Response: Interesting paper. It is now referenced in the opening paragraph of the revised manuscript (ref 4, line 34).
Reviewer 2 Report
This is a very interesting fMRI study addressing the cortical nature of suppression in amblyopia, in particular what the author called the long-range interocular suppression.
My only major concern is about the very small sample size (5 amblyopes compared to 5 controls). Although the authors recognize this limitation in the discussion, they try to argue that that the lack of significant differences for most of their results is not a type II error; I am not convince. The standards for fMRI studies in terms of sample size is much higher, even with a "simple" (low-level) task like here. Moreover, I am wondering about the validity of parametric stats analysis (3-way mixed ANOVA models) with such a low number of participants. To my opinion, this is a great study but this is a great pilot study that must be replicated on a larger number of amblyopes to be able to draw any conclusion. Therefore, I recommend to modify the title by adding at the end ... : A pilot fMRI study.
Minor points:
-Typo error on line 28 (This should be this)
-Line 76: This last sentence really breaks the rhythm of what comes before. Either move it somewhere else or clarify the point.
-Line 129: Please provide more details of the MRI scan including its location (since the affiliations of the authors are scattered
-Line 177: Why the authors their connectivity analysis are exploratory (whereas BOLD activation is not)?
Author Response
This is a very interesting fMRI study addressing the cortical nature of suppression in amblyopia, in particular what the author called the long-range interocular suppression.
My only major concern is about the very small sample size (5 amblyopes compared to 5 controls). Although the authors recognize this limitation in the discussion, they try to argue that that the lack of significant differences for most of their results is not a type II error; I am not convince. The standards for fMRI studies in terms of sample size is much higher, even with a "simple" (low-level) task like here. Moreover, I am wondering about the validity of parametric stats analysis (3-way mixed ANOVA models) with such a low number of participants. To my opinion, this is a great study but this is a great pilot study that must be replicated on a larger number of amblyopes to be able to draw any conclusion. Therefore, I recommend to modify the title by adding at the end ... : A pilot fMRI study.
Response: Thank you for this suggestion, the title has been simplified and modified as recommended. We have also added the following text to the concluding paragraph (lines 411-415). “We note that our sample was small and subtle group differences may have emerged with more participants. A larger study is required to replicate these initial findings. However, based on the available data, we speculate that interocular timing differences and/or higher level brain areas may underpin long range suppression in strabismic amblyopia.”
Minor points:
-Typo error on line 28 (This should be this)
Response: Corrected.
-Line 76: This last sentence really breaks the rhythm of what comes before. Either move it somewhere else or clarify the point.
Response: Agreed. The sentence has been removed.
-Line 129: Please provide more details of the MRI scan including its location (since the affiliations of the authors are scattered
Response: The location of the scanner has been added (line 183): “located within the Montreal Neurological Institute”.
-Line 177: Why the authors their connectivity analysis are exploratory (whereas BOLD activation is not)?
Response: The term exploratory has been removed.
Reviewer 3 Report
Overview
This paper aims to verify if long-ranged interocular suppression is measurable in individuals with strabismic amblyopia in fMRI. They measure interocular suppression of non-overlapping stimuli by comparing the BOLD signal change between monocular presentations and dichoptic presentations. They find no evidence of a statistically significant change in % BOLD change in individuals with strabismic amblyopia or their control group. Thus, they conclude that long-range interocular suppression cannot be measured in fMRI.
While the aims of this manuscript are interesting, however, it suffers from severe issues that prevent me from recommending this paper for publication. First, the sample size is too small for a paper using a sample of observers with amblyopia. Second, the authors argue for the null hypothesis in this manuscript. The authors allude to a few explanations as to why long-range suppression was not measured with fMRI. However, when one aims to argue that null effect is evidence for a non-effect, it is important to obtain additional empirical evidence that explains why a particular effect was not found. The authors could implement the following to improve the quality of the manuscript. First, the procedures of the psychophysical study should be clarified in order to properly relate the behavioral findings to the fMRI data. Second, there are neuroimaging methods that are temporally sensitive (EEG / MEG) that the authors should use to verify whether long-range suppression can be explained as a response lag effect from the amblyopic eye. Finally, arguing for or against a particular mechanism to explain the amblyopic deficit requires a computational model. There are multiple examples in the literature of models being applied to both psychophysical and neuroimaging data (see Richard, Chadnova, & Baker, 2018). A model of amblyopia should be implemented here to explain the psychophysical and neuroimaging effects.
Details
Introduction – Overall, the introduction could be improved. It is limited in the background that it offers on dichoptic masking studies, and measurements of monocular versus binocular vision with fMRI. Flow between paragraphs and sentences could also be improved, as many seem disjointed.
L28 – Delete Using function MRI
L36 – caused by strabismic amblyopia
L37 –The amblyopic deficit has been attributed to abnormal interocular suppression, but there is computational evidence that other factors may also contribute. For example, you cite Baker et al., 2008 [34], who demonstrated that interocular suppression is normal in individuals with strabismic amblyopia, and that the visual deficits can be explained with attenuation and increased noise in the amblyopic eye. You acknowledge this as a possibility in the discussion (L263) and it would be good to prepare the reviewer for this in the discussion.
L44 – Change the f in fMRI to Functional MRI.
L56 – 58 – This should be supported with references from psychophysical studies on dichoptic surround suppression (e.g., Petrov & McKee, 2006) and/or neurophysiological studies (e.g., Macknik & Martinez-Conde, 2004)
L76 – The justification for investigating functional connectivity with the methods presented here should appear prior to the stating that functional connectivity will be measured. The authors could also expand on why this is a relevant and important measurement to make in their study.
L78 – Why only five participants in each group? This is a very small sample size.
L86 – Change ETRDS to ETDRS
L98 – Stimuli are described as square patches of sinusoidal gratings but appear to be plaids in Figure 1.
L98 – Why was the experiment conducted at a single contrast value? Surely the suppressive effects will vary according to stimulus contrast. How long did the MRI experiment take to complete?
L98 - f these are sinusoidal gratings, shouldn’t the background be set to mean luminance? I imagine this would be particularly important in the dichoptic conditions as the dark background will rival with the stimulus presented in the opposite eye.
Figure 1 – Since the columns are numbered in text, it would be nice to have them numbered in the Figure as well.
L113 -114 – You mention here that the psychophysical task asked observers to match perceived contrast, but report mismatch values as cd/m2, which is luminance. Did observers match luminance or contrast? I understand this psychophysical task has been previously published elsewhere, but a few additional details here to keep this manuscript self-contained would be beneficial (i.e., stimulus to match in the fellow fixing eye, reference contrast value, number of matches…)
L114 -115 – You report observers completing a fixation task during scanning, but don’t report any accuracy values for the task.
L123 – I imagine the stereo display system was used for all stimulus conditions and observers wore the polarized glasses for all conditions as well?
L129 – This is a general comment in regard to referring to other previously manuscripts for details. While this is a great tool for brevity, the manuscript should never the less be self-contained and contain sufficient information for the reader to understand the experiment procedures. This applies to my comment on the psychophysical procedures, and the same applies for the retinotopic mapping procedures. Standard retinotopic mapping here refers to rings and wedges?
Figure 2 - Would it be possible to offer an example of the V1 / V2 / V3 ROIs in addition to the Right / Left eye localizers?
L159 – Change The to This
L169 – Statistical significance is set by the alpha level (a = 0.05). A result is considered statistically significant if p < a.
I’m not sure how a small sample size justifies not implementing any corrections for multiple comparisons. To me, this reads as an acknowledgement that not enough data was collected.
L177 – I recommend removing the functional connectivity analysis as it does not appear to offer any additional or meaningful information in regard to the goals of this study. However, if the authors were to keep it, it would be helpful to have a few more details on the analysis procedure within the manuscript.
L184 – You report luminance matching task for control and amblyopes, but you describe a contrast matching task in the stimulus section of the manuscript. Given your stimuli and the goals of the current study, the behavioral results would be most meaningful if contrast matching values were reported, no?
L192 – It would be good to report effect sizes for the ANOVA results (partial eta squared ()) and t-tests (Hedge’s g / Cohen’s d).
L197 – Effects are either statistically significant or not. A “marginally” statistically significant result is an inaccurate way of reporting statistics and can be misleading to the reader.
L198 – Change p < 0.05 to p > 0.05
L207 – Change Figure 2 to Figure 3
Figure 3 – In the control observers, dominant eye, two observers are labelled by an upright triangle.
All plots within a figure should have the same axis scale to ensure the magnitude of an effect is properly conveyed by the figure.
Figure 4 – The tick marks for B should be adjusted to match those of A C D
Line 228 – Change Figure 2 to Figure 3. Error bars should be shown in the charts, as they are an important tool to visually assess the statistical significance of an effect. For HRFs, these can be dashed lines which would not affect plot clarity.
L239 – Figure 3
Figure 6 – The axes for both plots should be of the same scale
L246 – I don’t fully grasp the importance of this analysis. The results are not particularly interesting, and the authors rely on a trend, which is not statistically significant, to convey some type of difference in connectivity strength between the amblyopic eye and the fellow eye across visual areas. It’s also unclear how a difference in connectivity strength between V2 and V3 for controls, but not amblyopes is relevant to the manuscript, particularly since this analysis is not mentioned in the discussion.
L263 – If you acknowledge here that interocular interactions in amblyopes may be normal, it would be good to present this in the introduction as well. In addition, the authors may want to describe predictions of the attenuator model for their study.
L266 – 270 – This is arguing for the null hypothesis, which statistical analyses are incapable of doing. In the current scenario, they test against the null and are unable to reject it. However, this does not demonstrate that the null hypothesis is true. The absence of an effect is not evidence for no effect. If the authors want to conclude that long-range suppression is not measurable with fMRI, they would need to run additional control studies that explain why they were unable to measure the desired effect.
L271 – If the authors knew the experiment was underpowered, then why not collect data from more observers? 5 observers with amblyopia is definitely too small a sample size to obtain any meaningful measurements of effects.
L278 – This would be a relatively simple experiment to conduct and is necessary for the authors to conclude that long-range suppression is not measurable with fMRI.
L284 – This is also a possible explanation, however there is neurophysiological evidence for dichoptic suppression in V1 (see Macknik & Martinez-Conde, 2004). It would be informative to the reader if the authors could reconcile their current argument with findings of dichoptic suppression in V1.
L303 – 307 – I am not clear on the argument the authors are trying to make here as the attenuator model of amblyopia, which is used as an explanation for the absence of long-range suppression in fMRI data, predicts minimal impairments on suprathreshold contrast perception. The authors may want to clarify here what other mechanism they believe is responsible if attenuation is an unlikely candidate.
As a general comment, 23 of the 40 references are previous publications from the authors of this manuscript. While they have contributed greatly to our current knowledge of strabismic amblyopia, the manuscript would benefit from acknowledging other work on the effects they aimed to measure here.
Author Response
First, the procedures of the psychophysical study should be clarified in order to properly relate the behavioral findings to the fMRI data.
Response. Excellent point. We have now provided specific details of the psychophysical procedure and how the psychophysical results were verified directly prior to scanning (new secion 2.4, lines 160-178).
Second, there are neuroimaging methods that are temporally sensitive (EEG / MEG) that the authors should use to verify whether long-range suppression can be explained as a response lag effect from the amblyopic eye.
Response: We have now identified that these methods should be used in future studies of long range suppression: “it is conceivable that a relatively delayed neural response to the amblyopic eye could contribute to perceptual suppression under dichoptic viewing conditions. Further studies using techniques with a high temporal resolution such as MEG will be required to address this possibility.” (lines 375-378).
Finally, arguing for or against a particular mechanism to explain the amblyopic deficit requires a computational model. There are multiple examples in the literature of models being applied to both psychophysical and neuroimaging data (see Richard, Chadnova, & Baker, 2018). A model of amblyopia should be implemented here to explain the psychophysical and neuroimaging effects.
Response: This is an interesting suggestion. Contrary to our hypothesis, we did not observe an effect of amblyopic eye suppression in our fMRI data. Unfortunately, this means that modelling the data is unlikely to informative.
Details
Introduction – Overall, the introduction could be improved. It is limited in the background that it offers on dichoptic masking studies, and measurements of monocular versus binocular vision with fMRI. Flow between paragraphs and sentences could also be improved, as many seem disjointed.
Response: We prefer to keep the introduction focussed on long range interocular suppression in amblyopia as this is the issue that is addressed by our experiment, however we agree that dichoptic masking studies should be mentioned. We have added the references suggested in the L-56-58 comment below (line 65). We have also edited the introduction to improve flow as requested - paragraphs 3 and 4 in particular.
L28 – Delete Using function MRI
Response: Corrected.
L36 – caused by strabismic amblyopia
Response: Corrected.
L37 –The amblyopic deficit has been attributed to abnormal interocular suppression, but there is computational evidence that other factors may also contribute. For example, you cite Baker et al., 2008 [34], who demonstrated that interocular suppression is normal in individuals with strabismic amblyopia, and that the visual deficits can be explained with attenuation and increased noise in the amblyopic eye. You acknowledge this as a possibility in the discussion (L263) and it would be good to prepare the reviewer for this in the discussion.
Response: Agreed. We have added the following text to the end of the second introduction paragraph: ” It is currently unclear whether suppression is caused by active suppression of amblyopic eye signals, weak and noisy inputs from the amblyopic eye to a normal interocular inhibition mechanism [e.g. 23] or a combination of both.” (lines 45-48).
L44 – Change the f in fMRI to Functional MRI.
Response: Corrected.
L56 – 58 – This should be supported with references from psychophysical studies on dichoptic surround suppression (e.g., Petrov & McKee, 2006) and/or neurophysiological studies (e.g., Macknik & Martinez-Conde, 2004)
Response: We agree. These references have been added: “In participants with normal binocular vision, both surround suppression and lateral inhibition occur for dichoptic stimuli [26,27].” (lines 64-65).
L76 – The justification for investigating functional connectivity with the methods presented here should appear prior to the stating that functional connectivity will be measured. The authors could also expand on why this is a relevant and important measurement to make in their study.
Response: Agreed. This section of the manuscript has been expanded (lines 87-100).
L78 – Why only five participants in each group? This is a very small sample size.
Response: We agree that the small sample size is a key limitation of the study and this has been highlighted in the manuscript (lines 366-368; 411-413). Our original sample was 6, but one participant was excluded because they fused the stimuli in the scanner. Recruitment for this study was challenging because it involved a long fMRI procedure (minimum of 2 hours in the scanner bore including set-up) and a specific patient group (adult strabismic amblyopia and no contra-indications for MRI). A number of potential participants declined to participate because the MRI procedure was long and they disliked confined spaces.
L86 – Change ETRDS to ETDRS
Response: Corrected.
L98 – Stimuli are described as square patches of sinusoidal gratings but appear to be plaids in Figure 1.
Response: This must have been an issue with figure rendering in the draft manuscript. The stimuli were square patches of sinusoidal grating.
L98 – Why was the experiment conducted at a single contrast value? Surely the suppressive effects will vary according to stimulus contrast. How long did the MRI experiment take to complete?
Response: The experiment required at least 2 hours in the scanner bore, partially because it took some time to set up the dichoptic presentation system to ensure zero cross-talk between the eyes. This long duration precluded the inclusion of additional variables within the experiment. We chose our stimuli to induce robust cortical activity but still be subject to interocular suppression.
L98 - f these are sinusoidal gratings, shouldn’t the background be set to mean luminance? I imagine this would be particularly important in the dichoptic conditions as the dark background will rival with the stimulus presented in the opposite eye.
Response: The dark background was used to match the background used in the luminance matching psychophysical experiments. Interestingly, during pilot testing, the dark background induced less rivalry than a mean luminance background, possibly due to the large luminance difference between the eyes.
Figure 1 – Since the columns are numbered in text, it would be nice to have them numbered in the Figure as well.
Response: Excellent point. Numbers have been added (revised figure 1)
L113 -114 – You mention here that the psychophysical task asked observers to match perceived contrast, but report mismatch values as cd/m2, which is luminance. Did observers match luminance or contrast? I understand this psychophysical task has been previously published elsewhere, but a few additional details here to keep this manuscript self-contained would be beneficial (i.e., stimulus to match in the fellow fixing eye, reference contrast value, number of matches…)
Response: Thank you for spotting this error. Participants matched luminance. Details of the psychophysical procedure have now been added to the manuscript (new section 2.4, lines 162-178).
L114 -115 – You report observers completing a fixation task during scanning, but don’t report any accuracy values for the task.
Response: Excellent point. Mean and SD percent correct values are now reported for the control and amblyopia groups; “Fixation task performance means (±SD) were 88.3±9.5% correct for the control group and 90.0±14% correct for the amblyopia group.” (Lines 270-271).
L123 – I imagine the stereo display system was used for all stimulus conditions and observers wore the polarized glasses for all conditions as well?
Response: Yes, during scanning the same display system and cross polarizers were used for all conditions.
L129 – This is a general comment in regard to referring to other previously manuscripts for details. While this is a great tool for brevity, the manuscript should never the less be self-contained and contain sufficient information for the reader to understand the experiment procedures. This applies to my comment on the psychophysical procedures, and the same applies for the retinotopic mapping procedures. Standard retinotopic mapping here refers to rings and wedges?
Response: Yes, rings and wedges were used. This detail has now been added to the manuscript: “Standard monocular retinotopic mapping was conducted in exactly the same way as previously described using wedge and ring stimuli [40,41].” (lines 180-181).
Figure 2 - Would it be possible to offer an example of the V1 / V2 / V3 ROIs in addition to the Right / Left eye localizers?
Response: Absolutely. Please see new figure 3 for example ROIs rendered on a flattened patch of cortex.
L159 – Change The to This
Response: corrected
L169 – Statistical significance is set by the alpha level (a = 0.05). A result is considered statistically significant if p < a.
Response: Thank you. Now corrected: “The alpha level of statistical testing was set at 0.05” (line 226).
I’m not sure how a small sample size justifies not implementing any corrections for multiple comparisons. To me, this reads as an acknowledgement that not enough data was collected.
Response: Agreed. This comment has been removed.
L177 – I recommend removing the functional connectivity analysis as it does not appear to offer any additional or meaningful information in regard to the goals of this study. However, if the authors were to keep it, it would be helpful to have a few more details on the analysis procedure within the manuscript.
Response: We have now added additional details relating to the connectivity analysis to the manuscript with a reference to a methods paper that describes the procedure in detai: “The analysis technique utilizes a nonlinear autoregressive exogenous model combined with a least squares statistical method to generate an F value for each connection for each participant that reflects the strength of connectivity between brain areas within a network (see [36,43] for a detailed description). The F values are then subjected to standard statistical analysis.” (Lines 234-238).
L184 – You report luminance matching task for control and amblyopes, but you describe a contrast matching task in the stimulus section of the manuscript. Given your stimuli and the goals of the current study, the behavioral results would be most meaningful if contrast matching values were reported, no?
Response: Thank you for pointing out this error. Luminance matching was conducted as reduced luminance is the perceptual experience of participants with amblyopia who do not experience complete long range suppression (Maehara et al., Invest Ophthalmol Vis Sci 2011, 52, 9011-9017). Refences to contrast matching have been removed from the manuscript.
L192 – It would be good to report effect sizes for the ANOVA results (partial eta squared ()) and t-tests (Hedge’s g / Cohen’s d).
Response: Thank you for this suggestion. These details have now been added to all statistical analyses.
L197 – Effects are either statistically significant or not. A “marginally” statistically significant result is an inaccurate way of reporting statistics and can be misleading to the reader.
Response: Agreed, these terms have been deleted.
L198 – Change p < 0.05 to p > 0.05
Response: Corrected
L207 – Change Figure 2 to Figure 3
Response: Corrected
Figure 3 – In the control observers, dominant eye, two observers are labelled by an upright triangle.
Response: Thank you for spotting this error. Now corrected.
All plots within a figure should have the same axis scale to ensure the magnitude of an effect is properly conveyed by the figure.
Response: Scales have been matched within the figures.
Figure 4 – The tick marks for B should be adjusted to match those of A C D
Response: Corrected
Line 228 – Change Figure 2 to Figure 3. Error bars should be shown in the charts, as they are an important tool to visually assess the statistical significance of an effect. For HRFs, these can be dashed lines which would not affect plot clarity.
Response: The figure number error has been corrected. Dashed lines indicating error bars span almost the entire figure because between subject variability is large. This is stated in the figure legend and readers are referred to figure 3 for the individual data. (lines 305-306).
L239 – Figure 3
Response: Corrected
Figure 6 – The axes for both plots should be of the same scale
Response: Corrected
L246 – I don’t fully grasp the importance of this analysis. The results are not particularly interesting, and the authors rely on a trend, which is not statistically significant, to convey some type of difference in connectivity strength between the amblyopic eye and the fellow eye across visual areas. It’s also unclear how a difference in connectivity strength between V2 and V3 for controls, but not amblyopes is relevant to the manuscript, particularly since this analysis is not mentioned in the discussion.
Response: Thank you for this comment. The motivation for this analysis is now provided more fully in the introduction and the results have been included in the discussion. We are not claiming any group differences relating to suppression from this analysis. (lines 89-100; 355).
L263 – If you acknowledge here that interocular interactions in amblyopes may be normal, it would be good to present this in the introduction as well. In addition, the authors may want to describe predictions of the attenuator model for their study.
Response: Good point. The possibility of normal interocular inhibition in amblyopia is now mentioned in the introduction: “It is currently unclear whether suppression is caused by active suppression of amblyopic eye signals, weak and noisy inputs from the amblyopic eye to a normal interocular inhibition mechanism [e.g. 23] or a combination of both.” (lines 45-48).
L266 – 270 – This is arguing for the null hypothesis, which statistical analyses are incapable of doing. In the current scenario, they test against the null and are unable to reject it. However, this does not demonstrate that the null hypothesis is true. The absence of an effect is not evidence for no effect. If the authors want to conclude that long-range suppression is not measurable with fMRI, they would need to run additional control studies that explain why they were unable to measure the desired effect.
Response: Agreed. We have softened the language in this section: “This lack of association between the BOLD data and perception suggests that we were unable to detect long range interocular suppression in strabismic amblyopia within V1, V2 and V3 using the specific paradigm that we employed” (lines 362-364).
L271 – If the authors knew the experiment was underpowered, then why not collect data from more observers? 5 observers with amblyopia is definitely too small a sample size to obtain any meaningful measurements of effects.
Response: Unfortunately, we were unable to recruit additional participants. However, the current data are of interest given the large difference in perception between the two groups in combination with no difference in fMRI results.
L278 – This would be a relatively simple experiment to conduct and is necessary for the authors to conclude that long-range suppression is not measurable with fMRI.
Response: We agree and have identified this as the next step in the text: “Further studies using techniques with a high temporal resolution such as MEG will be required to address this possibility.” (lines 377-378). We have also toned down the claim that long range suppression is not measurable with fMRI throughout the discussion.
L284 – This is also a possible explanation, however there is neurophysiological evidence for dichoptic suppression in V1 (see Macknik & Martinez-Conde, 2004). It would be informative to the reader if the authors could reconcile their current argument with findings of dichoptic suppression in V1.
Response: We have added a reference to Macknik 2004 in the relevant paragraph: “Finally, although weak dichoptic masking is evident in single cell recordings within V1 for non-overlapping stimuli in close proximity [27], long range suppression may involve brain areas outside of V1, V2 and V3 that have large visual receptive fields” (lines 385-387).
L303 – 307 – I am not clear on the argument the authors are trying to make here as the attenuator model of amblyopia, which is used as an explanation for the absence of long-range suppression in fMRI data, predicts minimal impairments on suprathreshold contrast perception. The authors may want to clarify here what other mechanism they believe is responsible if attenuation is an unlikely candidate.
Response: Thank you for identifying this point of confusion. The term attenuation has been replaced with suppression: “cannot easily be explained by an early, active suppression of the neural response to the suppressed eye.” (lines 407-408).
As a general comment, 23 of the 40 references are previous publications from the authors of this manuscript. While they have contributed greatly to our current knowledge of strabismic amblyopia, the manuscript would benefit from acknowledging other work on the effects they aimed to measure here.
Response: Agreed. A number of new references from different groups have been added as part of the revision process.
Reviewer 4 Report
I enjoyed reading this manuscript. The paper is succinctly and clearly written, with a clever experiment well targeted at the hypothesis. The results are clearly laid out and the discussion is appropriate in scope. The result described in this paper is an important one for the field. My primary recommendations for revision are (1) to clarify the both the purpose for, and the methodology used, in the connectivity analysis; and (2) clarify a few points in the methods/analysis that will help convey the details of the experiment.
p 2, end of introduction: the functional connectivity idea felt thrown in there as the very last sentences (74-76). Can you explain why you want to look at this in more details, aside from one sentence saying it's known to be “abnormal?” Plenty of studies have been done looking at functional connectivity in amblyopia, so it would be with adding a paragraph in the introduction summarizing what has been done and how your analysis will contribute (has anyone done FC under binocular viewing in amblyopia and is that normal? Is FC known to abnormal in visual cortex -> other areas, has anyone done V1-V2-V3 connectivity.. etc). Did you have an intuition beforehand if it would be feedforward, feedback, or both?
p 5, Analysis:
“Suppression was quantified for each eye separately by comparing the BOLD response of the eye-specific ROIs to the relevant monocular four quadrant condition and the dichoptic condition.” I can interpret this sentence in two ways: either you are using the localizer for both localization of the ROIs and as the baseline condition (2 conditions: monocular > localizer, dichoptic > localizer), or you are looking at the BOLD response in the areas defined by the ROIs for dichoptic > monocular and monocular > dichoptic. Can you please clarify?
I think it would be worth really spelling out in more detail how the ANOVA conditions correspond to your experimental conditions. I’m still not completely clear on this. When the factor is "eye" is this when this eye is being suppressed or when this eye is being presented with the information? ie for monocular, when eye is AE, is the AE being suppressed by information in the other eye, or is AE getting the information? [It’s clearer when you get to the results, but it is not clear here]
I think it's important to include more methodological details on the connectivity analysis. I know readers can look elsewhere for the nitty-gritty, but enough needs to be in here for us to understand what data is being used (mean BOLD? Timecourses?), how "connectivity strength" is quantified and how its range and magnitude can be interpreted. How does it differ feedforward/feedback? Is the F value the result of the statistical analysis? Or does a connectivity strength analysis give you an F per person which is further analyzed on the group level?
There are additional minor points of clarification/detail I suggest the authors implement:
Intro:
p. 2, (48-50): I think this point deserves more detail. This participant was 1 of 4 tested; was the only one with infantile esotropia, and was the only one who reported experiencing suppression. Please expand on what you mean by “attributed to interocular suppression”: I think what you are getting at is the interesting part of Conner et al’s result, the correspondence between perceptual report and BOLD response.
p. 2, (52-54): Can you be clearer about why this is definitely suppression and definitely not rivalry?
Methods:
p. 2, Control participants: Sorry to make you do this, but readers are curious.. what's normal vision? 20/20? close to 20/20? Do they have normal stereopsis? If not wearing contacts, anyone corrected with MRI glasses?
Table 1: Instead of sex this should probably say gender. If the journal's style will let you, just put the actual symbol beside the participants' initials, or in the column. Referring between the word description in a table to identify a shape in a plot is awkward. I think part of the table under “History, stereo” is cut off in the PDF I downloaded – I see “local stereopsis” and only some of them have a measurement. Maybe the data point column is too big relative to this column. I’d suggest making stereo its own column.
p. 3, Stimuli: I can’t tell if there were 2 ROIs per eye or 4 ROIs per eye in the localization stimuli. Left eye gets top right/bottom left quadrants; right eye gets top left/bottom right quadrants (2 ROIs per eye). This is always the case, right? It’s never reversed to get the other 2 ROIs in each eye?
p. 3, (111): did all participants need prisms (including ET/XT 1), were they the participants’ habitual prismatic correction?
p. 3 (112-114): “Prior to scanning, participants viewed the dichoptic stimulus configuration and attempted to match the perceived contrast of the patches shown to each eye using a previously described protocol” Why? for practice? To inform your choices in fMRI? for behavioural measures that you will describe shortly? (It feels out of place as its between two MRI descriptions)
p 4 (116-118): Why a fixation cross in one eye only? My instinct would have been to put it in both eyes, so as not to bias foveal activation or attention to only one of the two eyes (in controls anyway). Can you explain in the text why you made this choice?
P 4 (129): Retinotopic mapping was monocular I’m guessing - can you state this in the text?
Results:
I think some of your figure numbering is incorrect here, make sure you take another peek.
p 5 (190): I’m not sure how 40 cd/m2 compares to the people who couldn’t see it at all. What is the max luminance you could show the participants with full suppression? (Perhaps belongs in methods)
Figure 4: Can you be clearer here. It's AE viewing when it's the monocular condition, but both eyes are "viewing" in the dichoptic condition.
Discussion:
P 9, paragraph starting at 271: This is just a general comment about the claim being made here, not about the study itself: The biggest problem with samples as small as 5 is not that you are underpowered to do comparisons but that you haven't collected data from enough people for systematic differences between groups to emerge in the first place. In other words, "there are no systematic differences between groups" is not a good way to explain away the small sample problem. There is still the problem that you haven't sampled enough of the population to get an accurate estimate in the first place (before even doing comparisons). In other words, you're trying to convince readers that even if you got more people the mean differences you've reported still wouldn't be significant, but I would still be worried that with more people the means themselves would change. I do think your results are very compelling and I don't mind the small sample, but I do take issue with this type of argument for small samples. I’d recommend removing the claim “an inspection of the individual data (Figures 3 and 5) clearly supports the results of the statistical analyses – there are no systematic differences between groups.” Despite the small sample, the fact that your BOLD responses don't covary with the behavioural data you collected, THAT is really convincing and compelling.
Author Response
p 2, end of introduction: the functional connectivity idea felt thrown in there as the very last sentences (74-76). Can you explain why you want to look at this in more details, aside from one sentence saying it's known to be “abnormal?” Plenty of studies have been done looking at functional connectivity in amblyopia, so it would be with adding a paragraph in the introduction summarizing what has been done and how your analysis will contribute (has anyone done FC under binocular viewing in amblyopia and is that normal? Is FC known to abnormal in visual cortex -> other areas, has anyone done V1-V2-V3 connectivity.. etc). Did you have an intuition beforehand if it would be feedforward, feedback, or both?
Response: Thank you for this comment. We have now provided an overview of previous amblyopia connectivity studies and provided our predictions for how feedback and feedforward connectivity might be differentially affected by different suppression mechanisms. (lines 89-100).
p 5, Analysis:
“Suppression was quantified for each eye separately by comparing the BOLD response of the eye-specific ROIs to the relevant monocular four quadrant condition and the dichoptic condition.” I can interpret this sentence in two ways: either you are using the localizer for both localization of the ROIs and as the baseline condition (2 conditions: monocular > localizer, dichoptic > localizer), or you are looking at the BOLD response in the areas defined by the ROIs for dichoptic > monocular and monocular > dichoptic. Can you please clarify?
Response: Excellent point. We used the localizer for ROI definition only. This has now been clarified in the text: “The localization stimuli (Figure 1, columns 1 and 2) were used for ROI definition only.” (lines 221-222).
I think it would be worth really spelling out in more detail how the ANOVA conditions correspond to your experimental conditions. I’m still not completely clear on this. When the factor is "eye" is this when this eye is being suppressed or when this eye is being presented with the information? ie for monocular, when eye is AE, is the AE being suppressed by information in the other eye, or is AE getting the information? [It’s clearer when you get to the results, but it is not clear here]
Response: Good point, the factor “eye” refers to the eye-specific regions of interest. This information has now been added to the description of the ANOVA: “Data were analyzed using separate 3-way mixed ANOVA models for each visual area with factors of Eye (dominant eye ROI vs. non-dominant eye ROI)”. (Lines 223-224).
I think it's important to include more methodological details on the connectivity analysis. I know readers can look elsewhere for the nitty-gritty, but enough needs to be in here for us to understand what data is being used (mean BOLD? Timecourses?), how "connectivity strength" is quantified and how its range and magnitude can be interpreted. How does it differ feedforward/feedback? Is the F value the result of the statistical analysis? Or does a connectivity strength analysis give you an F per person which is further analyzed on the group level?
Response. Agreed, additional information has been added to the description of the connectivity analysis: “The analysis technique utilizes a nonlinear autoregressive exogenous model combined with a least squares statistical method to generate an F value for each connection for each participant that reflects the strength of connectivity between brain areas within a network (see [36,43] for a detailed description). The F values are then subjected to standard statistical analysis.” (lines 223-240).
There are additional minor points of clarification/detail I suggest the authors implement:
Intro:
p. 2, (48-50): I think this point deserves more detail. This participant was 1 of 4 tested; was the only one with infantile esotropia, and was the only one who reported experiencing suppression. Please expand on what you mean by “attributed to interocular suppression”: I think what you are getting at is the interesting part of Conner et al’s result, the correspondence between perceptual report and BOLD response.
Response: Agreed. These details have now been added: “One of the four participants tested had strabismic amblyopia due to infantile esotropia and reported suppression of the amblyopic eye. This participant exhibited an attenuated BOLD response in V1 for the mean luminance relative to the occlusion condition, a difference that could be associated with the participant’s perceptual experience of amblyopic eye suppression.” (lines 53-57).
p. 2, (52-54): Can you be clearer about why this is definitely suppression and definitely not rivalry?
Response: We have rephrased this sentence to read: “This experiment involved presenting a stationary checkerboard stimulus to the fellow eye while the amblyopic eye viewed briefly presented counter phase checkerboard stimuli.” (lines 59-60).
Methods:
p. 2, Control participants: Sorry to make you do this, but readers are curious.. what's normal vision? 20/20? close to 20/20? Do they have normal stereopsis? If not wearing contacts, anyone corrected with MRI glasses?
Response: Good point. These details have now been added to the participants section: “and five control observers with best-corrected logMAR visual acuity of 0.00 or better in each eye, and normal Randot stereopsis (mean age 35 years, range 25-58, two participants wore refractive correction in the scanner) participated” (lines 107-109).
Table 1: Instead of sex this should probably say gender. If the journal's style will let you, just put the actual symbol beside the participants' initials, or in the column. Referring between the word description in a table to identify a shape in a plot is awkward. I think part of the table under “History, stereo” is cut off in the PDF I downloaded – I see “local stereopsis” and only some of them have a measurement. Maybe the data point column is too big relative to this column. I’d suggest making stereo its own column.
Response: Agreed. The table has been modified to improve clarity.
p. 3, Stimuli: I can’t tell if there were 2 ROIs per eye or 4 ROIs per eye in the localization stimuli. Left eye gets top right/bottom left quadrants; right eye gets top left/bottom right quadrants (2 ROIs per eye). This is always the case, right? It’s never reversed to get the other 2 ROIs in each eye?
Response: That is correct, the localization stimuli were never reversed. The same 2 ROIs per eye were used for all conditions.
p. 3, (111): did all participants need prisms (including ET/XT 1), were they the participants’ habitual prismatic correction?
Response: Only participants GN, JG and AD required prismatic correction. No participants habitually wore prismatic correction. This information has now been added to the manuscript: “Prisms were used to cancel manifest strabismus for participants GN, JG and AS (no participants habitually wore prism correction)”. (lines 142-143).
p. 3 (112-114): “Prior to scanning, participants viewed the dichoptic stimulus configuration and attempted to match the perceived contrast of the patches shown to each eye using a previously described protocol” Why? for practice? To inform your choices in fMRI? for behavioural measures that you will describe shortly? (It feels out of place as its between two MRI descriptions)
Response: This was to assess the depth of the participants’ long-range suppression. This section has now been expanded to provide greater detail. (new section 2.4, lines 160-178).
p 4 (116-118): Why a fixation cross in one eye only? My instinct would have been to put it in both eyes, so as not to bias foveal activation or attention to only one of the two eyes (in controls anyway). Can you explain in the text why you made this choice?
Response: The following has been added to the text: “We did not present the fixation mark to both eyes because we wanted to maximize suppression of the amblyopic eye in the dichoptic condition.” (lines 148-149).
P 4 (129): Retinotopic mapping was monocular I’m guessing - can you state this in the text?
Response: Yes, now added: “Standard monocular retinotopic mapping” (line 180).
Results:
I think some of your figure numbering is incorrect here, make sure you take another peek.
Response: You are correct. The numbering has now been fixed.
p 5 (190): I’m not sure how 40 cd/m2 compares to the people who couldn’t see it at all. What is the max luminance you could show the participants with full suppression? (Perhaps belongs in methods)
Response: Thank you for the comment – the luminance values were incorrect and have now been corrected: “Participant AR had an interocular mismatch of 11.9 cd/m2 whereby the quadrants shown to the amblyopic eye were seen at a lower luminance than those shown to the fellow eye. Participant AS had a mismatch of 8.1 cd/m2.” (lines 266-268).
The maximum luminance for the psychophysical measurements was 60 cd/m2. This information has been added to the methods: “(maximum luminance was 60 cd/m2)”. (lines 167-168).
Figure 4: Can you be clearer here. It's AE viewing when it's the monocular condition, but both eyes are "viewing" in the dichoptic condition.
Response: Corrected to “regions of interest”, lines 302-303.
Discussion:
P 9, paragraph starting at 271: This is just a general comment about the claim being made here, not about the study itself: The biggest problem with samples as small as 5 is not that you are underpowered to do comparisons but that you haven't collected data from enough people for systematic differences between groups to emerge in the first place. In other words, "there are no systematic differences between groups" is not a good way to explain away the small sample problem. There is still the problem that you haven't sampled enough of the population to get an accurate estimate in the first place (before even doing comparisons). In other words, you're trying to convince readers that even if you got more people the mean differences you've reported still wouldn't be significant, but I would still be worried that with more people the means themselves would change. I do think your results are very compelling and I don't mind the small sample, but I do take issue with this type of argument for small samples. I’d recommend removing the claim “an inspection of the individual data (Figures 3 and 5) clearly supports the results of the statistical analyses – there are no systematic differences between groups.” Despite the small sample, the fact that your BOLD responses don't covary with the behavioural data you collected, THAT is really convincing and compelling.
Response: Excellent point. The statement has been removed.
Round 2
Reviewer 3 Report
The authors have made significant improvements since the previous rounds of reviews. I still have issues with sample size and the differences between the psychophysical and fMRI stimulus characteristics. The authors do acknowledge these limitations, and for the scope if the current study, I think this is sufficient.
