Assessing visuospatial biases (pseudoneglect) using pupillometry: A replication and extension of Strauch et al. (2022)

We thank the authors for the thoughtful revision and are curious for the final findings. We only have one suggestion left. The authors now limit the age range to 18-50 years. However, we recommend to, if they want to show pseudoneglect, include participants in the age range of 18-30 or 35. If they stick to 18-50, and the average age would be much higher as compared to the original study, the absence of pseudoneglect or even rightward bias could be explained by an age effect. See references below:

https://doi.org/10.1016/j.neuroimage.2013.10.014

https://doi.org/10.1016/j.neuroimage.2017.03.050

https://doi.org/10.1016/j.bandc.2011.04.002

Christoph Strauch & Teuni ten Brink

Burns and McIntosh describe their attempt to replicate our paper from 2022 in Cortex (Strauch, Romein, Naber, Van der Stigchel & Ten Brink), where we introduced the, what the authors call, split-field pupillometry task (which is a great name!), and where we showed that the pupil responds stronger to the brightness on the left than to the brightness on the right, indicating pseudoneglect. Burns and McIntosh, however, find rightward not leftward biases in pupil light responses. As they tracked the right eye, whereas we tracked the left eye in our Experiment 1, this suggests that effects of our original Experiment 1 are likely confounded by the differences in illumination to the eyes rather than attention. The authors went further by reanalyzing our old data, reproducing the overall leftward bias we observed using the average of both eyes in Experiment 2, which suggest that both attention and a physiological explanation contributed to the effects reported. As the left eye here showed a substantial leftward bias in Exp 2 whilst the right didn’t (resulting in overall leftward bias), this shows an important issue in the split-field pupillometry task: the average of both eyes has to be taken to investigate pseudoneglect. The authors clearly applied rigor in their replication attempt and we think that this manuscript will be helpful in developing the earlier proposed method further. Yet, we have a couple of points that will align the original and the now preregistered replication attempt more closely and will be necessary to allow for a closer replication than the original data reported here. 

-        As an exploratory analysis (we would leave this to the authors), we would be curious whether sensitivity is different if a longer interval is used instead of the constriction amplitude alone. In our second paper (with patients; Ten Brink et al., 2023) we found better sensitivity when considering more data points.

-        About the inclusion of a vertical condition in our paper, it is stated that “Strauch and colleagues had used this as a ‘control condition’ in their second experiment, perhaps on the assumption that there would be no bias for the vertical dimension.” However, we were aware that (very strong) effects were to be expected (described as pupil anisotropies). We did not focus on these, as the vertical condition was merely added as a sanity check, to confirm the absence of a correlation with the left-right greyscales bias. This was, unfortunately, not clearly communicated in our paper so we understand the interpretation of the authors. That being said, we did not assume an attentional bias towards the upper field. Attention might contribute here, but anisotropies in receptors or visual areas in the brain should not be overlooked here as potential contributors. For instance, the SC is topographically organized in a way that might contribute to such anisotropies (Wang & Munoz, 2018, PNAS; reviewed in Strauch, Wang, Einhäuser, Van der Stigchel & Naber, 2022, TINS on page 643). 

-        Related to this, for H3, it would be interesting (and extending the experiment beyond a replication) to add a rotated version of the greyscales task that allows to assess attentional up/downward biases. These could then be correlated with the up/down differences in pupil constriction. Whilst this doesn’t imply that all anisotropies in the light response are due to attention, any covariation with the greyscales task could establish such attentional differences for the upper/lower fields (even though, as the authors correctly state, the eyes are not vertically centered). 

-        The distance from eye position to monitor is different in the currently described study (79cm) from ours (67.5cm). This is important as pseudoneglect is described to diminish or even flip from left to right with more leftward biases reported closer by (e.g., Heber, Siebertz, Wolter, Kuhlen, & Fimm, 2010, Brain and Cognition). An adjustment to the same distance seems therefore sensible. Similarly, the monitor should be of a similar size as the original monitor to control for possible effects induced by the stimuli covering larger/smaller parts of the visual field. Reduced leftward biases at further distances could otherwise be an indication that the split-field pupillometry task captures pseudoneglect.

-        It should be discussed whether the tower mount tracker results in restricted view to the periphery and additional reflections relative to the table-mounted tracker we used in both Cortex papers. 

-        The authors state the number of trials that must be absolved per stimulus. We assume that these will be presented in random order across conditions, as in our original paper. This should be described.

-        It should be described whether age-restrictions are taken as an inclusion criterion. Preferably, the sample should be roughly age-matched as in the original study of Strauch et al. (2022). In the study of Ten Brink et al. (2023) the healthy control group was (partially) of older age than the group in the study of Strauch et al. (2022), and in Ten Brink et al., (2023), we found increasingly less leftward and more rightward pupillometric pseudoneglect estimates (albeit not significant) – in line with the literature on age-effects seen on others measures of pseudoneglect.

-        If the hypothesis on at least partial anisocoria is correct (looking at the reanalysis of our data and the data put forward for the right eye here, we would not be surprised), this necessitates caution when interpreting the effect sizes in two of our papers. Besides the Strauch et al. (2022) paper, where effect sizes might be overestimated, this would also imply that the effect size for the dissociation between patients with neglect and controls in Ten Brink et al. (2023, Cortex) would be underestimated. The latter as we here tracked the left eye for patients (with assumed rightward attentional bias) and the average of both eyes for controls. In our opinion, a short discussion hereof might also benefit those intending to use the method to assess neglect in clinical populations. 

Signed,

Christoph Strauch & Teuni ten Brink

Summary             

In this Registered Report, the authors focus on the recent findings by Strauch et al. (2022) that pupillary responses seem to be an index of leftward biases in visuospatial attention (pseudoneglect). They thoroughly explored the original data and conducted a preliminary replication, which did not replicate and in fact found a rightward bias. Based on this they suggested that the findings originally reported by Strauch et al. may be an artefact of the fact that they primarily reported data collected from the left eye whereas the current authors collected data primarily from the right. They suggest that ipsilateral contraction anosocoria based on heminasal retinal responsivity to light may explain the disparate pattern of results, yet cannot fully rule out whether there is still an influence of covert attentional bias. As such, the study proposed in this registered report will explore both of these possibilities.

Review

The proposed study is exceptionally well thought out. The introduction provides a clear overview of the phenomena. They carefully considered the original data from Strauch et al., exploring which pupil size measure should be used, and accounted for chance variation in pupil constriction over the course of a full trial. After reporting the contradictory results of their replication study, they subsequently re-ran the analysis for Strauch et al.’s data by separating out the two eyes in the experiments where both were collected. In doing so they highlighted somewhat conflicting findings, and I would like to commend the authors for such a thorough revisitation of the original data.

The hypotheses, and planned analyses, were also well considered. I thought their proposed method of scoring the pupillometry data to be clever. If they simply measured pupil constriction of the left eye versus right eye, they would not be able to get to the heart of the matter – that is, whether (and to what extent) ipsilateral constriction anosocoria may account for the previous patterns of results.

The stimuli, procedure, and planned sample size are all justified and reasonable, and I would not suggest any changes. Overall this stage 1 manuscript has a logical theoretical basis, sufficient methodological detail, a sound analysis pipeline, and consideration of what a mixed pattern of results may look like.

I have only two minor comments for the authors to address:

1.      In section 2.5, the authors note that the plan of analysis for the pupillary pseudoneglect hypothesis will involve calculating pupil change in the bright right vs. bright left conditions. However I could not find mention of whether this was overall pupil change across both eyes, or whether this would be calculated and examined separately for the left and right eyes (or, even less unlikely, only examined for one eye). I suspect the first option is the case but clarity on this would be appreciated.

2.      On page 16 the authors note the possibility that apparent pseudoneglect may not correlate with greyscales, and if this is the case it could weaken the validity of the pupillometry task as a measure of spatial attention. While I don’t necessarily disagree, what of the various findings that different measures of pseudoneglect often don’t correlate with each other? Attention biases may have a few different underlying mechanisms/routes (e.g., see Learmonth et al., 2015) and even if they do have a shared commonality this can be difficult to observe for a number of reasons (Chen et al., 2019). I would like to see some consideration of these matters in their report.

Chen, J., Lee, A. C., O’Neil, E. B., Abdul-Nabi, M., & Niemeier, M. (2020). Mapping the anatomy of perceptual pseudoneglect. A multivariate approach. NeuroImage, 207, 116402.

Learmonth, G., Gallagher, A., Gibson, J., Thut, G., & Harvey, M. (2015). Intra-and inter-task reliability of spatial attention measures in pseudoneglect. PloS one, 10(9), e0138379.z

Note to editor

I was delighted to review this stage 1 registered report, and provide an overall positive review. I had 2 minor comments which need some clarity or thought, but overall I am happy for the authors to proceed with their research and would be keen to review their next submission in the future.