NMDA receptor antagonist memantine selectively affects recurrent processing during perceptual inference

Reviewer #1 (Public review):

The authors investigate the function and neural circuitry of reentrant signals in the visual cortex. Recurrent signaling is thought to be necessary to common types of perceptual experience that are defined by long-range relationships or prior expectations. Contour illusions - where perceptual objects are implied by stimuli characteristics - are a good example of this. The perception of these illusions is thought to emerge as recurrent signals from higher cortical areas feedback onto the early visual cortex, to tell the early visual cortex that it should be seeing object contours where none are actually present.

The authors test the involvement of reentrant cortical activity in this kind of perception using a drug challenge. Reentrance in the visual cortex is thought to rely on NMDAR-mediated glutamate signalling. The authors accordingly employ an NMDA antagonist to stop this mechanism, looking for the effect of this manipulation on visually evoked activity recorded in EEG.

The motivating hypothesis for the paper is that NMDA antagonism should stop recurrent activity and that this should degrade perceptual activity supporting the perception of a contour illusion, but not other types of visual experience. Results in fact show the opposite. Rather than degrading cortical activity evoked by the illusion, memantine makes it more likely that machine learning classification of EEG will correctly infer the presence of the illusion.

On the face of it, this is confusing, and the paper currently does not entirely resolve this confusion. But there are relatively easy ways to improve this. The authors would be well served by entertaining more possible outcomes in the introduction - there's good reason to expect a positive effect of memantine on perceptual brain activity, and I provide details on this below. The authors also need to further emphasize that the directional expectations that motivated E1 were, of course, adapted after the results from this experiment emerged. The authors presumably at least entertained the notion that E2 would reproduce E1 - meaning that E2 was motivated by a priori expectations that were ultimately met by the data.

I broadly find the paper interesting, graceful, and creative. The hypotheses are clear and compelling, the techniques for both manipulation of brain state and observation of that impact are cutting edge and well suited, and the paper draws clear and convincing conclusions that are made necessary by the results. The work sits at the very interesting crux of systems neuroscience, neuroimaging, and pharmacology. I believe the paper can be improved in revision, but my suggestions are largely concerning presentation and nuance of interpretation.

(1) I miss some treatment of the lack of behavioural correlate. What does it mean that metamine benefits EEG classification accuracy without improving performance? One possibility here is that there is an improvement in response latency, rather than perceptual sensitivity. Is there any hint of that in the RT results? In some sort of combined measure of RT and accuracy?

(2) An explanation is missing, about why memantine impacts the decoding of illusion but not collinearity. At a systems level, how would this work? How would NMDAR antagonism selectively impact long-range connectivity, but not lateral connectivity? Is this supported by our understanding of laminar connectivity and neurochemistry in the visual cortex?

(3) The motivating idea for the paper is that the NMDAR antagonist might disrupt the modulation of the AMPA-mediated glu signal. This is in line with the motivating logic for Self et al., 2012, where NMDAR and AMPAR efficacy in macacque V1 was manipulated via microinfusion. But this logic seems to conflict with a broader understanding of NMDA antagonism. NMDA antagonism appears to generally have the net effect of increasing glu (and ACh) in the cortex through a selective effect on inhibitory GABA-ergic cells (eg. Olney, Newcomer, & Farber, 1999). Memantine, in particular, has a specific impact on extrasynaptic NMDARs (that is in contrast to ketamine; Milnerwood et al, 2010, Neuron), and this type of receptor is prominent in GABA cells (eg. Yao et al., 2022, JoN). The effect of NMDA antagonists on GABAergic cells generally appears to be much stronger than the effect on glutamergic cells (at least in the hippocampus; eg. Grunze et al., 1996).

This all means that it's reasonable to expect that memantine might have a benefit to visually evoked activity. This idea is raised in the GD of the paper, based on a separate literature from that I mentioned above. But all of this could be better spelled out earlier in the paper, so that the result observed in the paper can be interpreted by the reader in this broader context.

To my mind, the challenging task is for the authors to explain why memantine causes an increase in EEG decoding, where microinfusion of an NMDA antagonist into V1 reduced the neural signal Self et al., 2012. This might be as simple as the change in drug... memantine's specific efficacy on extrasynaptic NMDA receptors might not be shared with whatever NMDA antagonist was used in Self et al. 2012. Ketamine and memantine are already known to differ in this way.

(4) The paper's proposal is that the effect of memantine is mediated by an impact on the efficacy of reentrant signaling in visual cortex. But perhaps the best-known impact of NMDAR manipulation is on LTP, in the hippocampus particularly but also broadly. Perception and identification of the kanisza illusion may be sensitive to learning (eg. Maertens & Pollmann, 2005; Gellatly, 1982; Rubin, Nakayama, Shapley, 1997); what argues against an account of the results from an effect on perceptual learning? Generally, the paper proposes a very specific mechanism through which the drug influences perception. This is motivated by results from Self et al 2012 where an NMDA antagonist was infused into V1. But oral memantine will, of course, have a whole-brain effect, and some of these effects are well characterized and - on the surface - appear as potential sources of change in illusion perception. The paper needs some treatment of the known ancillary effects of diffuse NMDAR antagonism to convince the reader that the account provided is better than the other possibilities.

(5) The cross-decoding approach to data analysis concerns me a little. The approach adopted here is to train models on a localizer task, in this case, a task where participants matched a kanisza figure to a target template (E1) or discriminated one of the three relevant stimuli features (E2). The resulting model was subsequently employed to classify the stimuli seen during separate tasks - an AB task in E1, and a feature discrimination task in E2. This scheme makes the localizer task very important. If models built from this task have any bias, this will taint classifier accuracy in the analysis of experimental data. My concern is that the emergence of the kanisza illusion in the localizer task was probably quite salient, respective to changes in stimuli rotation or collinearity. If the model was better at detecting the illusion to begin with, the data pattern - where drug manipulation impacts classification in this condition but not other conditions - may simply reflect model insensitivity to non-illusion features.

I am also vaguely worried by manipulations implemented in the main task that do not emerge in the localizer - the use of RSVP in E1 and manipulation of the base rate and staircasing in E2. This all starts to introduce the possibility that localizer and experimental data just don't correspond, that this generates low classification accuracy in the experimental results and ineffective classification in some conditions (ie. when stimuli are masked; would collinearity decoding in the unmasked condition potentially differ if classification accuracy were not at a floor? See Figure 3c upper, Figure 5c lower).

What is the motivation for the use of localizer validation at all? The same hypotheses can be tested using within-experiment cross-validation, rather than validation from a model built on localizer data. The argument may be that this kind of modelling will necessarily employ a smaller dataset, but, while true, this effect can be minimized at the expense of computational cost - many-fold cross-validation will mean that the vast majority of data contributes to model building in each instance.

It would be compelling if results were to reproduce when classification was validated in this kind of way. This kind of analysis would fit very well into the supplementary material.

Reviewer #2 (Public review):

Summary:

In this paper, the authors investigate the role of NMDA-receptors in recurrent processing. In doing so, the authors present data from two studies, where they attempt to decode different stimulus features, namely contrast, collinearity, and illusory contours. The latter of which the authors claim relies uniquely on recurrent processing. Therefore, to test whether NMDA receptors are particularly involved in recurrent processing they administer a NMDA-antagonist to see whether the decoding of illusory contours is specifically perturbed, and leaves the decoding of other features intact. They further aim to disentangle the role of NMDA-receptors by manipulating visibility and task relevance of the decoded features

In the first experiment, the authors decode two targets, the first was always presented clearly, the second's visibility was manipulated by presenting it after a short lag rather than a long lag (inducing attentional blink), as well as masking the target on half the trials. First, they find for target 1 clear evidence for the NMDA-receptor increasing (rather than decreasing) decoding performance of illusory contours. They move on to analyse target 2 to explore the manipulations of lag and masking. Here they find that masking reduced decoding of all three stimulus features, but only the lag reduced decoding of illusory contours. Importantly, the NMDA-antagonist improved decoding only in the unmasked, long lag condition, in the cluster analyses. However, the interaction with the lag condition was not significant, and the effect on decoding was primarily present in the later decoding time window, and not significant when exploring the peak of the decoding time window.

The second experiment was highly similar, but got rid of the lag manipulation, and replaced it with a manipulation of task relevance. Notably, masking did not abolish the decoding of illusory contours completely, in contrast to the first experiment. More importantly, they find that the NMDA-receptor now clearly increases decoding of illusory contours, particularly when the illusory contours are not masked. No effect of task relevance is found.

Taken together the authors state that evidence is found for NMDA-receptors role in recurrent processing.

Strengths:

This is an interesting study using state-of-the-art methods in combination with drug manipulation to study recurrent processing. Their analysis methods are state-of-the-art, and the question that they are trying to address is topical and interesting to a wide research audience, encompassing both researchers interested in visual perception and consciousness, as well as those interested in perturbed vision as found in psychiatric disorders.

Weaknesses:

The experimental design is somewhat complicated, which can make it difficult to match the authors' claims to the actual evidence that is provided. I have some reservations about the paper which are born out of a few issues.
(1) The title, abstract, and introduction hide their counterintuitive finding of increased decoding, presumably as it was unexpected.
(2) Their analysis choices are sometimes unclear, making it difficult to assess whether the analyses are sensible.
(3) The appropriate tests for the interactions that the authors claim they found are often lacking.

To start off, I think the reader is being a bit tricked when reading the paper. Perhaps my priors are too strong, but I assumed, just like the authors, that NMDA-receptors would disrupt recurrent processing, in line with previous work. However, due to the continuous use of the ambiguous word 'affected' rather than the more clear increased or perturbed recurrent processing, the reader is left guessing what is actually found. That's until they read the results and discussion finding that decoding is actually improved. This seems like a really big deal, and I strongly urge the authors to reword their title, abstract, and introduction to make clear they hypothesized a disruption in decoding in the illusion condition, but found the opposite, namely an increase in decoding. I want to encourage the authors that this is still a fascinating finding.

Apologies if I have missed it, but it is not clear to me whether participants were given the drug or placebo during the localiser task. If they are given the drug this makes me question the logic of their analysis approach. How can one study the presence of a process, if their very means of detecting that process (the localiser) was disrupted in the first place? If participants were not given a drug during the localiser task, please make that clear. I'll proceed with the rest of my comments assuming the latter is the case. But if the former, please note that I am not sure how to interpret their findings in this paper.

The main purpose of the paper is to study recurrent processing. The extent to which this study achieves this aim is completely dependent to what extent we can interpret decoding of illusory contours as uniquely capturing recurrent processing. While I am sure illusory contours rely on recurrent processing, it does not follow that decoding of illusory contours capture recurrent processing alone. Indeed, if the drug selectively manipulates recurrent processing, it's not obvious to me why the authors find the interaction with masking in experiment 2. Recurrent processing seems to still be happening in the masked condition, but is not affected by the NMDA-receptor here, so where does that leave us in interpreting the role of NMDA-receptors in recurrent processing? If the authors can not strengthen the claim that the effects are completely driven by affecting recurrent processing, I suggest that the paper will shift its focus to making claims about the encoding of illusory contours, rather than making primary claims about recurrent processing.

An additional claim is being made with regards to the effects of the drug manipulation. The authors state that this effect is only present when the stimulus is 1) consciously accessed, and 2) attended. The evidence for claim 1 is not supported by experiment 1, as the masking manipulation did not interact in the cluster-analyses, and the analyses focussing on the peak of the timing window do not show a significant effect either. There is evidence for this claim coming from experiment 2 as masking interacts with the drug condition. Evidence for the second claim (about task relevance) is not presented, as there is no interaction with the task condition. A classical error seems to be made here, where interactions are not properly tested. Instead, the presence of a significant effect in one condition but not the other is taken as sufficient evidence for an interaction, which is not appropriate. I therefore urge the authors to dampen the claim about the importance of attending to the decoded features. Alternatively, I suggest the authors run their interactions of interest on the time-courses and conduct the appropriate cluster-based analyses.

How were the length of the peak-timing windows established in Figure 1E? My understanding is that this forms the training-time window for the further decoding analyses, so it is important to justify why they have different lengths, and how they are determined. The same goes for the peak AUC time windows for the interaction analyses. A number of claims in the paper rely on the interactions found in these post-hoc analyses, so the 223- to 323 time window needs justification.

Reviewer #3 (Public review):

Summary:

In this study, Stein and colleagues use a clever masking/attentional blink paradigm using Kanisza stimuli, coupled with EEG decoding and the NMDA antagonist memantine, to isolate putative neural markers of feedforward, lateral, and feedback processing.

In two elegant experiments, they show that memantine selective influences EEG decoding of only illusory Kanisza surfaces (but not contour continuation or raw contrast), only when unmasked, only when attention is available (not when "blinked"), and only when task-relevant.

This neatly implicates NMDA receptors in the feedback mechanisms that are believed to be involved in inferring illusory Kanisza surfaces, and builds a difficult bridge between the large body of human perceptual experiments and pharmacological and neurophysiological work in animals.

Strengths:

Three key strengths of the paper are
(1) The elegant and thorough experimental design, which includes internal replication of some key findings.
(2) The clear pattern of results across the full set of experiments.
(3) The clear writing and presentation of results.

The paper effectively reports a 4-way interaction, with memantine only influencing decoding of surfaces (1) that are unmasked (2), with attention available (3) and task-relevant (4). Nevertheless, the results are very clear, with a clear separation between null effects on other conditions and quite a strong (and thus highly selective) effect on this one intersection of conditions. This makes the pattern of findings very convincing.

Weaknesses:

Overall this is an impressive and important paper. However, to my mind, there are two minor weaknesses.

First, despite its clear pattern of neural effects, there is no corresponding perceptual effect. Although the manipulation fits neatly within the conceptual framework, and there are many reasons for not finding such an effect (floor and ceiling effects, narrow perceptual tasks, etc), this does leave open the possibility that the observation is entirely epiphenomenal, and that the mechanisms being recorded here are not actually causally involved in perception per se.

Second, although it is clear that there is an effect on decoding in this particular condition, what that means is not entirely clear - particularly since performance improves, rather than decreases. It should be noted here that improvements in decoding performance do not necessarily need to map onto functional improvements, and we should all be careful to remain agnostic about what is driving classifier performance. Here too, the effect of memantine on decoding might be epiphenomenal - unrelated to the information carried in the neural population, but somehow changing the balance of how that is electrically aggregated on the surface of the skull. *Something* is changing, but that might be a neurochemical or electrical side-effect unrelated to actual processing (particularly since no corresponding behavioural impact is observed.)