Pascal Mamassian

Everything you always wanted to know about color, but were afraid to ask.

A superb, highly didactic video, now in English!

Tout ce que vous avez toujours voulu savoir sur la perception des couleurs sans jamais oser le demander.

Une superbe vidéo très didactique d'Alessandro Roussel.

Experimentology cover: title and curves for distributions.

Experimentology is out today!!! A group of us wrote a free online textbook for experimental methods, available at experimentology.io - the idea was to integrate open science into all aspects of the experimental workflow from planning to design, analysis, and writing.

Surely the point of the reviewing process is not just to get published at any cost, but also to convey some novel, reliable, and meaningful information.

(6/6)

A final note about the advice to authors to run their manuscripts through these tools before submission. I suspect that at one point these tools will have a "fix it for me" button, and unscrupulous authors might be tempted to click it and not checking what has been done in details.

(5/6)

The production of an argument map (5) is nice and it might find its way in the future generations of word processors.

(4/6)

The point about originality (2) is of course important, but novelty is not enough. AI can tell us how similar a manuscript is to the existing literature, but probably not whether the way that it is novel is actually interesting for the field. New and useless is of little value.

(3/6)

I feel that 3 of the 5 listed points are borderline of what we ask to reviewers: (1) screening for missing elements, (3) conducting a citation integrity, and (4) checking data consistency could possibly be done either at the submission filtering stage or at the production stage.

(2/6)

Thanks for the very nice survey of different uses of AI for peer reviews! And thanks for re-asserting to be transparent about the use of these tools, I think this is critically important.

(1/6)

I don't deny that some aspects of the review process could benefit from LLMs (improving the English, toning down the comments, ...). But to appreciate whether a manuscript is providing original results that can benefit the community, I'm not convinced that these models can be very useful.

As editor-in-chief of a journal (Perception), my expectations from reviewers are actually quite high, because I (naively?) believe that reviews can improve the quality of a manuscript. But it's true that there is a wide range of quality of reviews out there.

Using AI for peer reviewing is like using a microwave to reheat an old meal - Pascal Mamassian, 2025

"As appealing as they can be, Large Language Models are as useful to scientific research as microwaves are to fine
cuisine."

doi.org/10.1177/0301...

Happy Birthday Gerald Westheimer!

And with a few days too early, Happy Birthday Suzanne McKee!

Amongst many other things, Gerald and Suzanne worked on one of the strongest binocular stereopsis effect, the "Westheimer-McKee phenomenon" (see e.g. doi.org/10.1016/j.vi...).

#VSS2025 Honoring Eileen Kowler’s Legacy

We’re saddened by the loss of Eileen Kowler, a past VSS President and longtime contributor. Join us in honoring her legacy at a special symposium: "Honoring Eileen Kowler: Eye Movements as Windows to the Mind", May 16, 5:15pm, Talk Room 2

modelling of continuous confidence judgments

The model can also fit continuous confidence ratings without binning these ratings.

Matlab code here: github.com/mamassian/cn...

(4/4)

estimation of confidence boost parameter

Critically, the model includes a parameter that reflects whether confidence relies on the same information that was used for perceptual decisions, or some new information. We describe two ways to estimate this Confidence Boost parameter.

(3/4)

effect of sensory criterion on confidence efficiency

The CNCB (Confidence-Noise Confidence-Boost) model can be seen as an alternative to the popular meta-d’ and M-ratio analyses. It can take more than two stimulus strengths and is a bit more tolerant to sensory biases.

(2/4)

relationship between sensory evidence and confidence evidence

If you work with perceptual confidence judgments, you may be interested in our CNCB model of confidence ratings. Joint work with Vincent de Gardelle.

Uncorrected proofs here:
dx.plos.org/10.1371/jour...

(🧵 1/4)

Novel color via stimulation of individual photoreceptors at population scale Image display by cell-by-cell retina stimulation, enabling colors impossible to see under natural viewing.

Impressive work to stimulate only the M-cones of the retina and make green appear greener than ever.

Coming (not too) soon to a theater near you!

www.science.org/doi/10.1126/...

Thank you Karl Gegenfurtner and Doris Braun for two fantastic talks last week in our department!

It’s a shame because the study is otherwise spectacular, but I guess the title would have been less attractive if it had been

“...high-order thalamic nuclei gate criterion setting…”

...like the will to please the experimenter, how attentive the participants were, or maybe even how thirsty they were. At the very least if there was only one contrast level, the authors could have separated sensitivity from bias, but that’s not possible here.
...

I think Andrei Gorea would have agreed that a threshold estimated with a yes/no paradigm using simultaneously multiple contrast levels is just uninterpretable. That threshold is a mixture of the stimulus contrast necessary to report “seen” *and* a whole lot of other things...

Hakwan, my point was simply that the authors “missed” a good opportunity to properly measure the threshold for seen/not-seen, and their discussion of consciousness may well be a “false alarm”.

Erroneous clock time readings from different LLMs

Why is it so difficult for Large Language Models to properly read time?

arxiv.org/abs/2502.05092

Human high-order thalamic nuclei gate conscious perception through the thalamofrontal loop Human high-order thalamic nuclei activity is known to closely correlate with conscious states. However, it is not clear how those thalamic nuclei and thalamocortical interactions directly contribute t...

Congratulations, clearly a tour de force!

However, some may regret that a yes-no task (“seen”/“not-seen”, subject to criterion setting) was used rather than a proper psychophysical paradigm.

www.science.org/doi/10.1126/...

Psychophysics has traditionally been more interested in sensitivity than bias, so in that respect, I guess the 2IFC is better for psychophysics.

But indeed, this requires more extensive discussions!

I like your key example, but as inattentional blindness studies have taught us, even when/where you look, you may not perceive. So maybe limited accumulation time (between 2 saccades) better reflects the “real” world.

Psychophysics has traditionally been more interested in sensitivity than bias, so in that respect, I guess the 2IFC is better for psychophysics.

True, the 1-interval is more interesting for metacognition, and the 2-interval for psychophysics!

For a temporal 2IFC, the timing of the intervals is determined by the experimenter, so maybe you can change this timing to give more or less time to the participants to integrate stimulus information.