ha, and now I notice a typo: should be instead "_of_ definitions _on_ the tasks" not "on definitions of the tasks". Curse you, attention, for this slip!

It is! This must have been a tremendous amount of work for you, so congrats! I'm eager to read the rest of the comments (as soon as I'm done with grading, so probably in a year or two)

Like other layman terms it could be useful in the classroom or in outreach, but it's not useful as a target for theory-building. And it's not about ambitions ("not everyone needs to work towards this ambitious goal" says the response) - it's like building "a theory of the brain". Makes no sense.

... I still stand by our point that "Focusing instead on definitions of the tasks that observers perform and building computational models of their decisions ... seem to be a more promising avenue". Attention (like perception or memory) is just too broad and vague to have a meaningful theory.

And the next step? Full voxel-level modeling.

Recent numerical advances cracked the scalability barrier. Voxel-level hierarchical modeling is now feasible, revealing just how punishing traditional multiple-comparison adjustments really are.
arxiv.org/abs/2511.12825

Its a first paper from Ekaterina Andriushchenko's PhD, so all kudos to her for the work! 5/5

We also found that prolonged maintenance strengthens serial dependence, which is nice. But the lack of an attention effect, despite better precision for cued items, creates problems for existing models and suggests that changes in item representation may be responsible. 4/n

It looks like all the previous studies looking at attention effects confounded attention with reporting. And we used 80% cueing paradigm comparing cued and (non)reported items with non-cued and (non)reported ones. The results show a reporting effect but not a cueing effect. 3/n

Our results challenge previous claims that prioritization via attentional cueing modulates serial dependence (SD). Multiple previous studies suggested attended items/features produce stronger SD, but in two experiments we found no effect of precueing or postcueing—and we think we know why! 2/n

Although I probably wouldn't be putting all these stats (like CIs for eff. sizes) and try to write it a bit differently. I don't know, feel quite readable to me, but maybe I'm biased.

I actually do write like that (or try to) and hate stats in tables. It's much easy to interpret things when you see the stats by the statements rather than deciphering which line in the table to look at. And for a bird's-eye view you have the figures and the intro/conclusions in each paragraph.

and a reply to this reply www.sciencedirect.com/science/arti...

ahhh, now I see why: the last author of the original paper replies to the reanalysis. I guess he wasn't in favor of the retraction.
link.springer.com/article/10.3...

at least a warning label would be nice