sampendu.bsky.social
Kiwified neuroscientist & perception researcher at the School of Optometry & Vision Science at Waipapa Taumata Rau | University of Auckland, Aotearoa New Zealand. Lab website: sampendu.net
#UltimaDragon
302 posts
1,502 followers
889 following
Getting Started
Active Commenter
comment in response to
post
And yes I'll be posting more about this in the coming months and years. This survey stuff is really only the beginning. Lots of experimental studies to follow eventually!
comment in response to
post
Yes there is a really fascinating theory of mind element to all this. We have recurring (nightmare) discussions on this place between cognitive scientists interested in this question & there is a significant contingent of our kind that remain unconvinced that people can differ so dramatically... 😏
comment in response to
post
I've been using three cups and a coin but the doors might be a good idea. I could ask my daughter to create this out of Lego complete with two goats and a car 😉
comment in response to
post
I've been running a little demo of it in my teaching during the past few years as an example of how our intuitions about probability are a mess... It's worked almost too well - if I believed in the gambler's fallacy I would expect it would have to fail this year 😏
comment in response to
post
Doesn't work on phones I'm afraid. Computer or laptop is ideal - not sure if they support tablets.
comment in response to
post
What I really find shocking is that some people apparently see things that aren't there when their eyes are open! I have mental images, but I don't -see- them.
comment in response to
post
This looks really cool - I'm looking forward to reading this when I finally have time for anything again... 😉
comment in response to
post
Anyway, going home - but that's a conversation to be continued, ideally over 🍺 in Brisbane (or maybe I'll even make it to Sunshine Coast sometime... 🌴)
comment in response to
post
That's again where the dot variant is easier to work with (but that still wouldn't help in this case as you need the side by side comparison configuration).
comment in response to
post
I think you need to test everything 😉. That would be another way to do it although I'd also be very surprised if the ML becomes more like the Delboeuf in that case. But yes worth testing!
comment in response to
post
But there could've been many reasons for that. I am sure it would be worth trying that again.
comment in response to
post
One suggestion we're making is to retest the ML with a circular design where you effectively have a crown of arrowheads. It would be interesting if this they correlates with circle width discrimination. The part this isn't mentioning is that I tried such a stimulus before & it was awkward...
comment in response to
post
And I reiterate that we did matched a lot of things here. For instance, the arrow head length in ML is effectively equivalent to the target-inducer distance for small inducers in Delboeuf and Ebbinghaus. Not sure how much more you can match these given how differently they are configured.
comment in response to
post
Yes there can be situations when signing could lead to a potential COI in itself. I had that come up some months ago - although I'd be generally inclined to just say no to reviewing then anyway. Of course, there is the other side of the coin with double-blind reviews where you're not allowed to sign
comment in response to
post
As in filtering my rectified RFs... One might even call it Fourier transform (as a certain person who emails me once a year with comments on my research claimed it's "all in the stimulus itself") - except that it is largely unaffected by inverting the polarity.
comment in response to
post
If you look at our Figure 1G you can see just how much stronger but far less variable it is than the Ebb or Del illusion. And the Ponzo is way weaker (although that had been well known for some time)
comment in response to
post
Yes similar to how we did it in Tasi's study you reviewed as part of his PhD. Of course, there have been extensive studies of the ML parameters already but never side by side with Ebb/Del afaik. On that note though it's also interesting just how whoppingly different they are in magnitude & variance.
comment in response to
post
Most of it is attractive contour effects.
comment in response to
post
Should also note that our own data suggest Ebbinghaus and Delboeuf (not so clear about Ponzo) are pooling effects. Inhibition may play a role but it's not as clear cut. Delboeuf seems to extend out for >20 dva so that could be extraclassical inhibitory RF effects perhaps. (Should model this again).
comment in response to
post
More importantly, this discussion is all creating a bit of a false dichotomy. I'd say all these illusions presumably involve multiple factors, both low & high-level. The dot variant of the ML I think is tapping more into low-level properties as we suggested here: doi.org/10.1038/s420...
comment in response to
post
We tested them at common retinal location though so not sure how this adds. Ponzo is an interesting one because it is clearly substantially affected by non-local parameters (as our next publication will show hopefully very soon... 😉)
comment in response to
post
Long ago I vowed not to play the reviewer-guessing game so I refrain from doing that here. Even though there are some good indications given how much they seem to know about the details in a certain author's papers... Although then again, they also didn't seem to know those papers all that well 😳
comment in response to
post
This is just one of the reasons I sign my own reviews. I'd say it makes my reviews more considered as it forces me to think how to support my statements. I also feel it makes my reviews more civil. But most critically you know it was me, so you can direct your ire accordingly if you feel I'm wrong.
comment in response to
post
...but even if it is true, you should tell us why that is so not simply claim it is. And $%@#ing -read- the manuscript so you don't ask questions that have clearly already been addressed in previous versions. </rant>
comment in response to
post
I will refrain from descending in a deep rant on this but just want to say if you peer review something you have strong reservations too, back up your statement with concrete points don't just complain about one-sidedness or lacking debate. We don't think this is true about our study but...
comment in response to
post
I must say I appreciated the even-handedness of this action editor in the face of what I think might have been the most ill--informed & biased peer reviewer I have encountered so far (and I've seen some doozies). I got the impression the editor was fed up with them too by the end of it.
comment in response to
post
While this review-mandated change post-prereg was messy, the final results speak for themselves. As it happens, this extra data allowed some exploratory comparisons that further underline that Adjustment tasks are worse than 2AFC tasks for measuring illusions. Puts previous studies in context...
comment in response to
post
That required some funky coding in a special Firefox extension that someone had set up for me. Unfortunately I don't remember exactly how this works any more 😏
comment in response to
post
I am not in favour of blocking her as I think that opens a pandoras box & it's hard to justify. For a while I had a filter setting in my browser that automatically hid her comments but at this point my brain just blanks them out... :P
comment in response to
post
Yes of course.
comment in response to
post
Vision Research - although if they're true to their word & indeed accept it with double the sample size irrespective of the outcome so you'd find it out soon anyway... 🙂
comment in response to
post
Afraid we cannot really do that. I mean we're pushing back in our response but the editor already said they won't send it back for review again, so the reviewer won't see it unless they read this thread & hang their head in shame as they should 😉
Anyway, we're up to N=19 now, just 5 more to go...
comment in response to
post
Agreed - that is why these forms should be changed. We changed ours about 7 years ago, and neither ethics committees nor participants seem concerned by being informed that MRI (or EEG/MEG) is potentially identifiable.
comment in response to
post
Thanks for that link - sounds interesting also scientifically! In a way that makes sense. Brains & connectomes should be individualised because we are all of us individuals 🙂
The thing that struct me most about our analysis is just how basic those data are (and how simplistic the analysis).
comment in response to
post
🤣
comment in response to
post
Diffusion is structural (sort-of?) & I think people tend to be more aware of that. With fMRI data I feel less so. And I agree some people are generally aware of these issues already - but it is my impression that a lot of people, even in cog neuro, are not.
comment in response to
post
We also considered doing a similar analysis on a larger resting-state dataset one of my coauthors collected. But apart from the energy/capacity limitations I already mentioned, I also decided that this actually already a grey area as those participants didn't really consent to my doing this.
comment in response to
post
Agreed, it goes beyond scans. In fact, in my feeble attempts to make this a bigger study that perhaps could be more publishable in its own right, I also did a comparable analysis using our EEG data. But that data set isn't big enough for this to be a particularly good example so I dropped it.
comment in response to
post
In my most recent fMRI ethics, I've also included a statement about this in the consent process. I believe in data sharing. But the standard forms we have (& it's not just here) contain phrases like "cannot be traced back to you". That is quite frankly a lie.
comment in response to
post
I should say that one inspiration for my actually trying this analysis in the first place was some conversations I've had in the past with smart & knowledgeable colleagues who said explicitly that normalising fMRI data removes any risk of reidentification. Clear that isn't true.
comment in response to
post
Yes & that is fine (provided reasonable steps are taken to minimise the potential for abuse). My main issue is that "informed consent" means that participants should be made aware of the risks so they can make an informed decision. And that many researchers I met seem blase about the risks.
comment in response to
post
Yes of course & this is also why we can even do the correlation analyses we did. But it seems to me many people intuitively assume that normalisation blurs patterns sufficiently to be hard to distinguish - as would probably be true when normalising more basic info (e.g. single-value measures).
comment in response to
post
I wouldn't say "impossible" - that would entail a lot more trying. But it would require a lot more work that I don't have capacity for. I feel it's more important to put it out there rather than dealing with lengthy discussion of minutiae in peer reviews & doing countless revisions etc...
comment in response to
post
(Though I entirely understand it wouldn't be a solution for the kind of anatomical info you've been working with. So this all really depends a lot on the research question & the nature of the data).