| Tweetovi |
| Micha Heilbron proslijedio/la je tweet | ||
|
Matthias Fritsche
@matfritsche
|
27. sij |
|
Excited to share the last project of my PhD "A Bayesian and efficient observer model explains concurrent attractive and repulsive history biases in visual perception" biorxiv.org/content/10.110… with @flodlan @EelkeSpaak @predictivebrain
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
24. sij |
|
|
||
|
Micha Heilbron
@m_heilb
|
23. sij |
|
Als je in de UK ernstig ziek wordt tijdens je PhD wordt je beurs mooi stopgezet. Is het echt overdreven om dat barbaars te noemen?
nature.com/articles/53949…
(Om over betaald zwangerschapsverlof etc nog maar te zwijgen...)
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
18. sij |
|
There is behavioural evidence for an effect of pronounceable nonwords.
We wanted to maximise our contrast so we compared frequent words with unpronounceable nonwords - confounding lexical and phonological info. in the process. Res. in fig 4 consistent w/ both interpretations
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
... why this is an important debate in its own right.
It's just that there is still a yawning gap between the type of important, but subtle theoretical questions debated between modellers, and the type of questions we can currently (even begin to) tackle with neuroscience 😔
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
stories about functional asymmetries between cortical layers and the ability of fMRI to pick those up in a meaningful way.
RE: interactivity, the interactive models I used for the simulations are not very stable, prone to hallucinations & runaway effects, so I can totally see...
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
... addressing the latency question. However, I don't see how it would get at the interactivity issue, because there is still no way to distinguish the content of the bottom-up and top-down signals.
One way you might is using laminar fMRI, but that does require buying into....
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
the effect occurs at 150 ms or 1500 ms after onset. This is an important limitation because if it were the latter, calling the representational enhancement a *perceptual enhancement* might become problematic...
Currently planning a follow up with MEG, which can help ...
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
...namely: can we find evidence for a neural feedback effect?
As to the nature of this feedback, I'm afraid the data don't speak to this, as we can't distinguish between info being sent forward and information merely represented
With the current data, we don't even know if....
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
Right - thanks for clarifying. I only have a passing familiarity with ART and didn't realise it was not 'interactive'
There are clearly multiple debates going on (e.g. whether there is feedback at all, what is the source of the feedback, etc). We ask the most basic question...
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
Hi! Thanks for the interest — and for the interesting question.
Not sure if I fully understood it though. Are you here contrasting interactive accounts with ART? If so, could you elaborate a bit more on this theoretical difference?
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
thanks Anil, means a lot :)
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
... (although the paradigm was perhaps not perfectly set up to measure these in the first place). So there was a decoding effect without amplitude difference - making it neither a classic 'prediction error' or 'sharpening' effect I suppose
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
Hi! Thank you. Yes, Blank & Davis '16 very relevant here. Can't believe we forgot to cite it!😬(It figured prominently in the grant application I wrote to start this PhD in the first place.) 1 thing to keep in mind re: the discrepancy is that we did not find univariate effects...
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
thank you :)
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
17. sij |
|
Thank you Florence :)
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
16. sij |
|
Thank you!
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
16. sij |
|
Hi! Thanks for the interest. I’ll get back to you via dm :)
|
||
|
|
||
| Micha Heilbron proslijedio/la je tweet | ||
|
Jaan Aru
@jaaanaru
|
16. sij |
|
Knowledge enhances perceptual processing. Here @m_heilb @D__Richter @MatthiasEkman and @flodlan show that word context boosts the representational fidelity of individual letters in early visual cortex & increases the coupling to the reading network. Great! nature.com/articles/s4146…
|
||
|
|
||
|
Micha Heilbron
@m_heilb
|
16. sij |
|
Alright, that’s the gist! Do check out the paper if you’re interested, there are many more details and some 11 supplementary figures I omitted for now 🙃
Many thanks to my collaborators and especially to my supervisor, @flodlan for his input and guidance during the project
|
||
|
|
||