Please enable JavaScript.
Coggle requires JavaScript to display documents.
ERC NEW FOCUS: Objective recognition, feature diagnosticity, (INDIVIDUAL…
ERC NEW FOCUS: Objective recognition, feature diagnosticity,
INDIVIDUAL DIFFERENCES -Christian suggests to keep this focused on autism, not to give the impression of a fishing expedition
autism, big five etc., Alexithymia
-
-
-
ONE POSSIBLE HYPOTHESIS: As we move up the visual hierarchy, representations of diagnostic visual features go up while representations of non-diagnostic features go down
PUT DIFFERENTLY: object representations in IT might entail the representation of the set of the object's most diagnostic visual sfeatures
Directly relates to the Sigala & Nikos K. Logothetis Nature paper (2002).
Their main conclusion: "Our finding that inferior temporal cortex neurons can selectively represent visual object features that are important for the task at hand provides a mechanism through which feature diagnosticity shapes the encoding and the perceptual interpretation of visual objects"
The relation of this finding to more recent fMIR studies is reviewed in "The Contribution of fMRI in the Study of Visual Categorization and Expertise" chapter by Sigala
So far, aside using the bubbles approach,
Fundamental idea can be linked to Attneave:
Enhanced encoding of diagnostic visual features from lower to higher visual areas can be seen as an "economical" information encoding strategy
Enhanced encoding of diagnostic visual features from lower to higher visual areas can be seen as an "economical" information encoding strategy
Tricky issue- which Task?
Feature diagnosticity can only be assessed in the context of a specific task: e.g. animate vs inanimate or "object naming". I need to think carefully about which task to use - which needs to be maximally relevant and feasible (in terms of the nr of trials)I'm tempted by the object naming task as it most directly relates to object recognition. Instead of naming it could also just be a Recognize/Dunno response. If R, participants need to answer a multiple choice question. Pro for naming, might be quicker. Con, I would need to record the sounds and analyze them (could try to get AI to do this, or a hiwi...) Using the 150 semantically dissimilar dataset is tempting (although I think I might drop about 50 items that are too easy based on shape - e.g. the pyramid and the lap)
- It would cover a big space of object / be condition rich
- Therefore, findings will be more representative for object recognition in general
- Having a lot of different stimuli might reduce a focus just on the stimulus contours (as there are too many to keep active in STM)
- It will allow me to also look into the relation to semantic representations (note, I can also look at semantic priming during the pschophysical experiment)
- Con, it will be a long psychophysical experiment spanning several days and sessions (minimum, ~10 hours, 5 sessions)
Use Kay Images as localizer and present full object images in an interleaved design (loc-obj-loc...). Diagosticity of each object feature (e.g. 1000) is determined after the fMRI experiment, behaviorally (rev corr). Based on the localizer, we can estimate for each voxel how much it prefers each of the 1000 features. Within each area we can assess if voxels selective to diagnostic features are activated more (or less?), and how this varies across areas.
Can I assess semantic feature diagnosticity? Of objects? E.g. by showing a list of words followed by two object images, then select best fitting visual object. Then use Rev Corr....