Discussion: stimuli/standard-oddball

[lrudelt]

Hi everyone,

let's use this discussion to gather comments and thoughts on the current stimulus design that came up during your analyses. The idea is to use our experiences from preliminary analyses to optimize the stimulus design.

Some things I noticed during analysis

1. The standard stimulus is always the same, only the deviant one is varied. If we would swap deviant and standard stimulus half-way in the standard-oddball (e.g., like Hamm et al. PNAS 2021), this would not only allow to compare standard/deviant/control for the same orientation, but also allow to compare deviant responses for different prediction context (a la Furutachi et al., Nature 2024).

2. Different prediction context could be very interesting for the contrast oddball, i.e., which ROIs show responses depending on which stimulus is the standard direction.

3. As far as I understand, there is not really a control condition for the motion deviant (static gradient in standard orientation). To understand how much predictions alter the response, it might be good to include a short static gradients condition.

4. 45 degree responses are very similar to 0 degree, thus it is hard to interpret (not clear if it is oddball, or redundant). Instead, we might want to preserve the same angular relationship between all three stimuli (standard, deviant 1, deviant 2). To this end, one could choose all three stimuli with a 120 degree orientation difference. I attached an image to illustrate this idea. (In this case, only 12 orientations should be tested during the orientation tuning component to go in 30 degree steps and hit the same stimuli used for the oddball).

5. I noticed a lot of instances where an oddball is immediately followed by a second or even a third oddball stimulus. Also, the oddball block in the 2025-05-08 experiment starts with a 90 degree oddball. Since we only have 40 trials per oddball, it might make sense to include a minimum for required repetitions of standards before an oddball is allowed to occur. (e.g., enforce three consecutive standards, after which a deviant can occur with a certain probability - see Hamm et al., PNAS 2021).

Illustration of an oddball design with similar difference in orientation and changing standard orientation after half of the block.

[jeromelecoq]

thanks for this detailed comment!
An important detail is that the oddball is meant to occur every 11s in all session types. This is very different than 1/8 of the drifting gratings. This was chosen to match oddball rates from more complex context that are already published.
I appreciate this is lower than Hamm et al but our intent is to facilitate comparison across all 4 sessions types : The same oddballs is presented at the same rate in standard oddballs, sensory motor oddballs and sequential oddball. It is also calculated to mirror the presentation rate in control blocks.

We discussed the need to switch the drifting grating and both Hamm's lab and our recent work with Andre Bastos and Alex Maier labs found the same results for the two different orientations. I recollect we asked this very specific questions to Jordan and we agreed having more different types of oddballs was more important than varying the baseline gratings.

In addition, it is harder to change this in sensory motor context where the vertical orientation is easier to integrate perhaps.

Do you have specific analysis that critically depends on this ?