Hello everybody!
We have a categorization task which loads heavily onto procedural learning systems, but we're looking to see if a manipulation we introduced recruits any additional neural regions. We've done a univariate analysis, and want to, now, do an exploratory, whole-brain searchlight analysis.
One thing that we can't decide on is what searchlight radius to use. We've scoured high and low, and can't seem to find a single article comparing searchlight radii, so we're not sure what to do!! Some papers use multiple sizes, some just use one, but it's not usually rationalized. We found the Kriegeskorte article in which they suggest 4 mm may be appropriate, however, the literature in our field seems to use much larger radii (8+ mm), and one paper in particular suggested 4 mm may only be appropriate for things like low-level visual processing, but not high-level conceptual information.
Any references or advice on how to pick an appropriate searchlight size would be greatly appreciated.
With much thanks,
Kade
There is no universally 'optimal' radius size, but you could set it based on what has been used in previous literature in your field, or by what works best on simulated data (e.g., as in Article Information-based functional brain mapping
) or define it based on different (independent) data. Other potentially useful papers:Article Searchlight analysis: Promise, pitfalls, and potential
Article A comparison of volume-based and surface-based multi-voxel p...
Sorry for the slow reply Tijl Grootswagers ! I really appreciate the feedback. I actually rather like these papers, but only the Kriegeskorte et al. one really provides any concrete advice on searchlight radius (4 mm). In hindsight, I think I should have provided more details in my question!
In the Etzel et al. paper you mention, for example, they suggest "no single searchlight radius will be universally optimal or sufficient" (although the Oosterhoof et al. paper does claim accuracy increases with larger searchlight size).
Therefore, I'm more interested in resources specific to the category learning literature - a token study, or perhaps a comparison of radii for the field, specifically. Anything to point me in the right direction! Thank you again!!
- Badges
- Science topic
- Similar topics
- Psychology
- Mental Processes
- Learning