🚨New paper alert! What would you call this object? Most folks would say "guitar". You could be less specific ("musical instrument") or more ("Fender Stratocaster"), but we tend to use the mid-level (the "basic level" coined by Eleanor Rosch). Why? 1/ #neuroskyence #visionscience - ThreadSky

mgreenephd.bsky.social • 53 days ago

🚨New paper alert! What would you call this object? Most folks would say "guitar". You could be less specific ("musical instrument") or more ("Fender Stratocaster"), but we tend to use the mid-level (the "basic level" coined by Eleanor Rosch). Why? 1/ #neuroskyence #visionscience

Comments

tsawallis.bsky.social•53 days ago

But what about the "cherry red, maple board strat with whammy-bar" level?

mgreenephd.bsky.social•53 days ago

In this work, we examined how the brain clustered objects at three levels of specificity (e.g., musical instruments, guitars, Fender Stratocasters) using regression analyses and time-resolved EEG. Participants were either attending to the basic-level object categories, or subordinate. 2/

mgreenephd.bsky.social•53 days ago

We found that, as with behavior, the brain favors the basic level. Although some task-related differences show up between 200-300 ms after stimulus onset, the brain shows far more basic-level grouping, even when people attend to the subordinate category! 3/

mgreenephd.bsky.social•53 days ago

Each subordinate level had 40 different exemplars, and we decorrelated the RDMs between the three category levels. Therefore, the very early basic-level grouping is unlikely to be driven by low-level visual features. Paper link: https://rdcu.be/d5oVo 4/4 #neuroskyence #psychscisky #visionscience

rsek.bsky.social•53 days ago

Michelle: do you expect something similar to happen when subjects are shown scenes?? That is, would basic level categorization take priority??

mgreenephd.bsky.social•53 days ago

It's an open question, but I suspect so. We did a series of experiments testing the automaticity of object and scene categorization using a Stroop-like paradigm. It was more difficult to classify object & scene names when placed on an incongruent image. https://jov.arvojournals.org/article.aspx?articleid=2193927

rsek.bsky.social•53 days ago

We are also doing some stuff with scenes from the SUN database. After my initial selection of candidate scenes (based on a version of Bainbridge’s definition of a scene), it became very clear that some scenes are very easy/fast to categorize at a basic level, but others are not.

rsek.bsky.social•53 days ago

Thanks, Michelle. I totally missed that paper. Incidentally, I second your suspicion about scenes.

thomasserre.bsky.social•53 days ago

Hi Michelle -- congrats on your paper! Many years ago, we showed that task difficulty is a significant confound in studies claiming that some categorization levels take precedence over others. https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1004456

thomasserre.bsky.social•53 days ago

As an illustrative example, we had shown that we could sample natural scenes that would make the superordinate arbitrarily faster or slower than the basic level.

thomasserre.bsky.social•53 days ago

Please note that I have no stake in this but is it possible that the same confound applies to your study?

mgreenephd.bsky.social•52 days ago

Thomas: thanks for bringing this up. Important issue, but I don't think it applies here. Our study examines the similarity of EEG signals at each category level (super, basic, sub) when the variability from the other levels has been accounted for.

mgreenephd.bsky.social•52 days ago

We main result is that neural signals cluster along basic levels early and primarily, and this is true even when observers are attending to the subordinate level category. If discriminability were driving the result, we'd expect more clustering for subordinate because they share more features.

mgreenephd.bsky.social•52 days ago

Behaviorally, observers were more accurate for the basic level, but this doesn't account for why the basic level grouping persists when folks perform the subordinate task. I'm happy to discuss more if this misses the mark.

andrewmckee13.bsky.social•53 days ago

Just curious. Do you know if expertise in a field would change the results?

mgreenephd.bsky.social•53 days ago

There's good reason to believe that it would. Some nice work in the 1980s by Pierre Jolicoer, Jim Tanaka, and other folks demonstrated that experts' basic-level is shifted towards the subordinate.

tinareynolds.bsky.social•52 days ago

📌

Comments

Posting Rules

Reply