You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

 
Poster C56, Thursday, November 9, 10:00 – 11:15 am, Harborview and Loch Raven Ballrooms

Clustering Abstract Concepts into Distinct Categories

Catherine Walsh1, Stephen J. Gotts1, Alex Martin1;1Laboratory of Brain and Cognition, National Institute of Mental Health

Object concepts fall naturally into easily identifiable and distinct categories such as animals, tools, furniture, and flowers. In contrast, our large corpus of abstract concepts do not easily fall into distinct categories.  In order to make progress on understanding the neural underpinnings of abstract concepts it seems that the first questions that must be addressed concern distinctions among abstract concepts.  Do abstract concepts show a categorical structure?  Do clear boundaries exist between these different categories?  In this study we sought to address these questions using a novel behavioral sorting paradigm and a measure of automatic semantic priming.   We completed two experiments. In the first experiment, subjects (n=20) performed a computerized clustering task using 35 words that referred to abstract concepts.  All words were rated very low on concreteness and imageability based on standard published norms (Coltheart, 1981; Brysbaert et al., 2014). Participants were presented with the words randomly distributed around a circular array on a computer screen and asked to group them in any way they saw fit. A k-means clustering analysis identified four consistent and robustly separated abstract word clusters, which we called: 1) cognition and thoughts, 2) negative qualities, 3) morality and personality, and 4) aesthetics and morals.  When compared to chance (determined through random shuffling of rows/columns for each individual subject in permutation testing, 5000 iterations), within-cluster distances were all significantly smaller than chance (P<.0002, for all) and between-cluster distances were larger than chance (P<.002, for all but cluster 2 with cluster 3, P<.13). These results indicated that abstract concepts can be reliably categorized.   We then used these clusters in an automatic semantic priming paradigm with 15 new subjects in order to determine whether these distinctions were processed automatically. Prime words were briefly presented (100 msec), pattern masked (50 msec followed by 100 msec blank screen), with the probe word presented for 250 msec and participants judging whether the probe was a concrete or abstract word. The prime-probe relationship did indeed influence response time, but, surprisingly, not in the expected direction: within-cluster pairs yielded significantly slower, rather than faster responses relative to the across-cluster pairs (paired t-test: t=3.259 (df=14), P<.01).   We are currently evaluating this inhibitory priming effect using different priming paradigms (lexical decision; pleasant/unpleasant judgements) and different subsets of abstract words.

Topic Area: Meaning: Lexical Semantics

Back to Poster Schedule