Abstract
Research suggests that visual working memory is sensitive to the semantic richness or meaningfulness of the stimuli used (see Brady, Konkle, & Alvarez, 2011). Differences in meaningfulness are often treated as categorical by contrasting abstract shapes with images of real-world objects. Can we instead quantify image meaningfulness? Gilman, Ware, & Limber (2010) asked participants to list associations that came to mind when viewing images from earlier working memory trials and found that color photos elicited significantly more and more-varied associations than grayscale drawings. Colorful shapes showed a similar advantage compared to grayscale shapes. Their stimuli were not assessed for recognizability, which could influence the retrieval of meaningful associations. The present study contrasted well-tested images of real-world objects for associational differences. Two image sets were created, each containing 22 color photos of big and small objects (Konkle & Oliva, 2012), 10 color drawings (Rossion & Pourtois, 2004), and 10 partial line drawings (Hayworth & Biederman, 2006). Objects featured in a color drawing in the first set were featured as line drawings in the second and vice versa. Following training, 41 participants named each image and provided their image-recognition confidence and related associations. Photos elicited the most associations (M=3.61), followed by color drawings (M=3.55) and line drawings (M=3.45); this difference was signficant, F(2,1666)=6.136, p< .01. Surprisingly, higher recognizability was weakly but significantly correlated with fewer associations provided, with the greatest effect (r=-.15, p< .01) for line drawings, the most variable image type (M=87, sd=18 versus M=95, sd=12 for color photos and drawings). These participants had no time restrictions, unlike Gilman, Ware, & Limber's (2010), and provided narrative associations not obtained in the earlier precedent. Despite the puzzling negative correlation between recognizability and associations offered, our results support using associations for quantifying image meaningfulness, preferably with specific requests for words evoked by the image rather than general associations.
Meeting abstract presented at VSS 2016