Semantic properties and categorization norms for the 260 Snodgrass and Vanderwart objects: A 45-year conceptual...

We present a new set of English property norms for 260 object concepts based on the standardized Snodgrass and Vanderwart (1980) picture set. For each object, 100 participants provided a basic-level label (dog), a superordinate category (animal), and three features (bark, tail, fur), yielding a dataset of 78,000 features. Our norms differ from other datasets in four important ways: they (1) probe basic-level information, (2) separate taxonomic and feature information, (3) use open-ended responses for natural descriptions, and (4) include a larger number of responses per object (100 vs. ~ 30 participants in other norms). We analyzed feature statistics such as frequency, distinctiveness, and co-occurrence, and contrasted our norms with those of McRae et al. (2005), CSLB (Devereux et al., 2014), and Hovhannisyan et al. (2021). Compared to picture-based norms, our data-derived from black-and-white line drawings-elicited more diverse features and aligned more closely with language-based norms, particularly CSLB. We assessed the generalizability of our norms using an object-property congruency task, where 144 participants judged whether properties (basic level, superordinate, and features) were related to objects. Objects were shown in three picture formats with increasing ecological validity: (1) colored line drawings, (2) realistic photographs, and (3) realistic photographs of objects in scenes. We then contrasted these data with those of Antal & de Almeida (2024) employing the original line-drawing set. Agreement rates for object-property pairs remained high across picture formats and property types. Bayesian inference revealed minimal variability in congruency judgments across picture formats, with responses tightly clustered around zero. Results show that our norms are generalizable to realistic visual stimuli. Norms are available at https://osf.io/c6brw/overview .