Gongting Wang^1,2 (), Ilker Duymaz², Matthew Foxwell⁴, Micha Engeser², David Pitcher⁴, Radoslaw Martin Cichy¹, Daniel Kaiser^2,3; ¹Department of Education and Psychology, Freie Universität Berlin, ²Department of Mathematics and Computer Science, Physics, Geography, Justus Liebig University Gießen, ³Center for Mind, Brain and Behavior, Justus Liebig University Gießen and Philipps University Marburg, ⁴Department of Psychology, University of York

When asked to draw a typical living room, which objects would you include? The objects drawn could depend on both visual experience (i.e., the objects present in living rooms previously encountered) and conceptual knowledge (i.e., the semantic relationship between the scene and its constituent objects). Here, we tested to which extent visual and conceptual factors predict object content when participants draw rea-world scene categories. We collected data from 156 participants, which were asked to draw typical exemplars of six scene categories: bathroom, bedroom, café, kitchen, living room, and office. We annotated all objects within these drawings and then computed object occurrence frequencies for each category. Next, we modelled these occurrence frequencies using two predictors: First, to quantify visual experience, we extracted occurrence frequencies for all the objects drawn in each scene category from the ADE20K dataset of segmented scenes. Second, to quantify conceptual knowledge, we computed similarities between object and scene concepts using a word2vec language model. A generalized linear model revealed that both factors uniquely contributed to the composition of drawings, with a combined two-factor model performing better than both single-factor models. This result was relatively stable across scene categories, but the visual predictor consistently yielded stronger predictions. Further, we show that even objects drawn less frequently are still diagnostic of a scene category, but less so than objects drawn more frequently, an effect predicted by both models. Together, our results demonstrate that visual and conceptual factors jointly determine which objects are included in typical scene drawings.