Abstract
Recent evidence suggests that the level of detail present in a scene image is inherently related to the memorability of that scene. Encoding of scene images is hence dependent upon a two-level memory process (Evans & Baddeley, 2019), in which the first level extracts the “gist trace”, and the second level extracts idiosyncratic details. It is these details which both aid later recognition of the scene, and allows remembered scenes to be distinguished from similar category distractors. Level of detail in scenes can be operationalized as scene complexity enabling comparisons between the complexity and memorability of a given scene. Finding a positive linear relationship between complexity and memorability suggests that scene complexity captures the “detail trace” of human memory (Kyle-Davidson et al, 2022). We hypothesise that detail and memorability have a nonlinear relationship given that when one explores the extremes on the image complexity continuum there may be reduced or enhanced memory for very simple and complex images. Linearity appears because the memorability of scene images used to investigate this relationship lie along a local tangent of the true nonlinear relationship; appearing linear because they capture a small part of the actual relationship. To explore this, we investigated the relationship between complexity (i.e, detail) and memorability modulating memorability in a repeat-recognition memory experiment by applying four different load levels during the encoding phase with scene images. We compared observer recognition performance after encoding with no cognitive load, low (digit repetition), medium (1-back counting) and high (3-back counting) load conditions for scene images with different ground-truth complexity values. We find the emergence of a nonlinear relationship between memorability and complexity across different load conditions; initially suggestive that low-complexity and high-complexity scenes display better performance under load even if not unique exemplars in a set, whereas medium-complexity scenes are more easily forgotten.