Files
Abstract
Interacting with nature is beneficial. Even brief exposures to natural environments or viewing sets of natural scene images can restore attention and memory (Berman et al., 2008). Despite the known benefits of nature, it remains unclear what aspects of viewing nature scenes lead to these improvements. While previous studies focus on the distinction between natural images and non-natural (urban) images, very few examine the connection between distinct properties of natural images and their cognitive benefits. One idea is that natural scenes are more fluently processed, making them less memorable, which in turn contributes to their benefits. A potential confound is that nature scenes may be more confusable than urban scenes, leading to their lower memorability scores predicted by the ResMem Neural Network (Needall & Bainbridge, 2021). The goal of this thesis was to test the hypothesis that image set similarity drives differences in memorability. To test this hypothesis, the present study calculated image statistics and image semantic labels into vectors and determined the Euclidean and cosine distances between pairs of natural and urban images. While the low-level visual features in natural and urban images differ, importantly, there were no significant differences in the average distances of every image pair. However, the semantic similarity analysis showed mixed results. The significant differences in semantic similarity were shown in both direction under different types of label coding techniques. This finding suggests the relationship between image similarity and memorability is more complex than a simple positive or negative relationship. In the future, studies can focus on the processing fluency and memorability of natural images to uncover the mechanism behind their restorative effect.