Designing a test battery for real-world visual search

Abstract

Abstract Although much research has investigated the mechanisms of visual search behaviour in laboratory-based computer tasks, there has been relatively little work on whether these results generalise to more naturalistic search tasks. In addition, work on real-world search has often been carried out by researchers working in very different disciplines, leading to a high degree of task fragmentation. Here, we detail our development of a ‘naturalistic search task battery’, which aims to provide a suite of open-source, reproducible, and standardised real-world search tasks, thus enabling the generation of comparable data across multiple studies and aiding theory and modelling in this area. We show that we can replicate findings from previous search tasks using Lego, demonstrating set-size effects and differences between search conditions. We also show effects of scene regularity using a ‘bookcase’ task where participants search for specific books on shelves. Finally, we use a ‘jigsaw’ task and a ‘Lego building’ task as examples of more complex real-world search tasks and show how we can use them to consider the effects of strategies. Finally, we consider whether performance on one task can predict performance on another, and thus to what extent similar cognitive capabilities underlie diverse ‘real-world’ search tasks.