Abstract
Recent studies using repeat-detection memory tasks have shown that images differ consistently in their memorability (e.g., Isola et al., 2011). In a previous study, we extended these results to a more traditional visual long-term memory task, with a separate study and test phase, and longer retention intervals (Goetschalckx et al., 2017). Here, we investigated an incidental type of memory task, with a surprise memory test, and asked how the incidental nature would affect the consistency and ranking of the memorability scores. If memorability is truly an intrinsic property of the image, then one should not expect large differences between incidental and intentional conditions. On the other hand, observers might process images differently when not anticipating a memory test, and therefore different results could be expected. The study list of our incidental memory task was identical to our previous, intentional study. Participants were to watch the images of this list carefully as they appeared on the screen one by one (free viewing). They were told that the study was about eye movements. An actual eye tracker was set up and calibrated to make the cover story credible. Afterwards, a surprise recognition memory test was administered, in which half of the studied images were shown again, mixed with an equal number of randomly selected lures. Incidental memorability scores were computed as the proportion of participants who correctly recognized the target images. Despite the free viewing instructions during the study phase, the incidental memorability scores still showed high levels of consistency across observers (mean split-half Spearman's rho of .44 of for 27 responses per image), although slightly lower than in the intentional study. In addition, there was large overlap between the incidental and intentional memorability ranking: Spearman's rho of .62. These results further support the idea of memorability as an intrinsic image property.
Meeting abstract presented at VSS 2018