Submission 165
A New Multinomial Model of Prospective Memory: Validating the Retrospective What Component
SymposiumTalk-05
Presented by: Fabian E. Gümüsdagli
Remembering to perform an intended action when a specified event occurs in the future is referred to as event-based prospective memory (PM; see Bayen et al., 2024, for an overview). Successful PM involves several cognitive processes: remembering that something must be done (prospective component), remembering when it must be done (retrospective when component), and remembering what must be done (retrospective what component). Smith and Bayen (2004) developed the multinomial processing tree model of PM, which allows us to disentangle the prospective and retrospective when components, but not the retrospective what component. We extended this model by introducing new parameters and response categories to disentangle all components of PM. To validate the new parameter representing the retrospective what component in line with the strong validation criterion (cf. Schmidt et al., 2025), we conducted an experiment (N = 256) designed to selectively affect this parameter while leaving all others unaffected. Participants performed an ongoing task and were additionally instructed to press one of two keys (F1 or F8) when a pre-learned word appeared, with specific words assigned to specific keys. We manipulated the encoding difficulty associated with the key by varying the learning sequence: One group first learned all words for one key (e.g., F1) and then all words for the other key (e.g., F8), whereas the other group alternated between keys during learning. Consistent with the strong validation criterion, only the parameter representing the retrospective what component was affected by the manipulation, thereby supporting its construct validity.