Assessing multisensory integration and estimating speed of processing with the dual-presentation timing task: Model and data

Our ability to detect temporal asynchrony is sometimes an obstacle to multisensory integration. A seamless multisensory experience occurs when the temporal misalignment of two signals is within the allowance for subjective synchrony, referred to as the temporal window of integration (TWI). The TWI is most commonly measured with the temporal-order judgment task or the synchrony judgment task, which are single-presentation methods that confound sensory and decisional determinants of performance. Thus, data collected with these tasks are unsuitable for elucidating whether changes in the TWI observed in studies on prior entry or temporal recalibration have a sensory or a decisional origin. The indeterminacy can be resolved with a dual-presentation timing (2PT) task in which observers report their judgment with a ternary response format: whether the first or the second presentation was more synchronous or else that they were equally synchronous. Yet, the analysis of 2PT data suffers from the lack of a process model that captures sensory, decisional, and response components of performance via separate parameters whose estimation can identify the cause of prior entry or recalibration effects. The main goal of this paper is to present and validate such a model. Simulation studies reveal that model parameters are identifiable and not confounded under the ternary response format. The empirical validity of the model is demonstrated by showing its capability to account for published data collected with the 2PT task under binary response formats and for new data collected with the ternary format. The new data relate to a study on differences in speed of processing between magnocellular and parvocellular visual pathways, and our analysis revealed that low-spatial-frequency information is processed 20-30 ms faster than high-spatial-frequency information. Measures of the TWI and the point of subjective synchrony are presented for their estimation from ternary 2PT data. Software in MATLAB and R to fit the model to ternary 2PT data is made available with this paper. (C) 2020 Elsevier Inc. All rights reserved.