Both manual and saccadic reaction time tend to be facilitated when stimuli from two or more sensory modalities are presented in spatiotemporal proximity and subjects have to respond to the stimulus detected first (redundant target AKA divided attention paradigm). It is commonly accepted that this enhancement is typically larger than predicted by the probability summation effect of a race model. Retaining the notion of a race among stimulus-triggered peripheral activations, the time-window-of-integration (TWIN) model (Colonius & Diederich, JCogN 2004) postulates a first stage of parallel processing followed by second stage of (neural) coactivation and response preparation. A necessary condition for crossmodal enhancement to occur is that the peripheral processes terminate within a given time window. TWIN has been tested in a series of studies (Diederich & Colonius, ExpBrRes 2007, 2008; Perc&Psyphys 2007) where stimuli from one modality were designated as targets and stimuli from the other could be ignored (focused attention paradigm), and the model accounted for variations in spatial configuration, stimulus onset asynchrony, and intensities of targets and non-targets. Here we demonstrate how the TWIN model, within the redundant target paradigm, permits a separate assessment of reaction time enhancement due to probability summation and due to “true� multisensory integration.