<p>Multitasking research has shown that individuals differ in whether they prefer a more serial or a more parallel mode of task processing at the level of whole tasks. Such preferences can be identified using the task switching with preview (TSWP) paradigm. This paradigm allows, but does not require, individuals to preview the stimulus of the next task switch in a predictable task switching procedure [AAABBB...]. Although several studies have shown that some participants consistently use the preview information, it is still unclear when exactly this information is used and, thus, how parallel processing takes place. The present study is an important step in clarifying this issue. In two experiments, I investigated when exactly individuals who prefer parallel processing during task switching use a preview to prepare for the next task switch. In Experiment <InternalRef RefID="Sec3">1</InternalRef>, the onset and thus the length of the preview was varied within participants. This allowed to disentangle whether parallel processing of the preview depends on the length of the preview (i.e., its likelihood increases with longer preview presentation), or occurs contingent on a single trial (i.e., is a rapid process). Strikingly, parallel processing occurred regardless of preview duration, suggesting that a short preview may be sufficient. In Experiment <InternalRef RefID="Sec16">2</InternalRef>, participants received the preview in discrete steps, i.e., with the same length throughout the sequence. There was a clear peak in the use of the preview immediately before a task switch. This suggests that although individuals who prefer parallel processing are able to process information in parallel throughout a task sequence, they clearly prefer to do so just before the task switch.</p>