Previous research has revealed that task-switch costs (worse performance for task switches than for task repetitions) at the first position of an explicit task sequence are eliminated or reduced when repeating or switching sequences. The authors hypothesize that such effects are restricted to points in the sequence representation that are associated with sequence-level processing such as chunk retrieval that changes the contents of working memory. In an experiment testing this chunk-point hypothesis, subjects memorized and performed explicit task sequences under different chunking instructions that induced chunk points at different positions within the sequences. Regardless of position, performance was slower at chunk points than at non-chunk points, providing direct evidence of chunking, and task-switch costs were reduced or eliminated at chunk points while they remained large and robust at non-chunk points. These findings support the chunk-point hypothesis and are discussed in relation to task-set inhibition and associative interference.