One of the most persistent arguments against the segment as the minimal planning unit is that the seemingly ubiquitous, thus, presumed obligatory, nature of anticipatory coarticulation (AC) effects favors the syllable or a larger unit. By contrast, we present the results of 3 experiments showing that AC is not ubiquitous, but graded and variable based on (a) phonological availability and (b) the specific criterion to initiate articulation adopted by a speaker. We further argue that phonological encoding is parallel. These results point to (a) the segment, and not the syllable, as the minimal planning unit and (b) a flexible planning scope. Implications with respect to the current formulation of AC regarding phonological availability and the minimal unit of speech articulation are discussed.