Willingness to communicate (WTC) has recently been researched as a dynamic variable, with some investigations viewing it as a complex dynamic system (CDS). One important property associated with CDS is the notion of attractor states, which are characterized by stable patterns of behavior. The present study employed an idiodynamic method to monitor per-second WTC changes with 20 participants during three-minute speaking tasks. Using idiodynamic graphs illustrating the WTC dynamics, patterns of stability were identified as the unit of analysis for WTC attractors. Temporal measures of the utterances that coincided with the WTC attractors were also examined. Results demonstrated that attractors were likely to emerge in both WTC and L2 fluency, and largely depended on two categories of content-related and linguistic-cognitive factors. Data analysis also indicates that WTC and L2 fluency exhibited parallel dynamics in most of the cases. The study's findings yield pedagogical implications that are discussed at the end.