Proceedings of the Annual Meeting of the Cognitive Science Society

The stream of thought has received less empirical attention than external cognition, with existing approaches relying on thought probes or continuous verbal reports rather than modeling internal representational dynamics. The Attention-to-Thoughts (Amir & Bernstein, 2022) model offers a dynamical-systems framework in which thought trajectories emerge from moment-to-moment interactions among lower-level components, capturing several thought patterns, including those consistent with repetitive negative thinking (RNT). However, the model uses simplified thought representations and does not capture the breadth of internal attentional selection: a dimension extensively studied in external attention and implicated in internal attention, too, in studies of creativity (broad) and rumination (narrow), for instance. We extend A2T by grounding thoughts in semantic embeddings and affective ratings, and introducing an attentional scope parameter modulating internal selection breadth. Simulations reveal systematic affective and dynamic differences between thought trajectories in broad and narrow attention runs. Control analyses confirm these effects arise from model dynamics rather than embedding structure alone. This gives us an empirically testable computational basis for understanding how the scope of internal attention shapes affective experience during spontaneous thought streams.