<p>When we observe an object, our visual system extracts both coarse and fine information to build a coherent percept. This process is often described as a coarse-to-fine integration, in which low spatial frequency (LSF) information provides an initial scaffold for interpreting high spatial frequency (HSF) detail. Here, we examined how this integration unfolds in face perception across two experiments. We leveraged the face inversion effect—where inverted faces are harder to recognize than upright ones—as a tool to probe processing flexibility. In Experiment 1, ten participants matched two familiar faces presented in rapid succession (template and probe), which were either spatial frequency-congruent (both LSF or both HSF) or incongruent (LSF–HSF or HSF–LSF). Accuracy was higher for HSF templates in congruent conditions, but in incongruent conditions, performance improved when matching LSF probes onto HSF templates—an effect limited to upright faces. In Experiment 2, twelve participants completed the same task with both familiar and unfamiliar faces. The pattern for familiar faces replicated that of Experiment 1; however, for unfamiliar faces, accuracy was generally higher in congruent conditions and was unaffected by template SF. These findings suggest that coarse-to-fine dynamics do not rigidly constrain spatial frequency integration in face recognition but instead reflect flexible processing strategies that adapt to stimulus familiarity, orientation, and task context.</p>