The context-specific proportion congruency (CSPC) effect denotes smaller congruency effects in a context containing mostly incongruent trials (MI trials) compared to a context with mostly congruent trials (MC trials). In this experiment, we investigated how the global context volatility, manipulated by the switch rate between MC and MI contexts, affected the size of the CSPC effect. We used a 4-alternative forced-choice face-name Stroop task, and the lower location was always the MC and the upper location the MI context. In different blocks, we measured CSPC effects when the global context volatility was low, medium, and high. The order of volatility levels either increased or decreased across the experiment and was counterbalanced between participants. Results showed a large CSPC effect in the low-volatility condition, which strongly decreased as volatility increased, exclusively for the group starting with the low-volatility context. Here, a significant CSPC effect was observed only at the low but not at the medium and high levels. For the group starting with the high-volatility context, the CSPC effect was not modulated by the different volatility levels. In fact, the CSPC effect was not reliably present in either the high or low volatility contexts but reached significance only at the medium level. These findings suggest that an initial experience of a stable context with few switches between MC and MI contexts supports the learning of context-control associations. Beginning with a high-volatility context, characterized by frequent switches between MC and MI, may limit the learning of these associations.