When making inferences about the mental lives of others (e.g., others' preferences), it is critical to consider the extent to which the choices we observe are constrained. Prior research on the development of this tendency indicates a contradictory pattern: Children show remarkable sensitivity to constraints in traditional experimental paradigms, yet often fail to consider real-world constraints and privilege inherent causes instead. We propose that one explanation for this discrepancy may be that real-world constraints are often stable over time and lose their salience. The present research tested whether children (N = 133, 5- to 12-year-old mostly US children; 55% female, 45% male) become "less" sensitive to an actor's constraints after first observing two constrained actors (Stable condition) versus after first observing two actors in contexts with greater choice (Not Stable condition). We crossed the "stability" of the constraint with the "type" of constraint: either the constraint was deterministic such that there was only one option available (No Other Option constraint) or, in line with many real-world constraints, the constraint was probabilistic such that there "was" another option, but it was difficult to access (Hard to Access constraint). Results indicated that children in the Stable condition became less sensitive to the probabilistic Hard to Access constraint across trials. Notably, we also found that children's sensitivity to constraints was enhanced in the Not Stable condition regardless of whether the constraint was probabilistic or deterministic. We discuss implications for children's sensitivity to real-world constraints.