Pain-related fear (kinesiophobia) is an important factor in the development of chronic pain and disability through the maintenance of maladaptive avoidance behaviors, which appear resistant to common exposure treatments. Studies employing an exposure paradigm across two dissimilar movements have failed to show generalization of corrected overprediction (i.e., the tendency to bring expectations of pain and harm in line with experienced pain and harm) in highly kinesiophobic participants, suggesting that learning is limited to the specific movement targeted for intervention. In the current study, 60 participants with chronic low back pain performed a graded reaching task, comprising two trials per four movements of increasing physical demand. During each trial, participants rated predicted and experienced pain and harm using a visual analog scale. In line with previous studies, a recurrent pattern of pain and harm overprediction was observed at Trial 1 of each movement, with correction apparent at Trial 2. However, analysis of high-kinesiophobic responses indicated that the magnitude of pain overprediction steadily declines across movements, eventually converging with the responses of low-kinesiophobic participants. In contrast to previous studies, this pattern is suggestive of successful generalization.