Background and objectives: Previous studies suggest that patients with panic disorder and agoraphobia (PD/A) tend to overestimate the associations between fear-relevant stimuli and threat. This so-called threat expectancy bias is thought to play a role in the development and treatment of anxiety disorders. The current study tested 1) whether patients with PD/A (N = 71) show increased threat expectancy ratings to fear-relevant and fear-irrelevant stimuli relative to a comparison group without an axis I disorder (N = 65), and 2) whether threat expectancy bias before treatment predicts treatment outcome in a subset of these patients (n = 51). Methods: In a computerized task, participants saw a series of panic-related and neutral words and rated for each word the likelihood that it would be followed by a loud, aversive sound. Results: Results showed higher threat expectancy ratings to both panic-related and neutral words in patients with PD/A compared to the comparison group. Threat expectancy ratings did not predict treatment outcome. Limitations: This study only used expectancy ratings and did not include physiological measures. Furthermore, no post-treatment expectancy bias task was added to shed further light on the possibility that expectancy bias might be attenuated by treatment. Conclusions: Patients show higher expectancies of aversive outcome following both fear-relevant and fear-irrelevant stimuli relative to the comparison group, but this does not predict treatment outcome.