A novel experimental paradigm for measuring state paranoia by means of signal detection theory was evaluated. A liberal response bias, indicating the tendency to recognize facial expressions as threatening, was expected to reflect paranoia. Against theoretical expectations, heightened paranoia questionnaire scores were associated with a non-liberal bias, which was not affected by negative emotion per se. However, subsequent analyses revealed that, if anxious, participants with heightened paranoia adopted a comparatively more liberal response bias. These findings corroborate the eminent role of anxiety in paranoia and demonstrate that state paranoia can be captured with the presented paradigm.