Stable differences in the tendency to attribute meaning and emotional value to experience may represent an indicator of liability to psychosis. A brief task was developed assessing variation in detecting affectively meaningful speech (speech illusion) in neutral random signals (white noise) and the degree to which this was associated with psychometric and familial vulnerability for psychosis. Thirty patients, 28 of their siblings, and 307 controls participated. The rate of speech illusion was compared between cases and controls. In controls, the association between speech illusion and interview-based positive schizotypy was assessed. The hypothesis of a dose-response increase in rate of speech illusion across increasing levels of familial vulnerability for psychosis (controls, siblings of patients, and patients) was examined. Patients were more likely to display speech illusions than controls (odds ratio [OR] = 4.0, 95% confidence interval [CI] = 1.4-11.7), also after controlling for neurocognitive variables (OR = 3.8, 95% CI = 1.04-14.1). The case-control difference was more accentuated for speech illusion perceived as affectively salient (positively or negatively appraised) than for neutrally appraised speech illusions. Speech illusion in the controls was strongly associated with positive schizotypy but not with negative schizotypy. In addition, the rate of speech illusion increased with increasing level of familial risk for psychotic disorder. The data suggest that the white noise task may be sensitive to psychometric and familial vulnerability for psychosis associated with alterations in top-down processing and/or salience attribution.