This article aims to validate the schizotypal personality questionnaire in a sample of French speaking adolescents. Because early schizotypal manifestations are predictive of psychosis-proneness, reliable self-report measures are crucial for early detection of vulnerability to schizophrenia during adolescence. Unlike most existing self-reports, the questionnaire de personnalité schizotypique (SPQ) assesses individual differences in all nine feature of DSM-IV schizotypal personality (i.e. ideas of reference, excessive social anxiety, odd beliefs, unusual perceptual experience, odd behaviour, no close friends, odd speech, constricted affect and suspiciousness). Furthermore, it yields dimensional scores concerning the main schizotypal factors, which represent valuable information for the clinician's case formulation and can be used as a screening instrument in the general population. Our sample consisted of 174 adolescents (98 girls) between 12 and 17 years old. All completed the SPQ 74-item self-report. Participants were recruited in secondary schools in Switzerland, and through the child and adolescent community outpatient psychiatric service (office médico-pédagogique) affiliated to the University of Geneva's Psychiatry Department and to the Canton of Geneva Education Department. A confirmatory factorial analysis was conducted on our sample to test nine competing models of SPQ. The 3-factor model of Raine et al. was compared to concurrent 2, 3, and 4-factor models. Simple structure models of Raine et al. and Stefanis et al. were also tested. The following observations were highlighted in our results: (1) goodness-of-fit indices are better for structures allowing cross loadings than for simple structures; (2) amongst the simple structures, the best goodness-of-fit index was obtained for the Raine model and (3) the fit between our data and the Raine model is improved by a cross loading for suspiciousness subscale. The latter seems problematic for the global data fitting. This led us to test simple structures models of Siever and Gunderson, Raine et al., and Stefanis et al., based on eight subscales rather than nine. Without suspiciousness subscale, goodness-of-fit indices are enhanced in these three models. The 3-factor model yields the clearest and most reliable results in comparison with other competing models. In summary, the best goodness-of-fit indices were obtained for the 3-factor Raine model. Goodness-of-fit indices could be improved by the exclusion of the suspiciousness scale. Consistent with earlier analyses by Raine et al. and Dumas et al., our data confirm the 3-factor model of the SPQ (cognitive-perceptive; interpersonal; disorganized) in a sample of French speaking adolescents. Our analyses confirm that two dimensions are insufficient to explain the structure of schizotypy during adolescence. These results further suggest the stability of a 3-factor structure during lifespan. We note that the inclusion of the suspiciousness subscales engenders statistical issues. Most studies to date have dealt with these issues by performing a cross-loading with this subscale, or by the inclusion of a paranoid factor which is linked with the negative and the cognitive-perceptive factors. We found that the most statistically sound strategy was reached without the inclusion of the suspiciousness subscale. Future studies with larger samples could investigate the SPQ structure at an item-level, which carries the benefit of reduced restrictions on the factorial analysis. In conclusion, the current study shows that the French version SPQ constitutes a reliable self-report questionnaire for the assessment of schizotypal trait expression during adolescence that may assist in the evaluation of psychosis proneness in youths.