The aims of this prospective study were to explore in subjects with psychosis participating in a rehabilitation program whether cognitive performances at baseline predicted (i) psychosocial functioning over a 15-16 month follow-up; (ii) improvement in psychosocial functioning over the rehabilitation program. Visuo-spatial tests from the Cambridge Neuropsychological Test Automated Battery (CANTAB) were administered to assess cognitive performance in 55 subjects with schizophrenia spectrum disorders who completed a rehabilitation program. The Multnomah Community Ability Scale (MCAS) was used to measure dimensions of community functioning. One subscale of the Client's Assessment of Strengths, Interests, and Goals (CASIG) provided a measure of subjective quality of life (QoL). Improvement was defined as a 15% or more increase in psychosocial scores between baseline and follow-up. Worse baseline sustained attention predicted better self-rated quality of life, and better baseline visual memory predicted better community functioning over the rehabilitation follow-up period, in particular, higher autonomy in activities of daily living, and less physical and psychiatric symptoms that could interfere with rehabilitation. Baseline cognitive performances predicted community functioning improvement during the follow-up period: visual memory predicted improvement in daily living autonomy and in social competence; sustained attention predicted improvement in behavioral problems (such as medication compliance, collaboration with treatment providers or impulse control) and social competence; planning performances predicted improvement in social competence. These cognitive functions could be specifically targeted in a rehabilitation program aimed at enhancing functioning in those particular dimensions.