In schizophrenia, alteration in the prefrontal cortex can induce some deficiencies of the executive functions, and among them errors in inhibition of prepotent responses. This type of inhibitory processes was called "restraint function" by Hasher et al. It implies a conscious and voluntary inhibition which demands attentional resources. Among the tasks exploring this function, the Hayling completion sentence task (Burgess and Shallice) appears to be the most specific. Moreover, healthy subjects performing this task in functional magnetic resonance imaging (fMRI) show activation of the prefrontal cortex. In this study, we investigated inhibitory processes in schizophrenic patients using two versions of the Hayling completion sentence task, a behavioural version and an fMRI version in order to assess both performance levels and brain correlates of inhibitory processes. Forty-eight schizophrenic participants according to DSM-IV, (mean age: 32.8, S.D. 7.7), stabilized for at least one month, receiving antipsychotic medication and with IQ higher than 70 (mean: 96.86, S.D. 20.67) and education level (mean: 11.15, S.D. 3.26) participated in the behavioural study. They were matched on age (mean: 33.8, S.D. 7.6) and education level (mean: 12.28, S.D. 2.87) with thirty-two healthy controls. Nineteen of schizophrenic participants (mean age: 33, S.D. 6.9 and IQ: 99, S.D. 10.74) were assessed by an fMRI adaptation of the Hayling task, matched with 12 controls (mean: 33.9, S.D. 7.3). All the participants had to perform the Hayling task and a speed accuracy task. The Hayling task consists in sentences for which the last word is missing. In the initiation condition, the participants had to complete the sentence with the appropriate word, whereas in inhibition condition the participants had to complete the sentence with inappropriate and unrelated words. Compared to controls, schizophrenics showed an increased number of errors in the inhibition of prepotent responses associated with increased reaction times, even when considering information processing speed. fMRI results showed fairly similar frontal activations in both groups. Nevertheless, schizophrenic patients presented principally large activations in dorsolateral and ventrolateral frontal cortex, the superior frontal sulcus, the frontal pole and the premotor cortex, and stronger activations (bilateral) in the posterior parietal cortex. Control subjects demonstrated a network of deactivated brain regions whereas the schizophrenics did not. Our results are in favour of poorer efficacy of restraint function, sometimes comprising impairment of inhibitory processes inducing errors in schizophrenics. This deficiency might be considered as insufficiency in attentional resources and/or in working memory. Hence patients cannot simultaneously restrain prepotent response and find appropriate controlled strategy for correct completion of the task. Moreover, bilateral patterns of parietal hyperactivation and absence of patterns of deactivation seem also in favour of an attentional hypothesis. The Hayling task might be interesting for assessment of inhibitory processes in schizophrenia.