We followed up a cohort (. n=. 35) of clients with an " At Risk Mental State" (ARMS) for almost 2 years (mean 21.3 months). At baseline, these clients had taken part in research looking at the relationship between reasoning biases, memory, personality styles and delusional ideation. During the follow-up period, clients underwent a package of intervention from a specialist early detection team. Eighty percent (. n=. 28) of these clients were successfully re-interviewed. There was improvement across the cohort as a whole, however five participants (17.9 had made the transition to psychosis at follow-up. Those who had become psychotic had lower levels of manic symptomatology at baseline than those who did not enter the first episode. Further, across the cohort, impaired working memory and delusional ideation at baseline combined to predict 45% of the delusional ideation at follow-up. These preliminary findings suggest that working memory impairments may be linked to the persistence of delusional ideation and that manic symptoms in someone with an ARMS may suggest that such an individual is less likely to develop a frank psychotic episode. © 2010 Elsevier Masson SAS.