Background
The outcome of first episode psychosis (FEP) is highly variable and difficult to predict. Cognitive insight measured at illness onset has previously been found to predict psychopathology 12-months later. The aims of this study were to examine whether the prospective relationship between cognitive insight and symptom severity is evident at four-years following FEP and to examine some psychological correlates of cognitive insight.

Methods
FEP participants (n = 90) completed the Beck Cognitive Insight Scale (BCIS) at illness onset, and associations between BCIS scores with symptom severity outcomes (4-years after FEP) were assessed. The BCIS scales (self-reflectiveness and self-certainty) were examined as a composite score, and individually compared to other cognitive measures (IQ and jumping to conclusions (JTC) bias).

Results
Regression analyses revealed that the cognitive insight composite did not predict 4-year symptom remission in this study while the self-reflection subscale of the BCIS predicted severity of symptoms at 4-years. Self-certainty items of the BCIS were not associated with symptom severity. Significant correlations between the JTC bias, self-certainty and IQ were found, but self-reflection did not correlate with these other cognitive measures.

Conclusions
Self-reflective capacity is a more relevant and independent cognitive construct than self-certainty for predicting prospective symptom severity in psychosis. Improving self-reflection may be a useful target for early intervention research.