This abstract summarises the thesis entitled Defensive Avoidance in Paranoid Delusions: Experimental and Computational Approaches, submitted by Michael Moutoussis to The University of Manchester for the degree of Doctor of Philosophy (PhD) in the faculty of Medical and Human Sciences, in 2011.The possible aetiological role of defensive avoidance in paranoia was investigated in this work. First the psychological significance of the Conditioned Avoidance Response (CAR) was reappraised. The CAR activates normal threat-processing mechanisms that may be pathologically over-activated in the anticipation of threats in paranoia. This may apply both to external threats and also to threats to the self-esteem.A temporal-difference computational model of the CAR suggested that a dopamine-independent process may signal that a particular state has led to a worse-than-expected outcome. On the contrary, learning about actions is likely to involve dopamine in signalling both worse-than-expected and better-than-expected outcomes. The psychological mode of action of dopamine blocking drugs may involve dampening (1) the vigour of the avoidance response and (2) the prediction-error signals that drive action learning.Excessive anticipation of negative events might lead to inappropriately perceived high costs of delaying decisions. Efforts to avoid such costs might explain the Jumping-to-Conclusions (JTC) bias found in paranoid patients. Two decision-theoretical models were used to analyse data from the 'beads-in-a-jar' task. One model employed an ideal-observer Bayesian approach; a control model made decisions by weighing evidence against a fixed threshold of certainty. We found no support for our 'high cost' hypothesis. According to both models the JTC bias was better explained by higher levels of 'cognitive noise' (relative to motivation) in paranoid patients. This 'noise' appears to limit the ability of paranoid patients to be inﬂuenced by cognitively distant possibilities.It was further hypothesised that excessive avoidance of negative aspects of the self may fuel paranoia. This was investigated empirically. Important self-attributes were elicited in paranoid patients and controls. Conscious and non-conscious avoidance were assessed while negative thoughts about the self were presented. Both 'deserved' and 'undeserved' persecutory beliefs were associated with high avoidance/control strategies in general, but not with increased of avoidance of negative thoughts about the self. On the basis of the present studies the former is therefore considerably more likely than the latter to play an aetiological role in paranoia.This work has introduced novel computational methods, especially useful in the study of 'hidden' psychological variables. It supported and deepened some key hypotheses about paranoia and provided consistent evidence against other important aetiological hypotheses. These contributions have substantial implications for research and for some aspects of clinical practice.