Specific phobias are the most common form of anxiety disorder. Two factors have influenced cognitively oriented research in specific phobias: (i) conditioning theory; (ii) the assumption that phobias are "irrational". Experimental investigation has largely focused on information processing paradigms, which have not advanced treatment. In other anxiety disorders, the cognitive approach, with its emphasis on cognitive products such as meaning and beliefs, has been at the forefront of therapeutic development. In this thesis, a series of experiments investigated the contribution of attention, memory and beliefs to the maintenance of specific phobias. These were hypothesised as contributing to the generic meaning of the phobic object, which is the level of meaning at which emotional change takes place. In the first experiment the content of phobic beliefs was examined and was found to form a logical framework for the maintenance of the phobia: phobics had a high level of belief that their phobic object would cause them physical harm or make them behave foolishly and that they would be unable to cope. As disgust has been raised as a possible factor in phobic acquisition and maintenance, cognitions and processes concerning disgust were examined in a separate series of experiments. Disgust beliefs were found to be present in phobics but did not contribute to an attentional bias and were not found to be closely linked to the phobic fear response. The third experiment examined attentional bias to threat words utilising a computer Stroop test in order to examine the merits of different hypotheses of attentional effects in spider phobia. Phobics were found to selectively attend to threat at the conscious, but not the pre-attentive level, but there was no evidence that threat information was subsequently suppressed (cognitive avoidance). The fourth experiment examined attentional bias to actual threat (a live Tarantula rather than word stimuli). Spider phobics showed prolonged attention to the threat stimulus. The results indicated that attention may be divided between threat and escape. The fifth and sixth experiments examined the effect of anxiety on memory in spider phobics for spider videos, first with a recognition test and second with a recall test. Spider phobics were not impaired in their memory for phobic stimuli: this is again consistent with the view that cognitive avoidance may not play a role in the maintenance of spider phobia. The seventh and final experiment was a treatment study. Spider phobics were given the Stroop test and questionnaires concerning beliefs, then given one session treatment for spider phobia. Compared to untreated spider phobic controls the treated phobics changed significantly in their negative beliefs about spiders after treatment, but were no different to the controls in their reaction time latencies to spider stimuli. It is the meaning that the threat object has for the phobic which causes them to attend to threat, rather than an automatic processing bias. It is the change in this negative meaning which is associated with a change in emotional response.