[Truncated abstract] It is well established that anxiety vulnerability is characterised by two biased patterns of selective information processing (Mathews & MacLeod, 1986; Mogg & Bradley, 1998). First anxiety is associated with an attentional bias, reflecting the selective allocation of attention to threatening stimuli in the environment (Mathews & MacLeod, 1985; MacLeod, Mathews & Tata, 1986; MacLeod & Cohen, 1993). Second anxiety is associated with an interpretive bias, reflecting a disproportionate tendency to resolve ambiguity in a threatening manner (Mogg et al., 1994). These characteristics are shown by normal individual high in trait anxiety (Mathews, Richards & Eysenck, 1989; Mogg, Bradley & Hallowell, 1994; Mathews & MacLeod, 1994), and by examining clinically anxious patients who repeatedly report elevated trait anxiety levels (MacLeod, Mathews & Tata, 1986; Mogg & Bradley, 1998). '...' Two alternative hypotheses regarding this relationship are proposed. One hypothesis is that attentional and interpretive biases are concurrent expressions of a single underlying biased selectivity mechanism that characterises anxiety vulnerability (the Common Mechanism account). In contrast, a quite different hypothesis is that attentional and interpretive biases are independent cognitive anomalies that represent separate pathways to anxiety vulnerability (the Independent Mechanisms account). The present research program was designed to empirically test the predictions that differentiate the Common Mechanism and Independent Mechanisms accounts. The general methodological approach that was adopted was to employ bias manipulation tasks from the literature that have been developed and validated to directly modify one class of processing bias (i.e. attentional bias or interpretive bias). The effect of these direct bias manipulation tasks on a measure of the same class of processing bias or the other class of processing bias was then examined. The Common Mechanism and Independent Mechanisms accounts of the relationship between attentional and interpretive bias generate differing predictions concerning the impact of directly manipulating one class of processing bias upon a measure of the other class of processing bias. The central difference between the alternate accounts is their predictions regarding cross-bias transfer, that is the transfer of training effects from direct manipulation of one class of processing bias to a measure of the other class of processing bias. Whereas the Common Mechanism account predicts that such cross-bias transfer will occur, the Independent Mechanisms account does not predict such transfer. A series of seven studies is reported in this thesis. There was some difficulty achieving successful bias modification using bias manipulation approaches established in the literature; however when such manipulation was achieved no cross-bias transfer was observed. Therefore the obtained pattern of results was consistent with the Independent Mechanisms (IM) account, and inconsistent with the Common Mechanism (CM) account. A more detailed version of the IM account is developed to more fully accommodate the specific results obtained in this thesis.