The importance of understanding and predicting behaviour has been a quintessential element of psychology, social sciences, and health research for some time, often driven by a desire by stakeholders to create theory-based behaviour change strategies. Yet, current evidence indicates many of the behaviour change strategies currently in use have produced sub-optimal results, and models to predict behaviour often leave large portions of variance unexplained and do not fully mediate the effects of past behaviour on future behavioural occurrences. Considering these limitations, there has been an increasing drive to extend and improve upon theories of behaviour. One school of thought which has experienced rapid growth since the late 90’s are dual process theories. Advocates of these dual process theories raise the concern that common theories, like the theory of planned behaviour, generally view behaviour as entirely volitional. However, as the processes by which consciously held beliefs influence behaviour are theoretically slow and cognitively inefficient, dual process researchers argue that relying entirely on a reasoned system of beliefs is unlikely. Instead, behaviour is theorised to be guided by both reasoned processes and highly efficient automatic and impulsive processes. To date, empirical evidence has provided support for the use of these dual process models given findings of unique effects of impulsive system constructs on behaviour beyond the effects of reasoned system constructs like attitude and intention. But, the use of these dual process models remains a topic of debate, fuelled by inconsistent findings and a likely over-simplification of the impulsive system in modelling. The presented program of research has two overarching aims. The first aim of the current research program was to provide a comprehensive synthesis of the predictive validity of the implicit attitude construct by assessing its relative effect on behaviour as compared to explicit attitude and testing a suite of potential moderators of the effects of implicit and explicit attitude. The second aim of the research project was to extend upon current dual process theory by assessing models in which the impulsive system was represented by multiple constructs. As a thesis by publication, three papers report three studies that address the aforementioned aims and comprise this PhD program of research. Study 1 addressed the first aim of this research program through a meta-analysis of dual process models which contained the implicit attitude construct (715 included articles, 1104 included samples). Multi-level meta-analysis found both implicit and explicit attitude had significant zero-order correlations with behaviour, but explicit attitude had stronger effects. Implicit and explicit attitude were also correlated. Similarly, implicit and explicit attitude both had significant effects when simultaneously regressed on behaviour in a multilevel meta-analytic structural equation model in the overall sample and in all bar one moderator group. Moderator analysis using multi-level meta-regression found the implicit attitude construct performed best in adult samples and when implicit attitude was assessed using homogenised measures. The findings support the use of the implicit attitude construct, however the combined effect of implicit and explicit attitude still explained only a modest portion of the variance in behaviour and a substantial amount of heterogeneity remained even after moderator analysis. Study 1 demonstrated the potential utility of the implicit attitude construct, but also highlighted the clear need to extend upon simple dual process models in order to make more meaningful predictions of behaviour. Noting the limitations in the field observed in Study 1, Study 2 aimed to test an extended dual process model on a sample of undergraduate students (N=205) in which both implicit attitude and habit were hypothesised to predict prospectively measured free-sugar limiting behaviour alongside intention and past behaviour. Using a Bayesian structural equation model it was found that implicit attitude and habit both had unique effects on behaviour beyond the effects of intention. The effects of past behaviour on prospectively measured behaviour were partially mediated through the reasoned system and habit, but not through implicit attitude. Further, past behaviour still had a direct effect on behaviour. These findings, combined with the weak correlation between the two impulsive system constructs provided strong evidence for the potential utility of improving dual process models by including a more in-depth conceptualisation of the impulsive system. To further investigate the interplay between habit and implicit beliefs Study 3 tested a model in which implicit beliefs and habit interacted in the prediction of prospectively measured behaviour in two undergraduate student samples (free sugar intake sample N = 205; Binge drinking sample N = 122). Using partial least squares structural equation modelling it was found that both habit and explicit attitude predicted behaviour in both samples. While implicit beliefs did not predict behaviour in either sample, a significant implicit attitude by habit moderation was observed in both samples, such that implicit attitude had stronger effects on behaviour when individuals rated that behaviour as automatic. Findings add to the evidence presented in Study 2 for the potential efficacy of dual process models which conceptualise the impulsive system as multi-faceted. Overall, the findings of this research program make an important contribution to dual process research and have both theoretical and practical applications. The findings support the use of the implicit attitude construct, but add to calls for more rigorous methodology. Despite these findings however, the research program also provides evidence that the implicit attitude construct alone may be an insufficient conceptualisation of the impulsive system in dual process models. Instead, the findings demonstrate that the impulsive system may be better represented by multiple constructs as habit made a significant contribution to predicting behaviour beyond implicit attitude and moderated the implicit attitude-behaviour relationship. Taken together, these findings support the use of dual process models, but highlight the need for further research particularly through enhanced methodology and a more in-depth conceptualisation of the impulsive system.