Background: Classification is integral to both psychiatric research and clinical practice. However, despite well over 100 years of research and revisions to the nomenclature, our ability to accurately classify even the most common of psychiatric disorders is questionable. Psychiatry continues to be dominated by a number of categorical nosologies, which propagate distinct diagnostic entities. While such systems have been integral to the advancement of research and practice, they are not without limitations. One of the key weaknesses of categorical nosologies is their inability to account for widespread comorbidity. Indeed, research indicates that comorbidity is the rule rather than the exception, and occurs at rates that far exceed chance. The comorbidity phenomenon challenges the putatively distinct nature of psychiatric disorders, which in turn raises questions regarding the validity of these categories. An alternative approach, the liability-spectrum model (Krueger & Markon, 2006) has been proposed in an attempt to address the issue of comorbidity. This model, which has its roots in child and adolescent psychiatry, suggests that psychopathology, rather than consisting of a multitude of discrete disorders, is more likely to reflect as a smaller number of broad dimensions. This model is empirically based and has been supported by a large number of factor analytic studies, which indicate that the variance shared amongst the most common psychiatric disorders can be accounted for by two latent factors; internalizing (INT) and externalizing (EXT). INT reflects emotional disturbance, and encompasses the many mood , and anxiety disorders. EXT involves behavioural disinhibition, and includes such disorders as ADHD, oppositional defiant disorder, and conduct disorder. While this approach has been hailed for its empirical support, researchers have noted consistently high correlations between the !NT and EXT factors. This undermines the liability-spectrum model, as it appears comorbidity remains an issue at the dimension level. Aims: By better understanding the phenomenon of comorbidity, we may improve our understanding of the fundamental nature of psychiatric disorder. The present thesis sought to test a number competing hypotheses that attempt to explain the correlation between the INT and EXT dimensions of psychopathology. Methods: Analysis was conducted on data from the A von Longitudinal Study of Parents and Children (ALSPAC, N=14,541), a community cohort of children in England. Maternal-report versions of the Development and Wellbeing Assessment (DA WBA)were completed by the study mothers at three time points; when children were aged 7.5, 10.5 and 14 years. The DAWBA is used to assess Axis-I disorders. Five common INT disorders were assessed; social phobia, specific phobia, major depression, PTSD, and generalized anxiety disorder (GAD). Three EXT disorders were also assessed; ADHD, oppositional defiant disorder, and conduct disorder. A series of competing models, comprised of both observed and latent variables, were fit to the data. Results: The structure of psychopathology was successfully modelled as; i) two latent variables (INT, EXT) subsumed by a general psychopathological factor (P), ii) four homogenous sub-populations characterised by different patterns of comorbidity, and iii) a network of causally associated disorders. Conclusions: Psychiatric comorbidity is incredibly complex and open to multiple interpretations. It may be difficult to unequivocally establish whether comorbidity is due to an underlying shared aetiological agent (i.e. p) or whether comorbidity arises due to complex networks of association between symptoms and disorders. There are strengths and weaknesses to both approaches, which may become more apparent in particular contexts.