The research for this thesis was commenced in 1992 following the completion of the survey of audit committee practices (Collier, 1992). The thesis comprises a general study of audit committee practices in major UK companies and in particular investigates the factors affecting the voluntary fonnation of audit committees by these companies. The study covers a JX>pulation which is defined as UK listed companies which are included in the top 250 of the Times 1000. The research began with a literature review which focused on the development of audit committees, their nature and fonn and factors involved in their formation. In the second stage a questionnaire was developed to gather information on the motives for having an audit committee. audit committee practices and other matters. The data collected enabled the resJX>ndents to be grouped in companies in the JX>pulation which had voluntarily formed an audit committee as defined for the research and those who had not. Given this background information the research investigates the proJX>sition that there are differences between companies which have voluntarily formed audit committees and those which have not. Testable hYJX>theses were derived initially from agency theory using variables that have been found to be significant in studies of voluntary accounting JX>licy choices. The premise was that audit committees may be viewed as voluntarily formed monitoring mechanisms, which have been established in situations where there is a separation of ownership and control. An audit committee will be fonned provided the associated costs of of an audit committee are lower than the agency costs associated with weaker monitoring. Therefore, the pressure for the formation of audit committees will be greatest in high agency cost situations. The variables identified as measures of the agency cost relationships are the proportion of non-managerial ownership, gearing. asset structure and size. The impact of board structure, choice of auditor, the use of internal audit and industrial classification are also examined. The hypotheses were tested using univariate and multivariate approaches applied to datasets established for 1991 and 1987. The univariate tests involved parametric and non-parametric methods, while multivariate testing involved the use of Logit analysis to overcome multi-collinearity problems amongst the variables. The results of univariate and multivariate tests revealed statistically significant differences between the two groups for variables representing (i) agency costs of equity Gi) agency costs of debt and (iii) director pressure to reduce information asymmetries. These effects were found to have existed in both 1991 and 1987, and to be relevant in explaining why companies formed an audit committee during this period.