Six main empirical studies are reported. Study 1 employed the "thinkaloud" procedure in the minimal group paradigm (MGP) and concludes that social categorization is insufficient to cause social identity or intergroup discrimination, and that no theoretical explanation of minimal group behaviour is adequate to explain the variety of strategies employed within that paradigm. Study 2 employed both "Tajfel matrices" and new "allocation grids" in the MGP and concludes that two distinct forms of intergroup discrimination need to be distinguished: one which maximizes in-group profit consistent with positive in-group distinctiveness, and another which maximizes positive in-group distinctiveness by accompanying in-group profit with out-group derogation. In Study 3 subjects completed three sets of Tajfel matrices in the MGP: individually, in "sub-groups" , then again individually. Mean intergroup discriminatory behaviour polarized and mean intergroup equitable behaviour depolarized between the first and the latter two conditions. The best account of the results was concluded to be a normative one. Study 4 demonstrated that the self-esteem hypothesis within social identity theory (SIT) is best tested using a state measure of specific social identity contingent self-esteem and concludes that this hypothesis has to date been both inadequately formulated and inadequately tested. Study 5 compared predictions from SIT with those from Tesser's self-evaluation maintenance model concerning the consequences of social comparison outcomes and concludes that a modified version of the former theory is best able to account for the results obtained at both group and individual levels of comparison. Study 6 investigated a host of issues within SIT and concludes that the theory is too simplistic in respect of many of its key notions and propositions. A general discussion argues that a modified version of SIT can be developed which improves on Tajfel's "original" social identity theory by more adequately specifying the processes by which group phenomena are manifest.