Amniotes comprise all mammals and reptiles, accounting for almost four fifths of all land vertebrate diversity. They share the ability to produce an amniotic egg, defined by a series of extraembryonic membranes, which serve to protect and nourish the embryo during development. Amniotes are a monophyletic group that first arose around 320 million years ago in the Carboniferous period, and soon diverged into the two crown- group clades, Synapsida, total-group mammals, and Sauropsida, total-group reptiles. The transition from a physiological reliance on wet conditions to a fully terrestrial ecology required changes in modes of respiration and locomotion, freeing early amniotes from the bauplan of non-amniote tetrapods, leading to morphological disparity that we can observe in the fossil record. The early diverging groups of amniotes retained many plesiomorphic features, with few diagnostic anatomical differences between them. One of the principle differences in morphology was the number of openings or fenestrae that evolved in the temporal region of the skull; one in synapsids and two in diapsids, a group of derived sauropsids. Although this feature has been fundamental to amniote classification for over a century, recent work has begun to question the relevance of temporal fenestration, as it is currently coded for, in amniote systematics. It is now widely accepted that patterns of temporal fenestration are highly variable in early amniotes, and fenestrae have been lost, regained or highly modified in all lineages. Parareptiles, a group considered as early diverging sauropsids in which temporal fenestrae were absent, are now recognised as possessing a wide-range of temporal openings, which may even be a synapomorphy of the group. Much of the new data on parareptiles and other early amniotes has not yet been considered in the wider context of early amniote evolution. Recent phylogenetic research has focused on resolving intra-clade affiliations rather than the interrelationships of major taxonomic groups. The relative incompleteness of existing phylogenetic character lists for early amniotes can only be remedied by detailed cross-clade assessment. One of the aims of my thesis is to compile a new character-taxon matrix, specifically designed to perform a broad-based phylogenetic analysis of early amniotes. In this thesis, I also explore the efficacy of temporal fenestration as a character in early amniote systematics. I will show that patterns of temporal fenestration among early amniotes are more variable than is implicit in the current system of coding, and that coding for the presence or absence of a ‘lower’ temporal opening fails basic tests of homology. I propose a series of new characters, which takes into account my observations of variation and similarity in temporal fenestration across a wide range of early amniotes. I include these new characters into my new character-taxon matrix, and compare the results against the same dataset with the traditional presence/absence coding. I conclude that my new coding, based on detailed observations of temporal morphology, can play a role in resolving the phylogeny of early amniotes. One of the earliest known diapsids, Orovenator mayorum from the lower Permian of Oklahoma, is redescribed using high-resolution μCT, revealing remarkable details of the skull anatomy. My findings are relevant to both palaeoecology (suggesting burrowing and nocturnality) and phylogeny. My research on Orovenator reveals a number of anatomical similarities with varanopids, a group typically assigned to Synapsida. Orovenator and other sauropsids share at least 16 character states with varanopids, many of which were not recognised by previous studies. These include a subnarial shelf of the premaxilla, a posterodorsal extension of the external naris, the asymmetrical bifurcation of the anterior vomer, and a prominent dorsomedial shelf of the surangular. This exceptional degree of similarity between Orovenator and vara- nopids questions our current understanding of relationships among early amniotes. I test this by including Orovenator in a phylogenetic data matrix used in an earlier study to differentiate between early diapsids and synapsids, and find a monophyletic clade of Orovenator + varanopids, within Diapsida. I therefore propose that Orovenator forms a clade with varanopids; raising questions about whether this clade belongs to Synapsida or Sauropsida. Finally, I present the results of phylogenetic analyses of my new character-taxon matrix of early amniotes, which is one of the largest and most comprehensive to date. My character list of 286 discrete characters was assembled by undertaking a compre- hensive review of key phylogenetic studies on stem- and early crown-group amniotes, and was augmented by new characters based on my examination of Orovenator mayo- rum. Since my objective was not to resolve the in-group relationships of early amniote subclades, but to resolve the relationships among higher taxonomic groups, my taxon sample of 60 taxa was focused on the basal chronologically early members of each clade. Phylogenetic analysis was performed using both maximum parsimony and Bayesian in- ference. Both methods were used in order to present a fully comprehensive series of results. My results confirm my initial hypothesis that varanopids and Orovenator mayorum form a monophyletic group. I also find this group to be positioned within Sauropsida, challenging the traditional view of varanopids as ‘pelycosaur’-grade early synapsids, which was established over 75 years ago. I also find a sister group relationship between neodiapsids and parareptiles, with Captorhinidae positioned stem-ward to this group within the total-group reptiles (Sauropsida). This topology conflicts with the conventional hypothesis of captorhinids as the sister taxon to diapsids within tilia. I also find an unexpected degree of uncertainty over the affinities of caseasaurs as synapsids, stem-amniotes or stem-reptiles. A series of additional parsimony analyses were run, which included topological constraints, to compare tree lengths of alternative phylogenetic hypotheses. Pairwise Templeton tests were used to statistically compare the topologically-constrained analyses to the most parsimonious trees. My results have implications for hypotheses concerning the transition between ‘pelycosaur’-grade synapsids of the early Permian and therapsid-grade synapsids of the middle Permian, and on patterns of evolution of temporal fenestration in early amniotes.