This thesis resembles the main part of my research between my PhD (2002) and this moment. I focussed on the study of verbal domains deprived of finiteness, non-finite or non-tensed domains. I include here any verbal form, or derived from a verb, which does not present any temporal inflection or personal agreement marker, but may present aspectual markers. This is something which characterises the infinitival domains, cf. to make vs. to have made, but which may charaterise equally certain deverbal nominals (for instance the Romanian supine nominal), and participial forms, as I show in the present thesis. The thesis itself contains 6 chapters, among which the final one presents conclusions, open questions and research perspectives. In each part, I present and discuss the results of the work resembled in the Annexes volume. After an introductory chapter, presenting the issues and problems, as well as the background, Chapter 2 is dedicated to the question of the categorization and of categorial change in the grammar. The absence of tense and personal agreement is indeed associated to categorial indeterminacy, which is a main characteristic of participial domains. I choosed to begin this presentation of my contribution to linguistic theory by the issue of lexical categories.The domains which I study are problematic for all lexicalist theory of categorization, because of the wide variety of mixed combinations of property they offer. I therefore argue for a flexible theory of categories, in a syntactic approach to word formation, where categorization is the result of dedicated functional levels, and ultimately to syntactic environment. The chapter proposes also a general theory of participles as truncated domains which contain a gap, and from which different domains, more or less reduced, can be built, having corresponding different structures and interpretations. Chapters 3 and 4 are dedicated to the issue of nominalizations, i.e. deverbal nominals such as construction or assessment, which accross languages may take different forms, sometimes deprived of an explicit nominalizer (i.e. suffixes like -age or -tion), and appear as default nominalizations of participial or event inflected verbal forms. Nominalizations are also mixed domains, bringing together in the same projection verbal and nominal properties, which explains the constant interest they triggered in linguistic theory. Conducting my research in a comparative perspective, I formulated cross-linguistic hypotheses on the issue of aspectual properties of nominalizations, which afterwards have been confirmed and extended, in a parametric approach (including for instance german and slavic language). In the end, I got back to the study of French derived nominals, which imposes to reformulate and refine some generalizations. The main properties that I studied are aspect, the projection of argments and their necessary interplay. Since we deal with domains where tense is not projected, but which are nevertheless similar to the sentence by certain properties, the presence of aspectual features is a privileged explanation of certain " more verbal " behaviours that some of the deverbal may have. I showed most importantly, that the more verbal nominalizations have a more complex structure, and some of them have a pluractional interpretation (the Romanian supine). My hypotheses and generalizations have been tested and refined by typological studies (germanic and slavic languages). I also contributed to a finer-graied typology of deverbals in general, and showed in which conditions eventive denotations can correspond to internal event structure of the verbal type (in the study on 'agentive'-eur nominals). Finally, the undestudied issue of stative nominals is also approached. The presentation of the results of my research is integrated in an overview of the evolution of the field, since nominalizations proper are an excellent prism accross which the very development of grammatical theory, and generative grammar in particular, can be followed. Chapter 5 is dedicated to an older an more reduced area of my research, but with a crucial importance for my research project, namely the one of non-finite verbal forms such as the infinitive and the participles, which can enter constructions of a verbal nature, such as verbal periphrases or predicative structures like reduced relatives or Tough constructions. In my work on infinitival relatives (of the type maison à vendre 'house to sell') also driven in a comparative perspective on data from Romanian, French and Italian, I show that these constructions are reduced relatives with a modal interpretation. I study in detail the functional structure of these relatives, articulated on a Pred head which attracts the closest syntactically active nominal to its specifier. This analysis is extended to Tough constructions, and I show that the difference between argumental and predicative infinitives in Romance languages is the same as the one between complete sentences and reduced clauses: the former include a subject position and have an satisfied EPP feature; as such, they qualify as " strong phases ", whereas object-gap infinitives, built through a movement operation, do not have these properties, and count as " weak phases". This distinction is correlated to the nature of functional categories proposed to account for these properties. The final part of the chapter is dedicated to the structure and the interpretation of participles and infinitives in modal-aspectual periphrases and non-finite complementation. In this part of my research I make an investigation of constructions with auxiliaries and reduced participial structures, nonfinite (PredP), with the aim of clarifying the different elements which contribute to the interpretation, in comparison with my research on nominalizations. The evolution of my research overviewed in this thesis can be followed and traced back in the papers assembled in the Annexes volume. This includes publications starting with early work, published immediately after my thesis and unfolding questions and issues which I raised in the dissertation, and finishing with recent papers and ongoing projects. The publications are grouped both thematically (nominalisations - non-finite predications) and chronologically, as far as possible. It is possible to trace back the main ideas which have guided my research and were present in nuce in my " early " papers, and continue to be elaborated to a fully articulated form in my recent papers; such is the case for instance with the nominal/verbal patterns of nominalization, the pluractional analysis of the Romanian supine nominal, but also for the analysis of reduced relatives. For the sake of theoretical coherence of the whole proposal, some novel fields have been studied, like the analysis of 'agent' nominals.
