This dissertation explores some of the traditionally labeled possessive relations, and proposes a basic syntactic structure that underlies them. The two nouns act as subject and predicate in a small clause, dominated by two functional projections, where reference/agreement and contextual restrictions are checked. Looking first at container-content relations, we propose that the container is always a predicate for the content. Because in our system selection is determined in the small clause and agreement is checked in an AgrP, selection and agreement need not be determined by the same noun. Selection also distinguishes between a container and a content reading. The evidence from extraction shows that container readings are more complex than content readings. We propose that the container reading adds a higher small clause whose predicate is the feature number. Number is thus a predicate, which type-lifts mass terms to count nouns, the way classifiers do in languages without number. Evidence from Spanish and Asturian shows a three-way distinction between absence of number (mass terms), singular and plural. We also propose that nouns are not divided into rigid classes, such as mass/count. Rather, any noun may be used as mass or count, depending on whether number is added to its syntactic derivation or not. An analysis of possessor raising to both nominative and dative in Spanish also supports the idea that nouns are not divided into rigid classes with respect to their ability to enter possessive relations. Relations such as part/whole, alienable and inalienable possessions, are all analyzed as small clauses where the possessor is the subject and the possessed is the predicate. Finally, we propose a universal principle: possessor raising can occur in languages that have a structural Case in a v-projection, in addition to the Case checked by the direct object. This predicts that causative verbs in languages with possessor raising should also allow the Case checking of both the object and the subject of an embedded transitive clause. The prediction is borne out, giving rise to four types of languages, according to their Case system.