In this dissertation, I analyze the linguistic constraints that condition the variation in Spanish between the null pronoun and the clitic lo referring to a proposition. Previous literature on Spanish has analyzed null objects referring to first order entities, mostly in varieties in contact with other languages. This dissertation contributes to the literature on anaphora in Spanish by establishing and analyzing the existence of propositional null objects in two monolingual dialects, Mexican and Peninsular Spanish.

A variationist approach was used to discover the significant constraints on the variation of the null pronoun and the overt clitic lo in Mexican and Peninsular Spanish. Following the generalizations from the previous literature on two separate areas of study, anaphora resolution and null objects (Chapter 2), several internal factor groups were included in the coding scheme. In Chapter 3, I provide an explicit statement of the envelope of variation and I specify the coding scheme employed.

Chapter 4 offers the results of the multivariate analyses of Mexican and Peninsular Spanish. These results show that some of the linguistic constraints conditioning the variation are shared by both dialects (presence of a dative pronoun, type of antecedent, sentence type), suggesting that the null pronoun has the same grammatical role in both dialects. Nevertheless, divergences in the conditioning of the null pronoun also emerge from the analysis and the overall rate of null objects in Mexican Spanish is significantly greater than in Peninsular Spanish.

Chapter 5 offers a discussion of the results and their theoretical implications for studies of anaphora. The accessibility of the referent hypothesis, operationalized with the factor groups referential distance, turn, and number of mentions, finds no support in the data, suggesting an important difference in the behavior of propositional versus noun-phrase anaphora. Instead, I provide qualitative evidence indicating that a crucial constraint on the variation is the degree of completeness of the referent: a complete proposition favors the overt clitic lo whereas null objects more likely refer to incomplete propositions. Finally, Chapter 6 explores the hypothesis of a change in progress and provides concluding remarks and suggestions for further research.