The aim of this thesis is exploring VP ellipsis (VPE) in Croatian from an informationstructural point of view. We focus on (i) the possibilities of realization of VPE in various syntactic and semantic contexts; (ii) the possibilities of realization considering informationstructural constraints, and (iii) the possibilities of interpretation considering the role of the information structure in discourse structuring. An overview of historical and contemporary definitions and taxonomies of elliptical phenomena is provided, followed by an overview of theoretical approaches to ellipsis considering the issues of (i) the content of the ellipsis site; (ii) the identity conditions on ellipsis; (iii) ellipsis licensing. Issues of determining equivalents of VPE crosslinguistically are discussed. Several types of predicate ellipsis which can be analyzed as VPE are detected in Croatian. By conducting VP constituency tests and examining distributional properties, we confirm previous analyses of “English style” VPE as instances of VPE in Croatian. Two aspects of information-structural constraints of VPE realization in Croatian are investigated. Firstly, we examine the information-structural status of the remnant, concluding that the obligatoriness of the contrastive interpretation of a remnant may be traced down to general relations of parallelism and contrastivity between clauses. We detect three information-structural patterns – in VPE in Croatian, at least one element in a remnant is realized as: (i) contrastive focus; (ii) contrastive topic; (iii) polarity / modality focus. The information-structural constraints on VPE are then tested experimentally. Three written acceptability judgment experiments have been carried out in order to test the contrastive focus constraint on VPE. Following López and Winkler (2000), Winkler (2005), Kazenin (2006), and Ha (2008), as well as relying on analysis of the remnant status in VPE in Croatian, we predict that the acceptability of VPE in maximally parallel an maximally contrastive coordinated structures depends on the realization of contrastive focus. In first experiment, we tested the acceptability of VPE in following structures: (1) a. Marija neće pročitati članak, a Ivana hoće. b. Marija neće pročitati članak, a Ivana će. ‘Mary won’t read the article, and Jane will.’ The prediction that the sentences in which contrastive polarity focus (indicated by the stressed form of the auxiliary) is not realized will be judged as unacceptable was experimentally confirmed. In second experiment, we tested the contrastive focus constraint on backwards ellipsis (Ha 2008), predicting that the sentences in which contrastive polarity focus on the pre-elided element is not realized would be judged as unacceptable: (2) a. Marija hoće, a Ivana neće pročitati članak. b. Marija će, a Ivana neće pročitati članak. ‘Mary will, and Jane won’t read the article.’ Since no statistically significant difference was found between the experimental conditions, the prediction was not confirmed. In third experiment, we additionally test the effects of (i) the order of the antecedent and the ellipsis site and (ii) contrastive focus symmetry on the acceptability of VPE: (3) a. Marija neće, a Ivana će pročitati članak. b. Marija neće, a Ivana hoće pročitati članak. ‘Mary won’t, and Jane will read the article.’ c. Marija će pročitati članak, a Ivana neće. d. Marija hoće pročitati članak, a Ivana neće. ‘Mary will read the article, and Jane won’t. While the main effects of the order of the antecedent and the ellipsis site and contrastive focus symmetry were not found, the interaction effect suggests there is a penalty on VPE acceptability which was not predicted by the contrastive focus constraint. The possibilities of interpretation of VPE in Croatian considering the role of the information structure in discourse structuring were investigated experimentally. Two hypotheses proposed by Frazier and Clifton (2005) were tested: (i) Main Assertion Hypothesis (MAH): Comprehenders prefer to relate material in a new sentence to the main assertion of the preceding sentence. (ii) Conjunction Domain Hypothesis: The processing of conjunction allows either the entire conjoined phrase to be processed or just the closer conjunct. The differences in the interpretation of ambiguous VPEs were measured in a comprehension task for two-sentenced discourses including (i) complex sentences and (ii) compound sentences: (4) a. Tina je otišla kući nakon što je popila piće. I Marija je. ‘Tina went home after she had a drink. So did Mary.’ b. Tina je otišla kući i popila piće. I Marija je. ‘Tina went home and had a drink. So did Mary.’ Following Frazier and Clifton (2005), we predicted (i) that the preferred antecedent of VPE in example 4a. was VP from the main clause, regardless of the clause order; (ii) the least preferred antecedent of VPE in 4.b. was the first conjunct. The results suggest that the dominant strategy in both cases is choosing the conjoined structure as an antecedent of VP ellipsis. However, the effects of both the main assertion and recency are observed in the interpretation of the ellipsis which follows complex sentences. The results of the preferred interpretation of elliptical sentences which follow compound sentences confirm the Conjunction Domain Hypothesis, i.e. the role of the clause order in processing.