How does ‘do so’ do? Testing the applicability of the ‘do so’ test in Croatian

The argument–adjunct distinction is tacitly present in many grammatical theories and has been supported by psycholinguistic research on various languages. One way of testing this distinction is through the ‘do so’ test, which involves replacing a projection of the verb with the pro-form ‘do so.’ The current study undertakes two experiments, an acceptability judgment task and a self-paced reading task; in both, the ‘do so’ test is applied to Croatian verbs that can take theme NP arguments in both the accusative case and the instrumental case. We tested sentences in which ‘do so’ strands an accusative or instrumental theme of such a verb. We also tested similar sentences with instrumental-marked noun phrases denoting instruments and prepositional accusative phrases denoting direction. Our first goal in conducting the study was to investigate whether the ‘do so’ test, applied to Croatian data, can discriminate between VP-internal arguments and adjuncts when both are instrumental noun phrases (based on a number of diagnostics, we took instrumental-marked themes to be arguments and instrumental-marked instruments to be adjuncts). Our second goal was to assess whether the results of the ‘do so’ test in Croatian would be better explained by appealing to the structural status of the remnant constituent (the argument–adjunct explanation) or to the lexical requirements of the verb ‘do,’ paired with an identity requirement on the theta roles assigned to the constituent following ‘do so’ and its antecedent (the lexical explanation). The experimental results we obtained did not support either approach. In both experiments, instrumental arguments and instrumental adjuncts behaved the same (both yielded grammatical results when stranded by ‘do so’ substitution), while accusative themes behaved differently from prepositional accusative adjuncts (accusative themes could not be stranded by ‘do so’ substitution, but prepositional phrases could).