A Note on Parallelism Effects in Processing Deep and Surface Verb-phrase Anaphora

Tanenhaus and Carlson (1990, experiment 1) reported that surface anaphors were more difficult to comprehend than deep anaphors when the antecedent for the anaphor did not form a syntactic constituent, i.e. when it was presented in a passive construction as compared to an active construction and the anaphor was in active voice. However, parallelism of the antecedent did affect the time it took readers to comprehend deep anaphors. This paper reports a reanalysis of that data and two experiments based on the reanalysis. The parallelism effects for deep anaphors were completely eliminated following short passives but not full passives. The results are interpreted as support for the claim that deep and surface anaphors access different types of representations. We also suggest that previous findings of parallelism effects for deep anaphors may be due to factors other than structural parallelism.