Against Sequence Priming: Evidence from Constituents and Distituents in Corpus Data

Structural priming, i.e., the tendency to repeat linguistic material, can be explained by two alternative representational assumptions: either as the repetition of hierarchical representations generated by syntactic rules, or as the repetition of lexical sequences. We present two studies that test these explanations by investigating priming effects in a dialogue corpus. We compare syntactic constituents with distituents, i.e., part-of-speech pairs that cross constituent boundaries. We find a reliable short-term priming effect for constituents, but no priming for distituents. This result supports the rulebased view of priming, which does not predict priming of distituents. The data are incompatible with a sequence priming analysis, which cannot distinguish between constituents and distituents. In a second corpus study, we study long-term priming and find priming effects for both constituents and distituents. This indicates that the mechanism underlying longterm adaptation differs substantially from short-term priming.