As a reader, think of sentence processing as a noisy, online computation rather than a static lookup. When listeners build structure incrementally, memory demands and retrieval operations unfold over time. Those dynamics can create longer processing times for distant dependencies regardless of whether grammar posits empty positions. The paper walks through how empty-category frameworks can predict the same dependency-distance signatures and points to empirical work that lines up with that perspective.
If you care about how linguistic theory connects to mind and learning, this discussion matters. It reshapes which empirical results count as decisive when evaluating competing grammatical models and nudges researchers to model the processing pipeline explicitly. Follow the full article to see how these ideas could change what counts as evidence for theories about human language, learning, and inclusive models of cognition.
Abstract
A long-standing question in cognitive science is whether psycholinguistic evidence can adjudicate among grammatical theories. This question arises in debates over the representation of linguistic dependencies: whether they are mediated by unpronounced elements (empty categories; ECs) or established without them. Some psycholinguistic research has taken dependency-distance effects—the observation that longer dependencies impose greater processing cost—as evidence in favor of EC-free over EC-mediated accounts, and this claim has recently been revived. In this letter, I argue that this conclusion is unwarranted. Once the incremental, real-time nature of sentence processing is taken into account, dependency-distance effects are expected under EC-mediated accounts as well. I show how EC-mediated accounts can explain these effects and review findings consistent with this view. Dependency-distance effects are, therefore, not diagnostic of whether dependencies involve ECs. More broadly, psycholinguistic data can be misleading when used to draw inferences about grammatical theories without reference to the real-time, incremental processes that give rise to those data.
