“…This proposal not only posits a distinction between two types of memory representations (representations of syntactic category and semantic features, respectively) required for computing filler-gap dependencies but crucially it also posits a distinction between how easily these two types of representations may be maintained in memory or how quickly they decay over time. As such, it could have important theoretical implications, both for theories of sentence processing (e.g., Lewis, Vasishth, & Van Dyke, 2006; Ness & Meltzer-Asscher, 2017; Santi, Friederici, Makuuchi, & Grodzinsky, 2015) as well as models of memory representations (e.g., Baddeley, 2010; Tulving, 1972).…”