A key question in language processing regards how we build up meaning representations as we understand a sentence. Within the relevant literature are various claims that the interpretation process may under certain conditions involve "shallow processing": roughly, extracting meaning in a superficial or unstructured fashion. Various phenomena broadly support the idea that meaning extraction is at times "gisty" or underspecified, but lack of explicit models leaves us unclear as to whether "shallow processing" should in fact be considered as a single monolithic concept.
We attempt to pull this concept apart, looking more closely at certain phenomena motivating ideas of shallow processing, and working to model the processes that may be giving rise to those phenomena. For modeling these processes, we introduce the use of tools borrowed from current engineering work in language processing (NLP), but which we use to represent ideas founded in long-established models of spreading activation and feature comparison. We demonstrate the use of this modeling technique in the case of one influential finding in the electrophysiological language processing literature (Federmeier and Kutas, 1999).