Contemporary accounts of language emphasize predictive processing as a major contributor to processing load during interpretation (Altmann & Kamide, 1999, etc.). Accounts differ with regards to the mechanisms that support prediction during language processing (Pickering & Gambi, 2018). Bayesian/Information-theoretic accounts view prediction as resulting from automatic spreading activation in lexical representations based on prior established patterns of association (Hale, 2006; Levy, 2008). These accounts lead to models including entropy and surprisal as key indices of processing load. Evidence for these accounts can be found in analyses of n-gram correlations with processing times (at the word level) and exposure effects on processing times (at the sentence level). We have been developing an alternative, two-factor account under which both predictive and integration processes contribute to processing load. Further, rather than responding reflexively to pattern information activated by resonance-like processes, we view comprehenders as flexibly adapting to cues available in the wider context. I will discuss recent eye-tracking and EEG/event-related potential (ERPs) experiments that evaluate hypotheses relating to predictive processing during language interpretation.