Late Positive Potentials as an Index of "Desirably Difficult" Learning Processes Engaged During Language Comprehension: ERP Evidence From Studies of Domain Knowledge.

Electrophysiological studies of language comprehension have primarily examined the kind of information that comes to mind, and when, as people process words and build message-level understanding. However, less is known about the factors that allow people to commit the message-level information to memory for future use. One promising marker of such explicit memory processes is the late positive component (LPC), an event-related brain potential (ERP) effect linked to recollection in the memory literature and predictive of memory performance at timescales ranging from minutes to months. Here, we examine LPCs, alongside N400 brain potentials (sensitive probes of implicit semantic processing), to investigate the hypothesis that domain knowledge influences explicit memory mechanisms during comprehension. We re-analyzed three existing datasets in which young adults with varying domain knowledge about a fictional book series read short descriptions of fictional "facts" about that domain while ERPs were recorded. As predicted, correct completions of these facts elicited larger LPCs as a function of individuals' overall domain knowledge. We also assessed item-level difficulty using completion norms from an independent peer group. More difficult facts engendered larger LPCs-but only in individuals with relatively greater domain knowledge. By contrast, N400 amplitudes, reflecting implicit, real-time lexico-semantic activation, were modulated by item-level difficulty for individuals with weaker knowledge but to a much lesser degree for those with stronger knowledge. These findings demonstrate that domain-specific knowledge shapes not only what information can be implicitly accessed in the moment, but also whether explicit memory mechanisms are immediately engaged. Consistent with the memory literature, we propose that explicit (possibly intentional) memory processes support deeper encoding when input is "desirably difficult" based on an individual's knowledge base. Because these effects appear only in knowledgeable individuals, we suggest they reflect a strengthening of relational memory between in-the-moment linguistic input and extant knowledge networks.