From surprise to sense: automatic and effortful processes in reading and learning
Abstract
Readers sometimes encounter moments in a story that force them to completely
rethink what they thought they understood. In Harry Potter and the Philosopher’s Stone,
for example, readers are led to believe that professor Snape is plotting to steal the
Sorcerer’s Stone, only to learn at the end that it was the timid professor Quirrell all along. This revelation disrupts the coherence of the mental model that readers have constructed throughout the story, compelling them to reinterpret earlier events in a new light. This dissertation examines how readers respond to such moments, when new and unexpected information conflicts with what they already know, tracing the process from moment-to-moment comprehension to longer-term learning. It focuses on three interrelated processes: (1) detecting inconsistencies (e.g., realising that not professor Snape, but professor Quirrell wants to steal the Sorcerer’s Stone); (2) adapting mental models (e.g., reinterpreting the behaviours of both Snape and Quirrell in light of this revelation); and (3) integrating the new information into long-term knowledge (e.g., the twist becomes part of the reader’s broader knowledge base). Although this example comes from a fantasy novel, the same processes are at play when people read any kind of text, from fictional narratives we enjoy in our free time to expository and informational texts we encounter in school or at work. Readers construct mental models (or mental representations; throughout this thesis, these terms will be used interchangeably) of the characters, events, places, and relations described (Johnson-Laird, 1983; Van Dijk & Kintsch, 1983; Kintsch, 1988; Zwaan & Radvansky, 1998). A mental model is a dynamic representation that readers build by integrating information from the text with relevant background knowledge. Constructing and maintaining such a coherent representation is central to successful reading comprehension (Kintsch & Van Dijk, 1978; Gernsbacher et al., 1990; Graesser et al., 1994; Zwaan et al., 1995; Van den Broek et al., 1999). Even coherent texts unfold in ways that require readers to revise and modify their mental models as new information becomes available. In the Landscape Model of reading comprehension (Van den Broek et al., 1999), this process is described as a dynamic cycle of activation and integration, in which relevant information is retrieved, reactivated, or suppressed to maintain coherence. As part of this process, readers continuously monitor their mental models for inconsistencies (Singer, 2013; Kendeou, 2014; Van Moort et al., 2018) and revise them when confronted with new or unexpected information (Kendeou et al., 2013). Crucially, readers also evaluate the plausibility and relevance of new information in relation to prior knowledge, a process that guides whether the information becomes integrated into the mental model or dismissed as implausible or irrelevant (Isberner & Richter, 2014; Singer, 2013). Yet maintaining coherence while reading is not always straightforward. Sometimes inconsistencies may go unnoticed, and even when detected, obsolete information can continue to influence understanding (Johnson & Seifert, 1994; Rapp & Kendeou, 2007, 2009). Understanding when and why these processes succeed or fail is crucial for theories of reading, memory, and learning, as it reveals how people maintain coherence, update their knowledge, and decide whether to accept or reject information that challenges existing beliefs. Beyond its theoretical significance, this understanding also has practical importance: in an information-rich world, readers are continually exposed to claims that conflict with what they believe, for instance when reading conflicting viewpoints or evaluating information encountered online. Understanding these processes helps explain not only how we learn from texts, but also why we sometimes fail to do so, providing a foundation for developing methods that help readers navigate conflicting viewpoints and misinformation. Together, the processes of detecting inconsistencies, adapting mental models, and integrating new information are at the core of how readers maintain coherence, update their understanding, and ultimately learn. However, it remains unclear to which extent these processes occur automatically or depend on effortful, controlled processing. Across the three empirical chapters, the dissertation addresses this question through complementary approaches. Chapters 2 and 3 examine the role of working memory by manipulating cognitive load during reading to test whether detecting inconsistencies and adapting mental models depend on limited cognitive resources or can proceed automatically. Chapter 4 broadens the focus from moment-to-moment comprehension to longer-term learning, exploring how automatic and effortful components, such as surprise and judgments of plausibility, jointly determine whether information is remembered and integrated into existing knowledge. The remainder of this introduction elaborates these perspectives in turn. The next section discusses the core processes of coherence monitoring and model modification. Later sections address theoretical debates about the role of working memory in comprehension, and finally evaluative reactions such as surprise and plausibility judgements and their consequences for memory and learning.
