Invited Speakers

Johanna Kaakinen, University of Turku

Eyes in Motion, Minds in Flux: Linking Eye Movements to Readers' Inner Experiences

Readers' thoughts fluctuate between immersion in a text and moments of mind-wandering. This talk examines how eye movements can serve as sensitive indicators of these changing mental states during reading. Drawing on studies conducted in our lab, I will discuss characteristic eye movement patterns that accompany different subjective experiences of engagement and disengagement. The findings demonstrate that eye movements not only reflect linguistic processing but also capture broader variations in readers' cognitive and experiential states. 

Erik D. Reichle, Macquarie University

Simulating the Eye-Movement Disturbances of Readers with Schizophrenia

Despite the societal importance of skilled reading and the potential for reading skill to impact socioeconomic function in schizophrenia (see e.g., Revheim et al., 2014), relatively little is known about how the latter disorder affects the former skill.  Recent experiments using eye tracking to better understand how readers with schizophrenia (SZ) might differ from healthy control readers (HC) are therefore important for understanding both the nature of schizophrenia and how the perceptual, cognitive, and motoric processes that support skilled reading are coordinated in healthy individuals. One study that has directly compared SZ and HC readers suggests that, relative to HC individuals, SZ readers exhibit reductions in reading fluency (or rate) that are disproportionate to reductions in comprehension or general cognitive impairment (Dias et al., 2021; see also Christofalos et al., 2024). This reduction in reading fluency is characterized by longer, more frequent fixations, a reduction in word skipping, and more regressions back to previously read portions of text.  To understand these differences, a computer model of eye-movement control in reading, E-Z Reader (Reichle, 2021), has been used to simulate these observed behavioural differences.  The results of these simulations suggest a “double hit” account in which the impairment associated with schizophrenia reflects an increased propensity to initiate corrective refixations that is disproportionate to the general slowing of lexical processing (Dias et al., 2021), with the increased refixations possibly reflecting impairment of the oculomotor system’s corollary discharge (e.g., see Spering et al., 2013).  The present talk will discuss these findings, along with new E-Z Reader simulations that were designed to localize the possible source(s) of the differences between SZ and HC readers.  Although these simulations are preliminary, they also support the “double hit” account in that the model parameters that best explain the differences between the eye movements of SZ versus HC readers result in the former group exhibiting both a general, high-level slowing in lexical processing and a specific, low-level increased propensity to refixate words.  The theoretical implications of these simulations will be discussed in relation to both schizophrenia and the development of skilled reading.

Lena A. Jäger, University of Zurich

Measurement Reliability of Individual Differences in Reading

In recent years, an increasing number of studies have investigated individual differences in sentence processing. This development stems from the recognition that human readers vary in the extent to which they show behavioral responses (e.g. eye-movement behavior, reaction times) to well-studied psycholinguistic manipulations in the language materials, such as those giving rise to garden paths or locality effects. Understanding the source of this variability requires identifying whether certain psycholinguistic phenomena, such as garden-path or locality effects, are modulated by an individual’s cognitive capacities. 

Most studies addressing questions about individual differences in sentence processing therefore correlate the magnitude of psycholinguistic effects with cognitive capacities, such as cognitive control or working memory capacity, thereby testing assumptions about the role of these capacities in sentence processing. However, these studies typically rely on behavioral data from one single measurement occasion. As a result, we know very little about whether the relationship between individual cognitive capacities and the effect of psycholinguistic manipulations on a reader's parsing behavior is stable over time. Moreover, because these studies typically investigate well-established psycholinguistic phenomena that are robust (i.e., replicable) at the group-level, they implicitly assume that the individual-level effects of psycholinguistic manipulations as well as the relationship between the psycholinguistic effect and the individual’s cognitive capacities also remain stable, i.e., they can be replicated across experimental sessions and across different behavioral methods. This assumption is challenged by the reliability paradox (Hedge, Powell, & Sumner 2018) which states that measurement reliability at the individual level is necessarily lower for predictors with high group-level replicability.

Thus, a crucial first step for a principled investigation of individual differences in sentence processing is to establish their measurement reliability, that is, the extent to which individual-level psycholinguistic effects correlate across multiple measurement occasions and methods. A second step is to explore the extent to which the reliability of these effects over time depends differently on distinct measures of an individual’s cognitive capacities.

In this talk, I will present a large-scale naturalistic reading dataset consisting of a total of 392 experimental sessions (measurement occasions) from 133 native speakers of German who were tested multiple times each, either with eyetracking only (2 sessions) or with both eyetracking and self-paced reading (4 sessions per participant). Half of the participants also completed a comprehensive battery of psychometric tests. We use these data to assess the measurement reliability of individual-level ,effects of word length, lexical frequency, surprisal, dependency length and number of to-be-integrated dependents. To this end, we deployed a two-task Bayesian hierarchical model to compute correlations of individual-level effect sizes obtained with i) the same method, but on two measurement occasions, and ii) two different behavioral methods (eyetracking and self-paced reading). Finally, we explored whether measurement reliability depends on   an individual’s cognitive capacities. 

I will show that across-session and across-method reliability varies substantially between the psycholinguistic phenomena under investigation. Moreover, I will also show that the replicability of psycholinguistic effects that exhibit high to moderate reliability at the individual level over time depends differentially on cognitive capacities, such as reading fluency and cognitive control. These findings underscore the importance of establishing measurement reliability before drawing inferences about individual differences in sentence processing.

Marloes van Moort, Utrecht University

Validation in Reading Comprehension: Text-based vs Knowledge-based Validation and the Role of Individual Differences

Successful comprehension requires readers to build a coherent, meaningful mental representation or situation model of a text. An essential aspect of building such representations is that readers monitor to what extent information they encounter in the text is both coherent with prior information in the text (i.e., congruent) and valid with respect to their background knowledge (i.e., accurate)—a process called validation. Validation processes are important for establishing coherence during reading and protecting the emerging mental representation against incongruencies and inaccuracies. In my talk, I will discuss the complex interplay between what we read (i.e., information from the text) and what we know (i.e., our own background knowledge) in validation and explore the potential role of reader-related factors (e.g., a reader's purpose for reading or their available topic relevant knowledge) in validation.