Neurocognitive processing of textual cues in literary reading

:speech_balloon: Speaker: Mesian Tilmatine

:classical_building: Affiliation: FU Berlin / Radboud University

Title: Neurocognitive processing of textual cues in literary reading

Abstract (long version below): When we read literary fiction, there are many factors that have an impact on our experience. In my PhD project, the central question is how to identify and scientifically control these factors when researching naturalistic literary perception in readers. In a proof-of-concept study, we re-analyzed an existing set of eye-tracking data with a focus on text-based predictions of the reader response. The next two studies were a closer investigation of the role of aesthetics and the larger narrative for the reader response, both in terms of rating data and reading behaviour. In the final study, we plan to use fMRI data to study mental actualizations and narrative structuring during literary reading.


tilmatine_igel_poster_23.pdf (940.2 KB)


:newspaper: Long abstract

When we read literary fiction, there are many factors that have an impact on our experience. In my PhD project, the central question is how to identify and scientifically control these factors when researching naturalistic literary perception in readers.

The act of reading can be regarded as a cognitive process of higher-level perception in the broadest sense: The environmental context and the visual circumstances of reading can easily be controlled in a lab setting, but this is much harder for every other way in which the reader interacts with the text. Is the text being understood? And how would such an understanding precisely be defined, to begin with? Is the reader emotionally engaged with the story or just skipping through the pages without caring for the content of the pages? Which parts of the text affect the reader in which way? There are ways to potentially get answers to all of these questions, but they may require an array of different methodologies and some solid modelling to make sense of the results.

After having reviewed the current state of science in the relevant sub-disciplines, my next step was to extend the Neurocognitive Poetics model of literary reading (NCPM) in a way that it would fit with the aforementioned factors while remaining empirically testable. Then, I first concentrated on the concept of reader engagement. In a proof-of-concept study, we re-analyzed an existing set of eye-tracking data that were recorded during the lecture of 4 excerpts from the German-language version of Stephen King’s novel The Shining. New to our re-analysis was a focus on text-based predictions of the reader response. The results showed that the distributional patterns of text features associated with the evocation of emotional valence and arousal are correlated with both reading speed and ratings. This indicates that the emotionalizing aspects of a literary text that lead to the reader’s engagement with the story are measurable in the form of continuous numerical variables.

Next, the focus shifted onto a closer investigation of the role of aesthetics and the larger narrative for the reader response. For that we ran two separate studies with the same stimulus material; one on ratings and another one on reading times. In both cases, the subjects read two excerpts of Johann Wolfgang von Goethe’s Faust I. In the rating study, subjects rated the 140 stanzas across the two excerpts on how much they enjoyed them (liking), how well they thought to understand them (comprehensibility), how positive or negative they made them feel (valence), and how exciting or calming they considered them to be (arousal). Unlike in established word-based studies, the text feature AAP was not a good predictor of the subjects liking a stanza, implying a contextual influence of higher-level text aspects like the narrative on the reader evaluation. Valence and arousal were easier to predict based on text features – implying an important role of aesthetic stylistics for the reader evaluation –, although narrative context influenced the evaluation here as well.

In the reading-speed study, two main questions were investigated. First, we took a closer look at the make-up of reading speed: How much of it is due to word skipping, and how much of it related to re-reading? And which textual elements are associated with each of these phenomena? Theoretically, while both leading to non-fluent reading, they would be expected to be associated with two different sorts of literary processing: Superficial skimming vs. intense engagement with the text. Second, we compared four different models of text feature combinations (surface readability, aesthetic stylistics, discrete emotion measures, and two-dimensional emotion measures) regarding their impact on the reading behaviour on four levels (individual words, text blocks, moving average measures of the narrative, and discrete measures of the narrative). We also explore a surprising lack of impact of comprehensibility ratings on the reading behaviour.

This all culminates in the design of a final study with functional magnetic resonance imaging (fMRI) of the brain activity during literary reading. Three questions cannot satisfyingly be answered without neural activity measures: (1) in which way readers use event segmentation to mentally structure the narrative, (2) to which extent literary processing is mentally actualized, and (3) to which extent textually evoked emotion activation is mentally actualized. All three of these questions can be answered by replicating the reading-speed study with the addition of recording fMRI data at the same time, as well as afterwards gathering rating data similar to the aforementioned rating study.

With all of the above, we obtain additional empirical data to found more holistic approaches to literary reception studies, which I would like to present in this poster.