Date Published: February 5, 2019
Publisher: Public Library of Science
Author(s): Sebastian Wallot, Jun Taek Lee, Damian G. Kelty-Stephen, Veronica Whitford.
Reading research uses different tasks to investigate different levels of the reading process, such as word recognition, syntactic parsing, or semantic integration. It seems to be tacitly assumed that the underlying cognitive process that constitute reading are stable across those tasks. However, nothing is known about what happens when readers switch from one reading task to another. The stability assumptions of the reading process suggest that the cognitive system resolves this switching between two tasks quickly. Here, we present an alternative language-game hypothesis (LGH) of reading that begins by treating reading as a softly-assembled process and that assumes, instead of stability, context-sensitive flexibility of the reading process. LGH predicts that switching between two reading tasks leads to longer lasting phase-transition like patterns in the reading process. Using the nonlinear-dynamical tool of recurrence quantification analysis, we test these predictions by examining series of individual word reading times in self-paced reading tasks where native (L1) and second language readers (L2) transition between random word and ordered text reading tasks. We find consistent evidence for phase-transitions in the reading times when readers switch from ordered text to random-word reading, but we find mixed evidence when readers transition from random-word to ordered-text reading. In the latter case, L2 readers show moderately stronger signs for phase-transitions compared to L1 readers, suggesting that familiarity with a language influences whether and how such transitions occur. The results provide evidence for LGH and suggest that the cognitive processes underlying reading are not fully stable across tasks but exhibit soft-assembly in the interaction between task and reader characteristics.
Research on reading has identified multiple levels at which the reading process operates, from the perceptual front-end to the integration and processing of higher-level semantic information. While there are competing hypotheses on what might constitute a particular level or sub-level, the most prominent ones are word identification, syntactic parsing, and higher-level semantic integration [1–4]. There is plenty of evidence for distinct processes that operate at these different levels, but the investigation of these three levels constitutes almost three different research areas. Research on these different levels is often distinguished by the use of different designs, manipulations, and preferred types of stimuli. For example, investigations of word recognition heavily rely on paradigms using individual words or sublexical features, for the most part not extending the scale of presenting two related words. Word-recognition researchers prefer particular tasks such as word naming or lexical decision. However, some studies of eye movements during reading might also be classified as primarily aiming at word recognition processes because eye movements are often modelled as to be primarily driven by word-level features (e.g., [5,6] but see ). Investigations of syntactic parsing prefer a sentence level scale, with studies often not extending beyond the scale of two sentences, because one sentence alone already provides most of the linguistic space needed to investigate syntactic structures . Here particularly, eye tracking and EEG-recordings are prominent, even though self-paced reading has been used as well. Finally, research on higher-level semantic integration necessarily needs to investigate longer text passages where the build-up of an overarching meaning is possible (e.g., into situation models), and the reading process is mostly measured using eye movements or self-paced reading.
The aim of study 1 was to test the outlined hypotheses of the presence of a phase-transition in the reading process between two reading tasks, namely random word list reading and ordered text reading. To that end, participants performed a self-paced reading task in which they were randomly assigned either to read a sequence of random words and then subsequently to read ordered text half-way through the task or, in another condition, to read words appearing as ordered text and then, half way through the task, to read words in random order.
The results from the O→R task switching condition are very much in line with what was expected from LGH. The findings provide us with a standard example of a phase-transition as a loss-of-temporal structure between the stable phases (i.e., the performances within each task). The positive slopes for the recurrence measures within each task suggests that there is still an increased adaptation of the cognitive processes with each reading condition, strengthening the notion that the reading process is not context-independent across different reading tasks but that reading reflects an interaction with the constraints of different reading tasks. Also, it is worth noting that the length of this transition spans–on average–about 190 words. That is, settling in from one reading task into another spans a range of the stimulus material that alone rivals or exceeds the size of text material that is currently used within the reading task to examine cognitive performance in that task.
The aim of study 2 was to investigate whether a more consistent pattern for phase-transitions in reading times could be observed with a sample of less-skilled readers who are less familiar with the language of the text. Our remarks preceding Study 1 and the results following from Study highlighted long-unaddressed context-sensitivity of reading. That is, reading is a complex process whose dynamics might be rooted in single-word recognition but quickly bleed from single-word levels to discourse processing at the sentence- or paragraph-level and again to the whole-text narrative level. All of that context sensitivity manifest in strictly monolingual readers. Second-language (L2) reading embodies a dramatic accentuation of that context sensitivity offering an excellent test case to compare with Study 1. Not only does L2 reading reflect all of the same interactions across level, from word to discourse and narrative contexts, within the L2 text, but L2 readers’ ability to navigate an L2 text depends further on their own individual differences in command with their first language (L1) [65–69].To that end, Danish native speakers who spoke English as their second language (L2) were recruited to perform the same reading tasks as in Study 1, using English texts. Testing second-language bilinguals potentially provides evidence for the generality of the phase-transition effect across languages and native and non-native speakers. Moreover, we collected subjective ratings of text meaningfulness in order to investigate the effect of the task manipulation on subjective meaning. We suspect that the ability of participants to extract or construct meaning from randomized words depends on their prior reading experience (i.e. the presence vs. absence of an intact text), and that this in turn might modulate aspects of the phase-transition-like behavior during task switching.
As in Study 1, we tested the hypothesis that switching between two reading tasks leads to phase-transition like behavior in reading times also in L2 readers. Moreover, we speculated to find a more consistent pattern of phase-transitions for both task orders, O→R and R→O, insofar as the presence of such a transition phase was influenced by the lower familiarity of L2 readers compared to the L1 readers. Our analysis shows that Model 3 implementing H3 was consistently selected as the best model throughout in Study 2, indicating that the addition of intercepts and slopes for the phase-transition regime improved model fit. For the O→R condition, we did find similar effects as in study 1, but contrary to our expectations, these effects were attenuated, not pronounced. However, we did now find effects of the phase-transition parameters in the R→O condition, but they were not consistently marked by negative signs, indicating a loss of structure, but also showed increased gain in temporal structure, albeit from a lowered level. So, in line with our expectations, some of the parameters for the phase-transition regime were now found to be significantly different from zero in the reading times of the L2 sample compared to the L1 sample in Study 1. However, also study 2 shows that the transitions from O→R compared to R→O are indeed asymmetrical.
In the current study, we tested two hypotheses about what happens at the switch-point between two different reading tasks. The first prediction was inferred mainly from the current practice of conducting experiments in reading research but also accompanying theory. This led us to suggest that the reading process–or more specifically, its different components–are thought to be relatively task independent and quick to adapt to new reading tasks, akin to a hard-wired processing architecture that will readily process any input that is encountered. In contrast, we derived a second prediction about what happens at the switch-point between two reading tasks from the language game hypothesis (LGH) of reading. Here, the reasoning is that the cognitive processes involved in reading are not hard-wired, but soft-assembled. That is, they are not constantly available in the background of the cognitive architecture, but emerge as an interplay between organism and (reading) task. Moreover, the hypothesis was motivated by problems from the philosophy of language, particularly the problem of how meaning in language can be grounded and be flexibly constructed. Here, we interpreted reading in terms of Wittgenstein’s  concept of language game, where a particular context of language use (here: reading) is not just dependent on an individual’s language capacities and the structure of a given language, but is inherently context dependent. Drawing from LGH, we predicted that the switch from one reading task to another entails a change in the way language is processed and a re-assembling of the cognitive architecture in accordance with the specific constraint of each task. Hence, we predicted the occurrence of phase-transition like behavior in reading times when switching from one task to another.