Abstract: An important question for research on reading across the lifespan concerns whether mechanisms of cognitive processing undergo only quantitative changes or also qualitative changes with aging. To process written language effectively, readers use their existing knowledge to make predictive inferences about linguistic information. Quite often this will facilitate the processing of newly acquired information but will sometimes incur a processing cost due to predictive error. As Older adults appear to rely more heavily on lexical prediction during reading (Zhao et al., 2019, 2021). However, it is currently unknown whether, like young adults, they experience a processing cost due to predictive error, and whether the magnitude of this cost differs across age adult groups. Accordingly, the present research aims to understand the processing consequence of predictive error in both young and older adults, using methods that can shed light on both the behavioral and neural bases of these effects. This will be achieved using novel co-registration methods that synchronize the recording of electroencephalographic (EEG) signals with eye movements, so that behavioral and neural indices of language processing can be acquired simultaneously, in real-time, during natural reading. In particular, this approach will enable the analysis of fixated-related potentials (FRPs), which are averaged EEG waveforms time-locked to a fixation on a target word in a sentence during normal reading.
Study 1 will manipulate whether a target word is predictable from the prior sentence context, using contexts in which the target word is predictable, ones in which it is unpredictable, and neutral contexts containing an unpredictable word. Crucially, comparisons of an unpredictable word in neutral compared to constraining contexts will provide a measure of prediction error, which is the cost incurred when the target word is unpredicted in a constraining context. The study will investigate the behavioral and neural correlates of this prediction error using a combination of eye movement measures and FRPs for target words. Moreover, by investigating age differences in these effects (i.e., for young compared to older adults) the study will reveal whether this prediction error differs across adult age groups.
Study 2 will test these effects further by examining both the contribution to the prediction error cost of parafoveal information availability and individual differences in visual, cognitive and linguistic abilities. To examine the contribution of these individual differences, we will comprehensively assess the visual, cognitive and linguistic abilities of young and older adult participants prior their taking part in experiments. We will obtain information about participants' educational background, vocabulary knowledge and recent reading experience to match participants in terms of formal educational experience and to obtain indices of linguistic experience. In addition, we will assess processing speed, working memory, and inhibition as measures of cognitive capabilities. The data obtained will be used for the linear mixed-effects modelling of Study 3. Experiment 1 will use the boundary paradigm to investigate age differences in the prediction error cost when parafoveal information is available or not. The aim of this experiment is to establish whether limiting the availability of parafoveal information about an upcoming word differentially impacts lexical prediction by young and older adults. Experiment 2 will use masking text paradigm to investigate the aging effects on prediction error cost under high or low working memory load conditions. The aim of this experiment is to explore the effect of working memory load on prediction processing mechanism of young and older readers. Finally, in Experiment 3, the older adult participants will be divided into good and poor reading skill groups to examine whether there is a difference in the prediction error cost for older participants with good and poor reading skills as compared to skilled young adult readers. This will reveal how reading skills mediates predictive processing by older adults.
Study 3 will use linear mixed-effects modelling and data-mining methods. All relevant factors will be included in the model analysis as covariates to investigate their effects on the prediction processing of older readers. Moreover, survival analysis and distribution analysis will be adopted to investigate the time course and individual differences of the above-mentioned effects (using data from Study 1 and 2).
The findings from these studies will provide important insights into the nature of effects of cognitive aging and individual differences in visual, cognitive and linguistic abilities on neural and cognitive indices of word prediction in reading, and will form the basis for future models of these effects in Chinese reading. Moreover, the findings will shed light on the contribution of parafoveal processing, memory load, and reading skill on the predictive abilities of older adult readers.

Key words: contextual effects, prediction error cost, Chinese reading, aging effects, co-registration of eye movements and EEG