Leigh Breakell Fernandez

The divergence point analysis using bootstrapping (DPA-B) has become an increasingly popular approach for analysing visual world paradigm data. However, several researcher degrees of freedom must be considered, and the impact of these decisions has not been systematically in-vestigated. We examine the influence of five degrees of freedom: (1) region of interest size, (2) inclusion/exclusion of incorrectly answered trials, (3) bin size, (4) significance window length, and (5) bootstrap iteration number. We illustrate these effects using two datasets collected from human participants. Because this data does not have a known divergence time, we additionally present a publicly available data simulator that allows users to specify a true divergence time, thus allowing us to more directly assess the impact of these degrees of freedom on DPA-B outcomes. Across both empirical and simulated data, we found that bin size and significance window length exerted the strongest influence, with 50 ms bins and a 200 ms significance window length producing the most stable and interpretable results. Additionally, we found that the DPA-B tends to overestimate divergence times; however, this bias was consistent across conditions, preserving its utility for cross condition comparisons. Based on our findings, we offer empirically grounded recommenda-tions for DPA-B implementation and emphasise the importance of transparent reporting, coupled with post-hoc flexibility tailored to individual datasets. Together, these results provide practical guidance to improve the quality, reproducibility, and comparability of research employing DPA-B.

We investigate timing and eye-movement behavior during semantic prediction in L1 and L2 speakers of English using the Visual World Paradigm, additionally exploring speech rate. We differentiate first-stage predictions, considered to be automatic and relatively cost-free, from second-stage predictions, which are non-automatic and more cognitively demanding, with differences between L1 and L2 speakers believed to arise in second-stage predictions. We found no differences in the divergence of looks to the target in first- or second-stage predictions across groups. However, speech rate played an important role. Both L1 and L2 speakers showed similar first-stage predictions at slower speech rates, but L1 speakers showed earlier predictions as the speech rate increased. L2 speakers showed reduced and more variable second-stage predictions, suggesting they were impacted during the more demanding second-stage prediction. This may indicate a wait-and-see strategy to help reduce costs associated with second-stage prediction.

While there is strong evidence that younger adults use contextual information to generate semantic predictions, findings from older adults are less clear. Age affects cognition in a variety of different ways that may impact prediction mechanisms; while the efficiency of memory systems and processing speed decrease, life experience leads to complementary increases in vocabulary size, real-world knowledge, and even inhibitory control. Using the visual world paradigm, we tested prediction in younger (n = 30, between 18 and 35 years of age) and older adults (n = 30, between 53 and 78 years of age). Importantly, we differentiated early stage predictions based on simple spreading activation from the more resource-intensive tailoring of predictions when additional constraining information is provided. We found that older adults were slower than younger adults in generating early stage predictions but then quicker than younger adults to tailor those predictions given additional information. This suggests that while age may lead to delays in first activating relevant lexical items when listening to speech, increased linguistic experience nonetheless increases the efficiency with which contextual information is used. These findings are consistent with reports of age having positive as well as negative impacts on cognition and suggest conflation of different stages of prediction as a basis for the inconsistency in the aging-related literature to date. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

Research has shown that first (L1) and second language (L2) speakers actively make predictions about upcoming linguistic information, though L2 speakers are less efficient. While prediction mechanisms are assumed to be qualitatively the same, quantitative prediction-driven processing differences may be modulated by individual differences We tested whether L2 proficiency and quality of lexical representation (QLR) impact the capacity of L2 readers to extract parafoveal information while reading, leading to quantitative differences in prediction. Using the same items as Slattery and Yates, we investigated the impact of predictability and length of a critical word on bottom-up parafoveal processing, measured by skipping rates, and top-down predictability processing, measured by reading times. Comparing our L2 English to their L1 English data, we found that L2 speakers skipped less and had longer gaze duration. However, both groups showed increased skipping rate and decreased gaze duration for predictable relative to unpredictable words and for shorter relative to longer words. We argue that L1 and L2 predictability mechanisms are qualitatively the same and quantitative differences stem from L2 speakers’ Reduced Capacity for Parafoveal Processing, the ReCaPP hypothesis.

Reading is an essential method for adults to learn new languages, but difficulty reading texts in a foreign language can increase learners’ anxiety. Identifying text difficulty from the reader’s perspective can aid language learning by tailoring texts to readers’ needs. There is little research focusing on L2 speakers or using a multimodal approach, i.e., using multiple sensors, to detect subjective difficulty. In this study (N = 30) we determined L2 speakers’ subjective difficulty while reading using language proficiency and objective text difficulty, combined with physiological data. We compared machine learning classifiers combining eye, skin and heart sensor data against models using each modality separately. Additionally, we assessed the effect on model performance of shifting the data to account for delayed physiological responses. The models detected 3 levels of subjective difficulty (low, medium, high) and were evaluated using leave-one-participantout (LoPo) and leave-one-document-out (LoDo) cross-validation. The results showed acceptable levels of generalization to new participants (Acc_LoPo = 0.434) and documents (Acc_LoDo = 0.521). Combining sensor data from all modalities improved predictions in both LoDo and LoPo cross-validation, compared to each modality in isolation. Shifting the data to account for physiological response delay did not improve model performance compared to not shifting the data. These findings support refining subjective difficulty detection models and their implementation in adaptive language learning systems. Finally, this work contributes to the field of cognitive science and technology by laying the foundation for innovative approaches to cognitive state detection.

Research has found that both first language (L1) and second language (L2) speakers make predictions about upcoming linguistic information, with predictive behaviour being impacted by individual differences and methodological factors. However, it is not clear whether a cost is incurred when a prediction is made, but not met. L2 speakers have less experience with their L2 and parsing can be cognitively demanding, which together may lead L2 speakers to incur prediction costs differently relative to L1 speakers. In this study using the visual world paradigm, we test whether L1 and L2 speakers predict in the same way, within the same time frame, and incur the same costs if predictions are not met. We also explore the role of proficiency and speech rate. We found that both groups predict in a similar way and within a similar time frame. In addition, neither group incurred a prediction cost when the target was the most likely alternative, though L2 speakers take longer to shift their attention to the target object when predictions are not met. We argue that this reflects a slowing of lexical access rather than a specific cost of prediction. We only found prediction differences when speech rate was included in the analysis, highlighting the importance of attending to speech rate in studies using the visual world paradigm. Overall, this study supports research showing that both L1 and L2 speakers may make multiple partial predictions about upcoming information rather than predicting one specific lexical candidate while inhibiting less likely lexical candidates.

The Uniform Information Density (UID) hypothesis proposes that speakers communicate by transmitting information close to a constant rate. When choosing between two syntactic variants, it claims that speakers prefer the variant distributing information most evenly, avoiding signal peaks and troughs. If speakers prefer transmitting information uniformly, then comprehenders should also prefer a uniform signal, experiencing difficulty whenever confronted with informational peaks. However, the literature investigating this hypothesis has focused mostly on production, with only a few studies considering comprehension. In this study, we investigate comprehension in two eye-tracking experiments. Participants read sentences of two different lengths, reflecting different degrees of density, containing either a dense structure (a nominal compound, NC) or a structure that spreads the information through more words (a noun followed by a prepositional phrase, PP). Favoring the UID hypothesis, participants gazed longer at text segments following the critical structure when it was an NC than when it was a PP. They also regressed more in sentences containing longer structures. However, the pattern of results was not as clear as expected, potentially reflecting participants’ experience with the denser structure or task differences between production and comprehension. These aspects should be taken into account in future research investigating the UID hypothesis for comprehension.

In a recent study, Fernandez et al. (2021) investigated parafoveal processing in L1 English and L1 German–L2 English readers using the gaze contingent boundary paradigm (Rayner, 1975). Unexpectedly, L2 readers derived an interference from a non-cognate translation parafoveal mask (arrow vs. pfeil), but derived a benefit from a German orthographic parafoveal mask (arrow vs. pfexk) when reading in English. The authors argued that bilingual readers incurred a switching cost from the complete German word, and derived a benefit by keeping both lexicons active from the partial German word. In this registered report, we further test this finding with L1 German–L2 English participants using improved items, but with the sentences presented in German. We were able to replicate the non-cognate translation interference but not the orthographic facilitation. Follow up comparisons showed that all parafoveal masks evoked similar inhibition, suggesting that bilingual readers do not process non-cognate semantic or orthographic information parafoveally.

Welcome to e-content platform of John Benjamins Publishing Company. Here you can find all of our electronic books and journals, for purchase and download or subscriber access.