ETD

• finger-tracking
• L1/L2
• predictive processing
• reading beahviour
• second language aquisition

Research on reading behaviour has traditionally relied on eye-tracking and electrophysiological techniques to investigate the temporal dynamics of language processing. This thesis explores the potential of finger-tracking as an innovative, low-cost alternative to capture reading behaviours in real time. The study compares a small group of native English speakers (L1) and of English L2 learners as they read short sentences presented on a tablet screen: 16 native English speakers recruited in Oxford and 16 native Italian speakers learning English as an L2. The two groups were comparable in age, gender distribution, and years of schooling. The L2 group spanned CEFR proficiency levels from A1 to C1, with an average of B2.
The stimuli include consisted of 16 medium-length English sentences. Each sentence had two versions: a grammatically correct version and a version containing a deliberate violation. The violations fell into two main categories (morphosyntactic and semantic) with a total of seven violation types. Most sentences contained morphosyntactic errors (e.g. phrasal verbs, Saxon genitive, word order, passive voice, double negation), while four contained semantic errors (lexical or idiomatic). These error types were chosen because they are characteristic of English and particularly challenging for L2 learners.
Behavioural measures derived from the finger trajectories are analysed and quantitatively modelled to identify possible differences between native and non-native readers, and between correct and anomalous sentences. The central hypothesis is that native readers rely more on predictive processing and will show earlier, stronger disruptions when expectations are violated, whereas L2 readers will show delayed or weaker effects, reflecting more integrative and less predictive processing.
To encourage natural reading and assess attention, yes/no questions followed some sentences. Two types of questions were used: error-detection questions (“Is the sentence correct?”) and comprehension questions, as distractors to reduce focus on grammaticality. This design ensured comparable conditions across participants while maintaining naturalistic reading behaviour.
Results at the sentence level showed in L1 readers clear, systematic speed changes in response to linguistic violations, while L2 readers displayed greater variability. At the word level, violations triggered sharp, localized slowdowns in L1 readers but more gradual, widely distributed effects in L2 readers. At the within-word (symbol) level, L1 readers exhibited early sensitivity to violations, whereas L2 readers showed delayed responses. Overall, the findings demonstrate both quantitative and qualitative differences in the temporal dynamics of reading between the two groups.
The study suggests that finger-tracking is a sensitive, continuous behavioural measure of real-time language processing that complements eye-tracking and ERP methods. By capturing fine-grained changes in tracking speed at sentence, word, and within-word levels, it reveals how readers respond to linguistic violations as reading unfolds. The results indicate that L2 readers tend to rely more on cumulative lexical and syntactic information, showing slower and more gradual responses to anomalies, whereas L1 readers exhibit earlier, more predictive sensitivity.
The alignment between symbol-level finger-tracking dynamics and established ERP findings supports the view that finger-tracking reflects behavioural manifestations of predictive versus reactive processing. Applied implications include the potential use of temporal reading profiles as markers of processing efficiency, with relevance for language assessment and educational interventions. Overall, the study highlights finger-tracking as a scalable and effective tool for studying real-time reading strategies.