My mistake! Production and perception of grammar anomalies

Abstract

This thesis examines the production as well as the perception of naturally occurring grammar anomalies, focusing on verb-third (V3) word order in written Danish (and Norwegian). In these verb-second (V2) languages, the norm prescribes that the finite verb must occur as the second constituent in the sentence. Sentences with V3 word order do not confine to this norm as they have the finite verb in the third position, e.g. *Nu han spiser ‘Now he eats’ for Nu spiser han. V3 is rare in L1 Danish, but is common in Danish as a second language (DSA). The thesis consists of three research articles: the first examines learners’ production of V2 and V3, while the second and third examine L1 users’ attention to V3 (and other anomalies), as well as their online processing of V3. In contrast to previous research on DSA, which has mainly been qualitative, this thesis is based on quantitative studies. As something new, it examines which factors especially have influence on V3 production, including crosslinguistic influence. As for perception, the thesis contributes knowledge about L1 users’ processing of anomalies produced by learners, and about allocation of attention to different types of grammar anomalies. Article 1 builds on a cross-sectional corpus study of texts from 217 learners of Danish (A2- B1 level). The study examined effects of the learners’ language background (V2 or non-V2), their CEFR level, and the complexity of the first three sentence constituents on correct V2 production. The study showed that V2 is not difficult for all learners, as learners with another V2 language as their L1 had significantly higher shares of V2. For non-V2 learners, the share of V2 significantly increased with CEFR level. Finally, V2 is not equally difficult in all contexts, as increasing complexity of the first constituent and the subject was negatively correlated with the share of V2 (both significantly). Article 2 is based on an error detection study, in which 211 Danish high school students read texts containing different types of errors: syntactic errors (V3), morphological agreement errors (verb inflections, gender mismatches in NPs) and orthographic errors. The study showed that different types of errors do not attract the same amount of attention, and that V3 is a prominent anomaly. In general, it was found that error rates in an L1 corpus could be used in the hypotheses to predict detection rates for different subtypes of grammar errors. However, frequency was not the only possible explanation. Finally, the study also showed that high scores in a grammar quiz and high levels of self-reported annoyance with errors affected error detection positively. Article 3 builds on an eye-tracking study of L1 Norwegian users’ online processing of V3 anomalies. Participants read sentences with sentence-initial adverbials, followed by either V2 or V3 word order. The study showed that L1 users reacted immediately to V3 word order, as indicated by increased fixation durations and more regressions out on the misplaced subject, and subsequently on the verb. Participants recovered quickly, already on the following word. This thesis can inform future eye-tracking and language processing models on allocation of attention to, and processing of, different types of grammar anomalies. It both presents a tentative model of factors in anomaly detection and a preliminary model of how perception patterns can be affected by what is common in L1 production (inspired by prediction theory). This may lead to more robust models which can accommodate naturally occurring, non-standard variation. From a didactic perspective, knowing which anomalies attract more attention, or require additional processing costs, can both help language instructors and school teachers prioritize grammatical focus areas in the classroom.