Sesssion 1D

Discourse competence, which typically develops at higher levels of language learning (Savignon 2002), includes aspects such as thematic development, coherence and cohesion (CEFR 2020: 138-142). Written production at B2 level entails the ability to mark "the relationship between ideas in clear connected text" (ibid. 67). This means that a B2 language user should be able to establish coherence and adequately use cohesive conjunctions in written texts. 
The goal of the present study was to analyse the coherence and the use of cohesive conjunctions in texts written by Croatian B2 learners of English and German as foreign languages. The corpus of 100 argumentative essays (50 in English and 50 in German) was analysed to answer the research questions referring to the following: (i) the occurrence of particular types of thematic progression and the frequency of coherence breaks in texts; (ii) the occurrence of particular types of cohesive conjunctions and the frequency of adequately used cohesive conjunctions; (iii) the relationship between the text coherence scores and the use of cohesive conjunctions. 
For the text coherence analysis, a method developed within the KohPiTekst project (2017 – 2020; https://pdb.irb.hr/project/3874) on the basis of the methods by Daneš (1970) and Lautamatti (1987) was applied. In the analysis of cohesive conjunctions the classifications by Pasch et al. (2003) and Kunz & Lapshinova-Koltunski (2014) were used. The quantitative analysis was conducted using the SketchEngine (http://www.sketchengine.eu). Descriptive and inferential analyses were run using IBM SPSS 20. 
The preliminary results revealed the occurrence of coherence breaks mainly within the text paragraph, and pointed to individual differences in the use of certain cohesive conjunctions (over- and underrepresentation), as well as to the pragmatically inappropriate use of some cohesive conjunctions. In addition, a positive but statistically non-significant correlation between coherence scores and the use of cohesive conjunctions was observed.
Key words:
coherence; cohesive conjunction; English as a FL; German as a FL.
References:
CEFR = Council of Europe (2020): Common European Framework of Reference for Languages: Learning, teaching, assessment. Companion volume. www.coe.int/lang-cefr. 
Daneš, František (1970). "Zur linguistischen Analyse der Textstruktur". Folia Linguistica 4, 72–78.
Kunz, Kerstin; Lapshinova-Koltunski, Ekaterina (2014): "Cohesive conjunctions in English and German: Systemic contrasts and textual differences". Vandelanotte, Lieven; Davidse, Kristin; Gentens, Caroline; Kimps, Ditte (eds.): Recent Advances in Corpus Linguistics: Developing and Exploiting Corpora. Amsterdam/New York: Rodopi, 229-262. 
Lautamatti, Lisa (1987). "Observations on the development of the topic of simplified discourse". Connor, Ulla; Kaplan, Robert B. (eds.). Writing across Languages: Analysis of L2 Texts. Reading, MA: Addison-Wesley Publishing Co., 87-114.
Pasch, Renate; Brauße, Ursula; Breindl, Eva; Waßner, Ulrich Hermann (2003): Handbuch der deutschen Konnektoren: Linguistische Grundlagen der Beschreibung und syntaktische Merkmale der deutschen Satzverknüpfer (Konjunktionen, Satzadverbien und Partikeln). Berlin: Walter de Gruyter. 
Savignon, Sandra J. (2002): "Communicative language teaching: Linguistic theory and classroom practice". Savignon, Sandra J. (ed.): Interpreting Communicative Language Teaching: Contexts and Concerns in Teacher Education. New Haven, CT: Yale University Press, 1-27.

Several studies have shown that null subject languages (NSLs) behave differently from each other in pronoun resolution. Studies with speakers of an NSL who acquire a non-null subject language (NNSL) as a second language (L2) have shown mixed findings: some report similar performance in L2 learners and L1 speakers (e.g., Cunnings et al., 2017; Contemori & Dussias, 2020), whereas others reveal differences between the two groups (e.g., Roberts et al., 2008). One question that arises is if the different outcomes are due to differences in the language-specific features of the L1 NSL. Yet no studies have investigated pronoun resolution differences between groups of L2 NNSL learners, who have different L1 NSLs. Our study aims to address this issue by investigating whether L1 speakers of Greek vs. Italian (NSLs) transfer the resolution preferences of their L1 to their L2 German (NNSL). 
In Italian and Greek, null pronouns usually refer to subjects, whereas overt pronouns preferably refer to objects. However, Torregrossa et al. (2020) claim that Greek is more flexible in the interpretation of null and overt pronouns compared to Italian. In the NNSL German, subjects are usually encoded with personal pronouns and objects with demonstrative pronouns (e.g., Schumacher et al. 2017). If participants transfer language-specific features from the L1 to the L2, learners with L1 Greek should show more variability in German pronoun resolution than learners with L1 Italian.
We tested pronoun resolution preferences in two groups of highly proficient L2 German learners (30 L1 Greek, 29 L1 Italian) with age-of-onset after 18 years and 74 L1 German controls using a picture-selection task with 32 items (16 experimental, 8 filler). In each item participants were presented with two sentences auditorily; the first one contained two referents, the second one either the personal pronoun er or the demonstrative pronoun der (both 3rd-person-singular-masculine pronouns), see (1) below.


1. Der Tigerj will den Igelk vorsichtig auf die Hand küssen. Aber erj / derk kann nicht stillstehen.
The tiger wants the hedgehog carefully on the hand kiss. But he cannot hold still.
'The tiger wants to kiss the hedgehog carefully on the hand. But he cannot hold still.'
 
Three pictures were presented simultaneously on a computer screen, representing the subject and object referents (competing pronoun antecedents) and a distractor (also mentioned in the sentence). After the sentences, participants answered a content question by selecting a picture, forcing them to resolve the pronoun towards one of the possible antecedents. 
In contrast to previous studies, both groups of highly proficient L2 learners differed significantly from the L1 speakers (p > .001) (see Fig.1). L1 Greek and L1 Italian participants resolved pronouns at chance and there was no significant difference between the two groups. This goes against the assumption that speakers transfer pronoun resolution features from their L1 NSL to the L2 NNSL, suggesting that earlier contradictory findings cannot be reduced to this explanation. We will discuss the role of age-of-onset, length of exposure, and proficiency in pronoun resolution.

Although learning plays a crucial role in language learning, insights from Learning Theory are rarely integrated into L2 studies (with some exceptions such as Zhao & MacWhinney, 2018). Taking the English article system as a case study, we explore the contribution insights from Learning Theory can make to help us determine what guides the choice of article and thus what c(l)ues should be taught to EFL learners to optimise their learning. Since research within Learning Theory has found that order of exposure affects learning, we argue that the choice of cues and the order in which they are learned/taught is crucial. First, we identify which elements in the context are the most informative for the choice of article. Then, we focus on order of exposure; that is, learners should first be presented with the most reliable elements that guide native speakers' choice of articles, which account for the bulk of article uses, and only progressively introduce less informative and more limited elements so as to avoid blocking effects (Ellis, 2007).
To identify which elements guide the choice of article in English, we manually annotated a dataset of 2000 discourse chunks extracted from the BNC (1000 spoken, 1000 written): each chunk contains seven sentences with the sentence containing the target article in the middle. We annotated this dataset with the five most frequently used variables, i.e., Hearer Knowledge (HK), Specificity of the referent (SR), countability, number and elaboration. Using the Widrow-Hoff algorithm (1960) we ran computational simulations of learning on our dataset. In the first simulation, we fed the algorithm our data with no particular order of exposure (random). We found that different articles are learned from different cues, with HK leading in the acquisition of a and the, but number:plural and uncountable being the most informative cues for Ø. We then tested order of exposure by creating two order-manipulated datasets in which the cues for the variable HK (1) or the cues for SR (2) are introduced halfway through training.
The results from the simulations suggest that HK is crucial in determining article choice and when introduced late it affects the learning process, i.e., it hinders the acquisition of the article system by relying solely on other, less informative cues that become less and less reliable as learning unfolds. Introducing SR before HK also affects the learning process. The SR cues get very salient at first, before inducing a blocking effect. These results thus confirm the importance of incorporating insights from Learning Theory into research on language learning.
Selected references
Ellis, Nick C. (2007). Blocking and learned attention in language acquisition. Proceedings of the Annual Meeting of the Cognitive Science Society, pp. 965-970.
Widrow, B., & Hoff, M. E. (1960). Adaptive switching circuits. Paper presented at the WESCON Convention Record Part IV.
Zhao, Helen & Brian MacWhinney. (2018). The instructed learning of form-function mappings in the English article system. The Modern Language Journal 102(1), pp. 99-119.