WhatsHardQuantitativeEvidenceforDifficultConstru.pptVIP

What’s Hard?Quantitative Evidence for Difficult Constructions in German Learner Data Amir Zeldes amir.zeldes@rz.hu-berlin.de Anke Lüdeling anke.luedeling@rz.hu-berlin.de Hagen Hirschmann hirschhx@rz.hu-berlin.de Research questions What’s hard/easy for L2 German learners, and how can we find this out? What do (advanced) learners do differently from natives? Why? Overview Operationalizing L2 difficulties Learner data and the Falko corpus Analysis of two case studies Summary and conclusions Approaches to L2 difficulty Use intuition / introspection as learner, teacher or native speaker Compose questionnaires for students or teachers (Diehl et al. 1991) Gather corpus data: Learner corpora (see Pravec 2002; Tono 2003; Granger, to appear) Comparable L1 corpora Corpus Data Learner corpora contain L2 learner data from essays, exercises etc. (see Granger 2002, to appear) Usually give metadata on learner level and background Some contain explicit error annotations (Corder 1981) Error annotation Essentially based on a target hypothesis: “what should the learner have said?” John goed home John went home anno=[irregular past tense form error] But things are not always so simple… Ambiguity of error annotation Lüdeling (2008) Target hypothesis: experiment 5 annotations for 17 sentences (one text) target hypothesis differs, annotation scheme identical Working with raw learner data Frequencies of word forms etc. in learner data Work on lexical density as an index of L2 competence (Halliday 1989; Laufer/Nation 1999) Studies using underuse/overuse compared to native data in the framework of Contrastive Interlanguage Analysis (see Selinker 1972; Ringbom 1998; Granger et al. 2002) Underuse and Overuse Simplified model of target language competence Learner’s interlanguage distributions as opposed to L1 distributions Underuse and overuse defined as statistically significant deviations from L1 control frequencies Underuse as an index of difficulty Phenomena that are und