Les débutants parlent-ils à l'infinitif? : Influence de l'input sur la production des verbes par des apprenants adultes du français

Sammanfattning: This dissertation deals with the characteristic variation found in adult French L2 oral production. In their production of regular verbs, beginners use a short form, like parle, and a long form, like parlE, in contexts of both the present (je parl/'je parlE) and the infinitive (je veux parlE/'je veux parl). This sometimes results in target-like uses and sometimes not. The question addressed in the thesis is if there is any systematicity in learners’ use of the verb forms. In attempting to account for the variation, some previous studies have pointed at specific linguistic factors while others have suggested that the observed variation can be more or less arbitrary. One kind of systematicity has been found in earlier studies focusing on the development of past tense morphology. These studies showed that production of the forms could be predicted from the prototypical lexical aspect of the verbs. Based on French L2 studies of this kind, I present a first hypothesis stating that, also when referring to non-past contexts, i.e. present and infinitive contexts, learners use the short form with static verbs and the long form with dynamic verbs. I further propose a second, contrasting hypothesis, namely that learners are influenced by the frequency of surface forms in the input. This hypothesis emerges from a usage-based perspective (Bybee, 2008) where input frequency is a key factor (Ellis, 2002). A first investigation of 16 regular verbs in different input sources (for example, corpora of spoken French and teacher talk) is conducted in order to obtain specific data for the second hypothesis. Firstly, this examination shows that the distribution of verb forms in the input is uneven: some verbs occur more often in the short form and others more often in the long form. A gradual scale can thus be established for how frequent each of the studied verbs is in the short and the long form respectively. Secondly, I found that lexical aspect and frequency are closely related. The verbs more frequent in the short form turned out to be static verbs while the verbs more frequent in the long form turned out to be dynamic verbs. The analysis of the same verbs in learner data from a corpus of oral production shows tendencies in favour of the influence of input frequency. This investigation is completed by using experimental data for 12 regular verbs from a free production task, an elicited imitation task and a (serial) recall task. The results of the imitation task, where the verbs are presented in sentences, showed a possible influence of input frequency from the beginning stages. An influence of lexical aspect could only be detected in the production by learners having passed the beginning stages. The recall task, where the verbs are presented in series of five at a time, reveals a weaker correlation with input frequency and no significant differences between the verbs with respect to the lexical aspect. In conclusion, the data show that lexical aspect and input frequency are closely related in input as well as in learner data and the results speak in favour of input frequency for the verbs where the two hypotheses can be distinguished.