Two key methods of perceptually training difficult L2 contrasts are the perceptual fading (PF) technique and the high variability phonetic training (HVPT) technique, and perceptual benefits from using both of these methods have also been found to transfer to pronunciation. However, these techniques have not been compared in their classic forms (PF with one speaker vs. HVPT with multiple speakers) with regard to perceptual gains, nor have they been compared with regard to gains in pronunciation accuracy or how any improvement is retained in the long term. Furthermore, whilst a number of studies suggest that motivation, the concern for L2 pronunciation accuracy aspect in particular, along with perception and/or pronunciation training may contribute to more nativelike pronunciation in late L2 learners, this has not been examined with specific reference to these training techniques. The present work compares these techniques for training native English speaking learners of French on difficult L2 French contrasts (/u/ vs. /y/ and /ɑ˜/vs./ɔ˜/),and assesses participant concern for pronunciation accuracy in order to ascertain an optimal training technique to improve the perception and pronunciation of less able learners. Experiment 1 of this thesis compares HVPT and PF using multiple and single speakers and found that the single speaker HVPT technique was significantly less effective than the others immediately after training. Testing again after at least one month suggested that training was best retained either through using PF with one speaker or HVPT with multiple speakers, that is, the techniques in their classic forms. Experiment 2 examines the benefits of these perceptual training techniques vs. pronunciation training vs. perception AND pronunciation training for both perceptual and pronunciation improvement. Undergoing multiple speaker HVPT + pronunciation training (over the same timescale as training in a single modality) appeared to be most beneficial for perception and pronunciation. Experiment 3 examines the relationship between average pronunciation improvement and participant concern for pronunciation accuracy as measured Elliott’s (1995) Pronunciation Attitude Inventory and found that a high concern for pronunciation accuracy is only related to greater improvements when specific, perhaps more monotonous, training techniques (using only one modality and speaker) are used. Overall, the present results provided no evidence of transfer of perceptual training benefits to pronunciation, and only slight evidence of transfer of pronunciation training benefits to perception, although there was a clear link between participant perception and pronunciation ability before training commenced. This is likely to be at least partly why some training in both modalities emerged as most successful in terms of improvements in both domains. It was therefore suggested that it may be prudent to consider the relationship between perceptual and production learning as distinct from any links between perception and production in general.

Cette thèse s’inscrit dans une réflexion sur les apports potentiels de la phonétique expérimentale à la didactique et à l’enseignement de la prononciation. Son objectif est d’examiner les préférences perceptives des locuteurs du finnois, de l’allemand et du français, pour les voyelles antérieures non-arrondies /i e ɛ/ ; et de prolonger cette démarche dans le cadre de l’enseignement de la prononciation du français, en s’interrogeant sur la nécessité de continuer à considérer le contraste existant entre les voyelles moyennes /e/ et /ɛ/. La première partie de ce travail dresse un portrait des trois langues concernées, et introduit les principales hypothèses de processus cognitifs impliqués dans la perception. La deuxième partie de ce travail aborde la perception des voyelles /i e ɛ/, à travers trois tests de perception permettant 1) de définir les prototypes privilégiés par chacune des populations, 2) de mesurer l’effet d’un aimant perceptif spécifique à la langue, et 3) d’évaluer l’importance accordée à l’abaissement de la mandibule pour le choix des exemplaires de chacune des catégories concernées. L’analyse d’un corpus de parole spontanée nous permet enfin d’examiner la robustesse du contraste /e/~/ɛ/, et de mesurer l’intérêt didactique d’enseigner la différenciation et l’acquisition du timbre ouvert et du timbre fermé de ces voyelles en français langue étrangère. La troisième partie de ce travail s’intéresse enfin à la place accordée à la phonétique dans l’enseignement des langues, et offre une revue des principales tendances méthodologiques qui ont contribué à sa diffusion. L’intérêt d’investir différents outils issus de la phonétique expérimentale, du support multimédia, ou de disciplines non-linguistiques, y est discuté. / This work has its place within the context of reflection on the potential input of experimental phonetics to pronunciation teaching. Its aim is to compare the perceptual preferences of native speakers of Finnish, German, and French, for the front non-rounded vowels /i e ɛ/ ; and to extend this approach to French pronunciation teaching, by analyzing the contrast between the middle vowels /e/ and /ɛ/. The first part of this dissertation presents a picture of the phonological systems and phonotactics of the three languages and recalls the main theories about the cognitive processes involved in the perception of phonetic categories. Based on this comparison, the second part examines the perception of the vowels /i e ɛ/. Three studies have been conducted, in order to 1) determine the category’s prototypes for the three populations studied, 2) measure the impact of a language-specific perceptual magnet [Kuhl, 1991], and 3) assess the involvement of mouth opening, to ensure the contrast between the categories in the three languages. The robustness of the contrast /e/~/ε/ through the acoustic analysis of 633 occurrences of /E/ in a spontaneous speech corpus by one French native speaker has been evaluated and the interest to systematically distinguish the sounds of these two vowels in French as a foreign language has been assessed. The diverse conditioning factors, as the trends highlighted in our corpus, are compared to a similar study carried out by Léon and Tennant [1990] on 100 occurrences of /E/ taken from television broadcasts of Bernard Pivot. The third part of this work finally deals with the place granted to phonetics in language teaching and provides a review of the main methodological tendencies which have contributed to its diffusion. The potential of applying different tools resulting from experimental phonetics research, multimedia or non-linguistic disciplines is also discussed.