Innovation in Language Learning

Non-native speakers usually display pronunciations which differ from the native norm resulting in a foreign accent (FA). FAs may have important consequences on communication such as decreases in intelligibility, speaker stereotyping or an increase in listener effort. Most foreign language (FL) speech learning theories (Flege 1995, Best 1995, Kuhl 1993) consider that non-native pronunciations are closely related to perceptual difficulties in the FL. In this study we investigate foreign-accented sounds in order to ascertain to what extent native and non-native (NN) listeners consider them conveyors of FA and whether this is due to a lack of perceptual acuity or to top-down phonological processing differences between natives and NNs. If NN listeners cannot readily perceive differences between foreign-accented pronunciations resembling their own and native renditions, pronunciation training can be a much harder task than if they are able to detect FA. We designed a  categorisation and discrimination study applied to FA. Waveforms for exemplars of six English consonants (/h v t r j dz/) which are problematic for Spanish learners of English were manipulated to produce a number of degrees of deviation from native English targets, leading to a 9-step continuum from foreign to native pronunciation. A group of native English listeners and a group of Spanish learners of English took part in both a discrimination and categorisation task. In the discrimination task, each stimulus was made up of two pronunciations of the same word, either both belonging to the same step along the FA continuum ("same" tokens) or separated by two steps in the FA continuum ("different" tokens). Listeners were asked to decide whether the two pronunciations were the same or different. In the two-alternative forced-choice categorisation task, listeners were asked to decide if manipulated word tokens had a foreign or native accent. Results indicate that non-native and native listeners had broadly comparable auditory discrimination abilities, i.e., both groups were similarly able to distinguish  items in the continuum. On the other hand, for half the tested sounds (/v j dz/), non-native listeners were noticeably less able than native listeners to detect the presence of FA. Interestingly, for these three sounds, the FA and the native pronunciations are possible realizations of the same phoneme in the learners' L1. Consequently, our results suggest that the perceptual problems underlying the acquisition of these sounds may be due not so much to atrophied perceptual capabilities but to top-down L1 phonological processing. These findings can be used to inform pedagogical decisions concerning acquisition possibilities and the amount of focus required for the learning of various non-native sounds. The outcome of the study suggests that NN listeners can be trained to distinguish foreign-accented renditions of sounds, even when these resemble their own NN pronunciations, but that particular emphasis is required for sounds which fall within their L1 categories.