How does the perception of a nonnative phoneme contrast develop? In answering this question we test two hypotheses: (i) Development through Acquired Distinctiveness: before learning a new phoneme contrast, differences between and within phoneme categories are hard to discriminate. Through training, the phoneme boundary is learned. (ii) Development through Acquired Similarity: before learning a new phoneme contrast, differences between and within phoneme categories are well discriminated. Due to training, only the phoneme boundary remains discriminable. The specific question that the present study dealt with is: how do Dutch listeners learn to perceive a length contrast from Finnish, although their native language does not use phonological length contrastively? On the basis of results from the literature and earlier findings within this project, we expect that the new contrast will be learned through Acquired Distinctiveness.
A phoneme training study with 28 native Dutch listeners (mean age = 23 years) was run. Half of them completed the full procedure of pretesttraining- posttest, whereas the other half formed the control group and only participated in pretest and posttest. As materials, the Finnish pseudowords 'ata' (/AtA/) and 'atta' (/At:A/) were manipulated into phoneme continua by varying stop closure duration from /t/ to /t:/ in productions from six speakers (3 males, 3 females). In pretest and posttest, classification and 2IFC discrimination of one speaker's continuum were administered. Training consisted of classification (with feedback) of the other five speakers' continua until participants reached a score of 90% correct responses in two subsequent 350-trial training sessions. Control data from six Finnish listeners were also collected.
Trained listeners learned to classify the new length contrast in a more native-like manner, whereas control listeners showed no such progress. With respect to 2IFC discrimination, neither trained listeners nor controls clearly showed the same perceptual sensitivity at the phoneme boundary as had been found for the Finnish controls. Trained listeners only tended to be more sensitive to the relevant region of the continuum. Thus, short-term training on a nonnative phoneme contrast may not successfully alter perceptual sensitivity. This means that nonnative listeners possibly need additional experience to become as sensitive to the phoneme boundary as native listeners are. However, comparison of Dutch and Finnish discrimination results leads us to the conclusion that learning most likely follows the first hypothesis: Acquired Distinctiveness.