In a series of experiments we investigated whether L2 learners could learn to perceive non-native phonological contrasts that are based on acoustic cues not used in the native language. We examined the perception of the Dutch contrast (/a:/-/A/) by native-speakers of Brazilian Portuguese (BP). This Dutch contrast involves spectral and durational differences and is not found (phonemically) in Portuguese. Forty-one subjects took part in the experiments: 10 Dutch listeners, and 31 BP learners of Dutch who had lived in The Netherlands for 6 to 216 months. Subjects perfomed three experiments:
Experiment 1 was an AXB experiment with natural Dutch words (/tAk/ 'branch' and /ta:k/ 'task'). Subjects heard three words in a row (/ta:k/ or /tAk/) and had to decide whether the middle word (X) was equal to the first (A) or to the third (B);
Experiment 2 was similar to Experiment 1, but with hybrid stimuli with synthetic vowels inserted between natural consonants. The vowel was any of the 49 tokens resulting from a 7-step spectrum and 7-step duration manipulation;
Experiment 3 consisted of a read-aloud task, and was used as an independent measure of learner's L2-experience/proficiency.
Our results show that several learners performed in a native-like manner, which suggests that their L2 perception has developed to match that of Dutch listeners. However, results from Experiment 1 also reveal that not all L2 learners were native-like. We examined two possible explanations for this divergence: (1) The learners have no categories for the L2 contrast; (2) Their L2 categories deviate from the native ones. Results of Experiment 2 show that all L2 learners were able to identify the vowels in a native-like manner when spectral and durational cues were manipulated at the same time. However, results of the same experiment reveal that, unlike Dutch listeners who always integrate both temporal and spectral cues when perceiving the contrast, several L2 learners typically integrate these cues in a non-native manner. This leads us to conclude that hypothesis (2) is borne. As for the differences in degree of native-like perception, we found no correlation between Experiment 1 or 2 and Experiment 3 (or any other independent measure of language experience/proficiency). Nevertheless, we can explain such differences by suggesting distinct categorization patterns at the onset of the learning process, which lead to the different developmental paths attested in the experiments.