Linguistics: The pronunciation paradox

One of the most difficult aspects of learning a foreign language has to do with pronunciation. Learners are typically prone to specific sets of errors, which differ depending on the learners first language. For instance, Germans typically have trouble articulating the initial 'th' in

English, as evidenced by the classical expression 'Senk ju vor träwelling' familiar to passengers on German railways. Conversely, native speakers of English tend to have difficulty with the German 'ü', which they tend to pronounce as 'u'. Many people laugh at these mistakes in pronunciation, even though they make the same mistakes themselves. But this reaction in itself points to a paradox: It demonstrates that learners register errors when made by others. Nevertheless, the majority of language learners finds it virtually impossible to eliminate these typical errors even after years of practice. A study carried out by LMU linguists Eva Reinisch and Nikola Eger, in collaboration with Holger Mitterer from the University of Malta, has now uncovered one reason for this paradox. "Learners have a tendency to overestimate the quality of their own pronunciation," says Reinisch. "As a rule, they believe that their English is better than that spoken by their fellow students at language schools, although they make the same set of errors." This exaggerated assessment of one's own ability is an important factor that helps explaining why it is so difficult to learn the sounds of a foreign language.

In the study, the researchers asked 24 female German learners of English to read out 60 short sentences, such as "The family bought a house," "The jug is on the shelf," and "They heard a funny noise." Several weeks later, the same learners were invited back to the lab and asked to listen to recordings of four learners -- three others and themselves. Specifically, they were asked to grade the pronunciation of each sentence. In order to ensure that participants would not recognize their own productions, the recordings were manipulated in such a way that the female speakers sounded like male speakers. "This element of the experimental design is crucial. It was essential that none of listeners would be aware that their own productions were included in the test sample; otherwise their assessments couldn't be taken as unbiased," says Holger Mitterer. The results of this test were unambiguous. In all cases, the listeners rated their own pronunciation as better than others did, even though they were unable to recognize that it was their own recording. "We were surprised that the experiment so clearly pointed to the significance of overestimation of one's own abilities in this context," says Reinisch.

There are several possible explanations for these findings. Previous research has shown that familiar accents are easier to understand than accents that are less familiar. "One is best acquainted with the sound of one's own voice, and has no difficulty understanding it," says Reinisch, who is at LMU's Institute of Phonetics and Language Processing. "Perhaps this familiarity leads us to regard our pronunciation as being better than it actually is." Another possible contributory factor is what is known as the 'mere exposure' effect. This term refers to the fact that we tend to rate things with which we are more familiar -- such as the sound of our own voice -- as more congenial.

The results of the study underline the importance of external feedback in language courses, because it increases the learners; awareness of deficits in language production and comprehension. "As long as we believe that we are already pretty good, we are not going to put in more effort to improve," Reinisch points out. A lack of feedback increases the risk of what researchers refer to as 'fossilization'. Leaners feel that they have already mastered the unfamiliar articulation patterns in the new language, although that is in fact not the case. They therefore see no reason why they should invest more time in improving their pronunciation. The authors of the new study are not likely to fall into this sort of error. They are already considering ways to improve the situation with the aid of apps that generate the necessary external feedback -- irrespective of how users rate their own performance.