Being understood, not sounding accent-free

In 2007, Ulla Hirschfeld and Jürgen Trouvain looked into where the problem lies. When the sentence Es regnet was spoken with the stress on its last syllable, more than half of the German listeners did not recognise it. With the right stress, 5 per cent missed it. Not a single sound was wrong, only the stress was. The recurring problems the two of them see in learners are all of this kind. They are word stress on the wrong syllable, short stressed vowels that come out too long, too little difference between stressed and unstressed syllables, the missing weak e in endings such as -en, -el and -em, too many stresses in a sentence, and pauses in the wrong place. It is not a single sound that decides, but the rhythm.

For you, this means learning where the stress falls in every new word, just as you learn its gender along with it. Pay attention to the length of vowels, because in German it tells words apart, as in bieten and bitten, Staat and Stadt, Miete and Mitte. And let the final syllable drop. Lesen has a long, clear e at the front and almost none at the end.

Hirschfeld and Trouvain note that hardly any adult learner repeats a sentence well without first having heard it consciously. They recommend that learners first listen and tell the sounds apart, then repeat, correct, listen again, and repeat once more with feedback. Yo Hamada, who has studied repeating in time with the original for years, recommended the same to beginners in 2019, first as listening practice and then as speaking practice. That is exactly why, in WieGesagt, the microphone only appears once the sentence has finished playing.

In 2015, Junkyu Lee et al. calculated across 86 studies that pronunciation teaching has a large effect, and a larger one when it lasts longer and when there is feedback. In 2024, Thuy Ngo et al. followed this up for practice with speech recognition. Explicit feedback, such as a score for each word or a comparison with a speaker, worked much better than a mere transcript or a simple right or wrong. One to four weeks of practice achieved almost nothing, and five to eight weeks achieved a great deal. And practising alone achieved less than practising with a teacher, because nobody says what exactly was wrong. In 2008, Ambra Neri and colleagues showed with thirty adults learning Dutch how narrow the effect is. Those who received automatic feedback on particular difficult sounds improved exactly those sounds and nothing else. Wrong automatic feedback, they warn, can also spoil learning. Speech recognition can be wrong, more often with beginners than with advanced learners. That is why WieGesagt compares not only the words but also the stress, rhythm and speed with the original, and you can repeat the sentence as often as you like.

In 2017, Jennifer Foote and Kim McDonough had sixteen adults repeat short dialogues on their mobile phones for eight weeks, at least four times a week for at least ten minutes. Afterwards, their free speech was rated as easier to understand and more fluent. Their accent had not changed. In 2011, Nel de Jong and Charles Perfetti found that fluency only lasts if you speak the same content repeatedly instead of new content each time.

In 1986, Elaine Horwitz and colleagues described what anxiety about speaking a foreign language is. Adults who see themselves as clever and articulate find that image threatened as soon as every sentence is judged by standards they have not mastered. In the study by Muzakki Bashori and colleagues in 2022, teenagers reported that they felt less anxious with speech recognition than with other people. It is not a solution, but it lets you start alone, with feedback, as often as you like, until you feel confident enough to speak with other people.