Table of Contents

Introduction

This is the second article in a series based on the eight-task model of language learning. Each article takes one part of the model and asks a practical question: which methods can a teacher use in class, and which can a learner use alone, to master that part?

The first article dealt with Task 1, knowing enough words. This one deals with Tasks 4 and 5 together:

Task 4, recognising words when they are pronounced by native speakers;
Task 5, pronouncing words so that native speakers recognise them.

Why treat these two tasks together

Tasks 4 and 5 are two closely connected sides of the same phonological learning. Both depend on the sound system of the language: the same inventory of sounds, the same relationship between sounds and spelling, the same rhythm and melody. Many of the most useful techniques train both at once, because listening and speaking are two directions through the same phonological knowledge.

They are also linked in the mind. Accurate perception generally supports accurate production, so perception tends to come before production and to help it. They remain, however, distinct: you can learn to tell two sounds apart before you can produce them well, and the reverse can happen too. What justifies treating them together is not that they are one skill, but that the operative techniques overlap so heavily.

Finally, keeping them together avoids two thin, overlapping articles. Tasks 4 and 5 are moderately difficult for an Italian learning German, not enormous, and there is more to say once about the two together than twice about each apart.

What Tasks 4 and 5 actually involve

Task 4 is not just hearing sounds, it is finding known words inside the stream of native speech. Three things make it hard. The language may use sounds that do not exist in your first language, and you literally do not hear distinctions your ear was never trained to make. The spelling may not map simply onto the sounds, so a word you know on the page arrives unrecognised through the ear. And in natural speech words are not pronounced one at a time: they run together, get shortened and blur at the edges, so a word you recognise perfectly in isolation vanishes at conversational speed. A learner who understands the teacher but not a film has mastered slow, clear speech and not Task 4.

Task 5 is producing words so that a native speaker recognises them. The bar here is intelligibility, not a native accent. As I said in the original article, for an Italian it is relatively easy to pronounce German well enough to be understood even if the pronunciation is far from perfect, and that is exactly the goal. Two things carry most of the intelligibility: the individual sounds, and, at least as much, the rhythm and stress. A word with the right sounds but the wrong stress is often harder to recognise than one with an approximate sound but the right stress.

Both tasks draw on the same knowledge that Task 1 depends on. This is why I insisted in the first article that a word should never be learned silently: learn it with its sound from the first day, and you are already training Tasks 4 and 5 while you build your vocabulary.

Principles that make listening and pronunciation improve

1. Perceive first, then produce, but train them together. You need to hear a contrast before you can dependably make it. In practice this means pairing every production activity with listening to a model.

2. Aim at intelligibility, not perfection. Chasing a native accent absorbs a great deal of effort for little communicative return. Being understood without the listener straining is the target. Effort spent beyond that point is better spent elsewhere.

3. Rhythm and stress matter as much as individual sounds. Word stress, sentence rhythm and intonation carry a large part of intelligibility. Work on the melody of phrases, not only on single sounds.

4. Train on real, connected speech, not only slow models. Understanding the classroom recording is not the same skill as understanding a native talking at normal speed. Bring natural speech in early, with support at first, and reduce the support over time.

5. Feedback is essential, because self-monitoring is unreliable. When you speak you tend to hear what you intended, not what you produced. Recording yourself helps, because it lets you listen without the load of producing at the same time, but you are still judging yourself with your own ear. Genuine external feedback comes from a teacher, a native speaker, or a speech recogniser, and only that can tell you whether you were actually recognisable.

6. Repetition and imitation, spaced. As with vocabulary, sounds settle through many spread-out encounters and attempts, not through one long session.

Techniques for the classroom

Model with plenty of native audio, not only your own voice. Your voice is one accent at one speed. Learners need many voices and speeds to build recognition that generalises beyond the teacher.

Drill the specific contrasts your learners cannot hear, with minimal pairs. Take pairs that differ by a single sound the learners’ first language does not distinguish, for German for example the vowels in Stadt and Staat, and first do perception (“did you hear the first or the second?”), then production. Work only on the contrasts that are actually problematic for this group, not on sounds they already have.

Use shadowing. Play a recording and have learners speak along almost simultaneously, a fraction behind it, copying the rhythm and intonation as much as the words. This near-simultaneous imitation trains perception and production at the same time and is one of the most efficient activities available. It is not the same as listen-and-repeat below, where the learner waits for the model to finish.

Do listen-and-repeat with real chunks, not nonsense. Here the learner hears a phrase, waits, then reproduces it: delayed imitation rather than the near-simultaneous imitation of shadowing. Repetition works better on meaningful expressions the learners will reuse than on isolated syllables.

Use dictation and its variants. Dictation forces fine perception, because the learner has to resolve exactly which words were said. Running dictation adds movement and repetition.

Work explicitly on stress and rhythm. Clap the rhythm of a sentence, hum its melody, and build long words and phrases from the end (back-chaining) so the stress falls into place. This is often neglected and pays off quickly.

Expose learners to varied speakers, speeds and accents. Recognition that only works on one clean voice is fragile. Variety is what makes Task 4 transfer to the real world.

Give feedback about intelligibility, not perfection. Tell learners when something would not be understood, and let the rest go. Correcting every deviation from a native model discourages without improving communication.

Techniques for self-directed learning

Learn every word with its sound. Use dictionary audio, a site like Forvo for native recordings, or text-to-speech as a fallback, and put the audio on your flashcards. Never let a word into your memory silently.

Shadow, with a transcript for support. In shadowing you speak along with the recording almost as it plays, a fraction behind, copying its melody and rhythm; a transcript can help at first but is not essential. This is close to, but not the same as, the reading-aloud activity in the original article, where you read a written phrase after hearing it. Both are worth doing. Choose material slightly above your level.

Listen to large amounts of real speech. Podcasts, videos, series and interviews at natural speed are what build Task 4. Start with subtitles or a transcript for support, then take the support away as you improve. The point is to train on connected, fast speech, not only on tidy course recordings.

Train your problem sounds directly. Minimal-pair listening exercises and pronunciation apps let you drill the specific contrasts your first language does not have.

Record yourself and compare. Say a phrase, play the model, play your recording, and adjust. This does not give you outside feedback, since you are still judging with your own ear, but it sharpens your self-monitoring by letting you listen without producing at the same time.

Use speech recognition as a rough check. As I suggested in the original article, if a speech-to-text tool repeatedly fails to understand a word you are trying to say, your pronunciation probably needs work. Treat this as a crude proxy only: these tools are trained on standard accents and make their own mistakes, so they are a rough guide, not a judge.

Give prosody its own attention. Imitate the rise and fall and the rhythm of whole phrases deliberately, rather than assembling sentences one correct sound at a time.

The particular case of German

For an Italian, German sound is largely accessible, which is why I rated these tasks 4 and 5 rather than higher. One real advantage helps both tasks: the relationship between spelling and sound in German is fairly regular, far more so than in English, so a word learned on the page usually behaves as expected through the ear and the mouth.

The trouble concentrates in a short, predictable set of features, which is exactly where a learner should aim their effort:

the rounded front vowels ü and ö, which do not exist in Italian;
the two “ch” sounds, the ach-Laut in Bach and the ich-Laut in ich;
the German r, which has several accepted realisations, including a back, uvular one unlike the Italian r;
vowel length, which distinguishes words such as Stadt and Staat;
final devoicing, so that Tag ends in a sound closer to “k” than “g”;
word stress, especially in compounds and separable verbs.

Because the set is small and known in advance, German pronunciation for an Italian is a targeted job, not a vague one.

How to tell you are getting there

For Task 4, keep the test narrow, so that it measures aural recognition and not the other tasks. Following a whole conversation also depends on Task 1 and on understanding combinations of words, so it is not a clean test. The clean test is this: when a native speaker says, at natural speed, words you already know from the page or in isolation, do you recognise them? If you do, Task 4 is progressing, whatever else you may still be missing.

For Task 5, the test is not whether you sound native. It is whether native speakers understand you without unusual effort and without having to ask you to repeat. Note that a native switching to English is not a reliable sign of failure, since many will switch at the first foreign accent even when they understand you perfectly. When you are understood without strain, Task 5 is done well enough, and any further polishing is optional.

As with Task 1, these tasks should not be trained in isolation. Learn each new word with its sound, and Tasks 1, 4 and 5 advance together. The remaining tasks, understanding and producing combinations of words, are taken up in the other articles of this series.

How to hear and pronounce the words of a foreign language: techniques for the classroom and for self-study (Tasks 4 and 5 of 8)