Counting syllables is the fastest way to see what has happened. Je ne sais pas is four syllables on the page and two in the air. Il y a is three and becomes one. Qu'est-ce que c'est is four and becomes two. Nothing has been dropped carelessly; the reductions are regular, there are about fifteen of them that matter, and every one can be learned deliberately in a week.
So the way into French listening is not more hours. It is to learn the reduced forms as forms, the way you learned the words, and to stop waiting for the written shapes to arrive.
What the sentences actually turn into
Six reductions of ordinary speech. None is slang and none appears in a course book, and between them they cover a large share of what goes missing in a real conversation.
- je ne sais pas becomes chais pas, I do not know. The ne is gone, which is true of nearly all spoken negation, and the rest collapses.
- il y a becomes ya, there is. Three words, one syllable.
- tu as becomes t'as, you have, and tu es becomes t'es.
- je suis becomes chui, I am, which is the form you will hear far more often than the written one.
- qu'est-ce que becomes kess, opening most spoken questions.
- il faut becomes ifo, it is necessary, with the l silent.
Where the boundaries actually are
French gives a listener no word-level stress to use as a landmark, so there is nothing inside a sentence marking where one word stops. What there is instead is a beat at the end of each group of words, and that is the only reliable signal in the stream. A sentence arrives as two or three of these groups, each ending in a slightly longer syllable, and the words inside a group are run together by elision and liaison until the joins are somewhere other than the spaces on the page.
This changes what to practise. Trying to catch words first fails, because the words have no edges. Catching the group boundaries first works, because they are audible, and each group is short enough to hold in memory while it is decoded. Listen to a sentence and count the groups before trying to identify anything in them: two, or three, or four. That sounds like a strange exercise and it is the one that moves this skill, because it replaces a search with no landmarks with a search inside a two second window.
A year of the news and still lost at a table
Here is the stall, and good material causes it. The graded news bulletins are excellent, they are free, they come with transcripts, and they are read from a script by one trained voice in a formal register. That register uses nous, keeps the ne of the negation, and asks questions by inversion. It is, in other words, spoken written French.
Twelve months of it produces somebody who follows the news comfortably and understands almost nothing at a dinner table, and the diagnosis usually offered is speed. It is not speed. It is that the grammar at the table is a different grammar, the one described in formal and informal address, and getting it right, and no amount of the first will teach the second. The exit is to run unscripted conversation alongside the bulletins from the start, in small doses, and to accept understanding a fraction of it: ten minutes of two friends talking is worth an hour of a presenter.
Thirty seconds, four times
The order that makes a transcript useful
Reading first is the version almost everybody does, and it trains reading. The steps below cannot be reordered without losing the point of them.
- 01Listen cold, and failThe failure is the measurement. Do not shorten it
- 02Listen for the group boundariesCount them. Do not try to identify a single word
- 03Read the transcriptFind the two or three places where it disagrees with what you heard
- 04Listen againThe same thirty seconds, now with the disagreements known
The third step is where the learning is, and it is only worth anything because of the first two. A transcript read before listening answers a question the ear never got to ask.
Two sources with nothing smoothed out
Arte's website streams a large amount of French television free, with French subtitles on much of it, and it is not made for learners. It replaced waiting for French productions to turn up on a service you already pay for, and the difference matters because dubbed material has a subtitle track translated separately from the dub, so the two do not match and the mismatch teaches you to distrust your own ears. On original French material the subtitle is a transcript.
InnerFrench is the other one and it is the bridge. A French speaker talking about ordinary subjects for half an hour at a controlled speed, free, with transcripts, and crucially using spoken French rather than read-aloud written French: on for the first person plural, the negation reduced, questions in the spoken order. It replaced the textbook dialogue, which is written French performed by actors, and it is the material that makes the jump from the bulletin to the table survivable.
The mistake: reading the transcript first
The transcript is the most useful object in this subject and the order matters more than the object. Read it before listening and the sentence is decoded on the page, the audio then confirms something you already have, and nothing about your ear has changed. It feels efficient and it trains reading.
The order that works is fixed. Listen cold and fail. Listen again for the group boundaries only, without trying to identify words. Then read, and find the two or three places where what you heard and what is written disagree, because those are the reductions and they are the entire lesson. Then listen once more, and the sentence will separate. It is four passes over the same thirty seconds and most people take one, which is why the same clip can be listened to for a year without teaching anything. The failed first listen is doing the same work that a failed retrieval does in a deck, and for the same reason, which is set out in why you forget, and the schedule that beats forgetting.
What is left after the reductions
Once the fifteen reductions are automatic, what remains is ordinary vocabulary and the fact that a familiar written word may still be unrecognisable in speech, which is the liability described in the thousands of words English already borrowed. The processes that erase the boundaries are the same ones you have to perform when speaking, so this page and nasal vowels, silent endings, and liaison are two directions of one mechanism, and progress in either shows up in the other. And if the material is right and the hours are not there, the hours have to be converted rather than found, which is switching your phone, your music and your commute.