If you look at a spectral representation of speech, it’s impossible to see where one word ends and another starts. It takes hundreds of hours to train the brain to pick out words from this audio mush. Obviously a news broadcast by a national broadcaster will have clearly enunciated words, but in casual speech words are slurred together. I’ve spent close to 1,000 hours listening to French, and I still struggle with French films.
It’s common in Britain for school students to understand written French, but not spoken language.