WORDS and rules. That’s what language is, isn’t it? We have a mental lexicon, and we string words together with rules (grammar) to make sentences. So learning a foreign language involves stocking up on those dictionary-entries, and getting better and faster at applying the rules to them.
This is true, to a point. But another approach to foreign-language learning is duly earning converts. It involves learning short groups of words that often go together, but are shy of sentences. Unlike “lexicon”, “morphology” and “syntax”, this method has a pleasingly Anglo-Saxon name: chunking. (Chunks of language are also known as “collocations”, but “chunking” is more fun.) Learning these chunks of words together—and practising them frequently—so that they rush out instinctively, without having to be freshly constructed each time, is a key to achieving comfortable fluency when speaking a foreign tongue.
To take an example, if you said "close the light", as many languages do, in English, you'd probably be understood, but would sound like to a non-native: "turn off the light" is a chunk, and a common one that should flow from the tongue of a fluent speaker. Another example is the tricky problem of set prepositions, which are often idiosyncratic and must be learned as chunks: intime but bynoon or beforesix.
Other chunks are fully-formed pleasantries: “How are you?”, “Nice weather, isn’t it?”, “Won’t you sit?” and so forth. (They also have ritual replies: a German told me once how relieved she was to learn that the expected reply to “What’s up?” was “not much”.)
Chunks shouldn’t be confused with the full sentences prescribed in phrasebooks, a worry that Michael Swan, a language-pedagogy expert, expressed to Ben Zimmer in the New York Times: “formulaic expressions get more attention than they deserve, and other aspects of language—ordinary vocabulary, grammar, pronunciation and skills—get sidelined.” His concern is that the rush towards chunking in language pedagogy will encourage more people to neglect studying grammar. But his criticism of a "phrasebook" style of rote learning is misleading; chunks are largely not the stuff of travel books, “I’m sorry, haven’t you got anything cheaper?” and the rest. They are building blocks that can be shown to live in the brain in the same way as individual words. The best chunks (for language-learning) are phrases that native speakers cannot do without. Yet too many language classes never teach the likes of “I’m about to ___” or “He’s just finished _____ing”.
John McWhorter, a linguist at Columbia University and a keen language-learner (which is surprisingly atypical of academic linguists) has pieced together a list of hundreds of chunks over the years that he feels are needed for speaking any language: "all of a sudden", "I barely managed to" "He tends to" "oopsy-daisy!", "I couldn't even ____, much less ____", "that would come in handy", "who can blame him?" and many more. Learning a language without understanding these fundamental phrases is like creating a body with bones and muscle but no connective tissue.
Linguists know “chunks” are meaningful partly because they help computers translate. Old translation programs tried to break language down by rules, rebuilding the translated language according to a separate set of rules, with individual words slotted into their proper spots. This sounds sensible, but it was hopelessly difficult. New approaches (like Google Translate) are simpler: they scour parallel texts translated by humans, and work out how often X in English is translated as Y in French. They have gotten so much better not only because processors are better and data is bigger—but because “X” can be anything from a single word to a short series of them. The geeks call these n-grams (as in unigram, bigram, trigram, n-gram). Machine-translation engines also work out the probability that ABC in the source language will be XYZ in the target language. These are just a computer's version of chunks.
Some good books (Johnson is a fan of “Using French”, “Using Spanish” and so on, from Cambridge University Press) have a chunk-like approach. “Using French” calls them “complex verbal expressions”, but they are mostly quite simple chunks like veiller à ce que, “to see to it that”, or tenir à, “to be anxious [to do something]”. More books should do the same. The age of big data has given rise to a related useful language-learning tool, the frequency dictionary, in which 5,000 or so words are listed in order of their frequency in the language. It would be easy to add to future editions of such dictionaries the most common bigrams, trigrams and other n-grams. A good such book would please many learners, and so should make its publisher chunks of money.
