how do you stop translating everything in your head?
by meeting the same words and phrases often enough, with their meaning attached, that translating stops being the quickest way to reach what they mean. you cannot switch it off by deciding to. research on bilingual memory suggests that early learners lean on their first language and on what words look and sound like, while more practised speakers reach meaning more directly. even then the first language does not go quiet. what fades is the translating you can feel, the pause in which you hear the sentence, find the english, and only then understand.
why the translation comes first
when you learn a new word, you usually learn it as a partner to one you already have. that is how a word list works. in 1994 the psychologists kroll and stewart built a model of bilingual memory around that fact. in it, a new language's words begin tied more strongly to their translations than to meanings of their own, and the direct link to meaning grows with skill. parts of the model are still argued over, such as whether practised bilinguals pass through meaning when they translate back into their first language. the starting point it describes is less disputed.
a study by adrienne talamas, judith kroll and robert dufour in 1999 shows what that starting point looks like. english speakers with more and less spanish saw an english word followed by a spanish one, and had to say whether the spanish was its translation. the trick pairs were the interesting part. some were words that looked like the right translation but meant something else: spanish has a word for hunger that sits one vowel away from its word for man. others meant something close, such as the word for woman shown after man. the less practised learners were caught out more by the lookalikes. the more practised were caught out more by the near meanings, and the authors read the difference as a progression from form to meaning as a language grows.
the first language stays awake
it would be easy to take from that a promise that translation ends. a 2007 study by guillaume thierry and yan jing wu suggests it does not, at least not entirely.
they worked with chinese speakers living in britain, people who had started english at twelve, had been in the uk for about a year and a half, and used english every day. the task was in english only: decide whether two english words were related in meaning. what the participants did not know was that some pairs hid a repeated chinese character once translated. their answers showed no sign of it. their brain responses did. the same pattern turned up when the words were heard rather than read. the authors conclude that native-language activation goes on, unconsciously, while a second language is understood.
this changes what to aim for. the target is not a mind with no first language running in it, which the study gives no reason to expect. it is understanding that arrives before you have time to notice the translation, so that you no longer experience the step at all.
what makes it fade
nobody translates "gracias". english speakers who know almost no spanish hear it and understand it without passing through "thank you" first. that is not because they studied it. they met it so many times, in so many films, songs, holidays and restaurants, always with its meaning obvious from the situation, that the detour stopped being worth taking.
that is the pattern to copy, and three things seem to help it along.
meeting the same phrase many times, not just many phrases once. a phrase you have heard fifty times can be recognised whole. a phrase you have heard once has to be worked out.
hearing phrases whole rather than word by word. a short, common sentence is easier to take in as one unit than five separate words, each needing its own translation.
having the meaning there each time, without having to produce it. a translation beside the line, a situation that makes it clear, a scene you know. at first the english does the work. later it arrives after you already know what the line meant, and becomes a confirmation rather than a step.
none of this has a fixed timetable, and the research above does not offer one. it measures groups of more and less practised speakers, not how long the change takes for any one listener.
catching the moment
because the change is gradual, it is easy to miss. a simple way to notice it is to keep a small set of lines in front of you for a while.
choose ten or so short phrases that you will hear often: greetings, replies, things people say every day. listen to them in the same order for a week or two, with their meanings beside them. each time, pay attention to one thing only. does the meaning arrive before the english does, or after? at first it will be after. if the listening is working, there will be a day when, for one or two of the phrases, it comes before. those are the ones you have stopped translating, and you can count them as they come.
it is also worth noticing which phrases go first. they are usually the shortest and the most often heard, which is one more reason to start with the plain ones rather than the clever ones.
the plain ones first
driftspeak is subtitles for your ears: a line at a time, in its language and then in yours. the spanish ladder is one path through spanish, from first words to deeper expression, in latin american spanish. its first track, survival and social glue, is made of exactly the kind of phrases that fade first: "¿qué tal?" ("how's it going?"), "más o menos" ("so-so"), and "gracias" itself.
listen to a sample of the spanish ladder, whose first three tracks have free samples, and notice whether "hola." ("hi / hello.") still waits for the english that follows it.