You read a sentence, and before the meaning lands, there's an extra step: the sentence quietly becomes a native-language sentence in your head, and only then do you understand it. That extra step feels invisible most of the time, but it's the specific thing making your reading slower and more effortful than it needs to be, and it's fixable, though not in the way "just stop doing it" implies.
Why mental translation happens in the first place
Mental translation isn't a bad habit you picked up by accident. It's the natural strategy your brain reaches for when a word or a structure isn't yet strongly linked to its meaning directly, only linked to it through your native language as an intermediate step. Early on, this is genuinely useful. Looking up a word and mentally attaching its native-language equivalent is often the fastest way to learn what it means in the first place.
The problem is that this intermediate step, useful at the beginning, becomes a habit that outlives its usefulness. Once you've met a word enough times that you could recognize its meaning directly, the mental-translation habit keeps routing through the native-language step anyway, purely because that's the pathway you built first and reinforced through repetition. The extra step isn't necessary anymore, but nothing has told your brain to stop taking it.
Why this is a speed problem more than a comprehension problem
Mental translation usually still gets you to the correct meaning, eventually. That's part of why it's a hard habit to notice and break: it doesn't feel wrong, it feels like understanding, just slower. The real cost is processing time. Each extra translation step adds real cognitive overhead to every single word or phrase, which compounds across a whole sentence, and across a whole article, into reading that takes far longer than it needs to and feels far more tiring than direct comprehension would.
This is also why mental translation tends to break down specifically on longer or more complex sentences. A short phrase is easy enough to translate on the fly. A long sentence with multiple clauses, translated piece by piece and then reassembled into a native-language version in your head, is genuinely difficult to hold together, which is a large part of why complex real sentences feel disproportionately harder than their individual vocabulary would suggest.
The actual fix: building direct association, not willpower
You can't simply decide to stop mentally translating through effort of will in the moment, because the habit operates faster than conscious control in most cases by the time you notice you're doing it. What actually works is building a direct link between the word or phrase and its meaning, strong enough that the native-language step becomes unnecessary rather than forbidden.
Extensive reading at a slightly easier level than your ceiling. Reading a large volume of text where you already know almost all the vocabulary gives your brain repeated, low-effort opportunities to encounter words being directly understood rather than translated, which is exactly the repetition that builds a direct association. Reading constantly at the edge of your ability, by contrast, keeps forcing you back into effortful, translation-based processing for nearly every sentence, because too much is genuinely unfamiliar to process any other way.
Reading the same short passage more than once. The first pass through an unfamiliar passage often does require some translation-based work. A second or third pass through the same material, once the vocabulary and structure are no longer new, is a chance to practice direct comprehension on content you already understand, reinforcing the direct pathway rather than the translated one.
Noticing when you catch yourself doing it, without judgment. You won't eliminate mental translation by noticing it in the moment, since by the time you notice, the sentence is usually already processed. But tracking roughly which kinds of sentences trigger it, long ones, ones with unfamiliar connectors, ones on unfamiliar topics, tells you where to focus deliberate, easier-level reading practice, which is a more useful signal than trying to catch and stop yourself mid-sentence.
Reading aloud at a natural pace, without stopping to translate. This forces you to keep moving forward through a sentence rather than pausing to work out a native-language equivalent, since stopping to translate breaks the pace you're trying to maintain. It's uncomfortable at first precisely because it exposes how much translation-based processing you're actually doing, but that discomfort is the useful part: it's showing you exactly where the direct pathway isn't built yet.
A concrete way to tell if you're still doing it
A useful, low-effort test: pick a sentence you just read and ask whether you can recall its meaning without being able to recall the exact native-language sentence you translated it into. If the meaning only comes back to you attached to a specific translated wording, that's a sign the translation step is still doing real work, not just running invisibly in the background. If you can recall roughly what the sentence said without a translated version attached to it at all, that's direct comprehension, the thing you're actually building toward. Running this check occasionally, not on every sentence, gives you an honest read on how far the shift has actually progressed, since the habit is otherwise hard to observe from the inside.
Where this connects to reading speed and vocabulary work
The mental-translation habit and reading speed are closely linked, because the extra translation step is a large part of why reading in a new language takes longer than the equivalent native-language reading, beyond what raw vocabulary size alone would explain. As the direct pathway strengthens through the kind of repetition described above, reading speed tends to improve as a side effect, not because you're trying to read faster directly, but because fewer sentences require the extra translation step that was slowing you down.
The same repetition also strengthens the other side of this: building vocabulary through context works specifically because context-based learning skips the translation step from the start, linking a new word directly to a situation and a meaning rather than to a native-language equivalent. Vocabulary learned this way tends to need less deliberate untranslating later, since it was never routed through the native-language pathway to begin with. This is one real argument for building new vocabulary through reading itself, rather than through bilingual flashcards, once you're past the earliest beginner stage where an explicit translation is still the fastest way to learn a brand-new word.
Why the goal isn't zero mental translation
Even highly proficient bilingual readers sometimes translate mentally, particularly for genuinely unfamiliar vocabulary or unusually complex structures, and that's a reasonable, temporary strategy in those specific cases, not a failure. The realistic goal isn't eliminating the habit entirely. It's shrinking how much of your reading depends on it, so that familiar vocabulary and common sentence patterns get processed directly, and mental translation becomes the exception you reach for on genuinely new material, rather than the default step every single sentence has to pass through first.