You often hear, either as humble-bragging or as an apology, that Japanese HS graduates may not have much speaking or listening skill, but they are good at reading and writing.
This is false. The more advanced students are good at applying a conventionalized system of translation to English, and reading is the only one of the 4 skills that gives them enough time to do this. The less advanced students are left in a total fog because they lack the skill or focus to apply this very concentration-heavy set of techniques to English text and lack the practice to understand written English on its own terms.
And as for writing, well, I don’t think anyone who has worked with Japanese undergrads believes that part to begin with.
You may be wondering how application of translation skill differs from just reading. They may look very similar indeed when practiced at a high level. When practiced at a middling level though, translation produces different errors than reading as we usually understand it. For example, consider a phrase with a very context-dependent meaning such as “Could you come with me?” To high-level English users, that phrase doesn’t appear very context-dependent at all – one struggles to try to see it as anything other than a request. However (and ELT veterans in Japan already see where I’m going with this), a graduate of a Japanese high school is likely to see it as an inquiry as to whether the listener was able to come with the speaker previously, because “could” is conventionally translated as できた dekita, the past tense of できる dekiru – the usual translation of “can”, which is overwhelmingly more often taught as a reference to ability (e.g., “can you juggle?”) than requests. Hence a student equipped with some but not all of the tools for accurate translation will reach for the tool that they have instead of looking at context or other clues for what this utterance could mean. This is the fate most high school graduates, who by definition are of average skill, are consigned to.
Gaius Octavian Caesar: At best I’ll be a middling swordsman.
Titus Pullo: It’s better than nothing.
Gaius Octavian Caesar: There you are wrong. The graveyards are full of middling swordsmen. Best not to be a swordsman at all than a middling swordsman.
Translation, unlike reading, is inevitably a very processing-heavy ability. When learning to read, particularly in languages like English or Japanese where spelling relates to pronunciation on a case-by-case basis, teachers rely on extensive practice to automatize skills like word recognition, prediction and collocation. Making as many skills as possible automatic is crucial for reading if you want readers to focus on the message of what they’re looking at, not laboring over each word like a preschooler. Translation, however, demands that words or short chunks (e.g. “at home”) be individually translated and the entire sentence held in memory in order to be recoded in Japanese before it is understood. It is about as intuitive and conducive to understanding as having to do one long division problem every new line of a book. Luckily, students can give themselves the time to do this additional labor when reading, which is obviously not possible in listening or speaking where the pace is set by someone else. The requirements of mental calculation from translation can only be met when reading, which makes it look as if reading is the one skill that Japanese students learn well; when actually the only skill they’ve learned is translation, and reading is the only forum in which they can apply it.
So if you run an extensive reading program at your university or other institution, that program will likely be the first time your students have been called upon to practice actual reading. Don’t be fooled by protestations that reading is something they already know how to do – in actuality they’ve never done it before.