Even if you know all the words in the sentence, it doesn’t mean you know what the sentence means. Moreover, even if a grammatically correct translation is attainable, doesn’t mean that it carries the same discoursal value. Such is the curse of polylingualism (by which I mean, “the phenomena of having more than one language in existence”).

So in English, we have this fascinating effect of doubling negatives in order to soften the intent. Wonderfully documented in Laurence R Horn’s The Expression of Negation, Horn notes that the two negatives do not completely cancel each other, and instead end up producing a weaker positive than the original positive statement. Consider:

1. Mary’s happy. Or at least, not unhappy.

2. Bart: Dad, are you licking toads?
Homer: I’m not NOT licking toads…

3. A: I thought you were friends with her?
B: Well, I’m not NOT friends with her…

In (1), the second sentence softens the claim from the prior sentence. The speaker has clearly hesitated to bear the burden of proof of Mary’s happiness, but is more comfortable being able to attest that at the very least, Mary is not depressed. If the depression-happiness scale were from 0 to 10, with 0 being depressed, and 10 being happy, then “Mary is happy” connotes 8~10, while “not unhappy” connotes a 5~7. Similarly, in (2), Homer is not attesting that he’s licking toads (100% certainty), but that he denies having avoided licking toads (60~80% certainty). And in (3), the inference of “not NOT friends” could mean anything from “strangers” to “close acquaintances”, but definitely not “rivals” or “enemies”.

It should also be noted for the grammatically queasy, that (2) and (3) can be rephased from “I’m not NOT …” to “it’s not {that/like} I’m not ….”, for similar effect of softening the original statement.

In effect, we could rank statements from best to worst as follows: good > not bad > not good > bad. We might say “good” is 10; “not bad” is 7; “not good” is 3; and “bad” is 0. I’d go so far as to say that this is a pretty stable ranking among English speakers.

On the other hand, Japanese has a more rigid way of handling negation. For one thing, even though the grammar technically allows for two nots (e.g. ?よくなくない), they simply do not employ it as English speakers would in the above 3 examples. Instead, they can only opt for a periphrastic construction, similar to the grammatical gloss “it’s not that/like I’m not…” (…ない { わけじゃ/ことは} ない [ けど ] ).

The ~なくない is a construction that can only be used in interrogatives that are naturally formed with negated verbs. In English, we have the “is that not ___ ?” formula, which can take a negated verb phrase: “is that not not good?” Similarly, Japanese can legitimately only use the direct double negative in questions like that: 「それはよくなくない?」 “Is that not [a] not good [thing]?”

In English, this soft negation is consistent even with relatively complex ideas like “half”. If “half bad” were 4~6, then “not half bad” would be around a 8-9, but also potentially to be a 1-2 (in the twisted interpretation that “not half-bad” implies “completely bad”). “not entirely fractional” might imply integers, or possibly fractions that are very close to integers (say, 1.95, but not 1.45)

Now, as was tacitly made clear, English has the phrase “not half bad”, and uses it to mean “not terrible; pretty good”. However, a deceptively similar Japanese phrase (半端じゃない, lit. “not half-way”; “not half-arsed”) doesn’t seem to follow the “rule”, and in fact breaks with the intuitive double negatives. According to my English sensibilities, that would place the expression as somewhere between “thoroughly complete” and “completely half-way about everything”.

Instead, 半端じゃない “not half-arsed” basically means “super complete”, even though in my head, it basically means “semi-half-arsed — 3/4ths proficiency”. So With our original scale of 半端 being 4~6, then 半端じゃない would be like, 11+ (clearly breaking the scale).

But, to add another spanner in the works, this errant Japanese phrase is equally confusing to more mature Japanese speakers; from what I’ve gleaned from my online sleuthing, 半端じゃない is essentially a kind of younger slang that’s used by people born in/after 1989 — so-called “heisei kids”. Apparently, even for some “showa kids” (ie, those born before 1989), the phraseology of 半端じゃない can be confusing.

In fact, this last fact is gratifyingly reassuring, as it essentially confirms for me that like in English, Japanese negation also has a softening quality, which is why the outlier slang 半端じゃない is confusing for older speakers. But like any given slang or phrase, sometimes the connection between the phrase itself and its current meaning are extremely tenuous, and require historical context in order to learn (or indeed acquire) the phrase to any satisfaction.