One of the distinctive features of the Chinese language is the use of what are called chengyu (成語), fixed idiomatic expressions which consist of four characters and are usually telegraphic to the point of being ungrammatical — for example 孟母三遷 “Mencius mother three moves,” which expresses the importance of finding a good environment in which to raise children (because the philosopher’s mother moved three times in order to find such a place), but which is hardly grammatical Chinese. There are thousands of such expressions, and they are very common. Japanese also has these, having borrowed them from the Chinese. English idioms, on the other hand, are almost always fully grammatical, often making them much longer than their Chinese counterparts. For instance, where English has “to kill two birds with one stone,” a grammatical verb phrase which can only be used if you conjugate it and put it in a sentence, the Chinese equivalent is the telegraphic 一石二鳥 — “one stone, two bird.” (I’ve translated it as “bird” instead of “birds” to give some idea of just how ungrammatical it is in Chinese. In any other context, “two birds” would have to be 兩隻鳥. The character 二 is used only for counting; with a noun, you have to use 兩 plus the appropriate classifier.) My Chinese-speaking students are often surprised at the wordiness of English idioms, and I have to explain that we don’t really have anything like chengyu in English — or do we? There’s “long time no see,” but that’s a bit of a special case, being a deliberate imitation of American Indians’ broken English or perhaps even a calque of the Chinese chengyu 好久不見. Once I started thinking about it, though, I came up with several other expressions that could be considered English chengyu:
- Like father, like son.
- Another day, another dollar.
- Monkey see, monkey do.
- First come, first serve(d).
- Same shit, different day.
- Two Jews, three opinions.
- Garbage in, garbage out.
This kind of thing obviously isn’t as common as in Chinese, but the list is long enough to show a pattern. What intrigues me is the fact that, just as in Chinese, ungrammatical idioms tend to be four words long. It seems unlikely that this would be due to the direct influence of Chinese, so I wonder if it’s a pattern that comes naturally to people and turns up in many different languages.
Boisterous beholding 
I certainly don’t know. Good question for a linguist. Still, I do notice that all the four-word sayings you cited employ only three patterns: (1) substantive+verb, substantive+verb; or (2) modifier+substantive, modifier+substantive; or (3) verb+preposition, verb+preposition. In (2) the verb is implied; in (3) the subject is. Would you say that makes them still more grammatical than the Chinese chengyu?
Here’s one in French: comme ci, comme ca (like this, like that).
Here’s one in Greek: ou polla, alla polu (not many, but much).
Maybe it’s an Indo-European thing?
Actually four words is not the lower limit in IE languages: Here’s a two-worder from the dialogues of Plato: gnothi sauton (know thyself). Or, meden agan (nothing to excess).
Thanks for the additional examples, Consultus. Although I studied linguistics, my focus was on theory and didn’t involve actually learning many different languages, so I’m not specially qualified to answer the question I posed.
Your two-word examples don’t really qualify because they are grammatical. “Know thyself” is a complete sentence, and “nothing to excess” is a normal noun phrase. I’m looking for fixed expressions which break with the ordinary syntax of the language. That’s why I didn’t include phrases like “The more, the merrier” in my English list; it’s a four-word idiom, but not in the sense I have in mind.
A Latin example might be “fiat justitia ruat caelum.” It seems like this isn’t really a grammatical sentence, that a conjunction or something has been elided, but I don’t really know Latin, so I can’t say for sure. Judging from you handle, I assume you do have some familiarity with Latin. What do you think?
Of course you’re right about gnothi sauton, which clearly is a complete sentence.
My little Latin is very rusty, but I’ll see what I can do. The translation of “fiat justitia ruat caelum” ought to be close to, “Let there be justice, though the sky fall.” The inflected nature of Latin contributes to the appearance that the four Latin words are not a sentence. In fact, the subjunctive verbs carry the freight, so you really do have a grammatically complete sentence.
Here’s another try: “O tempora! O mores!”
My mother has two more to add to the list: “been there, done that” and “same time, same place.”
Also — how did I not think of this one before? — “so far, so good.”
Consultus, I think that pair of vocatives still probably belongs to the domain of ordinary grammar (cf. “Ah Bartleby! Ah humanity!”). How about “Et in Arcadia ego”? Is that ungrammatical? So many things can be omitted in Latin, one can never really be sure.
Read it, “Et in Arcadia ego sum. The verb “to be” is customarily implied.
Now I think of it, some of the other examples require only a single verb to become grammatically complete. For instance, “garbage in [yields] garbage out.” Or, “first come [is] first served.”
Sometimes the subject is implied: “[I have] been there, [I have] done that.”
On the other hand, it takes more than an implied subject or verb to make a complete diagram of “like father, like son.” So there’s another distinction.
I guess the real question isn’t actually whether the structure of a given expression is “grammatical” in the sense of being diagrammable, but whether it is linguistically productive — that is, whether and to what extent the same syntactic template can be used to create novel expression.
For example, “The more, the merrier” may seem ungrammatical, but form “the [comparative], the [comparative]” to express covariance is very productive. Anyone can coin a new expression of this form — say, “the wider, the more expensive” — without it seeming grammatically odd. The same goes for “no pain, no gain,” since “no [noun], no [noun]” is almost endlessly productive.
The structure of “long time, no see,” on the other hand, is not productive. You can replace “see” with another word (“long time, no blog,” etc.), but this will always be perceived as a joking allusion to the original chengyu, not a normal expression in its own right. If the form were truly productive, it would be possible to say something like “six hours, no eat” to mean “I haven’t eaten for six hours.”
Do you know if “et in Arcadia ego” represents a productive structure in Latin? I’m guessing it does, since “Num et Saul inter prophetas?” seems to be the same sort of thing, with an implied copula.