Noncompositional

We all know Google can be the syntactician’s (or semanticist, or pragmaticist) best friend when it comes to carrying out powerful searches of various constructions and verifying or complicating theories and predictions about grammaticality. But there is one situation when I absolutely cannot use Google, and instead prefer Altavista.

That situation comes with search strings containing multiple wild card slots: that is, more than one *. On numerous occasions, I’ve found that having multiple of these guys in a Google search results in basically useless results — or results so polluted with non-hits that it’s tedious to find the good ones. To give one (slightly contrived) example, to prepare for a presentation on funky-a constructions recently, me and my colleagues wanted to get some examples from the internet.

Let’s say we were interested in what could fill the [brackted] slots of

an [adj] one thousand [noun] came

If you try this with Google, you get some good results, but also some…well, not so useful stuff:

an [annual income of] one thousand [gulden; and from Holland there] came

an [order for] one thousand [shekels. Bezalel] came

And so on. That is, perhaps acceptably, the * stands for strings longer than one word. Still annoying sometimes.

In other cases, you get something far less acceptable, both for a syntactician and (I dare say) for any searcher using multiple *s: that is, two words without a * between them become separated. At the moment the best I can do to replicate that result is to search for “tired people * and * with.” You’ll find “tired of people…” Perhaps this issue has been mostly dealt with. (or maybe this is useful for someone?)

In any case, where Google does unexpected things, I turn to Altavista. A * is one word, and I have yet to find an exception.