Here’s the picture I try to paint for them, using Google Maps images as examples:
Grammar rules are like maps–and landscapes are like languages–in several important ways.
First of all, maps are not the landscape itself; they are a representation of that landscape. Likewise, grammar itself is not the language, but rather a way to describe language–to lay a grid over it so that we can examine and discuss its component parts. When we study a map of Japan, we are aware that we are studying Japan; however, we are equally aware that we are studying just a few aspects of Japan–its shape or geographic features. We know that there’s a lot about Japan that isn’t as easy to explain in map form, and we certainly don’t expect to be able to get around Japan without trouble just because we have a map in hand.
The second important point is that both landscapes and languages are massive, shifting, complex, organic things, with no distinct beginning or end, and few clear meaningful boundaries. In other words, it’s just as difficult to say where English begins and Spanish ends, or how large each one is, as it is to say where exactly the Indian Ocean meets the Atlantic or the Pacific, or to measure the surface area of any of them. So what we do is we observe as well as we can, for as long and with as much care as we can–and then we make what are often arbitrary boundaries which are, nevertheless, meaningful to us at the time. The problem is that we sometimes forget just how arbitrary some of these boundaries can be, and start thinking of them as natural, inherent qualities of the world. Regarding landscapes, this can lead to outdated ideas such as that all people of Type X belong in Country X. In linguistics, it can lead to a desire to prescribe stylistic conventions as rules, or to overgeneralize from one aspect of the language (i.e.: “English shares many characteristics with Latin”) to another (i.e: “therefore one shouldn’t split infinitives”).
But the third, and most crucial, point of analogy is that one can “zoom in” on both maps and on grammar–and the more one zooms in, the less reliable the description can become. So for example, from space, the coastline of Japan is pretty easy to describe. It doesn’t change noticeably over time, at least not the timescales we are concerned with. It’s the same with general grammar descriptions: one could examine a large corpus and say, with confidence, that subjects precede predicates in English X percent of the time. However, the closer one gets, the more detailed, the less one can be confident that the representation is accurate in all cases or over time. “Zooming in” on a particular writer doesn’t tell us very much–it’s not very useful as a map of the language–since idiosyncrasies of style vary greatly between writers.
The more we zoom in, the more discrete features we can see, but the less reliable their descriptions can become, and the more exceptions become apparent.
I am currently sitting in my office, which should be situated in the empty rectangular field at the center of this Google Map satellite photo. It’s not, of course, because the building I’m in is relatively new. Maps at this scale are often full of such mistakes, since they are relatively detailed, and features at this level change more often. With current technology, and depending on the area, Google maps can be several years out of date–just like very precise grammars and lexical frequency lists can become obsolete quickly as well.
Zooming in to sentence level reveals even more divergence, since individual utterances are often marked in relation to standard grammar usage–and anyway, are also affected by other factors including local usage, social context, etc. Languages change, and the closer one gets to real, meaningful language in use, the more drastically and quickly they can change. To the point, even, that worrying too much about grammar rules at “street level” begins to seem a bit mad.
We use maps as tools to learn how to get around–but we also get around to learn how to get around. We don’t endlessly study maps and then expect to get from A to B without a problem, because we understand that no map can ever be perfectly detailed. There will always be features of the real world that cannot be described easily, or that change too often, for maps to be our only tool. The best way to learn how to get around is to combine map-reading knowledge with the actual experience of getting around… just like the best way to learn a language is to be able to refer to grammar when needed, yes, but also to just get out there and communicate when that is the best option–which is more often than we may tend to think.