Prescriptive grammar has (or claims to have) normative force. It tells people how they should speak and write. Traditional language instruction (both first and second language) uses the prescriptive strategy. Prescriptive grammar is the domain of most school language teachers and most newspaper "language maven" columns. Descriptive grammar is what most linguists do. It concerns itself with describing what language users actually do. No native English speaker would ever produce the sentence Cat chimney up, and only an extremely inadequate descriptive account of English would fail to note this. A prescriptive grammar of English would say, "The sentence It 's me is ungrammatical; you must say It is I." A descriptive grammar of English (and Geoff Pullum and Rodney Huddlestone have just produced a gem) will note that in fact almost nobody says It is I, while almost everyone will produce and accept It's me without blinking. Both prescriptive and descriptive grammars would reject Cat chimney up, the former because it is wrong, and the latter because no native English speaker would ever produce such a sentence. Both traditions use both rules and examples. In the prescriptive tradition, rules are meant to be real rules that you are supposed to follow, like rules of etiquette and traffic laws. In descriptive grammar, rules are hypotheses that try to capture regularities in people's actual usage. Linguists usually believe that we have rules in our heads (see Pinker on regular verbs, for example), but that we are mostly unconscious of them. If you will forgive an extended example, English has three regular plural suffixes, all spelled -s or -es, but pronounced differently: the plural of cat is cats, the plural of dog is (pronounced) dogz, and the plural of finch is pronounced finchiz. We use these three distinct plural endings, -s, -z, and -iz, almost flawlessly and with near-perfect agreement between speakers. Nobody ever says doggiz, and nobody pluralizes play as place (using an s sound instead of a z). And yet not one speaker in a hundred can articulate the rule that allows them to choose the correct suffix. A descriptive account notes the regularity, takes a lot of data (examples), and often proposes a rule to explain the observed behavior. But if later data contradicts the rule, the rule will be modified or discarded. None of this is intended to belittle the value of the distinction you made in your original post. Those two approaches to language-learning, one based on lots and lots of examples and immersion, and the other based on rules and careful analysis, both definitely exist, and people do, absolutely, need to think about which approach (or mixture of approaches) matches their own cognitive styles. It's just that the two approaches can't be called prescriptive and descriptive, because language scholars have already appropriated those two words and made them mean something else. I still haven't come up with better words than inductive (for the example-based approach) and deductive or maybe analytic (for the rules-and-thinking approach). We definitely all learn our native languages inductively, while most second-language programs are pretty heavily analytic. One big reason for that is that little kids have a magical-seeming ability to suck up language by example, but that ability seems to fade during adolescence, so that the inductive approach becomes more frustrating. OK, this is my favorite topic so I'm blithering on and on. Sorry. Hope I wasn't too incoherent, and thanks for your patience.