I think that a failure of statistical thinking is the major intellectual shortcoming of our universities, journalism and intellectual culture.

Whenever you speak to someone, you are presuming the two of you have a certain degree of familiarity - which your words might alter. So every sentence has to do two things at once: convey a message and continue to negotiate that relationship.

I think this confusion leads intellectuals and artists themselves to believe that the elite arts and humanities are a kind of higher, exalted form of human endeavor.

Consistent style prevents presentation from getting in the way of content.

All over the world, belief in the supernatural has authorised the sacrifice of people to propitiate bloodthirsty gods, and the murder of witches for their malevolent powers.

Gossip is certainly one of the things that language is useful for, because it's always handy to know who needs a favor, who can offer a favor, who's available, who's under the protection of a jealous spouse. And being the first to get a piece of gossip is like engaging in insider trading: You can capitalize on an opportunity before anyone else can.

When a scientist considers all high-tech mental machinery needed to arrange words into ordinary sentences, prescriptive rules are, at best, inconsequential little decorations. The very fact that they have to be drilled shows that they are alien to the natural workings of the language system. One can choose to obsess over prescriptive rules, but they have no more to do with language than the criteria for judging cats at a cat show have to do with mammalian biology.