One of the most striking aspects of human language is the complexity of the meanings that it conveys. No other animal possesses a mode of expression that allows it to articulate intricate emotions, describe distant times and places, study molecules and galaxies, or discuss the production of sophisticated tools, weapons and cures. The complex meanings of natural language make it an efficient, general-purpose instrument of human thought and communication. But what are meanings? And how does language convey them?

To illustrate one aspect of the problem, let us consider a phrase in one of Bob Dylan's famous love songs. The phrase opens the song's refrain by describing a woman, whose identity is not disclosed. It goes like this:

If we want to restate the meaning of this phrase in simpler terms, we can do it as follows:

Without doubt, this way of paraphrasing Dylan's verse robs it of much of its poetic value. But at the same time it also highlights a remarkable property of meaning in natural language. When we hear a long expression like (1.1), we immediately draw from it all sorts of simple conclusions. This happens even in cases where we miss information that is important for understanding the “true meaning” of what is being said. Dylan's song only gives vague clues about the identity of the lady. Yet upon hearing the refrain we unfailingly draw from (1.1) the conclusions in (1.2). The converse is true as well: when Dylan invented his description of the sad-eyed lady, he must have implicitly assumed the statements in (1.2) as part of its meaning. This kind of back-and-forth reasoning occurs whenever we think and converse. When we hear, utter or think of an expression, we instinctively relate it to other phrases that we consider obvious conclusions. Drawing such trivial-looking inferences using our language is one of the abilities that characterize us as linguistic creatures. No other animal has this linguistic ability, and no current technology can accurately mimic it.