We have seen evidence suggesting that news media can play an important role in promoting associative reasoning by systematically supplying affectively charged information about political objects. Such evidence makes more credible the claim that long-term patterns in political discourse encourage the development of implicit attitudes. But how can we know that these implicit attitudes actually exist and matter for political decisions?

Several obstacles stand in the way of answering this question. But arguably, the most fundamental of these quandaries is this: How does one detect attitudes that are spontaneous, hard to control, and nonverbalized – in a word, implicit? If, as I have explained, implicit attitudes are beyond introspection, then measuring them by asking people about them is a nonstarter. A more creative measure for this task must be found. Yet this measure must be more than just innovative. It must also demonstrate that it can tap into something that other measures cannot. That is, it must capture the evaluative associations about political objects that implicit attitudes are said to be. Yet, just like other measures, it must also show that these implicit evaluations can systematically explain people's explicit political choices.

The goal of this chapter is twofold: first, to propose a solution to this measurement conundrum; and second, to show that what I capture with this proposed measure is politically consequential. My tool of choice is formally known as the Implicit Association Test (IAT). Rather than asking people to evaluate objects, the IAT enables researchers to infer the presence of implicit attitude, not from what people say, but from what they do and how fast they do it. Specifically, the IAT gauges the evaluative associations individuals have about objects by timing people's ability to rapidly sort exemplars of objects (e.g., Insects, Flowers), positive attributes (e.g., wonderful), and negative attributes (e.g., terrible) – all while using different classification schemes on a computer screen. The assumption here is that people will sort exemplars more quickly when their mental associations about objects match the classification scheme they are instructed to use. If, for example, a person has a more negative attitude toward insects than flowers, she should find it easier to sort words when exemplars of insects and negative attributes are classified using the same response key on a computer (e.g., “E”) than when flowers and negative attributes share that same key.