We left the last chapter with a brief glance at the way in which performance on one reasoning problem, the THOG task, can be radically affected by the kind of content used in it, and by the way in which the task is presented. Historically, reasoning tasks have usually been introduced in abstract, (supposedly) content-free form with a view to assessing performance uncontaminated by pre-existing knowledge, beliefs, and motives. Underlying this approach is an assumption that there is some kind of “pure” reasoning ability that we can only access by using abstract tasks. This assumption is questionable on at least two grounds. First, performance on abstract tasks has been shown to be subject to a number of biases, the existence of which implies that such tasks do not tap directly into any human logical ability-unless that ability is fundamentally biased. Second, abstract contents are still contents, and an alternative possibility is that the form and content of problems are not psychologically separable, so that how you think cannot be divorced from what you are thinking about.