Words have meanings; on that much, psycholinguists generally agree. However, the issue of what ‘meaning’ is, and why a word’s semantic content affects how easily it is recognised, are matters of less consensus. Studies of visual word recognition typically ask participants to perform one of two key tasks: deciding whether a letter string is a valid word (lexical decision) or reading a word aloud (word naming). On the face of it, one should not have to access meaning in order to perform either of these tasks: knowing that ‘fabric’ is a real word but ‘fabnic’ is not, or being able to pronounce it correctly, does not have an obvious semantic requirement. Nonetheless, the meaning of a word affects how quickly it can be processed. ‘Fabric’, for example, is recognised more quickly, and with fewer errors, than ‘factor’, even when other word-level variables such as length, frequency, and so on, have been controlled.