guage evolved in two stages1. His second stage is language as we know it (‘modern language’). He calls the first stage ‘protolanguage’; for now, one can think of it as modern language minus syntax. Bickerton’s interesting claim is that protolanguage is still present in modern humans, surfacing in the course of language learning and when normal language is disrupted. Thus evolution did not throw a Good Idea away; rather it built on it. Here, I want to elaborate on Bickerton’s idea: one actually can discern in the structure of human language a substantial number of distinct innovations over primate calls, some prior to Bickerton’s protolanguage, and some later. Like Bickerton, I will look for traces of these stages in degraded forms of modern language, and relate these stages to what apes have been trained to do. But in addition, in some instances I will be able to demonstrate ‘fossils’ of earlier stages of language in the modern language itself, offering a new source of evidence on the issue. The consequence will be that the language capacity can be conceived of as having evolved incrementally, rather than appearing all at once in an undecomposable bloc. It will no longer be meaningful to ask ‘Does primate P and did hominid H have language?’ We can only ask ‘What elements of a language capacity might primate P have, and what elements might hominid H have had?’ This helps defuse a long-running dispute. On one side are those such as Chomsky2 who have advocated a complex, innate, and unified language capacity that would seem difficult to explain through natural selection; they have therefore been forced to devalue evolutionary argumentation3. On the other side are those who insist on evolutionary justification and are therefore inclined to deny or at least minimize an innate language capacity4. The position proposed here – a complex language capacity that evolved incrementally – helps define a middle ground that I hope will be a useful contribution to discourse. I will assume without justification that any increase in explicit expressive power of the communicative system is adaptive, whether for cooperation in hunting, gathering, defense5, or for social communication such as gossip6–8. I will also take it for granted (although it has been disputed) that linguistic adaptation arose first in the interest of enhancing communication and secondarily in enhancing or refining thought9. Finally, I will assume that the evolution of language proceeded through the vocal–auditory channel, though nothing in my argument precludes an initial stage of gestural (sign) language. (All these assumptions are no doubt controversial, but that is a topic for a different article.) In this discussion I take my cue from the observation of Wolfgang Köhler10 that cognitive steps which appear to us altogether natural might decompose into some parts that are natural for another organism and some parts that are very difficult. The evolutionary counterpart of this observation is that it is not inevitable that evolution should immediately chance upon apparently natural and adaptive aspects of cognition. Thus we must not take it for granted, as some researchers do11, that, for example, an organism with hierarchically organized behavior is therefore poised to invent syntax. The steps I propose are summarized in Box 1.