In their paper "Why we may not find intentions in the brain," Uithol et al. (2014) convincingly argue that "the processes underlying action initiation and control are considerably more dynamic and context sensitive than the concept of intention can allow for." Their paper could be seen as a critical note to the widespread tendency to search for identifiable neurocorrelates of mental concepts. Their more specific suggestion is that the absence of clear neural correlates undermines the traditional understanding of intention. In this paper I will try to take their argument a step further. First of all, I will argue that our folk psychology leaves room for various understandings of intentions, and that the concept of intention discussed by Uithol et al. is an academic concept that has its roots in the causal theory of action and in functionalist approaches to cognition. I will argue that both these paradigms are contested, and that there seems to be theoretical wiggle room for alternative understandings of intention. Subsequently I outline such an alternative perspective based on Wittgensteinian philosophy of psychology, emphasizing the regulative role of intention talk. However, the proposed understanding raises the question how to think about neural realization: is intention talk "just" talk, or do intentions really exist? I will propose that intention talk should be understood as a form of pattern recognition, and that the patterns involved are extended in both space and time. The conclusion outlines some important implications for the neuroscientific investigation of intentions.