Phases play an important part in Minimalist theories of locality. But why do phases give rise to locality effects? In this presentation, I discuss a theory recently put forward by Grano and Lasnik (2018) (see also Barros and Frank 2017), which connect locality effects to phi-feature valuation, motivated by bound pronoun subject effects, such as the contrast in More people said {they / *their friends} like Pepsi than Coca-Cola. I sketch revisions to Grano and Lasnik's system, so that whether a phase imposes locality restrictions depends on whether phi-features on its head are valued or not, to model contrasts such as More people claimed that there was a problem with the economy than with illegal immigration (Barros and Frank 2017). Following Fox and Pesetsky (2003, 2005), I suggest that phi-feature valuation on a phase head is a necessary condition for linearization of the phase. This proposal allows us to derive locality effects from linearization, in turn eliminating the need to appeal to structural notions such as phase edge or domain. Lastly, I contrast this proposal with two recent proposals about phases developed by Richards (2007, 2011).