Last summer I evaluated Lift vs Play 1.x and was leaning heavily toward Play. I wrote some toy apps in each. We chose Lift only because it turned out that Play was incompatible with our Maven setup. (I have heard it has improved in 2.0 but I have not looked into it myself.)

I found Play to be fairly intuitive and it matched my mental model pretty well. It was lightweight compared to Lift and the other Java frameworks I looked at. Play 2.0 seems to have made improvements in the areas I found weak in Play 1.x. i.e. modularity, Scala support (specifically for continuations)

Lift is a fairly heavy framework, but you can ignore big parts of it if you need to. In Lift it took me a while to grasp the HTML transforming concept, partly because the documentation seemed to out of date. It felt both cumbersome and elegant at the same time. Sadly I didn't get very far on the project so I have no idea how Lift has changed/improved since then.

Today I would choose Play in a heartbeat if it fit the parameters of my project.