Preferences for natural and urban environments can be framed in terms of (1) beliefs about the likelihood of psychological restoration during a walk in each type of environment and (2) the evaluation of restoration given differing restoration needs. We conducted an experiment to test hypotheses about restoration as a basis for environmental preferences. Imagining themselves as attentionally fatigued or fully refreshed, participants (N=101) evaluated recovery, reflection, and social stimulation outcomes. Next, they viewed slides simulating a walk through a forest or an urban center, then rated the likelihood of recovery, reflection, and social stimulation outcomes following such a walk. This procedure was repeated with the second environment. Preference for the forest over the city was twice as strong given attentional fatigue. The greater likelihood of restoration in the natural environment in conjunction with more positive evaluation of recovery when fatigued appears to explain this pattern. The results have implications for environmental preference conceptualizations and our understanding of the relationship between preference and restoration.