We have determined the optimal placement of individual transmembrane helices in the Pyrococcus horikoshii Glt(Ph) glutamate transporter homolog in the membrane. The results are in close agreement with theoretical predictions based on hydrophobicity, but do not, in general, match the known three-dimensional structure, suggesting that transmembrane helices can be repositioned relative to the membrane during folding and oligomerization. Theoretical analysis of a database of membrane protein structures provides additional support for this idea. These observations raise new challenges for the structure prediction of membrane proteins and suggest that the classical two-stage model often used to describe membrane protein folding needs to be modified.