We present cw and pulsed EPR experiments on V4+ in the rutile and anatase phases of TiO2. For the rutile phase, the cw data confirm earlier results, but the relaxation data are different from those previously reported. No earlier results for V4+ in the unreduced pure anatase phase exist. We show that a simple point-charge model can be used to interpret the values of the g tensors, but that models considering only nearest neighbors will give erroneous results. We conclude that the observed V4+ is substitutional in the rutile, but interstitial in the anatase phase. We propose a relaxation mechanism through phonon-modulated hyperfine coupling to explain our T1 data in the rutile phase.