It has been reported previously that in thetranscinnamic acids and some substituted butadienes the crystal structure determines whether or not photodimerization occurs and the molecular structure of the dimer, if any. Similar results have been obtained also with chalcones. However, in anthracene and its 9-substituted derivatives the results are qualitatively different. Thus anthracene undergoes dimerization, if slowly, even though potentially combining atoms are 6 Å apart. 9-Cyano and 9-chloro-anthracenes, in structures with short axes of 4 Å, react to give head-to-tail dimers, whereas if structural effects were dominant the head-to-head dimers would be formed. It is concluded that in the last three examples, at least, the reaction does not occur in the bulk (i.e., at ordered sites). Two alternatives are suggested and discussed; the first is that reaction occurs in the vapor phase, the second that reaction occurs at defect sites, some of which are present initially, whereas others are produced during the dimerization. The factors which are responsible for the different behaviors of the anthracenes and of the other systems are discussed.