When RNA secondary structures predicted by free energy minimization programs are compared with the structures obtained by phylogenetic comparison it is found that there are often substantial differences. Here we consider a sample of large RNA molecules including 16S and 23S ribosomal RNA, and RNaseP RNA, and compare the minimum free energy and phylogenetic structures for domains of different size within the molecule. The pattern of differences between the structures is consistent with the idea that the native structures are influenced by the kinetics of the folding process, rather than solely by free energy minimization. It is found that the free energy of small domains in the phylogenetic structure of size 100 bases or less is usually much lower (i.e., more stable) than the average value of the minimum free energy of typical domains of corresponding size, whereas the reverse is true for large domains. We interpret these results using a picture of the folding process where short range secondary structure elements form first, which gradually rearrange to form domains of larger and larger size. Once low foe energy domains of a moderate size are formed, these structures will remain trapped, because free energy barriers associated with structural rearrangement will be too large. The result is a structure containing small low free energy domains linked by a few longer range helices which fit in as best they can at a later stage of folding. We also discuss other possible causes of error in the predicted minimum free energy structures, and conclude that the difference between the minimum free energy and the phylogenetic structures cannot simply be put down to errors in the free energy parameters used in the model.