The main-chain bond lengths and bond angles of protein structures are analysed as a function of resolution. Neither the means nor standard deviations of these parameters show any correlation with resolution over the resolution range investigated. This is as might be expected as bond lengths and bond angles are likely to be heavily influenced by the geometrical restraints applied during structure refinement. The size of this influence is then investigated by performing an analysis of variance on the mean values across the five most commonly used refinement methods. The differences in means are found to be highly statistically significant, suggesting that the different target values used by the different methods leave their imprint on the structures they refine. This has implications concerning the actual target values used during refinement and stresses the importance of the values being not only accurate but also consistent from one refinement method to another.