Different cross-dating methods are compared. All methods considered are based on the t statistic procedure introduced by Baillie & Pilcher (1973), and are distinguished by the way the data are transformed before cross-dating. Matches are quantified in terms of the probability, P, of achieving that match purely by chance. This is calculated by first deriving the probability of achieving an observed correlation coefficient from a single matching experiment, and then allowing for the fact that, when two series are compared at many overlap positions, a multitude of tests is performed. The best match (lowest P value) can be compared to other matches as an additional means of assessing cross-dating strength. Since cross-dating depends on matching the high-frequency elements of a sample against a master chronology, various methods are explored for removing the low-frequency variance in ring-width series before they are compared. The results show that a range of such "pre-whitening" methods can usefully be employed, and no single method is universally superior. The danger in attempting to date samples with relatively few rings, regardless of how the data are treated, is emphasized by these results. Even in relatively straightforward cases, all methods employed are sometimes found to produce spurious dates or to fail to identify a known correct match.