The biotas of a suite of neighboring patches of remnant vegetation often form a series of nested sub-sets, in which the species present in species-poor patches are non-random sub-sets of those present in richer patches. There has been recent interest in ways in which this knowledge may be used to aid conservation. We focus here on whether nested patterns can be used predictively. If nestedness in a fragmented system increases over time through biotic relaxation, locations where particular species may become extinct or are likely to colonize might be predictable and this could be useful in threatened-species management. We used the Temperature Calculator of Atmar and Patterson (1995) to arrange a matrix of bird species' occurrences in a series of buloke Allocasuarina leuhmannii woodland remnants so that nestedness was maximized. Probability bands generated by the calculator were used to predict possible colonization and extinction events. We then re-surveyed the avifauna of the fragments after a seven-year interval to test these predictions. Although nestedness increased between the two survey periods, there was no linear relationship between the generated probability of extinctions or colonizations and the accuracy of the predictions. The predictions derived from the calculator were no more accurate than a second set of predictions generated by use of a simple non-nested model. Despite the increase in nestedness, the arrangement of sites in each of the two maximally packed matrices was substantially different. For the nestedness matrix to generate accurate predictions, an increase in nestedness must be due to a minimization of unexpected species presences and absences rather than an extensive redistribution of species among remnants, as we found. The potential utility of nested patterns in predicting systematic colonization and extinction events should be further evaluated in other, less dynamic, fragmented systems such as those undergoing biotic relaxation.