but have declined in species richness, geographical range and abundance (2-5). Previous studies have assessed the roles played by habitat destruction and loss of flower resources (4,5), and pesticides (6) over relatively modest time scales and geographical ranges.Analyses of regions are rare (7-10) and our understanding of the effects of humanmediated actions over longer periods is limited. Here we assess the bee and flowervisiting wasp species that have gone extinct in Britain, using 494,117 records held by the Bees, Wasps and Ants Recording Society (BWARS), probably the most detailed available for a single country. We define extinct species as those that have not been recorded for at least 20 years following their last observation, despite extensive efforts by members of BWARS and other naturalists.Twenty-three bee and flower-visiting wasp species have become extinct in Britain (Table 1), including formerly widespread species. We exclude single early records that cannot be verified as representing stable breeding populations, but include one species which has recolonized Britain after an absence of six decades (see Supplementary Materials).Since the mid 19 th century the pattern of British bee and wasp extinctions has been characterized by intervals of relative stability, in which few species were lost, interspersed with times when over three species per decade went extinct ( Figure 1, Table 2). These data indicate a period of relatively sustained extinctions from the late 1920s to the late 1950s, with other isolated extinction peaks before and after this time. These features are confirmed in Figure 2, where the average gradient indicates the relative extinction rate over a period, and the period of sustained extinctions is evident as the phase of maximum gradient during the mid 20 th century.The varying rates of extinctions were quantified by applying breakpoint analysis to the cumulative record. In this analysis, a piecewise linear model is fitted to data to reveal periods of approximately constant extinction rate, separated by breakpoints where the rate changes. The analysis was iterated for up to 10 breakpoints and the Akaike Information Criterion (AIC), confirmed by coefficient of determination (multiple-R 2 ), was used to establish the best model (see Supplementary Materials). For these data, changes in AIC and muliple-R 2 level off for two models having four breakpoints (Table S2). These are very similar, sharing the latter three breakpoints, and revealing effectively identical periods of approximately uniform extinction rate for the majority of the 20 th century (Table 2).Both models must be interpreted with caution as the data for 'year last recorded' may not equate to 'year last living'. Declines in populations due to habitat changes may mean a species went unrecorded for some years prior to the actual extinction. The robustness of the breakpoints to this potential ambiguity of the probability of the 'year last living' has been assessed and, whilst there is some sensitivity in the timing of the e...