and Kariukstis). Verification is usually done against an independent set of predictors. The reader is urged to review the reference sources listed in Table 1 for specific details regarding each paleotemperature reconstruction.
Another feature of long-term reconstructions from tree rings that should be borne in mind is that the composition of tree-ring samples that make up each yearly value is variable. In general, an index value for a given site is composed of a number of tree samples from several individual trees. These may vary in age, and in some cases may come from both dead as well as living trees. The number of samples making a particular ensemble average will vary with time, although this is more generally the case near the beginning and end of a particular chronology. These extraneous factors can introduce "noise" and spurious fluctuations into the reconstructions. Nevertheless, the tree-ring indices considered here, while suffering to varying degree from the above-named shortcomings, comprise a high-quality set of proxy temperature records that are of value to compare.
Figures 6 through 8 show decadal-mean values of various tree-ring reconstructions, generally representing summer (or growing-season) mean temperature. For the most part, these tree-ring reconstructions do not sample the 1980s; nevertheless, it is clear from the degree of interdecadal variability that it would be hard to point to any particular recent period as being unique or exceptional. The November-to-April
temperature reconstruction for Tasmania (Cook et al., 1992) does indicate that the last two decades of that record (1970s and 1980s) exhibit significantly greater tree growth (Figure 8) and hence higher reconstructed mean growing-season temperatures than previously, but it would seem premature to conclude that a "climatic change" has taken place in this region of southern Australia. We also note that, unlike the North American and Scandinavian reconstructions, this one is based on tree samples from only one location.
Some unique features are evident for a few of the temperature indices. For example, the curve of D'Arrigo and Jacoby (1992) is a reconstruction of annual mean temperature based on several sites along the northern North America treeline, stretching from Alaska to the Northwest Territories, plus a site located in northern Quebec Province. This reconstruction exhibits a pronounced warming from a very cold period during the mid-1800s to a maximum around 1940. By comparison, the two other high-northern-latitude sites discussed here—the indices from northern Scandinavia and from the northern Urals (Figure 7)—do not exhibit quite the same type of variations. The reader should note, however, that a much longer time period is sampled by these other two reconstructions, and that a seasonal (April to August for the Tornetrask index, and June to July for the