SOIL CORES Meadow history was examined at three sites representing different communities and a range of microsite types within the current meadow. At each site (RBA, RBB, and RBC), a column of intact soil (10 cm x 8 cm cross section) was removed, wrapped in plastic and fit tightly in a metal canister. Basal soils at each site were unconsolidated, and not sampled. Texture was described at 2 cm intervals in the field using the 'ribbon' method (Brady, 1990). A subsample (5 cm depth) of the RBB core was frozen immediately after removal from the field and later measured for pH. In the laboratory, soil columns were wrapped with a large piece of pottery clay to keep the profile intact, and carefully sectioned in 0.5 cm thick layers. Loss on ignition (550ø C) was determined on 5 cm3 subsamples of each soil layer (size fraction < 1 mm). Subsamples for pollen analysis (1 cm3) were removed from the center of each 0.5 cm layer (1 cm intervals for RBB), and known quantities of Lycopodium spores (Stockmarr, 1972) were added before processing to determine pollen concentrations. Pollen was counted as for surface soil samples, with the addition that the number of well-preserved (Cushing, 1967) and unidentifiable grains were also tallied. Pollen records at each site were zoned using stratigraphically constrained cluster analysis (CONISS; Grimm, 1987).
RELATIONSHIPS BETWEEN POLLEN AND PLANT ABUNDANCE Relationships between pollen and plant abundance were examined at two levels: 1. Presence /Absence An association index (A) was calculated to reveal the degree of co-occurrence between pollen and corresponding plant taxa in vegetation plots: A = B0 / (B0 + P0 + P1) where B0 is the frequency of plots containing the pollen and corresponding plant taxon, P0 is the frequency of plots containing the pollen but not the plant taxon, and P1 is the frequency of plots containing the plant but not the pollen taxon (Davis, 1984). Values of 1 mean that a pollen and corresponding plant taxon always co-occur, values of 0 mean the pollen and plant taxon never co-occur. For Asteraceae, Scrophulariaceae, and Rosaceae comparisons were made at the family rather than genus level because in most cases pollen could not be identified at the genus level in these families.
2. Quantitative relationships The degree of correspondence between pollen abundance and plant cover within vegetation plots was examined by linear regression, using 3 expressions of pollen abundance and 3 expressions of plant cover. Pollen of meadow plants was expressed as: 1) percent of total pollen, the conventional measure of pollen abundance, which has the disadvantage of being influenced by pollen input from nearby forests; 2) percent of the herbaceous pollen, which may be more closely correlated with meadow vegetation, and 3) ratio of meadow pollen types to Tsuga heterophylla pollen, a measure that avoids the inherent interdependency of pollen percentages by expressing meadow pollen relative to "background" pollen assumed to be constant at all meadow locations (Beaudoin, 1986; Sugita, 1993b). Tsuga heterophylla is a regionally dispersed pollen type in the Olympic Mountains (Brubaker and Gavin, in prep.; Heusser, 1969) that should be evenly deposited across the meadow surface. The nearest Tsuga heterophylla trees occur in small stands more than 2.5 km from Meadow Ridge. Plant cover was expressed at three spatial scales to assess the source area of pollen: 1) neighboring plant cover (NPC), calculated as the percent cover within 0.5 m of the sample site, 2) distance-weighted plant cover (DWPC), calculated as the sum of 67% of percent cover within 0.5 m and 33% of the percent cover from 0.5 to 1 m radius, and 3) unweighted plant cover (UPC), calculated as the total percent cover within 1 m. The formulation of DWPC corresponds to weights of 1/dý, where d is the distance to the center of each distance class, which approximates the distance weighting used in theoretical (Sugita, 1994) and applied (Calcote, 1995) studies of pollen dispersal. For each expression of pollen abundance, linear regressions were performed on 3 expressions of plant abundance. For each plant taxon, coefficients of determination (rý) were compared to assess which relationship was strongest. If correlations increased with larger source areas (NPA to UPA), the relevant pollen source area (the radius at which pollen best represents vegetation) would likely be > 1 m. Conversely, if the highest correlation was achieved with NPA, then the pollen source area should be within 1 m. I removed single outlier samples from Cyperaceae and Polemanium, which would have greatly increased the correlation coefficient.
RADIOCARBON-DATING AND FOSSIL POLLEN I obtained AMS radiocarbon dates on pollen extracts purified from soil. For these analyses, one 50 cm3 sample near the base of each core and one at mid-depth of core RBA was purified using a series of sieving and chemical treatments, slightly modified from Brown (1994). Pollen was examined for purity by counting pollen and non-pollen fragments in subsamples at 400X magnification. To test if pollen had moved downward in the soil column for four samples, I also obtained AMS radiocarbon dates on charcoal fragments (1 - 5 mm in diameter) at the same depths as the pollen. The underlying assumption of this comparison was that the charcoal pieces represent plant material burned at the soil surface and are sufficiently large that they would not be moved down the soil column. Pollen grains also originate at the soil surface, but have potential for downward movement due to their small diameter (20 - 120 mm) (Dimbleby, 1985; Havinga, 1963; Russell, 1993). In order to compare dates from pollen and charcoal, radiocarbon dates were calibrated using CALIB 3.0.3b (Stuiver and Reimer, 1993), using a 200-year moving average on the calibration curve for the pollen extract, because the pollen extracts likely represent > 100 yr of pollen deposition. Calibrated dates for pollen and charcoal at the same depth were compared using the reported counting error and the t-test subroutine in CALIB.