Four months after Hurricane Gilbert we resampled 10 habitats that were sampled previously in December 1987. Overall, we found no change in the total number of species nor in the mean number of individual birds detected. The mean number of individuals declined in three montane habitats (cloud forest, pine plantation, and coffee plantation), where structural damage to tree trunks and branches was often severe, and trees were still defoliated. In the mountains, higher proportions of nectarivores and fruit/seedeaters declined than insectivores. We found increased mean numbers of individuals in two lowland sites (wet limestone forest and mangroves), where structural damage to trees was also severe, but where new foliage was present. Mean numbers of individuals did not change in five other lowland habitats, despite varying levels of vegetation damage. Populations of several species declined in some habitats and increased in others, a pattern consistent with interhabitat migration. Population declines in montane habitats were related to diet, suggesting that Hurricane Gilbert's greatest stress on Jamaica's montane bird populations occurred after its passage rather than during its impact. Frequent hurricanes may contribute to some of the commonly observed characteristics of the Caribbean avifauna. Received 5 March 1991, accepted 10 July 1991.

'Institute of Tropical Forestry, Southern Forest Experiment Station, USDA Forest Service, P.O. Box B, Palmer, Puerto Rico 00721; and Department of Biology, University of Puerto Rico, Cayey, Puerto Rico 00633, USA; and 2Terrestrial Ecology Division, Center for Energy and Environment Research, GPO Box 363682, University of Puerto Rico, San Juan, Puerto Rico 00936, USA. IN SOME tropical regions, hurricanes occur with sufficient frequency to be important fac- tors in determining the structure and species composition of biotic communities (e.g. Wads- worth and Englerth 1959, Odum 1970). For bird populations in particular, hurricanes can have both direct and indirect effects (for recent re- view, see J. Wiley, unpubl. manuscript). Direct hurricane effects include death from exposure to high winds and rain (Kennedy 1970). Indi- rect effects of hurricanes on bird populations include destruction of food supplies (or nesting, roosting, and foraging substrates) by the storm's high winds (Jeggo and Taynton 1980). Further- more, storm-weakened birds may be at greater risk to predation, particularly in the absence of vegetative cover or roosting sites for protection (Engstrom and Evans 1990). Hurricanes can dis- rupt normal migration patterns and, in some instances, change the geographic distributions of species (Thurber 1980). Following hurri- canes, humans may kill weakened birds and accelerate the rate of habitat destruction (Thompson 1900). The difficulty with much of the previous lit- erature on hurricane effects on bird populations is its anecdotal nature and the absence of prior baseline data for quantitative comparison after the storm's impact (but see Lynch 1991a, b, As- kins and Ewert 1991, Waide 1991). Another problem is that a variety of habitats or sites is rarely surveyed and, thus, it is difficult to assess overall population or community changes after a hurricane. Our study takes advantage of base- line samples of bird populations and vegetation structure gathered in 10 habitats in December 1987, prior to the arrival of Hurricane Gilbert in Jamaica. We replicated our baseline mea- surements in these same sites in January 1989, four months after the storm's passage, to eval- uate the short-term effects of the hurricane on vegetation structure and terrestrial bird popu- lations. Vegetation measurements provided us with an indication of the storm's damage at each of the sites where we counted birds. Here we describe the effects of a hurricane on avian pop- ulations before major plant successional changes took place. Hurricane Gilbert reached hurricane force on 10 September 1988 approximately 363 km southeast of the Dominican Republic. It side- swiped Puerto Rico and Hispaniola, and then struck Jamaica on 12 September. Hurricane Gil- bert arrived at 1700 with minimum sea-level pressure of 960 mbars and sustained winds of 205 kph, with gusts of up to 226 kph. Over 700 mm of rain fell over the interior mountains of the island from 10-12 September. After leav- ing Jamaica, Hurricane Gilbert became the most powerful storm recorded in the Western Hemi- sphere in this century (Lawrence and Gross 1988). Damage caused by Gilbert to Jamaican forests was widespread and severe (Jamaica Natural Resources Conservation Department, unpubL report; Varty 1991, Bellingham et aL 1992), causing local observers to believe that bird populations might have been seriously re- duced (Haynes-Sutton 1988, Gosse Bird Club 1989). Although Gilbert was an intense hurri- cane, Gupta (1975, 1988) has shown that storms of this magnitude are common, and even larger ones have occurred in Jamaica's past. METHODS Vegetation.--We sampled I0 typical Caribbean hab- itats (Appendix I). We used two to four 16-m-diameter circular plots (0.02 ha) to quantify vegetation in each habitat. Two plots were placed in vegetation repre- sentative of areas in which bird point counts were made. Occasionally, additional plots were located near the mist nets or in areas representing the extremes of vegetation structure in variable habitats. After the hurricane, we attempted to measure vegetation in the same locations as the prehurricane plots. Pre- and posthurricane plots were considered to be well matched when the original center tree could be found or, alternatively, if: (I) there was greater than 80% similarity in the number of trees in every diameter class; (2) the number and diameters of trees recorded by species or life form were the same (e.g. standing dead, palms, tree ferns, miscellaneous, dicots); and (3) the understory species or life forms were the same (i.e. grasses, forbs, ferns, bamboo, ginger and cacti). We obtained matched pairs of pre- and posthurricane plots in montane cloud forest (4 of 4 sites), montane pine plantation (I of 2 sites), montane coffee (2 of 2 sites), dry limestone forest (2 of 2 sites), dry limestone ruinate (I of 3 sites), wet limestone forest (I of 3 sites), lowland coffee with mimosaceous overstory (2 of 2 sites), lowland second growth forest (I of 3 sites), and mangroves (2 of 4 sites). Stems of all standing and hurricane-felled trees and saplings greater than or equal to 3 cm diameter were measured 1.3 m from the base (dbh) and were re- corded in diameter classes (dbh) of: 3-8 cm; >8-15 cm; >15-23 cm; >23-38 cm; and >38 cm. We also classified each tree according to life form or species as noted above, and. according to the type of structural damage from the hurricane. Damage classes (from most to least severe) were: trunk broken; trunk down (uprooted with trunk or branches on the ground); trunk leaning (partially uprooted); and presence of major branch breaks (branches >10 cm diameter; branch breaks recorded in all habitats except montane cloud forest). If an individual tree had more than one type of damage, then only the most severe damage was used in the analysis. Damage data were pooled from all posthurricane plots in each habitat, except in wet limestone forest where damage varied signif- icantly among plots. Foliage-height profiles were determined at 20 points located at !.6-m intervals along the north, south, east and west radii of the circular plot (after Schemske and Brokaw !98!). A 3-m pole (2.0 cm diameter) marked at 0.5-m intervals was placed vertically at each sample point. We recorded the presence or absence of foliage touching the pole within each height class. For height intervals above 3 m, we sighted along the pole and recorded the presence/absence of foliage in each of the following estimated height intervals: 3- 4, 4-6, 6-8, 8-10, 10-12, 12-15, 15-20, 20-25, and 25- 30 m. We measured heights of !0 canopy trees in the plot with optical range finders. According to these measurements, we overestimated the heights of the tallest trees using the canopy-profile method in 2 of the 42 plots; the data from the upper two categories of the canopy-height profiles in those plots, therefore, were moved to the next-lower height interval. For each height interval, we calculated percent cover by dividing the number of points in which foliage was present in that height interval by the total number of sample points (n = 20) and multiplying by I00. Pre- and posthurricane foliage-height profiles are shown on the same graph only if they represent a matched pair. When there was more than one matched pair of pre- and posthurricane measurements, only one plot is shown as determined by a coin toss. Foliage-height profiles were used to analyze changes in both the amount of foliage and its distribution. Friedman tests (nonparametric equivalent of an ANOVA) were used to determine if there were dif- ferences among all plots in the amount of foliage present for a given habitat, including pre- and post- hurricane measurements (two-tailed tests). If signif- icant or suggestive differences (P < 0.1) in the amount of foliage were found, then we quantified possible hurricane-induced changes in the amount of foliage by comparing the number of height intervals in which the frequency of foliage intercepts increased or de- creased in a plot. We used one-tailed sign tests for each plot (matched pre- and posthurricane pairs) to evaluate the null hypothesis that there was no de- crease in the amount of foliage after the hurricane. Understory vegetation was somewhat weighted in this analysis, because there were more height classes in the understory than in the overstory. Possible hur- ricane-induced changes in the distribution of foliage within a plot were analyzed by comparing pre- and posthurricane foliage-height profiles for each plot, using contingency chi-square tests. This test is insen- sitive to the amount of foliage present. Bird censuses.--We surveyed from 5-23 December 1987 before the hurricane, and 22 January to 2 Feb- ruary 1989 after the hurricane. We modified the fixed- radius point count method of Hutto et al. (1986) after taking preliminary counts to determine the effective- ness of the technique in three different habitats in Puerto Rico. A single observer recorded all birds seen and heard during a 10-min period at each point. Counts were initiated at sunrise and terminated before 1200, with most counts completed before 1100. Each point was at least 100 m from all others and not closer than 25 m from a habitat edge. We attempted to complete 30 point counts per habitat, but frequently the size of the habitat limited the number of point counts (summarized in Appendix 2). Two observers, moving in opposite directions, each independently made 15 point counts on the same morning in the same habitat. In dense habitats, we sampled from trails or roads and, frequently, used trails to travel through a habitat. In open habitats, we used a compass to follow a tran- sect. For each bird observed during a point count, we estimated the minimum distance from the observer. Those birds that were heard but not seen were tallied in one of two categories: -<25 m from the observer; or >25 m from the observer. Before sampling in Ja- maica, we chose 25 m as the radius within which we could detect all individuals in all but the most dense habitats. We calculated the mean number of detections per 25-m-radius point count for each species in each hab- itat both before and after the hurricane. We compared the before and after samples for each species in each habitat with a Mann-Whitney U-test. For each habitat we tallied the number of species that increased, showed no change, or decreased after the hurricane. The diet of each species was classified as either nectar, fruit/seeds, or insects on the basis of Lack (1976) and Faaborg (1985). This information was analyzed in a three-way table using a log-linear model (Sokal and Rohlf 1981) to test for the presence of a three-factor interaction among habitat (10 habitats), diet (two diet types: plant or insect), and population trend (two trend categories: increase/no change or decrease). The log- linear model was then used to test for conditional independence, as well as two-factor interaction. A probability of type I error of 0.05 or less was accepted as significant, but we show greater values for descrip- tive purposes. Throughout the text, we use standard errors to describe variation around the mean. Appropriate control sites to monitor changes in bird populations in the absence of a hurricane between 1987 and 1988 could not be located on Jamaica, be- cause all sites on the island sustained at least some hurricane damage. Therefore, as a control to examine changes in bird populations in the absence of a hur- ricane between 1987 and 1988, we used two sites on the island of Puerto Rico. Wunderle conducted point counts in a Puerto Rican lowland pasture and mon- tane second-growth forest in November 1987 and 1988. Point counts in the control sites were run in the same manner as in Jamaica, except that two consecutive mornings were required per site (15 counts/morn- ing). The lowland-pasture control site located in Salinas, Puerto Rico, was drier and contained more mesquite (Prosopis julifiora) than the lowland pasture in Jamaica. However, the two pasture sites have similar avifau- nas, with 12 genera and 8 species in common. The montane second-growth forest site at 720 m near Car- ite, Puerto Rico, was at a lower elevation and had more shrubby vegetation than the montane cloud for- est at Hardwar Gap, Jamaica. Similarities exist in the avifauna of the two montane sites in that they share 16 genera and 8 species. Mist-net sampling supplemented the point counts in 6 of the 10 habitats in Jamaica. We used 12-m nets with 30-mm mesh and four shelves, set to a height of 2.5 m, usually in a continuous line. The number of nets ranged from 15 to 20 per site. We usually set nets during the afternoon and then opened them at sunrise on the following day; nets were kept open until late afternoon (with the exception of the man- grove site after the hurricane). Morning netting was simultaneous with point counts in each habitat. After the hurricane, we attempted to set nets in the exact location as earlier, but this was impossible in some cases because of extensive damage to the site. In wet limestone forest and mangroves, we set nets 15-20 m from their prehurricane positions. We used mist-netting data to estimate population abundance before and after the hurricane in each of the sampled habitats using a modification of the method of Terborgh and Faaborg (1973). For each habitat and each netting session, we regressed the number of new captures per 15 net hours against the cumulative net hours of sampling effort. From the Y-intercept we obtained an estimate of the projected capture rate. In addition, we show the slopes of the regression before and after the hurricane in each hab- itat. The slope of the regression provides a measure of overlap in foraging range; a steep negative slope indicates little overlap. We compared the mist-netting results with those of the point counts, but did not analyze them statistically because of small sample sizes. RESULTS Damage to vegetation.--Hurricane Gilbert swept the length of Jamaica. The eye of the storm passed over all sampled habitats except for the dry limestone forest at Portland Ridge (Fig. I), where damage to structure (branches and trunks) was slight. Among the sites crossed by the eye of the storm, second-growth lowland 0 30 RIDGE Fig. 1. Path of Hurricane Gilber over Jamaica on 12 September 1988. Dashed lines indicate approximate width of hurricane eye, and names indicate approximate locations of sampled habitats: Hardwar Gap, montane cloud forest, montane pine plantations, and montane coffee; Portland Ridge, dry limestone forest; Discovery Bay, dry limestone ruinate; Windsor, wet limestone forest and lowland coffee; Negril, lowland secondary forest, lowland pasture, and mangroves. habitats with short stature were least damaged (dry limestone ruinate, Fig. 2A; lowland sec- ondary forest, Fig. 2E). Structural damage was most variable in the hilly limestone karst area around Windsor, where damage to the wet limestone forest was most severe in the flood- plain (Fig. 2H), and least severe in a saddle between two hills (Fig. 2G). Coffee trees were relatively undamaged compared to the shade overstory trees in the lowland coffee plantation, in contrast to the montane open-grown coffee in which an estimated 60-80% of the coffee trees were blown over. The most severe structural damage to large-diameter trees was in man- groves (Fig. 2F) and montane pine plantations (Fig. 2J). Thus, topography, tree stature, and species differences contributed to the variation in plant structural damage among habitats. Structural damage accounted for most of the reduction in foliage in the lowland habitats, as these habitats had refoliated by the time of our survey (Figs. 2 and 3). In the lowlands, signif- icant reductions in the total amount of foliage were found in only one plot each in lowland coffee and mangroves (Table 1). Although total amounts of foliage were unchanged in most lowland plots, significant changes in the dis- tribution of foliage indicate that regrowth of understory foliage compensated for loss of can- opy foliage in wet limestone forest, lowland coffee, mangroves, and dry limestone ruinate. In contrast, we found significant reductions in the amount of foliage in all plots in montane cloud forest and montane coffee, and a sugges- tive (P = 0.07) but statistically nonsignificant decrease in montane pine plantations (Table 1). Most of the foliage loss in montane habitats also was attributed to structural damage, with the exception of montane cloud forest. The severe foliage reduction in montane cloud forest (Fig. 3B) was due primarily to the absence of new foliage as damage to branches and trunks was moderate (Fig. 2B). Control sites.--At our two control sites in Puerto Rico, we detected little change in the bird populations over a one-year period (No- vember 1987 to November 1988). For example, the average number of individuals per point count did not change significantly in montane second-growth forest or lowland pasture (Table 2). Only slight decreases were noted in total numbers of species in each control habitat after one year, and the species turnover also was low (Table 2). None of the species in the control habitats had significant or suggestive popula- tion changes between censuses. Species turnover.--On Jamaica, we detected 58 species of birds in the point counts before the hurricane and 58 species afterwards. Of these, 50 species were encountered before and after, leaving a turnover of 8 species observed only before and 8 species observed only after the hurricane. Species observed only before the hurricane were infrequent (Crested Quail Dove, 150 A Dry limestone ruinate (2 plats} lOO 50, .D Dry limestone forest (2 plats) 75- 5O 25, o 15- lO- G Wet limestone forest (1 of 2 plots) Fig. 2. E Lowland secondary (2 plots) --] Undamaged :[] Branch break '.[] Trunk lean  Trunk down I Trunk break F Mangroves (4 plots) H Wet limestone forest (1 of 2 plots) J Pine (2 plots) i DIAMETER CLASSES (CM DBH) Major structural damage to trees in Jamaica caused by Hurricane Gilbert by diameter class. Trees classified by most severe damage sustained in following order: trunk break; trunk down (completely uprooted); trunk lean (partially uprooted); branch breaks (only those > 10-cm diameter); and undamaged. (A) Dry lime- stone ruinate (Discovery Bay); (B) montane cloud forest (Hardwar Gap); (D) dry limestone forest (Portland Ridge); (E) lowland secondary forest (Negril); (F) mangroves (Negril); (G-H) wet limestone forest (Windsor); (J) montane pine plantation (Hardwar Gap). Data pooled for similar 16-m-diameter circle plots; number of plots noted. Data not pooled in wet limestone forest because of heterogeneity in damage. Number of trees on Y-axis shows pooled sample sizes of trees in each diameter class. 25-30 20-25 15-20 12-15 10-12 ß ß 8-10 6-8 4-6 3-4 2.5-3 2-2.5 1.5-2 1-1.5 0.5-1 0-0.5 A Dry limestone ruinate B Montane cloud forest C Lowland coffee 25-30  D Dry limestone forest 20-25 L 15-2o r 12-15 10-12  8-1o ........ .......... ,,,,. 2.5-3  .......... I 2'2.51 1.5-2 1. ....... ,, ß 1-1.5 0.5-1 0-0.5 , E Lowland secondary F Mangroves 25-30 ;=z=a G Wet limestone forest 20-25 - ............. 15-20 ' ................. 12-15 ' ................. 8-10 ' 4-6 3-4 2.5-3 ...... I' fl 2-2.5 0.5-1 0-0.5 ................ Fig. 3. H Wet limestone forest ß [] BEFORE ß AFTER 0 20 ' 4) 6) ' 8 PERCENT CANOPY COVER J Pine i 0 20 40 60 80 Representative foliage-height profiles for various vegetation types throughout Jamaica before and after Hurricane Gilbert. (A) Dry limestone ruinate (Discovery Bay); (B) montane cloud forest (Hardwar Gap); (C) lowland coffee (Windsor); (D) dry limestone forest (Portland Ridge); (E) lowland secondary forest (Negril); (F) mangroves (Negril); (G-H) wet limestone forest (Windsor); (J) montane pine plantation (Hardwar Gap). For each 16-m-diameter plot shown (except in H), canopy-height profiles recorded along 20 vertical transects in December 1987 and again in January 1989. Percent cover (X-axis) shown for each height interval in meters on Y-axis. Height intervals increase with height on (Y-axis). T^BIE 1. Summary of probability values for tests of similarity in amount and vertical distribution of foliage before and after Hurricane Gilbert in Jamaica. Changes analyzed by tallying number of increases or decreases in foliage present in 15 height classes for each plot. Changes in vertical distribution of foliage analyzed by comparing overall vertical pattern of foliage distribution. Total numbers of vegetation plots shown, but comparisons of foliage before and after hurricane restricted to matched-plot pairs. Habitat Number Foliage amount Foliage of plots distribution (before: after) Among all plots a Before vs. after b before vs. after 1/2 Montane cloud forest 4:4 Montane pine plantations 2:2 Montane coffee 2:2 Wet limestone forest 2:2 Lowland coffee 2:2 Lowland secondary forest 2:2 Mangroves 2:4 Dry limestone ruinate 2:2 Dry limestone forest 2:2 <0.001 >0.100 <0.001 0.020 <0.005 0.020 <0.005 <0.001 >0.500 0.068 0.073 <0.005 <0.001 0.004 >0.100 0.004 <0.005 <0.001 0.500 <0.020 0.056 0.254 <0.005 0.054 < 0.005 0.175 0.500 >0.500 0.087 < 0.001 < 0.010 0.508 <0.050 0.552 0.500 <0.005 0.140 0.131 >0.500 0.194 > 0.975 Two-tailed Friedman test. One-tailed sign test. Contingency chi-square. n = 2; Zenaida Dove, n = 1; Red-necked Pigeon, n = 1; Vetvain Hummingbird, n = 1; White- eyed Thrush, n = 1; Tennessee Warbler, n = 1; Hooded Warbler, n = 1; Chestnut-sided War- bier, n = 6). Most species observed only after the hurricane were also rare and foraged alone (Ring-tailed Pigeon, n = 1; Ruddy Quail Dove, n = 1; Mangrove Cuckoo, n = 3; Greater Antil- lean Elaenia, n = 1; Cape May Warbler, n = 1; Yellow-throated Warbler, n = 3), but two spe- cies were found in flocks (Guiana Parrotlet, n = 4; European Starling, n = 5). The mean de- tections per habitat for each bird species are summarized in Appendix 2. Species turnover within individual habitats after the hurricane was frequently high (Table 2). For instance, 12 species were lost from low- land-coffee and dry-limestone-ruinate habitats. Fewer were lost in the following habitats: mon- tane coffee (11 species), montane cloud forest (10 species), and montane pine plantation (10 species). On average, the loss of species from a habitat was greater than the addition of new species to the same habitat following the hur- ricane, though this difference was only sugges- tive (paired t-test, t = 1.95, P = 0.08; Table 2). The average number of species per habitat also showed a suggestive, but not significant, de- cline (paired t-test, t = 1.95, P < 0.08; Table 2). The average number of habitats occupied by a species declined significantly (paired t-test, t = 2.41, P = 0.02) from 3.7 + 0.4 habitats per species before the hurricane to 3.2 + 0.4 habitats per species afterward. Despite the high spatial vari- ation in damage within some habitats, the vari- ance in individuals detected per point count was generally similar in the pre- and posthur- ricane counts. Only in the montane pine plan- tations were significant differences (Fmax = 3.13, P < 0.01) found in the variance of individuals detected per point count before (9.4) and after (3.0) the hurricane. Abundance.--The mean number of individu- als detected per point count changed signifi- cantly after the storm's passage in 5 of the 10 habitats (Table 2). We found significantly fewer individuals per point count in the three mon- tane habitats--cloud forest, pine, and coffee plantations. Netting results in montane cloud forest and montane coffee were consistent with the point counts as indicated by dramatic de- clines in total captures. The projected capture rates, consequently, were low within the two habitats (Table 3). In contrast, we detected sig- TABLE 2. Changes in mean number of individuals, total number of species, and species turnover in fixed- radius point counts in 2 habitats in Puerto Rico (Controls) and 10 habitats in Jamaica before and after Hurricane Gilbert. Comparison of mean values before and after hurricane for each habitat made with a Mann-Whitney U-test. Paired t-test used to determine significance level for total means. Individuals per point count (œ + SE) Total no. species Before After Habitat Before After Before After only only Species turnover Controls (Puerto Rico) Montane second-growth forest 5.6 + 0.3 6.1 + 0.4 21 18 4 1 Lowland pasture 4.8 + 0.4 4.7 _+ 0.3 18 17 5 4 Treatments (Jamaica) Montane cloud forest 3.3 _+ 0.3 2.3 _+ 0.4 b 27 21 10 4 Montane pine plantation 5.0 + 0.6 2.1 _+ 0.3 a 23 19 9 5 Montane coffee 2.9 +- 0.3 1.5 _+ 0.3 a 18 9 11 2 Wet limestone forest 3.8 _+ 0.6 5.0 + 0.6 c 28 32 3 7 Lowland coffee 3.0 _+ 0.4 2.5 _+ 0.6 26 20 12 6 Lowland secondary forest 4.8 + 0.5 4.8 _+ 0.4 27 29 7 9 Mangroves 4.4 _+ 0.3 5.5 + 0.4 c 22 23 5 6 Lowland pasture 4.5 _+ 0.6 4.3 _+ 0.6 25 23 7 5 Dry limestone ruinate 3.7 _+ 0.4 3.4 + 0.5 24 20 12 8 Dry limestone forest 3.0 +_ 0.4 3.5 _+ 0.4 21 20 4 3 Mean total 3.7 3.4 24.1 21.6 a 8.0 5.5 a SE 0.3 0.4 1.0 1.9 1.0 0.7  P - 0.001. b p - 0.003.  P = 0.05. d p = 0.08. nificantly more individuals per point count in the wet limestone forest and mangroves. Net- ting results supported the point-count results in the wet limestone forest, but not in the man- groves. This inconsistency in the mangrove re- suits may occur because the 1987 netting was initiated in the late afternoon and the 1989 net- ting was initiated in the early morning. The mist-net and point-count results were consis- tent in lowland secondary forest and dry lime- stone forest, where we found no significant changes. When making a large number of statistical comparisons, some observations are expected to deviate significantly from expected values by chance alone. For example, we performed 211 individual tests for population changes after the hurricane, of which 10% (21.1) and 5% (10.6) were expected to differ by chance at significance levels of 0.10 and 0.05, respectively. However, we found that 56 populations changed at P -< 0.10 and that 37 populations changed at P -< 0.05 (Table 4). Furthermore, eight populations showed changes at the level of 0.001--a very unlikely random event. We conclude that the majority of the population changes after the hurricane were not artifacts of the large number of statistical comparisons. Eighteen species showed significant or sug- gestive declines in mean number of individuals per point count in one or more habitats (Table 4). These declines could result from either hur- ricane-induced mortality, movement to other habitats, or decreased detectability. Sixteen spe- cies showed significant or suggestive increases in mean number of individuals per point count in one or more habitats (Table 4). These increas- es could result from movement into a habitat, because the birds were easier to detect as a result of increased foraging activity, or because pre- viously obscured canopy dwellers became more apparent. Eight species decreased in some hab- itats and increased in other habitats (Table 4). Such observations are consistent with a be- tween-habitat migration, but factors such as changes in detectability could also be involved. To examine changes in detectability, we used the methods of Hutto et al. (1986) to calculate detectability ratios for each species both before and after the hurricane. The ratio is equivalent to the number of point counts at which a given species was recorded only beyond the 25-m ra- dius, divided by the total number of counts at which the species was recorded. Of 48 species with adequate sample sizes before and after the storm, 23 species were more detectable after the storm, 17 species were less detectable, and 8 species were unchanged. The number of species TABLE 3. Summary of mist-net capture results before and after Hurricane Gilbert in six Jamaican habitats. Slope and Y-intercept derived from regression of captures per net hour against net hours. Total Individuals ! Habitat Net hours Species Captures 100 net h Slope Y-intercept Montane cloud forest Before 195.5 23 130 66.5 -0.001 0.79 After 293.7 16 54 !8.4 -0.00! 0.34 Montane coffee Before 176.9 18 74 41.8 - 0.002 0.59 After 200 10 32 16.0 -0.001 0.24 Wet limestone forest Before 170.5 17 56 32.8 -0.002 0.50 After 178.0 27 89 50.0 - 0.004 0.91 Lowland secondary forest Before 192.5 21 79 41.0 -0.005 0.91 After 148.5 23 80 53.8 -0.004 0.98 Mangroves Before 121.0 20 122 100.8 -0.019 2.67 After 172.6 25 151 87.5 +0.002 0.86 Dry limestone forest Before 283.8 16 46 16.2 - 0.00! 0.32 After 140.0 17 39 27.9 -0.001 0.37 that were more detectable was not significantly different from the number of species that were less detectable (chi-square goodness-of-fit test, X 2 = 0.90, df = 1, P < 0.50). Four species were significantly (P < 0.05) more detectable after the hurricane--Streamertail, Greater Antillean Bullfinch, Yellow Warbler, and Arrow-headed Warbler. The Olive-throat- ed Parakeet and Jamaican White-eyed Vireo were significantly less detectable. However, no obvious relationship was found between changes in a species' detectability ratio and changes in the average number of individuals per point count after the hurricane. Thus, sig- nificant changes in detectability ratios were found in some species after the hurricane, but it is unlikely that these changes caused a con- sistent bias. Overall patterns.--We found a strongly sug- gestive (G = 16.64, df = 9, P = 0.055) three-way interaction of habitat, diet, and population trend based on the log-linear analysis. Tests for con- ditional independence were run on two factors at a time given the level of a third factor. For instance, a significant interaction was found be- tween diet and population trend for specific habitats (G = 21.18, df = 10, P = 0.02). A sig- nificant interaction was found between habitat and population trend for a given diet type (G = 35.39, df = 18, P = 0.008). There was no sig- nificant interaction between habitat and diet for a given population trend (G = 20.86, df = 18, P = 0.34). However, a test of independence in- dicated a significant (G = 42.68, df = 27, P < 0.05) interaction between diet and population trend. Previous hurricane studies have suggested that montane bird populations are affected more than lowland populations (J. Wiley, unpubl. manuscript). To test this possibility, we com- pared the trends (i.e. decrease versus increase/ no change) of each species' population in each montane habitat (habitats above 700 m; n = 3) with the trends of populations in the lowland habitats (n = 7). This comparison indicated a significant difference (G = 4.45, df = 1, P = 0.04) in which 63% (49 of 78) of the montane popu- lations declined compared to 48% (100 of 210) of the lowland populations. As noted previ- ously, this pattern was also evident in the av- erage number of individuals per point count (Table 2). We expected the pattern of population de- cline in the montane regions to vary with diet, with those species relying directly on plants for food suffering most (J. Wiley, unpubl. manu- script). To examine this possibility, we catego- rized species into four diet categories (nectari- TABLE 4. Species that declined or increased (P -< 0.10) in mean detections per point count after passage of Hurricane Gilbert in Jamaica. Habitats where declines or increases detected are listed, as are significance levels based on comparison of pre- and posthurricane point counts using Mann-Whitney U-test. Location of decline (P) Location of increase (P) Dry limestone forest (0.063) Montane cloud forest (0.001) Montane coffee (0.006) Montane pine plantation (0.001) Lowland secondary forest (0.10) Lowland coffee (0.08) Caribbean Dove Streamertail Loggerhead Kingbird Black-faced Grassquit Montane cloud forest (0.03) Montane coffee (0.001) Montane pine plantation (0.013) Yellow-shouldered Grassquit Dry limestone ruinate (0.04) Gray Catbird Lowland secondary forest (0.086) Black-throated Blue Warbler Montane coffee (0.03) Wet limestone forest (0.08) Lowland coffee (0.08) Dry limestone forest (0.08) Montane pine plantation (0.02) Lowland coffee (0.04) Lowland coffee (0.08) Ovenbird Black-and-white Warbler Chestnut-sided Warbler Bananaquit Orangequit Stripe-headed Tanager Montane pine plantation (0.001) Montane coffee (0.04) Lowland pasture (0.05) Montane pine plantation (0.001) Montane coffee (0.04) Montane cloud forest (0.10) Montane coffee (0.04) Lowland secondary forest (0.09) Montane coffee (0.04) Lowland secondary forest (0.02) Montane pine plantation (0.09) Montane cloud forest (0.07) Montane pine plantation (0.08) Wet limestone forest (0.001) Lowland coffee (0.020) Dry limestone forest (0.018) Wet limestone forest (0.09) Lowland coffee (0.02) Dry limestone forest (0.018) Greater Antillean Bullfinch Montane pine plantation (0.09) White-chinned Thrush Lowland coffee (0.04) Blue Mountain Vireo Jamaican Oriole Montane cloud forest (0.06) Dry limestone ruinate (0.04) T^BI,œ 4. Continued. Location of decline (P) Location of increase (P) Lowland coffee (0.02) Lowland pasture (0.006) Dry limestone ruinate (0.04) PrairieWarbler Montane cloud forest (0.09) Mangroves (0.003) Jamaican Woodpecker Mangrove Cuckoo Dusky-capped Flycatcher Jamaican Euphonia Dry limestone forest (0.05) Mangroves (0.08) Wet limestone forest (0.09) Lowland secondary forest (0.001) Dry limestone ruinate (0.08) Dry limestone forest (0.04) Lowland secondary forest (0.05) Jamaican White-eyed Vireo Dry limestone forest (0.03) Yellow Warbler Northern Parula Yellow-faced Grassquit Mangroves (0.001) Lowland pasture (0.08) Dry limestone ruinate (0.04) vore, fruit/seedeater, resident insectivore, migrant insectivore), and the number of pop- ulations that decreased was compared to those that either increased or showed no change. We found a significant interaction between diet type and population trend in montane habitats (G = 10.69, df = 1, P = 0.01), indicating that nectar- ivores and fruit/seedeaters were more likely to have declining populations than insectivores in montane habitats. For instance, declines in montane habitats occurred in 80% of the nec- tarivore populations (n = 10), 83% of the fruit/ seedeater populations (n = 23), 50% of the res- ident insectivore populations (n = 22), and 44% of the migrant insectivore populations (n = 23). Even with the exclusion of nonnative montane habitats (pine and coffee), the pattern of pop- ulation decline and diet remained. For example, in montane cloud forest 67% of the nectarivore populations (n = 3), 90% of the fruit/seedeater populations (n = 10), 50% of the resident in- sectivore populations (n = 10), and 29% of the migrant insectivore populations (n = 7) de- clined. Population decline in these montane habitats may be related to foraging height within the habitat. Therefore, we used the observations of Lack (1976) to classify the montane species into those confined to the forest understory and those that are primarily canopy dwellers. It was dif- ficult to find canopy-only species, because many canopy dwellers frequently descend to the un- derstory via gaps and edges (e.g. Wunderle et al. 1987). Our analysis of population trends based on foraging location indicated that 60% of mon- tane populations confined to the understory (n = 20) had declining population trends. The re- maining 40% had trends indicating increase or no change. The same pattern (i.e. 60% declin- ing, 40% increase/no change) was also found for canopy dwellers (n = 58), suggesting that the pattern of population decline in montane habitats was unrelated to foraging height. We examined the possibility that the ob- served increases in the number of individuals in two lowland habitats (wet limestone forest and mangroves) after the hurricane might be attributed to an influx of montane nectarivores and fruit/seedeaters. The number of lowland populations that changed (decrease/no change versus increase only) was tallied for species in different diet categories. Fifty percent of the lowland nectarivore populations increased, 54% of the fruit/seedeater populations increased, and both the resident (67%) and migrant (47%) in- sectivores increased. No significant interaction occurred between diet type and population trend in the two lowland habitats (G = 0.13, df = 3, P = 0.99). Thus, nectarivores and fruit/ seedeaters were not any more likely than in- sectivores to have increasing populations after the hurricane in the two lowland habitats. Increases in bird populations in the two low- land habitats after the hurricane might also be attributed to increased detectability of canopy species that were foraging at ground level. We found that 62% of the canopy/understory pop- ulations (n = 47) had increased in contrast to 38% of the understory populations (n = 16). Although this trend was suggestive, it was not possible to test statistically because of the con- founding effect of diet on foraging location. For instance, 56% of the species relying upon plants for food (n = 27) were canopy species, in con- trast to 80% of the insectivores (n = 40), a sig- nificant difference (G = 4.50, df = 1, P < 0.05). DISCUSSION Altitude and diet.--The hurricane's impact on bird populations was much less severe in low- land habitats than in the montane habitats. Al- though the storm's intensity was probably greatest in the uplands, damage to trunks and branches of trees in some lowland habitats (e.g. mangroves and wet limestone forest) was ac- tually greater than in the montane cloud forest. Furthermore, many of the lowland habitats (wet limestone forest, lowland secondary forest, mangroves, lowland pasture, dry limestone ru- inate) reportedly had been extensively defoli- ated (including loss of flowers and fruit). How- ever, a major difference between lowland and montane vegetation was the rapidity with which the vegetation recovered. For example, most of the lowland sites had reportedly refoliated within several weeks, yet montane trees were still defoliated and showed the greatest differ- ences between pre- and posthurricane canopy profiles four months later. Even after 10 months, many of the trees at high altitude were just beginning to produce new leaves (Varty 1991). Slower recovery of montane vegetation also was reported in Puerto Rico after hurricanes San Felipe (Bates 1930) and Hugo (pers. observ.), and in Dominica following Hurricane David (Lugo et al. 1983). Slow rates of recovery from disturbance may be typical of montane cloud forests (Weaver 1986). Our findings that population declines in montane habitats were related to diet suggest that the greatest impact of Hurricane Gilbert on Jamaica's montane bird populations was from indirect effects that occurred after its passage, rather than from direct exposure to the storm. It seems unlikely that resident insectivores in the mountains were somehow less vulnerable to the hurricane's high winds and rains than most fruit / seedeaters and nectarivores residing in the same place. It is possible that humming- birds, because of their small mass and high met- abolic rate, suffered considerable mortality from exposure to the hurricane. Indeed, Streamer- tails (6 g) decreased significantly in more hab- itats (n = 4) than any other species, but the larger nectarivorous Bananaquits (11 g) and Or- angequits (16 g) also declined significantly in some habitats. Moreover, the small (6.4-g) in- sectivorous Jamaican Tody showed no popula- tion changes after the storm, despite its pre- hurricane abundance. These findings are more consistent with a posthurricane-food-stress ex- planation than with the possibility of differ- ential mortality arising from different vulner- abilities to hurricane exposure. Furthermore, most migrants were probably absent when Gil- bert struck on 12 September (Lack and Lack 1972), yet both resident and migrant insecti- vores displayed similar population declines (50% and 44%, respectively). This also indicates that the indirect effects of Gilbert on terrestrial birds were greater than the direct impact of the storm. Gilbert's most important direct effect apparent- ly was on the vegetation, which affected bird populations secondarily by limiting food after the storm's passage. In the montane region the degree of damage inflicted by the hurricane on bird populations varied among the different food-resource types. The winds which defoliated most habitats un- doubtedly also stripped flowers and fruits from plants, as reported in other hurricanes (Bates 1930; N. F. R. Snyder and H. A. Snyder, unpubl. 1979 report; Wunderle, unpubl. data). Indeed, flower and fruit abundance still appeared low in the montane sites at the time of our surveys, four months after the hurricane. Arthropod populations were probably reduced by the storm, but because of their shorter life cycles and rapid reproductive rates, they may have recovered more quickly than the vegetation. In addition, portions of some insect populations (e.g. Diptera, Lepidoptera, Coleoptera) likely survived the storm as larvae or pupae in rela- tively protected sites in the soil, in leaf litter, and under bark (Wolcott 1932; B. Freeman, pers. comm.). Also, some insect populations might increase after the storm due to decreased pre- dation and increases in their resources (J. Tor- res, pers. comm.). Thus, a reliance on a food source with a high diversity and turnover rate (i.e. arthropods) could buffer montane insectiv- orous birds from hurricane-induced food short- ages. The decline in montane nectarivores and fruit/seedeaters could have resulted from movement out of the montane habitats into oth- er less damaged sites. As foragers on patchily distributed and temporarily available food re- sources, tropical nectarivores and frugivores range widely and often migrate seasonally and altitudinally in some tropical regions (e.g. Stiles 1983). Tropical insectivores, which feed on more predictable resources, tend to wander less than nectarivores, frugivores and seedeaters. For in- stance, under drought conditions in the Carib- bean, insectivorous birds tend to have the most stable populations while nectarivores and fruit/ seedeaters are most likely to decline (Faaborg et al. 1984). Presumably, some of these declines result from migration or nomadism. The reli- ance on widely distributed and temporary food resources could preadapt nectarivores and fruit/ seedeaters to respond to localized hurricane- induced food shortages through emigration. Despite potentially higher levels of flowering and fruiting in the lowlands, there was no de- tectable overall influx of montane nectarivores and fruit/seedeaters into the lowlands. Our in- ability to detect an increase in these foraging types in the lowlands may have resulted from surveys in the wrong lowland habitats or lo- cations, or the inability of point counts to ac- curately detect relatively small increases (5- 10%). The montane sites were in east-central Jamaica, whereas most of the lowland sites were in western Jamaica. Furthermore, montane hab- itats above 1,000 m elevation comprise only 4- 6% of the total land area in Jamaica, and it would be difficult to detect an influx from the rela- tively small montane populations into such an extensive lowland area. However, some species decreased in some habitats and increased in oth- ers. This implies localized movements, where birds moved from a damaged to a less-damaged site. For example, the Bananaquits probably em- igrated from montane habitats (significant de- clines in two habitats) and into lowland habitats (significant increases in three habitats). Bana- naquits feed on nectar from a high diversity of flowers (Snow and Snow 1971, Lack 1976) and may be very sensitive to flower abundance, as evidenced by drastic declines in dry forest un- der drought conditions (Faaborg et al. 1984). The White-chinned Thrush, a widespread fru- givore, also fit a pattern consistent with inter- habitat movement. They disappeared from montane coffee and lowland secondary forest, and increased in lowland coffee. Local inter- habitat movements within the lowlands also occurred. For example, the Prairie Warbler, an overwintering migrant insectivore, decreased significantly in three lowland habitats and in- creased markedly in mangroves. Prairie War- biers appeared after the hurricane in the middle of the defoliated montane cloud forest, a habitat in which they had not been encountered pre- viously. Several other species had posthurri- cane population shifts consistent with an in- terhabitat movement pattern, though the differences were not as marked. Foraging substrates and cover.--Whereas post- hurricane food loss explains most observed population declines, factors such as changes in foliage profiles and vegetation structure could affect certain species. For example, loss of high- canopy foraging substrates undoubtedly ex- plains the absence of Black-and-white Warblers from montane pine plantations and their de- cline in lowland coffee plantations. The over- story trees in both of these habitats suffered substantial loss of branches and twigs, where this migrant warbler normally gleans insects. Extensive damage to mimosaceous trees in the overstory of lowland coffee, lowland pasture, and dry limestone ruinate scrub probably con- tributed to the decline of overwintering mi- grant Prairie Warblers, which commonly glean insects from their leaves (Lack and Lack 1972). Many forest-floor species are accustomed to for- aging in a dark forest understory and are likely to leave areas with reduced canopy cover. Such a response was suggested by the declining post- hurricane Ovenbird populations in wet lime- stone and lowland coffee. Ovenbirds are insec- tivorous migrants that normally feed on the forest floor only where a canopy is found over- head (Lack and Lack 1972). Thus, structural damage to vegetation produced by hurricanes could eliminate foraging substrates, as well as the vegetation characteristics used by some spe- cies as proximal cues for habitat selection. Hurricane-induced changes in foliage distri- butions could also disrupt the stratification of foraging normally associated with different fo- liage layers (e.g. MacArthur et al. 1966). This disruption should be most apparent in forests of tall stature in which prehurricane separation of canopy and understory foraging zones was most distinct. Such was the case in the aftermath of Hurricane Hugo in Puerto Rico, where can- opy destruction caused many former canopy dwellers to forage at ground level (Wunderle unpubl. data). Increased numbers of individu- als in Jamaican mangroves and wet limestone forest may have resulted from displacement of canopy dwellers, although we were unable to test this hypothesis statistically. Furthermore, frequent hurricanes may, over the long term, make it difficult for Caribbean forest-dwelling species to segregate on the basis of foraging height (J. Faaborg, pers. comm.). This may ac- count for the findings of MacArthur et al. (1966), in which Puerto Rican species appeared to rec- ognize two vertical layers in the forest while those in Panama recognized four layers. The tremendous loss of overstory canopy and the overall reduction of foliage in some patches of mangroves may account for the apparent posthurricane increase in Yellow Warblers. Al- though this resident warbler will sometimes forage in dry scrub bordering mangroves, we found them only in mangroves. Yellow War- biers have one of the narrowest habitat breadths of any Jamaican species (Lack 1976), and their "reluctance" to use other habitats may have compressed their populations into the remain- ing mangrove fragments. Density increases in remnant mangrove fragments also could ex- plain the increased number of aggressive in- teractions and vocalizations detected after the hurricane. Thus, some species with narrow hab- itat preferences may respond to hurricane dam- aged habitats by remaining at higher densities in remnants of their original habitat rather than by moving into other nearby habitats. The con- sequences of remaining in fragments of original habitat, at higher than normal densities, as op- posed to moving into other habitat types are unknown. However, in the case of Yellow War- biers, the high primary productivity of man- groves (Lugo and Snedaker 1974) may have al- lowed a rapid recovery of insects, their primary food source. This may have allowed the war- biers to remain in their traditional habitat de- spite substantial vegetation damage. Some characteristics of Caribbean birdlife.--It is reasonable to expect that species restricted to a single habitat type are most at risk of extinction from habitat damage by hurricanes. Our results indicate that this is even more likely for nec- tarivores and fruit/seedeaters restricted to high montane forests, particularly when undamaged lowland habitats are unavailable. This could oc- cur on mountainous islands with a limited coastal plain, or on islands in which lowland vegetation has been destroyed for agriculture. This was the situation on St. Kitts, where all but the mountain summit was planted in sugar cane in the late 19th century (Beard 1949). The Puerto Rican Bullfinch (Loxigilla portorecensis grandis), a fruit/seedeater, was abundant on that island in the late 1800s, but it was confined to high-elevation forest on one mountain. In 1899, two hurricanes hit the island and may have reduced the bullfinch population to a level from which it could not recover (Raffaele 1977). After the hurricanes, the bullfinches were probably unable to find food in the sugar-cane dominated lowlands and, as our study suggests, it is likely that the montane fruit and seed crops were de- stroyed. Although the population persisted un- til at least 1929 (Olson 1984), it is likely that it was no longer viable. Thus, in this instance, a combination of hurricane destruction of a food resource in an area with slow recovery and an absence of alternative habitats or food sources may have contributed to extinction of a bird species. Perhaps montane nectarivores and fruit/ seedeaters in the Caribbean face higher risks of hurricane-induced extinctions, particularly as more lowland forests are destroyed for agri- culture. If the slower recovery of hurricane-damaged montane vegetation places stress on montane bird populations for longer periods than low- land populations, then hurricanes may contrib- ute to some of the commonly observed patterns of avian distribution in the Caribbean. For ex- ample, throughout the Caribbean, montane for- ests support fewer bird species than lowland forests (e.g. Kepler and Kepler 1970, Ricklefs and Cox 1972, Lack 1976). Also, more endemic species are found in montane forests than in the lowland forests of the Caribbean (Ricklefs and Cox 1972, Lack 1976). Hurricane-related ex- tinctions of one or several disjunct montane populations may contribute to endemism among montane species. In addition, island birds tend to use a wider range of habitats than continental species (e.g. Crowell 1962, MacArthur et aL 1966), and this trait appears to favor posthur- ricane survival. Thus, hurricanes could contrib- ute to these avifaunal characteristics on Carib- bean islands by reducing the number of species surviving in montane habitats, increasing en- demism and extinction rates of montane spe- cies, and favoring the survival of species ca- pable of using a broad range of habitats. ACKNOWLEDGMENTS Ariel Diaz Perez and Luis E. Migenis Lopez pro- vided excellent field assistance and suffered a variety of indignities with good humor. Marcel Anderson and Elaine Foster of the Natural Resource Conser- vation Division (Ministry of Agriculture) and Roy Jones of the Department of Forestry kindly provided permits and assistance. Erroil Ziadie and R. Campbell provided assistance and permission to work on their land. Advice was provided by Alexander Cruz, Au- drey Downer, Brian Freeman, Richard Holmes, Chan- dler Robbins, Robert Ross, Ann Haynes-Sutton, Rob- ert Sutton, and Jeremey Woodley, and members of the Gosse Bird Club. The manuscript benefited from the comments of Robert Askins, John Faaborg, James Lynch, Fred Scatena, Noel Snyder, Juan Torres, Peter Weaver, James Wiley, Michael Willig, and Glen Woolfenden. Financial support was provided by the World Wildlife Fund-U.S., University of Puerto Rico, and the National Science Foundation (BSR-8811902) through the Luquillo Long-term Ecological Research program. LITERATURE CITED ANONYMOUS. ]987. Jamaica: Country environmental profile. Government of Jamaica, Ministry of Ag- riculture, Natural Resources Conservation Divi- sion, and Ralph M. Field Associates, Kingston, Jamaica. ASKINS, R. A., AND D. N. EWERT. 1991. Impact of Hurricane Hugo on bird populations in Virgin Islands National Park. Biotropica 23:481-487. BATES, C.G. 1930. Hurricane damage to Porto Rican forests. J. For. 28:772-774. BEARD, J. S. 1949. The natural vegetation of the windward and leeward islands. Clarendon Press, Oxford. BELLINGHAM, P. J., V. KAPOS, N. VARTY, J. R. HEALEY, E. V. J. TANNER, D. L. KELLY, J. W. DALLING, L. S. BURNS, D. LEE, AND G. SIDRAK. 1992. Catastroph- ic disturbance need not cause high mortality: The effects of a major hurricane on forests in Jamaica. J. Trop. Ecol. In press. CROWELL, K.L. 1962. Reduced interspecific compe- tition among the birds of Bermuda. Ecology 42: 75-88. ENGSTROM, R. T., AND G. W. EVANS. 1990. Hurricane damage to Red-cockaded Woodpecker (Picoides borealis) cavity trees. Auk 107:608-610. FAABORG, J. 1985. Ecological constraints on West In- dian bird distributions. Ornithol. Monogr. 36:621- 653. FAABORG, J., W. J. ARENDT, AND M. S. KAISER. 1984. Rainfall correlates of bird population fluctuations in a Puerto Rican dry forest: A nine year study. Wilson Bull. 96:575-593. GOSSE BIRD CLUB. 1989. Post-hurricane observations. Gosse Bird Club Broadsheet 52:8-9. GUPTA, A. 1975. Stream characteristics in eastern Jamaica, an environment of seasonal flow and large floods. Am. J. Science 275:825-847. GUPTA, A. 1988. Large floods as geomorphic events in the humid tropics. Pages 301-305 in Flood geo- morphology (V. R. Baker, R. C. Kochel, and P. C. Ballon, Eds.). John Wiley and Sons, New York. HAYNES-SUTTON, A.M. 1988. Hurricane Gilbert strikes Jamaica's unique birdlife. World Bird- watch 10:3-4. HUTTO, R. L., S. M. PLETSCHET, AND P. HENDRICKS. 1986. A fixed-radius point count method for nonbreeding and breeding season use. Auk 103: 593-602. JEGGO, D. F., AND K. M. TAYNTON. 1980. The effects of Hurricane Allen on the status of the St. Lucia Parrot Amazona versicolor. Dodo (journal of Jersey Wildlife Preservation Trust) 17:11-18. KAPOS, V. 1986. Dry limestone forests of Jamaica. Pages 49-48 in Forests of Jamaica (D. A. Thomp- son, P. K. Bretting, and M. Humphreys, Eds.). Jamaican Society of Scientists and Technologists, Kingston, Jamaica. KENNEDY, R.J. 1970. Direct effects of rain on birds: A review. Br. Birds 63:401-414. KEPLER, C. B., AND A. K. KEPLER. 1970. Preliminary comparison of bird species diversity and density in Luquillo and Guanica forests. Pages E183-191 in A tropical rain forest (H. T. Odum and R. F. Pigeon, Eds.). National Technical Information Service, Springfield, Virginia. LACK, D. 1976. Island biology illustrated by the land birds of Jamaica. Univ. California Press, Berkeley. LACK, D., AND P. LACK. 1972. Wintering warblers in Jamaica. Living Bird 11:129-153. LAWRENCE, M. B., AND J. M. GROSS. 1988. Annual summaries: Atlantic hurricane season of 1988. Monthly Weather Rev. 117:2248-2259. LUGO, A. E., M. APPLEFIELD, D. J. POOL, AND R. B. McDoNALD. 1983. The impact of Hurricane Da- vid on the forests of Dominica. Can. J. For. Res. 13:201-211. LUGO, A. E., AND S.C. SNEDAKER. 1974. The ecology of mangroves. Annu. Rev. Ecol. Syst. 5:39-64. LYNCH, J. F. 1991a. Distribution of overwintering Nearctic migrants in the Yucatan Peninsula. II: Use of native and human-modified vegetation. In Ecology and conservation of Neotropical mi- grant land birds (J. Hagan and D. W. Johnston, Eds.). Smithsonian Institution Press, Washington D.C. In press. LYNCH, J.F. 1991b. The impact of Hurricane Gilbert on a dry tropical forest in the northeastern Yu- catan Peninsula: Birds. Biotropica 23:488-496. MACARTHUR, R. H., H. RECHER, AND M. CODY. 1966. On the relation between habitat selection and species diversity. Am. Nat. 100:319-332. ODUM, H.T. 1970. Rainforest structure and mineral- cycling homeostasis. Pages H3-52 in A tropical rainforest (H. T. Odum and R. F. Pigeon, Eds.). National Technical Information Services, Spring- field, Virginia. OLSON, S. 1984. The last St. Kitts Bullfinch, Loxigilla portoricensis grandis (Emberizidae) and the extinc- tion of its race. Bull. Br. Ornithol. Club 104:121- 124. PROCTOR, G. R. 1986. Cockpit country and its veg- etation. Pages 43-48 in Forests of Jamaica (D. A. Thompson, P. K. Bretting, and M. Humphreys, Eds.). Jamaican Society of Scientists and Tech- nologists, Kingston, Jamaica. RAFFAELE, H.A. 1977. Comments on the extinction of Loxigilla portoricensis grandis in St. Kitts, Lesser Antilles. Condor 79:389-390. RICKLEFS, R. E., AND G. W. Cox. 1972. Taxon cycles in the West Indian avifauna. Am. Nat. 106:195- 219. SCHEMSKE, D. W., AND N. BROKAW. 1981. Treefalls and the distribution of understory birds in a trop- ical forest. Ecology 62:938-945. SNOW, B. K., AND D. W. SNOW. 1971. The feeding ecology of tanagers and honeycreepers in Trin- idad. Auk 88:291-322. SOCAL, R. R., ^ND F. J. ROHLF. 1981. Biometry, 2nd ed. W. H. Freeman, New York. STILES, F.G. 1983. Introduction to birds of Costa Rica. Pages 502-530 in Costa Rican natural history (D. H. Janzen, Ed.). Univ. Chicago Press, Chicago. TANNER, E. V.J. 1986. Forests of the Blue Mountains and Port Royal Mountains of Jamaica. Pages 15- 30 in Forests of Jamaica (D. A. Thompson, P. K. Bretting, and M. Humphreys, Eds.). Jamaican So- ciety of Scientists and Technologists, Kingston, Jamaica. TERBORGH, J., AND J. FAABORG. 1973. Turnover and ecological release in the avifauna of Mona Island, Puerto Rico. Auk 90:759-779. THOMPSON, L. 1900. Notes on the St. Vincent Parrot. Aviculture Mag. 6:144-149. THURBER, W. A. 1980. Hurricane Fifi and the 1974 autumn migration in E1 Salvador. Condor 82:212- 218. V^RTY, N. 1991. The status and conservation of Ja- maica's threatened and endemic forest avifauna and their habitats following Hurricane Gilbert. Bird Conserv. Int. 1:135-151. WADSWORTH, F. H., AND G. W. ENGLERTH. 1959. Ef- fects of the 1956 hurricane on forests in Puerto Rico. Caribb. For. 20:38-51. W^IDE, R. B. 1991. The effect of Hurricane Hugo on bird populations in the Luquillo Experimental Forest. Biotropica 23:475-480. WEAVER, P.L. 1986. Hurricane damage and recovery in the montane forests of the Luquillo Moun- tains. Caribb. J. Science 22:53-70. WOLCOTT, G.N. 1932. The effect of the hurricane of San Cyprian on insects in Puerto Rico. Insect Pest Surv. Bull. 12:409-410. WUNDERLE, J. M., JR., A. DIAZ, I. VELAZQUEZ, AND R. SCHARRON. 1987. Forest openings and the dis- tribution of understory birds in a Puerto Rican rainforest. Wilson Bull. 99:22-37. APPENDIX 1. Site descriptions. All sites have been described previously in detail (Lack 1976, Kapos 1986, Proctor 1986, Tanner 1986, Anonymous 1987). Montaae cloud forest.--Very Wet Ridge Forest (Tanner I983) on Mr. Horeb and Mr. Oatley in Port Royal Mountains (I,350 m elevation; mean annual rainfall 1,900 mm). Nets set along Shelter Trail in Hardwar Gap Forest Reserve, and vegetation measured in two plots near Shelter Trail and two near Mr. Horeb Trail. Overstory (up to 20 m tall) dom- inated by Cyrilla racemifiora, Podocarpus urbanii, Hedyosemum arborescens, Mecranium purpurascens, Clethra occidentalis, Alchornea latifolia, and Den dropanax pendulus, with tree ferns (Cyathea spp.) understory. Montane coffee.--Shadeless coffee (3-4 m height) with an understory of grasses and forbs on Mr. Horeb near Hardwar Gap (ca. 1,300 m) in Port Royal Mountains. Montane pine plantations.--Small plantations sampled between Har- dwar Gap and Irish Town on south side of Port Royal Mountains (ca. 1,000-1,300 m elevation; mean annual rainfall 1,900 mm). Even-aged Pinus caribaea plantations (canopy 12-25 m) had a few Eucalyptus, with an understory of ginger, bamboo, and ferns, as sampled in vegetation plots at Woodside Estate in Greenwich. Wet limestone forest.--Forest sampled at Windsor (ca. 300-600 m ele- vation; mean annual rainfall 2,000-2,500 mm) in limestone karst region known as Cockpit Country. Vegetation plots located near mist nets in floodplain forest (property of Mr. Campbell), and in saddle next to Windsor Cave. Canopy height reached 15-20 m in uplands and 25-35 m in lowlands. Cockpit Country has list of 46 tree species, including many in the Sapotaceae and Lauraceae with prominent endemics such as Terrainalia arbuscula and Manilkara excisa. Lowland coffee.--Small plantations with a shade overstory (Mimosa- ceae; 15-20 m canopy) and coffee understory (2-3 m) at Windsor. Lowland secondary forest.--Mist netted and sampled vegetation at site APPENDIX 1, Continued. near Negril Cabins in Negril (<10 m elevation; mean annual rainfall 1,900-2,540 mm); point counts made here and to north along main road. Canopy heights of 6-I0 m for dominant Pithecellobium sp. in these disturbed second-growth forests. Mangroves.--Nets and two vegetation plots located in mangroves near Negril Municipal dump, and two additional vegetation plots in Green Island. Point counts made in both areas. Overstory (canopy 12-15 m) dominated by Avicennia nitida and Rhizophora mangle, while Hibiscus tiliaceus and coconut palms common on higher ground. Lowland pasture.--Point counts made in pastures east of Green Island (near sea level), but no vegetation or mist-net samples obtained. Pasture contained scattered large leguminous trees (Pithecellobium sp. and Hae- matoxylum sp.). Diversity of large (10-I5 m tall) leguminous trees in fence rows bordering pastures. Dry limestone ruinate.--Censuses taken at Discovery Bay (10-30 m elevation; mean annual rainfall 1,000-I,500 mm). Censuses and vege- tation plots located south of Columbus Park and at University of the West Indies Marine Biology Station. Dry scrub vegetation disturbed by cutting (canopy 6-10 m). Common tree species include Bursera simaruba, Metopium brownil, Fagara spinosa, Coccoloba sp., and Tabebuia sp.; a colum- nar cactus (Stenocereus hystrix) also common. Dry limestone forest.--Relatively undisturbed dry forest sampled at Portland Ridge (100-150 m elevation; mean annual rainfall 1,270-1,900 ram). Mist nets and vegetation plots located in well-developed forest (canopy 10-16 m), but canopy height lower in many of point-count areas on ridges. Flora similar but more diverse than at Discovery Bay. Largest trees (up to 80 cm dbh) were Bursera sirearuba, while Tecoma stans, Thrinax parvifiora, and Bauhinia divaricata present in understory. APPENDIX 2. Mean number of individuals per point count before (upper row) and after (lower row) Hurricane Gilbert for 10 habitats in Jamaica. Acronyms for habitats and number of point counts in each as follows: MCF, montane cloud forest (n = 43); MPP, montane pine plantations (n = 30); MC, montane coffee (n = 33); WLF, wet limestone forest (n = 30); LC, lowland coffee (n = 28); LSF, lowland secondary forest (n = 30); Ma, mangroves (n = 30); LPa, lowland pasture (n = 30); DLR, dry limestone ruinate (n = 30); DLF, dry limestone forest (n = 30). Letters for diet indicate: F, fruit or seed; N, nectar; I, insects. Significance level for before-and-after comparisons indicated by footnotes. Habitats Species Diet MCF MPP MC WLF LC LSF Ma LPa DLR DLF White-crowned Pigeon F 0.00 0.00 0.00 0.00 0.04 0.00 0.07 0.00 0.00 0.00 (Columba leucocephala) 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 0.00 0.00 Red-necked Pigeon F 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 (C. squamosa) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Ring-tailed Pigeon F 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 (C. caribaea) 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 Zenaida Dove F 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 (Zenaida aurita) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 White-winged Dove F 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 (Z. asiatica) 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 Common Ground Dove F 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.10 0.03 0.00 (Columbina passerina) 0.00 0.00 0.00 0.00 0.14 0.07 0.03 0.10 0.00 0.00 Caribbean Dove F 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.37 (Leptotila jamaicensis) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.07 b Ruddy Quail Dove F 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 (Geotrygon montana) 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 Crested Quail Dove F 0.02 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 (G. versicolor) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Olive-throated Patakeel F 0.00 0.00 0.00 0.30 0.00 0.07 0.07 0.13 0.30 0.00 (Aratinga nana) 0.00 0.00 0.00 0.33 0.07 0.20 0.10 0.10 0.00 0.00 APPENDIX 2. Continued. Species Diet MCF MPP Habitats MC WLF LC LSF Ma LPa DLR DLF Guiana Parrotlet F 0.00 0.00 0.00 (Forpus passerinus) 0.00 0.00 0.00 Mangrove Cuckoo I 0.00 0.00 0,00 (Coccyzus minor) 0.00 0.00 000 Chestnut-bellied Cuckoo I 0.02 0.00 0.00 (Hyetornis pluvialis) 0.00 0.00 0.00 Jamaican Lizard Cuckoo I 0.00 0.00 0.00 (Saurothera vetula) 0.00 0.00 0.00 Smooth-billed Ani I 0.00 0.00 0.00 (Crotophaga ani) 0.00 0.00 0.00 Jamaican Mango N 0.00 0.00 0.03 (Anthracothorax mango) 0.00 0.00 0.00 Streamertail N 0.77 0.67 0.24 (Trochilus polytmus) 0.15 f 0.17  0.00 a Vervain Hummingbird N 0.00 0.00 0.00 (Mellisuga minima) 0.00 0.00 0.00 Jamaican Tody I 0.09 0.03 0.06 (Todus todus) 0.11 0.10 0.09 Jamaican Woodpecker I 0.02 0.03 0.00 (Melanerpes radiolatus) 0.00 0.00 0.00 Jamaican Becard I 0.00 0.00 0.00 (Platypsaris niger) 0.00 0.00 0.00 Loggerhead Kingbird I 0.00 0.07 0.03 (Tyrannus caudifasciatus) 0.02 0.00 0.03 Stolid Flycatcher I 0.00 0.00 0.00 (Myiarchus stolidus) 0.00 0.00 0.00 Dusky-capped Flycatcher I 0.02 0.03 0.00 (M, barbirostris) 0,02 0.03 0.00 Rufous-tailed Flycatcher I 0.00 0.00 0.00 (M. validus) 0.00 0.03 0.00 Greater Antillean Pewee I 0.02 0.07 0.00 (Contopus caribaeus) 0.00 0.03 0.00 Greater Antillean Elaenia I 0.00 0.00 0.00 (Elaenia fallax) 0.02 0.00 0.00 Yellow-crowned Elaenia I 0.00 0.00 0.00 (Myiopagis cotta) 0.00 0.00 0.00 Gray Catbird F 0.00 0.00 0.00 (Dumatella carolinensis) 0.00 0.00 0.00 Northern Mockingbird F 0.00 0.00 0.03 (Mimus polyglottos) 0.00 0.00 0.00 Rufous-throated Solitaire F 0.02 0.07 0.00 (Myadestes genibarbis) 0.00 0.00 0.00 White-chinned Thrush F 0.07 0.00 0.12 (Turdus aurantius) 0.09 0.03 0.00  White-eyed Thrush F 0.02 0.00 0.00 (T. jamaicensis) 0.00 0.00 0.00 Yellow-faced Grassquit F 0.02 0.00 0,00 (Tiaris olivacea) 0.00 0.00 0.00 Black-faced Grassquit F 0.26 0.47 0.70 (T. bicolor) 0.04  0.10 c 0.18  Yellow-shouldered Grassquit F 0.00 0.00 0.00 (Loxipasser anoxanthus) 0.00 0.00 0.00 Greater Antillean Bullfinch F 0.16 0.03 0.12 (Loxigilla violacea) 0.07 0.20 b 0.00 c Stripe-headed Tanager F 0.51 0.30 0.03 (Spindalis zena) 0.30 b 0.07 0.00 Jamaican Euphonia F 0.02 0.03 0.00 (Euphonia jamaica) 0.02 0.00 0.00 Orangequit N 0.26 1.00 0.18 (Euneornis campestris) 0.21 0.03 f 0.00  0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.13 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.10 b 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.67 0.00 0.00 0.00 0.00 0.00 0.00 0.67 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.04 0.03 0.00 0.43 0.00 0.00 0.00 0.07 0.00 0.03 0.43 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.03 0.27 0.00 0.00 0.00 0.00 0.03 0.00 0.13 0.57 0.25 0.27 0.07 0.00 0.40 0.17 0.57 0.32 0.10 b 0.13 0.03 0.07  0.13 0.00 0.04 0.00 0.00 0,00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.23 0.25 0,00 0,00 0.00 0.00 0.03 0.47 0.18 0.07 0.00 0.00 0.00 0.07 0.23 0.04 0.20 0.07 0.03 0.00 0.07 0.43 0.00 0.27 0.13 0.03 0.00 0.30 c 0.13 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.14 0.13 0.03 0.47 0.13 0.00 0.03 0.00  0.13 0.13 0.37 0.17 0.03 0.07 0.00 0.03 0.00 0.07 0.03 0.20 0.10 0.00 0.00 0.00 0.00 0.00 0.33 0.03 0.00 0,03 0.07 0.00 0.00 0.00 0.20 0.04 0.40  0.07 0.03 0.10 b 0.13 c 0.07 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.10 0,00 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.04 0.00 0,00 0.00 0,00 0.00 0.07 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.37 0.00 0.00 0.00 0.00 0.00 0.00 0.10 ' 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.03 0.37 0.07 0.07 0.00 0.00 0.10 0.10 0.47 0.00 0.23 a 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.10 0.04 0.17 0.03 0.03 0.00 0.00 0.20 0.25  0.00  0.00 0.00 0.07 0.00 0.00 0.00 0.00 0.00 0.00 0,00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.37 0.03 0.00 0.00 0.03 0.10 0.00 0.30 0.33  0.00 0.00 0.00 0.00 0.00 0.07 0.00 0.00 0.00 0.00 0.00 0.00 0.43' 0.03 0.00 0.00 0.04 0.07 0.00 0.00 0.17 0.07 0.00 0.00 0.00 0.00 0.00 0.00  0.03 0.10 0.00 0.20 0.00 0.00 0.07 0.53 0.07 0.00 0.03 b 0.00 0.00 0.43' 0.43 0.07 0.00 0,00 0,00 0.00 0.23 0.00 0.07 0.29' 0.00 0.00 0.00 0.47 0.10  0.00 0.07 0.13 0.00 0.07 0.00 0.00 0.00 0.00 0.33  0.00 0.03 0.03 0.00 0.03 0.07 0.00 0.00 0.00 0.10 0.00 0.20  0.00 0.03 0.00 0.00 0.03 0.00 APPENDIX 2. Continued. Species Diet MCF MPP Habitats MC WLF LC LSF Ma LPa DLR DLF Tennessee Warbler I 0.00 0.00 0.00 (Vermivora peregrina) 0.00 0.00 0.00 Northern Parula I 0.00 0.03 0.00 (Parula americana) 0.00 0.00 0.03 Yellow Warbler I 0.00 0.00 0.00 (Dendroica petechia) 0.00 0.00 0.00 Chestnut-sided Warbler I 0.00 0.00 0.00 (D. pensylvanica) 0.00 0.00 0.00 Magnolia Warbler I 0.00 0.00 0.00 (D. magnolia) 0.00 0.00 0.00 Cape May Warbler N 0.00 0.00 O00 (D. tigrlna) 0.00 0.03 0.00 Black-throated Blue Warbler I 0.30 0.47 0.49 (D. caerulescens) 0.30 0.37 0.21' Yellow-rumped Warbler I 0.00 0.00 0.00 (D. coronata) 0.00 0.00 0.00 Black-throated Green Warbler I 0.00 0.00 0.00 (D, virens) 0.00 0.00 0.00 Yellow-throated Warbler I 0.00 0.00 0.00 (D. dominica) 0.00 0.07 0.00 Prairie Warbler I 0.00 0.00 0.09 (D. discolor) 0.07 b 0.07 0.18 Palm Warbler I 0.00 0.00 0.00 (D. palmarum) 0.00 0.00 0.00 Arrow-headed Warbler I 0.07 0.13 0.91 (D, pharetra) 0.15 0.20 0.00 Black-and-white Warbler I 0.07 0.17 0.00 (Mniotilta varia) 0.09 0.00' 0.03 American Redstart I 0.21 0.07 0.09 (Setophaga ruticilla) 0.15 0.03 0.12 Worm-eating Warbler I 0.00 0.03 0.00 (Helmitheros vermivorus) 0.02 0.00 0.00 Ovenbird I 0.05 0.03 0.00 (Seiurus aurocapillus) 0.00 0.00 0.00 Northern Waterthrush I 0.00 0.00 0.00 (S. noveboracensis) 0.00 0.00 0.00 Common Yellowthroat I 0.07 0.10 0.46 (Geothlypis trichas) 0.13 0.20 0.64 Hooded Warbler I 0.00 0.00 0.00 (Wilsonia citrina) 0.00 0.00 0.00 Bananaquit N 0.07 0.77 0.12 (Coereba fiaveola) 0.07 0.10  0.00' European Starling I 0.00 0.00 0.00 (Sturnus vulgaris) 0.00 0.00 0.00 Jamaican White-eyed Vireo I 0.00 0.00 0.00 (Vireo modestus) 0.00 0.00 0.00 Blue Mountain Vireo I 0.23 0.20 0.00 (V. osburni) 0.15  0.03  0.00 Jamaican Oriole I 0.07 0.10 0.03 (Icterus leucopteryx) 0.00 b 0.00  0.00 Jamaican Blackbird I 0.07 0.00 0.00 (Nesopsar nigerrimus) 0.00 0.00 0.00 Greater Antillean Grackle I 0.00 0.00 0.00 (Quiscalus niger) 0.00 0,00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0,07 0.10 0.20 0.00 0.00 0.03 0.07 0.07 0.13 0.27 0.10  0.03 0.00 0.00 0.00 0.00 0.33 0.00 0.00 0.00 0.00 0.00 0.00 1.17  0.00 0.00 0.00 0.00 0.11 0.00 0.00 0.03 0.03 0.00 0.00 0.00  0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.04 0.07 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.27 0.29 0.13 0.03 0.07 0.03 0.03 0.33 0.14 0.07 0.00 0.13 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 O00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.21 0.03 0.00 0.00 0.00 0.00 0.00 0.07  0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.18 0.07 0.03 0.30 0.17 0.07 O00 0.00 c 0.10 0.33 e 0.03 d 0.00 ' 0.03 0.00 0.04 0.00 0.00 0.27 0.03 0.00 0.00 0.00 0.00 0.03 0.50 0.07 0.00 0.10 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 0.00 0.00 0.00 0.00 0.00 0.10 0.29 0.07 0,40 0.13 0.03 0.10 0.13 0.04' 0.03 0.37 0.03 0.03 0.03 0.17 0.11 0.57 0.57 0.10 0.10 0.13 0.27 0.04 0.43 0.57 0.03 0.03 0.17 0.07 0.00 0.00 0.00 0.00 0.00 0.03 0.03 0.00 0.00 0.00 0.00 0.03 0.13 0.10 0.21 0.07 0.07 0.03 0.10 0.10 0.00 b 0.04 b 0.00 0.00 0.00 0.07 0.00 b 0.00 0.00 0.00 0.87 0.00 0.00 0.00 0.00 0.00 0.03 0.83 0.00 0.00 0.00 0.00 0.07 0.27 0.37 0.33 0.03 0.00 0.03 0.18 0.43 0.37 0.43 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 0.83 0.37 0.33 1.23 0.23 0.53  0.25' 0.67 0.20 0.10 0.93 0.57  0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.33 0.00 0.00 0.17 0.00 0.30 0.07 0.03 0.20 0.13 0.30 0.00 0.17 0.13 0.00 0.10 0.43  0.07 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.10 0.04 0.27 0.13 0.10 0.00 0.10 0.03 0.00 0.23 0.07 0.067 0.17 c 0.10 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.14 0.00 0.23 0.03 0.03 0.00 0.00 0.07 0.07 0.10 0.00 0.00 0.00 P < 0.15. bP < 0.10. cP < 0.05. d P < 0.01. eP < 0.005. P < 0.001.