Ground beetles ( Carabidae ) as seed predators

The consumption and preferences of polyphagous ground beetles (Coleoptera: Carabidae) for the seeds of herbaceous plants was determined. The seeds were stuck into plasticine in small tin trays and exposed to beetle predation on surface of the ground. In the laboratory the effect of carabid (species, satiation) and seed (species, size) on the intensity of seed predation was investigated. The consumption of the generally preferred Cirsium arvense seed by 23 species of common carabids increased with body size. Seed of Capsella bursa-pastoris was preferred by small carabids and their consumption rates were not related to their size. The average daily consumption of all the carabid species tested (0.33 mg seeds . mg body mass-1 . day-1) was essentially the same for both kinds of seed. Because of satiation the consumption of seed of C. arvense provided ad libitum to Pseudoophonus rufipes decreased over a period of 9 days to 1/3—1/4 of the initial consumption rate. Preferences of P. rufipes (body mass 29.6 mg) and Harpalus afifiinis (13.4 mg) for the seeds of 64 species of herbaceous plants were determined. The small H. afifiinis preferred smaller seed than the large P. rufipes. Predation of seed present on the ground in the field was studied in 1999—2000, at PrahaRuzyne (50°06' N 14°16'E). Seeds were placed in stands of different crops as in the laboratory experiments and vertebrate predation was excluded by wire mesh cages. Pitfall traps placed near the cages revealed that carabids were the only seed predators active in the area. Rates of removal of seed of 6 weed species varied with crop, season, seed and site. Average rate of removal in June-August was 2.5 seeds.day—1.tray—1 and was smaller before and after this period. The rates of removal increased with increasing activity den­ sity of the carabids and paucity of seed from naturally occurring weeds, which may have satiated the carabids. In stands of winter wheat, millet and soybeans there were significant differences in the rates of removal of the seed of 43 herbaceous species. The field preferences were correlated with those established in the laboratory. Predation of seed on the ground in arable fields can be as high as 1000 seeds.m—2.day—1 and may selectively influence the quantity of seed of particular herb species that enters the soil seed bank. Seed predation thus may be an effective component of weed control on arable land, particularly at low weed densities.

In the temperate zone probably the most important seed eaters are ground beetles (Carabidae: Coleoptera).In the Czech Republic this family is represented by more than 550 species (Hurka, 1996).The species characteristic of arable land are capable of withstanding agricultural prac tices (Luff, 1987;Holland & Luff, 2000), often disappear when the fields are abandoned (Van Dijk, 1987) and their distribution is patchy and temporarily unstable (Kinnunen & Tiainen, 1999;Thomas et al., 2001).
Carabid diet has been studied in the field and laboratory, and by stomach dissections (Davies, 1953;Skuhravy, 1959;Dawson, 1965).Ground beetles accept a range of animal prey but some of them also eat plant material (leaves, fruits and pollen) and fungi (Johnson & Cameron, 1969).A number of investigations (reviewed by Thiele, 1977;Hengeveld, 1980a;Luff, 1987;Lovei & Sunderland, 1996;Kromp, 1999;Toft & Bilde, 2002) have distinguished two categories of phytophagous ground beetles, those that are normally carnivorous but supplement their diets with the vegetative parts of plants, and those that mainly eat seeds (Brandmayr, 1990).
Despite the interest in this topic there is no comparative study of the preferences using many carabid and seed spe cies.We therefore made this study of the seed preferences and consumption under controlled and field conditions.
The assumption that body size of the beetles and size of the seed determine the rates of seed consumption and preferences of carabid species was tested in the laboratory.To test the role of seed and carabid body size we examined (i) the rate of consumption of the seed of 2 species in relation to the body mass and level of satiation of several carabid species, and (ii) preferences of 2 carabid species of different body size for seed of different mass.In the field the role of carabids as predators of shed seeds was studied in order to establish whether (iii) carabids remove a significant proportion of the seed on the surface of the soil, (iv) the quantity of seed removed varies with carabid abundance and season, and (v) the predation on different seed species is selective.

Collecting and offering seed to carabid
Seeds of 65 broadleaved herbaceous species, particularly weeds of agricultural crops, were used in the experiments (Appendix 1).The seeds were collected in 1996 at localities in Bohemia (Martinkova et al., 1997), air dried, and stored at 24-26°C and 40 % relative humidity until required.Appendages or perianths present on some seed at dispersal were removed.The quality of seed was thus "standardized" and removal experiments used "naked" seed similar to that present on the ground surface at the time of entering the soil seed bank.Seed mass was determined by weighing 100 oven dried "naked" seeds taken randomly from samples used in the experiments.Seed was exposed to carabid predation in small tin trays, 28 mm in diameter and 6 mm deep (6.2 cm2).The trays were filled with white plasticine (JOVI®, Barcelona) and the seeds inserted into the plasticine.The number of seeds exposed on a tray depended on the size of the seed.Each tray contained 15 (Arctium lappa, Bidens tripartita, Galium aparine) or 30 seeds (other species).The trays were placed on the ground so that the plasticine sur face was level with that of the soil.The seeds were thus acces sible to beetles walking across the surface.After exposure to predation the numbers of intact seeds in each tray were counted using a lens with a 10 times magnification.Missing seeds and those of which > 50% was consumed were considered eaten.

Laboratory experiments
Carabids.Laboratory experiments were performed with mature adults collected in the field at the time of their maximum activity, using pitfall traps.Before the start of an experiment beetles were stored in the dark for 1-4 days, in 0.5 l plastic bot tles filled with folded moist filter paper, at temperatures of 5-7°C.This prevented cannibalism and standardised the level of hunger of the experimental beetles.The beetles were not sexed because we were interested in estimating the predation rates in the field, which might be influenced by the sex ratios of the populations of the various carabids.All experimental indi viduals of a species were assumed to be of a typical size, with the average body length cited by Hurka (1996) and average dry body weight calculated using the formula of Jarosik (1989).To prevent repeated measure data each individual was used only once.
No choice experiments.Experiments were done in July-August 1999 and 2000.The temperature ranged between 25-27°C and the relative humidity inside the experimental vials was 100%.The photoperiod was natural (i.e.decreasing from c. 17.5L : 6.5D to c. 15L : 9D) but the vials were screened from direct sunlight.The experimental arena consisted of a cylin drical glass tube 10 cm in diameter and 10 cm high, covered with a glass lid.Each tube contained a 2 cm of sieved soil (mesh diameter 4 mm), which was dug from a depth of > 0.5m and did not contain any seeds.The soil was moistened with 15 ml tap water and a piece of moist cotton wool was provided as source of drinking water for the experimental animals.A tray of either Capsella bursa-pastoris or Cirsium arvense seed was placed in the center of an arena.One adult beetle was placed in an arena and its seed consumption recorded daily, for 3 days.When trays were emptied of their seeds they were replaced and the beetles thus provided with an abundant food supply during the whole experimental period.Twenty three species of carabid (Appendix 2), which were abundant in fields, were tested.The numbers of individuals of particular species used in the experiment were different because some species were scarce and/or laboratory space was limited.
To determine the effect of carabid satiation on seed consump tion the same experimental design as described above was used to record the consumption of C. arvense seeds by Pseudoophonus rufipes over a period of 9 days, August 11-20, 1999.Ten arenas were used in the laboratory and 10 in the field, placed at ground level and shaded from direct sunlight and rain by a wooden construction.Each arena contained one adult bee tle.Laboratory conditions were as cited above, the average field temperature over the period of the experiment was 20.2 ± 2.6°C, with a daily minimum of 11.4 ± 2.1°C and maximum of 27.8 ± 4.4°C at ground level.Table 1.Arrangement of the seeds in the laboratory experi ments.Species indicated by numbers and as listed in Appendix 1. Italics: "reference" species used alternatively in experiments with H. affinis and P.pubescens (see Material and Methods).

Field experiments
Experimental area.Field experiments on seed removal were conducted at Praha -Ruzyně (50°06'N 14°16'E) where the average monthly temperatures in April to October are 7. 9, 12.7, 16.3, 17.6, 16.8, 13.2 and 8.2°C, and the average monthly pre cipitation are 36, 57, 64, 70, 60, 32 and 33 mm, respectively.The experimental grounds were situated within a 1 km2 area located on a south facing slope (4% inclination), at an altitude of 330-360 m.The experimental area is a mosaic of production and experimental fields divided by ridges and hedges, with a uniform soil quality.The crop rotation in the production fields was irregular, including different small grain cereals, oilseed rape, peas and alfalfa.All crops were grown using the standard practices recommended in the Czech Republic.No insecticides were used throughout the period of investigation.
Presenting seed.The seed placed in the open was presented in trays identical to those used in the laboratory except for a 25 mm long nail attached to the convex side, which prevented hori zontal movement (Fig. 1A).The seeds were stuck in plasticine and in the same numbers as in the laboratory.The trays of seed were exposed in groups of six, each group was covered with a 18 x 18 x 9 cm wire mesh cage (mesh diameter 9 mm, wire 1 mm thick), whose sides were inserted into the soil to a depth of 4 cm (Fig. 1B).The roof of each cage was shielded by 18 x 18 cm plastic plate wrapped in aluminium foil, which protected the trays from rain and direct sunlight.These cages prevented the access of vertebrates.Each tray in a cage contained seed of a different species.
Seed removal rates.Seasonal and between crop variation in removal rates were studied by exposing groups of trays with a "standard" series of seed of six species (Table 1, Fig. 1C).Series of 4, 6 (mostly) or 8 cages were placed simultaneously in a crop and exposed to seed predation for different periods of time.Nine series were exposed between July 19 and September 14, 1999, in crops of winter wheat, winter rape, maize, millet and on abandoned field.Thirty two series were exposed between April 18 and October 11, 2000 in 2 stands of winter wheat, 1 of winter rape and 1 of soybeans.To prevent bias when calculating the daily consumption rates the exposure of cages was always terminated before all the seeds of the preferred spe cies were eaten.The length of exposure therefore varied between 3-20 days depending on the intensity of seed removal.
Activity density of carabids.The activity density of the carabids around cages where seed was exposed was determined using 1 or 2 pitfall traps placed within 1 m of each cage.The pitfall traps were plastic cups, 7 cm in diameter (38.5 cm2 outlet area) and 8 cm deep.The cups were placed in the soil, with the rim at the soil surface, and screened from rain and direct sun light by a dish wrapped in aluminium foil.No bait was used.A few lumps of soil at the bottom of the cups provided shelter for the trapped arthropods.The traps were emptied at 2 or 3 day (weekends) intervals.The beetles were determined to species, Fig. 1.Schematic representation of the method of offering seed to carabids in the field.A -tin tray filled with plasticine into which seeds were inserted; B -a section through a wire mesh cage inserted into the soil, protected by a shield, and with trays inserted into the soil; C -position of the trays in a field cage; D -arrangement of cages in "Experiment 1" and "Experi ment 2" to test carabid preferences (each cage is labelled with the number of the series i -iv of seeds they contain, see Tab. 1); E -arrangement of cages in "Experiment 3".counted and immediately released.A list of the carabid species caught is presented in Appendix 2.
Weed abundance.The abundance of seed producing weeds was determined in winter rape and winter wheat stands, where seeds were placed between June 2-26.Within each crop during this period the number of seed producing weed plants in 0.16 m2 (0.4 x 0.4 m) plots were counted.In winter rape weeds were counted in 7 randomly selected plots.To compensate for their low abundance the weeds in winter wheat were counted in 20 plots.

Data analysis
Seed preferences.In the laboratory the rates of removal of seed were determined for either 30 or 15 seeds.The time of seed removal (in days) was the response variable, seed and carabid species factors, and replicates used as statistical blocks.Seeds remaining at the end of an experiment were censored.Censoring is a statistical method that extrapolates the observed trends in seed consumption beyond the period of observation.Removal rates were compared by likelihood ratio tests (Pyke & Thomp son, 1986) in which the course of seed removal was fitted by an exponential function, which is characteristic of a constant removal rate, and a Weibull function, which is characteristic of a continuous increase or decrease in removal rate (Appendix 3), and the fit tested using y2-tests.The appropriate likelihood test (exponential or Weibull) was selected based on a comparison of the residual deviance and the explanatory power of the models.The structure of the models was checked following Crawley (1993, p. 340), using error-checking plots for censored exponen tial and Weibull data on age at death (see Aitkin et al., 1989).The mean time to seed removal (consumption time, CT50) was calculated as the time to when 50% of the seeds were removed.To determine the relationships between seed preference and seed size in P. rufipes and H. affinis, the mean removal rates of the different seeds (CT50) were response variables, the species of carabid factors, and natural logarithms of seed weight covariates.The linear and quadratic term of seed weight were used to test whether the relationship had a parabolic shape.The calculations were made using the general linear model in the commercial statistical package GLIM® v. 4 (Francis et al., 1994).The aim of each analysis was to determine the minimal adequate model.In this model, all parameters were significantly (p < 0.05) different from zero and from one another.This was achieved by a step-wise process of model simplification, begin ning with the maximal model (containing all factors, interac tions and covariates that might be of interest), then all non significant terms are eliminated (using deletion tests from the maximal model) and significant terms retained.
In field experiments the preferences were evaluated using the difference in consumption of seed of particular species exposed simultaneously in a crop (see below).

Seed consumption.
In laboratory experiments seed consump tion was expressed as number of seeds removed.individual (or unit body mass)-1 .day-1.Regressions (y = b0 + b1x) and Pearson correlation coefficients were calculated using carabid body mass as an explanatory variable and seed consumption as a response variable.Coefficients of determination (R2) were calculated to determine the proportion of variance explained by a given rela tionship.
In field experiments seed consumption was expressed as number of seeds eaten day-1 tray-1.Average consumption of a particular kind of seed was the arithmetic mean (±SE) of the daily consumptions in the different cages.Carabid activity den sity was the number of individuals trap-1 day-1.In experiments establishing the seasonal variation in seed consumption, differ ences in seed removal rate were related to the seed species or position of the seed within the crop (cage).The significance of these differences was tested using one-way ANOVA, with seed or cage as a factor and average seed consumption as the response variable.The combined effects of cage location and seed were tested in 2000 in crops where seed was placed on sev-  (34 d).To compensate for temporal variation in seed consumption the analysis of covariance ANCOVA was used, with cage and seed as factors, average seed consumption in each particular period of seed exposure as covariate, and seed removal from particular trays as response variable.
In the field preference experiments the effects of both cage location and seed were determined using two-way ANOVA with cage and seed as factors, and average seed consumption as the response variable.When differences between cages were significant the correlation between carabid activity density and average seed consumption per cage was calculated.All calcula tions were made using STATISTICA for Windows (StatSoft, 1994).
Comparing preferences.To compare the field preferences established in different experiments, and field and laboratory preferences the results were standardized.Standardized removal rates for the three field experiments were expressed as a fraction of difference between the rejected (the least consumed) and pre ferred (the most consumed) ofthe seeds.Standardized consump tion rates for the laboratory experiments were calculated using the times required for 50% of the seed to be removed (CT50).CT5o for each kind of seed was calculated as an arithmetic mean for P. rufipes and H. affinis.The preference for each kind of seed was then expressed as a fraction of the difference between the rejected (longest CT50) and the preferred (shortest CT50) of the seeds.The standardized preferences, calculated from both seed consumption rate and CT50, then fell between 0 (rejected) and 1 (preferred).Pearson correlation coefficients between (i) standardized preferences for particular seeds in the different field experiments, and (ii) the average standardized preferences obtained in the field experiments and the laboratory were calcu lated.A crrelation was also calculated for the seed consumption of carabid species established in no choice experiments in this study and those of Goldschmidt & Toft (1997).The results of both studies were expressed as no. of seeds consumed .indi vidual-1 .day-1.

Seed consumption in the laboratory
The consumption of 23 species of carabids offered either seed of C. arvense or C. bursa-pastoris was recorded (Table 2).Six species (Anchomenus dorsalis, Calathus fuscipes, Dolichus halensis, Poecilus cupreus, Pterostichus melanarius, Stomis pumicatus) were reluc tant to eat either of the seeds.Two species (Calathus ambiguus, Trechus quadristriatus) mainly consumed C. bursa-pastoris seed.Anisodactylus signatus, of which only 1 individual was available, was offered C. arvense seed, but probably would have accepted C. bursa-pastoris seed, and the remaining 14 species accepted both kinds of seed.
The rates of consumption varied with kind of seed and carabid size.The seed of C. arvense was universally accepted (Table 2) and the consumption rate of 15 species of the tribes Zabrini and Harpalini (seed "consumers") increased with carabid size (Fig. 2).Their average con sumption was 0.32 ± 0.04 mg of seed .mg body mass-1.day-1.The seeds of C. bursa-pastoris were preferred by small species and rejected by large species of carabid and the consumption of the seed "consumers" was not related  to their body size.The average consumption of 14 seed "consumers" was 0.33 ± 0.04 mg seeds .mg body mass-1.day-1, i.e. essentially the same as that for C. arvense seed (pt = 0.9131).

Seed preferences in the laboratory
Consumption rates (CT50) varied with both seed and carabid species (Fig. 4).To evaluate the differences in the consumption of particular seeds the variation in the pref erence for the reference species was analysed first.Con sumption (CT50) of the reference species (the seeds of C. arvense in the experiments with P. rufipes and those of C. bursa-pastoris in those with H. affinis) in particular groups of seeds differed significantly (ANOVA for seeds of C. arvense: x2= 256.29, df = 6, p = 1.45E-52;ANOVA for seeds of C. bursa-pastoris: f = 49.9, df = 6, p = 4.92E-09).This indicated that the preference of both carabid species for the reference seed was affected by what other seeds were offered.The effect of replication within particular seed groups was also significant (seeds of C. arvense: % 2= 20.26, df = 2, p = 3.99E-05; seeds of C. bursa-pastoris: f = 7.2, df = 2, p = 0.027E-52), sug gesting an unknown effect other than that attributable to seed choice.
Despite the differences between seed combinations and replicates the results revealed general trends in the seed consumption of both carabid species.Both preferred par ticular seeds (Fig. 4).A large proportion of the variation in preference was explained by variation in seed size.The relationship between CT50 and seed size was significant Table 3.Rates of removal of seed (no seeds .day-1 .tray-1) of C. bursa-pastoris (13), C. arvense (21), Descurainia sophia (25), Lepidium ruderale (40), Sisymbrium loeselii (53) and Taraxacum officinale (57) placed in a fallow field (FAL), and in stands of maize (MAI), millet (MIL), winter rape (RAP), soybean (SOY) and winter wheat (WHT, WH1, WH2).Mean for each species in a crop and the average for all species in a crop (MEAN).N -number of periods (duration 3-20 days) during which seeds were exposed in each crop; CAG -number of cages per crop; STA -day when seeds were placed in a crop; DUR -total number of days for which the seeds were exposed in each crop.for seed series i (R2= 39.4%, p < 0.05), iv (R2= 50.4%, p Seed consumption in the field < 0.005) and vi R2 = 58.9%,p < 0.005), which contained seeds of contrasting size, and for the pooled data (R2 = 31.0,p < 0.001).The preferences varied with the size of the carabid, and the interactions between con sumption rates and carabid species in individual groups were highly significant (p < 0.001, df = 9).P. rufipes readily consumed medium sized seeds but was more reluctant to eat the smaller and larger seeds, while H. affinis preferred smaller seeds than P. rufipes (Fig. 4).

CROP
The calculated preferred seed size was 1.00 mg for P. rufipes and 0.32 mg for H. affinis.The differences in seed size explained a significant proportion of the variance in the seed consumption of both P. rufipes (R2 = 24.88%)andH.affinis (R2 = 33.09%).In the field the rates of removal of C. bursa-pastoris, C. arvense, D. sophia, L. ruderale, S. loeselii and T. offici nale seed from the ground under different crops (Table 3) varied with season, crop, cage location within a crop and seed.There was a marked seasonal variation in the overall rate of seed removal (Fig. 5).It was low before Julian day 150 (May 29) and after Julian day 250 (Sep tember 6), when the average rate of seed removal was 1.1 ± 0.2 seeds .day-1 (n = 8 series of exposed seed samples).Between Julian days 150-250 the average rate of seed removal was significantly (pt = 0.0011) greater and more variable (2.5 ± 0.3 seeds .day-1, range 0.01-6.2,n = 31).During this period there was no significant trend in the rate of seed removal (regression: daily seed removal = -0.001Julian day + 2.659, R2 = 0.6%, N.S.).The annual average rate of seed removal was significantly (pt = 0.0001) greater in 2000 (3.2 ± 0.4 seeds .day 22) than in 1999 (0.6 ± 0.2 seeds .day-1, n = 9).The traps were placed in different fields in the two years.
Between Julian day 150-250 the average rate of seed removal varied between crops.In 1999 the differences were not tested because the number of cage series (n = 9) and crops (n = 5) was small.In 2000, there was no sig nificant difference (pt = 0.2824) between the average rate of seed removal in wheat (3.1 ± 0.4 seeds .day 16) and soybean crops (2.4 ± 0.4, n = 12), although in both crops the tests were carried out consecutively.The differ ences in the rate of seed removal in winter wheat and winter rape crops, where seeds were exposed in parallel, are discussed in the next section.
The seeds of different species were removed at dif ferent rates.Since the rates of removal varied between cages of the same series, the importance of both factors was investigated in 2000, using data from 2 stands of winter wheat and 1 stand of soybeans (Table 4).The dif ferences between the rates of removal of different seeds were significant in two of the three cases analysed but not in "wheat stand 1", where seed removal rates were low.----------------------2---------- Location of a cage within a field significantly affected seed removal rates in all cases.There were thus signifi cant differences in rates of seed removal between micro sites within crops.

Factors affecting seed consumption
The activity density of seed eating carabids varied with season and crop, but P. rufipes and H. affinis were always dominant (Table 5).A significant relationship between the activity density of the seed eating carabids and the average rate of seed removal (Fig. 6) was established in a soybean stand where carabid activity density declined with the onset of autumn (Fig. 5).In other crops the num bers of cages and/or variation in carabid activity density were too small to reveal a significant relationship.
Differences in seed removal rates paralleling to the availability of naturally occurring seed were established between June 2-26, 2000, in adjacent winter rape and winter wheat stands (Table 6).The distance between cages in both crops was ca. 12 m.The rate of seed removal in the wheat stand (3.8 ± 0.5 seeds .day-1) was 5.5 times greater than in the rape stand (0.7 ± 0.1 seeds .day-1).The difference was associated with the species of weeds that produced seed at this time (C.bursa-pastoris, Lamium amplexicaule, S. media, Veronica persica, Viola Table 5. Carabid activity density (individuals .trap-1 .day-1) at the time seed was placed out in the field and the seed prefer ence experiments PRF1 -PRF3 were done in the field.Con sumers: Activity density of all seed eating species (A. aenea, A. aulica, A. consularis, A. familiaris, A. littorea, A. ovata, A. similata, A. signatus C. ambiguus, H. affinis, H. atratus, H. distinguendus, H. signaticornis, H. tardus, O. azureus, P. rufipes, T. quadristriatus) and the activity and percentage of the con sumers made up of P. rufipes and H. affinis.Non-consumers: Activity density of species not eating seeds (A. meridianus, A. dorsalis, B. lampros, B. obtusum, B. explodens, C. fuscipes, C. cancellatus, C. granulatus, L. ferrugineus, L. pilicornis, M. minutulus, N. pallustris, P. cupreus, P. melanarius, S. pumicatus).Date of exposure, duration and number of traps (identical to number of cages used for seed exposure) -see Table 3 1.24 (84%) 0.53 WH1 0.58 0.28 (48%) 0.04 (7%) 0.32 (55%) 0.98 WH2 0.76 0.61 (80%) 0.08 (11%) 0.69 (91%) 0.8 PRF3 1.2 0.67 (56%) 0.00 (0%) 0.67 (56%) 1.1 AVER 2.31 1.67 (72%) 0.13 (6%) 1.80 (78%) 1.36 arvensis), which were present in winter rape but virtually absent in winter wheat.The seed eating carabids were more abundant in the winter rape than the winter wheat.Despite the high carabid activity density in the winter rape stand the rate of seed removal was lower there than in the wheat stand, probably because the naturally occur ring seed in the rape stand satiated the carabids.Differ ences in weed abundance, carabid activity and seed removal rates were all significant (Table 6).Although the differences in seed removal were largely associated with crop and season (Table 3), there was also  variation in the removal rate of particular seeds.The average rates of removal of particular seeds in different crops in both years were significantly correlated (Table 7).

Seed preferences in the field
Differences in seed removal in the field were investi gated in 3 experiments (Table 8).There were significant differences in the rates of removal of the different seeds in all experiments (Table 9), indicating variation in the preference for particular seeds in the field.The standard ized removal rates for the seed of A. retroflexus, C. album, G. parviflora, L. amplexicaule, M. maritima, M. pratense, S. media, T. arvense and V. persica from winter wheat and millet crops in 1999 were not significantly cor related (df = 7, R2 = 3.09%, N.S.) probably because of differences in the composition of the carabid community.However, the removal rates for particular seeds from millet (1999) and winter wheat stands (2000) were corre lated (df = 3, R2 = 87.24%, p < 0.05) ) (Fig. 7).

Comparison of preferences obtained in the laboratory and field
The preferences for the seed of particular herbaceous species obtained in the field and laboratory were com pared using the standardized data.They were signifi cantly correlated (Fig. 8).The scatter in the data increased with increasing preference.The increase in scatter means that the seed of species rejected in the laboratory were    also rejected in the field, but the consumption of preferred seed in the field and in the laboratory varied greatly.

DISCUSSION
Seed removal vs. consumption.The seed removal rates recorded in the laboratory varied.Preferred seeds were consumed shortly after the start of an experiment whereas those of non-preferred species were eaten reluc tantly, probably as a consequence of the increasing hunger of the beetles.The carabids removed the seeds from the plasticine, and took them away, often carrying them in their mandibles around the experimental arena for several seconds, before crushing the testa and eating the seed's contents.The inspection of seed remains indicated that all the seeds that were removed were eaten by the large species, P. rufipes (body mass 29.6 mg).The small species, H. affinis (13.4 mg), removed but later left uneaten the seeds of some species, e.g.G. aparine.The seed used in this study experienced a long after-ripening period and intact seeds of most species were capable of germinating under laboratory conditions (Martinkova, unpubl.).Seed which was removed from the plasticine but not eaten would thus germinate in the moist soil of the experimental arenas.However, germination was rarely observed, which means that most of the seeds that were removed were eaten.
Causes of variation in seed removal.The trend of increasing consumption with body size was confounded by preferences for particular seeds.The most important character determining the consumption of a particular kind of seed in the multichoice experiment was its mass, which explained 25% of the variance in seed consump- Comparing laboratory data.Although the consump tion of seed by different carabid species was studied pre viously by several authors (Goldschmidt & Toft, 1997;Tooley et al., 1999;Lietti et al., 2000) comparison of the consumption rates they recorded is difficult due to differ ences in the species of seed offered, temperature and pre treatment history of the carabids.Goldschmidt & Toft (1997) provide data on the consumption of C. bursapastoris seed by 7 of the carabid species used in our study.The seed consumptions they report are similar (Fig. 9), except for T. quadristriatus, which consumed substantially more seed in our study.Both studies revealed similar seed preferences and consumption rates for populations of carabids from different regions of Europe.
Seed consumption by carabids in the field.This study revealed that a high proportion of the small seed of herbaceous plants present on the ground in arable fields in central Europe may be eaten by invertebrates, specifi cally carabid beetles.We found that several species of carabid eat large quantities of seed of a range of species.Our results thus amplify the results of earlier studies (Introduction) that mostly used a small number of preda tors and/or kinds of seed.What might be the role of other potential seed predators?Pitfall traps revealed the activity density of the arthropod fauna in the vicinity of the cages containing seed.Of the arthropods caught in the pitfall traps only the carabids are likely to eat seed.Ants (Lasius spp.) were scarce and their numbers always made up < 1% of the catch of adult carabids.Also, during visual inspections and servicing of the pitfall traps and seed cages, few ants were observed, probably because of the distance of the cages from ridges (mostly > 40 m).Crick ets, typical inhabitants of steppe localities in the Czech Republic, were not present.Some millipeds (Blaniulus spp., Polydesmus spp.) and beetles (Silpha spp.) present in pitfall catches did not eat seed in laboratory experi ments (Honek, unpubl.).Slugs will eat the plasticine and the seeds.However, there was no evidence of slug (Dero ceras spp.) activity (mucus trails and characteristic scars on plasticine) in our experiments.Carabids were thus the most important invertebrate granivores.The importance of granivory for polyphagous carabids may be greater than previously stated by Thiele (1977), who reported they eat mainly young green vegetation and fruit.This is not surprising since the energy content of seed is greater than that of other plant parts.
It is not possible to decide what proportion of the seeds were removed by carabid adults and larvae, respectively.Larvae (not identified to species) were rare in our pitfall catches.However, pitfall traps may underestimate their densities because of their different size and speed of movement (Adis, 1979).Therefore, their role in seed mortality may be greater than indicated by their activity density.Moreover, experiments of P. Saska (pers. comm.)revealed that most larvae were unable to remove seed from plasticine.The combined adult and larval predation may thus be greater than revealed by our study.
Variation in seed removal in the field.Changes in the composition of carabid community may affect the relative mortality experienced by small and large seed.The pref erences for particular seeds changed with carabid body mass.As the composition of carabid communities varied during the course of a season the proportion of the seeds of a particular size that were removed also changed.Thus in 2000, small seed consumers (Amara spp., Harpalus spp., Ophonus spp.) dominated the carabid community until early August.Later on the average size of the seed eating species increased due to a dramatic increase in the activity of P. rufipes.Other large seed eaters (Amara aulica, A. signatus) also became active.After mid-September the activity density of large species decreased and the community of seed eaters was again dominated by small species, particularly T. quadristriatus.Carabid communities made up of mainly small species may prefer small seeds, and those composed of large species mainly large seeds.
Seed predation.Our data may be used to estimate seed removal in the field.A minimum estimate is that obtained for the winter rape crop in 2000, where seeds produced by naturally occurring weeds were available in excess and the carabids were satiated.As a consequence, the carabids probably removed the experimental seeds at a similar rate to the naturally occurring seeds.From the area of the trays (6.2 cm2) and the daily removal rates (0.7 seeds .tray-1) the daily removal rate can be calculated and is c. 1150 seeds .m-2.This figure may decrease when the activity density of seed eating carabids decreases: it was rather high in our case, 3.7 individuals .trap-1 .day-1.On the other hand, consumption rates may increase with increase in carabid hunger -as demonstrated for P. rufipes.Average seed removal rate, in June-August, over the 2 year period (1999)(2000), was 2.5 seeds .tray-1 .day-1.Accepting this estimate of seed predation increases consumption to c. 4000 seeds .m-2. day-1.Greater con sumption rates may occur in crops where there are few weeds or they are patchily distributed, or they are at the beginning of seed production.This corresponds with the general conclusion that seed consumption is important when seed production is low (Harper, 1977).Average seed consumption did not change substantially between Julian days 150 and 250.This indicates that the potential pressure of carabids on seed on the ground may be rather constant in a particular year.Seed consumption by carabids decreases the number of seeds of herbaceous plants that enter the soil seed bank and may be an impor tant factor controlling weeds on arable land.

Fig. 4 .
Fig. 4. Regressions of the time to when 50% of the seed was removed (CT50) on seed mass for P. rufipes (above) and H. affinis (below).Data plotted on log scale.Seed of plants indi cated by numbers as listed in Appendix 1.

Fig. 3 .
Fig. 3.The average (±SE) rate of consumption of C. arvense seeds by individuals of P. rufipes kept for 9 days at a constant temperature of 26 ± 1°C (Lab, n = 10) and under fluctuating temperatures in the field (Field, n = 10).

DATEFig. 5 .
Fig. 5.The average rate of seed removal (seeds .day-1.tray-1) from field cages containing seed of six herbaceous species of plant, placed in different crops (MAIZ -maize; MIL -millet; RAPE -winter rape; SOY -soybean; WHT -winter wheat) and a fallow field (FAL), in 1999 (99) and 2000 (00).Each point represents a mean value for a series of cages exposed for a period of 3-20 days.On the abscissa the points are placed in the middle of the exposure period.

Fig 8 .
Fig 8.The relationship between the standardized field prefer ences and standardized laboratory preferences for the seed of 42 weed species (some species are shown more than once because of their repeated exposure in the field).R2 = 34.85%,df = 58, p < 0.001.

Table 2 .
The consumption of C. arvense and C. bursa-pastoris seeds by carabids in no choice experiments.The table indicates the body mass (Mass, mg), number of individuals tested (N), mean and SE of the seed consumption during the 3 day experiment, and daily seed consumption (C, mg seed.mg body mass-1.day-1) of each carabid.

Table 4 .
The effect of the position in a crop of the cages containing the seed and kind of seed (species) on the rate of seed removal.ANCOVA -model: see Material and methods. .

Table 6 .
The differences in weed density (plants .m 2), carabid activity density (individuals .trap 1.day 1) and seed removal rate (seeds .day-1) in rape and wheat stands, and the significance of the differences (one-way ANOVA).

Table 9 .
The differences (one-way ANOVA) in the removal rate of different seeds in field experiments 1-3 (see Table8).The high residual variation may be due to morphological and/or bio chemical differences between seeds, e.g. in the thickness and/or consistency of the seed testa and the hardness of a seed's contents (cotyledons).This is indicated by the dif ference in the consumption of similarly sized seed.Both P. rufipes and H. affinis readily accepted seeds of A.