No relationship between range size and germination niche width in the UK herbaceous flora


†Author to whom correspondence should be addressed. E-mail:


  • 1It is widely assumed that species with broad niches will be commoner (have larger ranges) than species with narrow niches, but attempts to demonstrate this in plants and animals have generally been unsuccessful.
  • 2We used seasonal seedling emergence data to define the width of one important niche axis for 175 UK herbaceous species. We then compared niche width with UK range, using species as independent data points and also phylogenetically independent contrasts.
  • 3We found no evidence of any relationship between niche width and range. Possible reasons for this are: (i) mature plant traits are more important than seedling emergence in determining plant ranges; and (ii) ranges of most plants are strongly dispersal-limited.


Rare species (those with small ranges) may differ from related common species (those with larger ranges) in a variety of biological traits (Kunin & Gaston 1997). One such hypothesized difference is niche breadth, on the grounds that species that have broad environmental tolerances and are able to use a wide range of resources will be able to survive in more places, and hence over a larger area (Brown 1984). Unfortunately, empirical studies of interspecific relationships between niche breadth and range size are often confounded by sample-size effects (Gaston 1994). If environmental breadth is determined for more individuals or at more sites for widespread species than for restricted species, then a positive but artefactual correlation between niche breadth and range size would be expected. For example, Burgman (1989) found that the habitat volumes (environmental tolerances) of a suite of plants from southern Western Australia were not significantly different for regionally scarce and ubiquitous species once sample bias was taken into account, although they were significantly different if sample bias was ignored. Another problem is that tests of the hypothesis are frequently based on measures of realized rather than fundamental niche breadth. For example, Thompson et al. (1998) found that a habitat specialism index, calculated from the pattern of habitat occupancy, was strongly correlated with range size in the UK flora. However, from this type of study it is impossible to tell if rare species have genuinely narrow niches, or whether they are excluded from many suitable habitats by better adapted or more competitive species. Much of the evidence is consistent with the latter explanation (Keddy 1989).

In an attempt to overcome this problem, Thompson et al. (1999) examined the relationship between geographical range in the UK and range of germination temperature. This range is one measure of ‘regeneration niche’ (Grubb 1977), which is itself one expression of fundamental niche width. Logically, species with wider germination temperature ranges should be able to exploit a wider range of opportunities for regeneration, both spatially and temporally, and should therefore come to occupy more sites and consequently have larger ranges. In fact UK range was unrelated to germination temperature (Thompson et al. 1999), but this study was inconclusive for at least two reasons. First, it is by no means obvious that range of germination temperatures, measured in the laboratory, is a good predictor of breadth of germination niche in the field. Second, Thompson et al. (1998) included all the species for which data were available, irrespective of life history and habitat. For many long-lived perennials, establishment of new individuals from seed may occur only rarely, and therefore germination temperature may be a relatively unimportant component of niche breadth.

Here we report a new analysis that largely overcomes both these problems. We re-analysed the large body of data accumulated by H.A. Roberts (see Data sources and methods), in which periodicity of germination of many species was examined under standard conditions outdoors. Many of the species studied were annuals and/or arable weeds, species in which establishment from seed is obligatory and therefore any linkage of range and breadth of germination niche should be most apparent. For example, at the simplest level, a species able to germinate at any time of year should be able to exploit both spring- and autumn-sown crops, while a species with a narrower germination window may be confined to just one type of crop. All being equal, the species with the broader germination niche should come to occupy more sites and have a larger range. There is certainly empirical evidence that the composition of weed communities is profoundly influenced by crop-sowing times (Milberg et al. 2001).

data sources and methods

Data on the periodicity of seedling emergence were extracted from several papers by H.A. Roberts and collaborators (Roberts 1964; Roberts 1979; Roberts 1986; Roberts & Boddrell 1983a; Roberts & Boddrell 1983b, Roberts & Boddrell 1984a; Roberts & Boddrell 1984b, Roberts & Boddrell 1985a; Roberts & Boddrell 1985b; Roberts & Chancellor 1979; Roberts & Feast 1973; Roberts & Neilson 1980; Roberts & Neilson 1981). Roberts’ standard method was to sink open-ended glazed earthenware or metal cylinders in the ground outdoors, so that their rims were 8 cm above ground level. Steam-sterilized sandy clay loam was added so that, after settling, the surface was 7·5 cm below ground level. Freshly collected seeds (two replicates of 1000 seeds each) were mixed with a volume of sterile loam which, when placed in the cylinder, brought the soil level up to that outside. On three occasions each year the soil was thoroughly mixed to stimulate germination. All seedlings were counted and removed as they appeared, and monthly totals derived. For most species separate experiments were begun in two or more different years. Emergence was observed and seedlings counted for 5 years.

Data were published as monthly percentages of the annual total. In the few cases where data tables were not published, data were extracted from graphs (Roberts 1964; Roberts & Boddrell 1984a). For each species we calculated mean monthly emergence totals over all sowing and emergence years. When few seedlings were recorded in any one month (less than 0·5% of the total), the data were converted to zero. For each species we summed the monthly percentages to check that they equalled 100%. When small errors (< 5%) were found, the difference was averaged over the months in which some germination was recorded. If the error exceeded 5%, the data were rejected. Data were extracted for 175 species (108 genera, 29 families): 78 annual, 86 biennial or perennial. Another 11 species could be either annual or sporadically/sometimes perennial, but they were considered as annual for the purpose of this study.

A Shannon equitability index, H/Hmax where Hmax = log12, was calculated to measure the breadth of germination niche of each species. The index takes a value between 0 (minimum equitability in the seasonal distribution of germination: all germination in one month) and 1 (maximum equitability: equal germination in every month). Species range was defined as the number of 10 × 10 km grid squares (hectads) occupied by a species in Great Britain. The Shannon index was normally distributed but range was not, and no simple transformation corrected this, so relationships between variables were analysed by non-parametric Spearman correlations.

We also conducted a separate, phylogenetically independent analysis of the relationship between germination niche width and range, using a modified version of the caic (Comparative Analysis by Independent Contrasts) package (Purvis & Rambaut 1995). caic calculates the difference (or contrast) in the traits between extant pairs of species and between internal nodes of the phylogeny. As we do not know what the ancestral species at these nodes were like, values at nodes are averages of the species (or nodes) that evolved from them. In principle it is possible to weight these averages by the branch lengths, but in practice we assumed that all branch lengths were equal. A dichotomous phylogeny with n species yields (n – 1) contrasts, but in practice phylogenies contain polytomies and therefore yield fewer than (n – 1) contrasts. The important point is that each contrast is independent of all the others. We used a molecular phylogeny, which is therefore quite independent of both range and niche width. Further details of the program and of the phylogeny employed can be found in Hodkinson et al. (1998). We then analysed the standardized linear contrasts by linear regression, forcing the regression through the origin (Purvis & Rambaut 1995), although the intercept was in any case very close to zero.


Shannon indices encompassed almost all the available range, from 0·08 to 0·94. To give some idea of what these values mean in terms of seasonal distribution of germination, the extreme values are plotted in Fig. 1.

Figure 1.

Mean seasonal distribution of germination in the species with the broadest (H/Hmax = 0·94, Lathyrus pratensis) and narrowest (H/Hmax = 0·08, Odontites vernus) germination niches in a database of 175 species. Odontites data are means of 3 years; Lathyrus are means of 2 years. For data sources see text.

We first analysed the relationship between germination niche and range for all species combined. We then repeated the analysis, first for annuals only and second for arable weeds only, based on habitat descriptions given by Stace (1997). None of these correlations was significant (Table 1). We also repeated the analysis after further dividing the data taxonomically. First we looked at the relationship between germination niche and range for all six families with at least 10 species in the data set, then at the relationship between family means of both variables. None of these correlations was statistically significant (Table 1). The regression of contrasts in range size on contrasts in germination niche width was also not significant (Fig. 2). There was therefore no evidence of a significant relationship between niche width and range size, either between all species, any subset of those species subdivided by life history or habitat, or at nodes in the phylogeny.

Table 1.  Spearman correlations between range and germination niche width for 175 UK herbaceous species and several subsets based on habitat, life history and taxonomy
All species1750·030·74
Annuals 890·130·23
Arable weeds 630·120·35
Apiaceae 13−0·320·29
Asteraceae 430·040·79
Brassicaceae 15−0·030·91
Fabaceae 11−0·060·85
Polygonaceae 100·590·07
Scrophulariaceae 120·390·22
Family means 29−0·240·21
Figure 2.

Phylogenetically independent contrasts in range plotted against contrasts in germination niche width for 175 UK herbaceous species. n = 98 contrasts, r2 = 0·003, NS.


These results indicate that niche breadth (at least as measured by seasonality of germination) does not influence range size in the UK herbaceous flora. This finding is consistent with previous work, which has generally failed to demonstrate strong or consistent relationships between seed traits and plant ranges. For example, attempts to link range size to dispersal ability (Rabinowitz 1978; Rabinowitz & Rapp 1981; Oakwood et al. 1993; Kelly et al. 1994; Peat & Fitter 1994; Edwards & Westoby 1996) or seed size (Aizen & Patterson 1990, Aizen & Patterson 1992; Jensen 1992; Oakwood et al. 1993; Edwards & Westoby 1996) have proved inconclusive. Moreover, there is disagreement about the form any relationship between range size and seed size might take. Rees (1995) suggested that large-seeded dune annuals are uncommon because of dispersal limitation, while Mitchley & Grubb (1986) suggested that the competitive superiority of large-seeded species allows them to become common, while small-seeded species are fugitive occupants of relatively rare microsites.

Owing to a shortage of standardized data there have been few attempts to relate range directly to germination niche breadth. Baskin & Baskin (1988) provide much the best example. They reviewed their data on germination phenology and response to temperature in controlled environments for 274 herbaceous species. This large data set contained many common/rare congeneric pairs, and in some cases the rare species were local endemics. The data revealed many significant patterns related to phylogeny, habitat and life history, but in every case the germination biology of the rare or endemic species was very similar to widespread members of the same genus. Further work by Baskin & Baskin since the 1988 review (Baskin et al. 1997; Walck et al. 1997a; Walck et al. 1997b) has not altered this conclusion. Attempts to relate niche breadth to variation in abundance or range in animals have been equally unsuccessful (Kouki & Hayrinen 1991; Gregory & Gaston 2000; Cowley et al. 2001; Gaston & Spicer 2001).

The implication is that seed traits in general, and breadth of germination niche in particular, are not very important determinants of plant range. Two factors may contribute to this lack of relationship. One possibility is that seed and mature plant traits are not closely related (Grime et al. 1997; Thompson et al. 2002), and the latter are more important in determining plant range. For example, local extinction of plants in the UK seems to be predictable from mature plant traits (Preston 2000). Thompson (1994) showed that, in the densely populated countries of western Europe, currently increasing or decreasing status could be predicted from traits of the mature plant, and was apparently unrelated to seed size, persistence in the soil or wind dispersal. Sometimes it is not easy to separate the effects of seed and mature plant traits. For example, although Stocklin & Fischer (1999) found that species with long-lived seeds were less likely to suffer local extinction in fragments of calcareous grassland, short-lived seeds were also correlated with stronger restriction to this type of grassland.

Second, numerous studies involving the experimental sowing of seeds into plant communities provide strong evidence for limitation of abundance by seed availability (Tilman 1997; Ehrlén & Eriksson 2000; Jakobsson & Eriksson 2000; Turnbull et al. 2000). In other words, the distribution of most plants is apparently limited more by the availability of seeds than by the availability of suitable niches. However, if most plants are dispersal-limited, why does dispersal ability itself appear to be unrelated to range (Thompson 1994; Thompson & Hodgson 1996; Thompson et al. 1999)? Partly to blame is our poor understanding of, and consequent inability to predict, dispersal capacity. In particular, prediction of dispersal ability on the basis of dispersule morphology is still a very inexact science (but see Nathan et al. 2002). Recent experimental evidence demonstrates that many species lacking obvious adaptations for exozoochory are dispersed on the outside of animals (Fischer et al. 1996; Graae 2002), while species with no obvious morphological adaptations for dispersal may be spread very effectively inside animals (Gardener et al. 1993; Malo & Suarez 1995; Pakeman et al. 2002; Sánchez & Peco 2002). A related problem is that humans may now be the dominant dispersal vector in densely populated landscapes, and different modes of human-mediated dispersal may involve highly disparate sets of traits. Some of the most rapidly spreading aliens in the UK flora, for example, fail to produce seeds at all in the UK (Hodkinson & Thompson 1997).

However, it is not obvious that even a substantial improvement in our understanding of dispersal would entirely solve the problem, as differences in dispersal ability do not appear to be the primary determinant of what Bullock (2000) called ‘gap attainment ability’. In a recent review, Bullock et al. (2002) conclude that while dispersal limitation is almost ubiquitous in plant communities, the identity of colonists depends more on proximity than on dispersal ability or other life-history traits. Whatever the cause, it seems increasingly likely that an inability to demonstrate correlations between range and niche width stem not from inadequate data or a failure to look at the right species, but most probably from the genuine absence of any such relationship.


Kevin Gaston and two anonymous referees made helpful comments on an earlier version of this paper.