Effects of marginality on plant population performance

Authors


Abstract

Aim

Populations at the edge of a species' distribution range may differ substantially from central populations. Peripheral populations may have either a high evolutionary potential or be prone to extinction, but the processes driving these outcomes are still unclear. Peripheral plant populations have been the subject of numerous studies and reviews, with many focusing on their genetic characteristics. In this review, we consider the effect of marginality on demographic species-specific traits.

Location

World-wide.

Methods

We reviewed the literature based on direct comparisons between central and peripheral plant populations. Strict inclusion criteria were applied to avoid biased analysis that may arise as a result of inaccurate boundary considerations or inappropriate comparisons. We inferred from the published data whether a certain trait had a better performance in central or peripheral populations (reliability of the abundant centre hypothesis, ACH).

Results

There have not been enough studies on plant performance to allow for generalizations on the effects of marginality on plants. ACH expectations were not met in most cases and specific responses to marginality were observed at the species and population levels. Population and plant size more often met the ACH assumptions, suggesting that most geographically peripheral populations are also ecologically marginal. The availability of resources, the reproductive strategy, the level of ploidy and the ability to cope with interspecific competitors seem to drive the numerous exceptions to the ACH expectations.

Main conclusions

The large numbers of exceptions to the ACH expectations suggest that a new comprehensive theory is needed to explain the effects of marginality in plants and to identify any general patterns. From the theoretical point of view, we propose that population history and dynamics should be considered when attempting to explain the processes that occur in peripheral plant populations.

Introduction

From the beginning of biogeography as a discipline in the 19th century, naturalists have recognized the importance of species population distribution patterns (Watson, 1835; Wallace, 1876) and, subsequently, the evolutionary significance of peripheral and isolated populations (e.g. Darwin, 1859; MacArthur & Wilson, 1967).

Species range limits typically arise as a consequence of biotic and abiotic factors and their interactions (see Sexton et al., 2009, for a review of the models explaining limit formation). Classically, most authors recognize two types of marginality: ecological marginality and geographical marginality (Soulé, 1973). In the case of ecological marginality, peripheral populations experience a different ecological context with respect to the species' optimum condition (Hoffmann & Blows, 1994; Herrera & Bazaga, 2008). In this case, limits arise because mortality exceeds birth rate, mainly as a consequence of habitat unsuitability and interspecific competition. Geographical marginality is determined by the geographical position of a population at the edge of a species' natural range, and, depending on dispersal barriers, may or may not overlap with ecological marginality (Lesica & Allendorf, 1995). Many examples can be found among alien species, whose spread is often limited by the presence of barriers. For instance, many marine species have been spreading beyond their original geographical distribution limits throughout the Mediterranean since the opening of the Suez Canal (e.g. Katsanevakis et al., 2013).

Most of the studies on peripheral populations acknowledge the assumption of the abundant centre hypothesis (ACH). The ACH postulates that the population abundance of a species is greatest at the centre of its geographical distribution and declines towards the edges of its range (Brown et al., 1995; Sagarin & Gaines, 2002). Moreover, populations occurring at the edge of a range tend to be smaller, less dense and have lower genetic diversity than central populations (Gaston, 2003; Reed, 2004), so they are thought to be less viable and more prone to extinction than central populations (Gaston, 2003; Channell, 2004).

Empirical data, although still sparse, often contradict the patterns predicted by theoretical models (Channell & Lomolino, 2000; Mandák et al., 2005; Sagarin et al., 2006). For instance, Eckert et al. (2008) demonstrated that while about 60% of peripheral populations have lower genetic variability than central ones, this pattern is not an absolute rule, nor does it clarify whether edge populations are really less viable than those in the core area.

Marginality is often associated with reduced gene flow (Sexton et al., 2009). This may result in genetic drift that, in combination with local selective pressures, may enhance the genetic divergence between peripheral and central populations (García-Ramos & Kirkpatrick, 1997). Furthermore, when peripheral populations are exposed to harsh conditions, the environment may select for individuals that are adapted to local conditions, which increases adaptive phenotypic plasticity (Sexton et al., 2009). When this happens, peripheral populations may acquire an evolutionary potential for adaptation and speciation (Levin, 1993). For these reasons, some peripheral populations may be prone to extinction, while in other cases they may be a source for speciation (Levin, 1970) or cause range shifting in response to radical environmental changes (Aizen & Patterson, 1990; Parmesan, 2006).

Over the last decade, many authors have pooled data in order to review the genetic and ecological data at range edges, thus improving our understanding of the effects of marginality (see Channell, 2004; Eckert et al., 2008; Sexton et al., 2009; see also Oikos, 2005, 108(1)). However, substantial physiological, morphological, genetic and reproductive differences between animals and plants make generalizations difficult when comparing the effects of marginality between these kingdoms. For instance, low genetic variation in plants is not always associated with low fitness or low success (Lammi et al., 1999). Given that many factors influence genetic diversity and demographic performance across the distributional range of plant species, the underlying causes of marginality effects remain unclear. Aspects of plant biology that contribute to these unexplained causes include: plant reproduction may be sexual and/or asexual; plants may have complex patterns of gene flow mediated by long-distance dispersal (pollen and seeds); some groups of plants have a very long lifespan; plants can produce fertile hybrids in the contact zone using relative congenerics (Thompson et al., 2010) and they can show different levels of ploidy within the same species. This complexity makes it difficult to detect unambiguous performance behaviour, the evolutionary fate and the conservation value of peripheral plant populations (PPPs) (Lesica & Allendorf, 1995; Holt & Keitt, 2005; Abeli et al., 2009; Gentili et al., 2011).

The aim of this study was to review empirical investigations that compared central versus peripheral plant populations for demographic, reproductive and morphological traits in order to verify whether ACH expectations were met. We might expect that peripheral populations will have lower or negative growth rates in comparison to central populations. The reasons for this pattern may be high individual mortality throughout the plant's lifespan, low reproductive performance and/or low fitness. In particular, as a consequence of unbalanced resources in the peripheral range areas, we might expect that peripheral populations will be characterized by smaller plants (at least for some traits), lower flower production and/or lower individual survival (at least for some life stages) than for central populations. Low pollen/seed viability and germination may also be expected because of inbreeding depression in small peripheral populations. In addition, we used plant-specific traits to explain patterns and exceptions to ACH assumptions in order to gain a better understanding of the mechanisms driving the survival or decline of geographically peripheral plant populations and to contribute to new research questions.

Materials and methods

Literature review

We reviewed the literature through a search (2011–2013) of the online Web of Knowledge database (Thomson Reuters) using the following query: central versus marginal OR central versus peripheral OR central marginal distribution OR central marginal population OR peripheral population OR marginal population AND plant. We also reviewed papers dealing with plant species quoted in the reviews of Sagarin & Gaines (2002), Channell (2004), Eckert et al. (2008) and Sexton et al. (2009).

Using the methods employed by the latter-cited authors, we decided to include only those studies that directly compared central versus peripheral populations. Articles were considered suitable for inclusion in this review when they met the following criteria:

  1. They contained an explicit comparison between geographically central and peripheral populations.
  2. At least one demographic parameter was examined.
  3. Exhaustive methodological details and distribution information were provided.

We also included articles that did not directly test for central versus peripheral populations, but had a strong geographical structure, such as central–peripheral range clines. Studies considering continuous versus isolated populations were also included, but only when the isolated populations corresponded to marginal populations.

We did not consider studies dealing with genetics only [see Channell & Lomolino (2000), Channell (2004) and Eckert et al. (2008) for a comprehensive insight on this topic] or studies where a species' native range was compared to the area where a species had been introduced.

We organized the literature data into the five categories listed below.

  1. Population: studies that compared population features such as population size (number of plant individuals within a population), density (number of individuals over a surface unit) and abundance or frequency (number or frequency of populations occurring at the range centre or boundary).
  2. Reproduction: studies that compared traits related to reproduction, such as flower production, fruit set, seed set, seed germination, etc.
  3. Demography: studies that compared population structure, plant survival and growth rate.
  4. Morphology: studies that compared morphological traits, such as leaf and flower size, plant height, etc.
  5. Plasticity: studies that compared phenotypic plasticity in response to one or more treatments.

The central versus marginal populations of each species we analysed were compared by recording the trend of a certain trait (e.g. population size, density, morphology, etc.) and noting whether it showed a better performance at the centre or edge of a species' distribution or when no differences were detected.

Results

Literature survey

We found 59 studies directly comparing central versus peripheral populations of 59 species. Ten studies (18%), involving 13 species, were biased by placing them in inappropriate range boundaries, such as local or national boundaries. One study (2%), covering three species, did not provide enough information to confirm that the species range margin was correctly identified and another study stated that no morphological differences were found between central and peripheral populations without producing any further details. A total of 48 studies (80%), covering 42 species, delineated true range boundaries and so only these studies were considered for further analysis (see Appendix S1 in Supporting Information).

Almost all the studies (39; 81%) could be included in the ‘population’ approach. Reproductive performance was the subject of 24 studies (50%), 7 (15%) concentrated on population demography, 16 (33%) included comparisons of morphological traits, and 3 (6%) focused on plasticity (see Appendix S2).

Most of the species studied were European and North American species (22: 52% and 14: 33%, respectively), a few were South American and Asiatic species (2: 5% and 3: 7%, respectively) and one was an Australian species (2%). Several studies took into account more than one boundary for the same species, with the northern boundary being the most researched (52%), followed by the southern boundary (25%), while the western (13%) and eastern boundaries (10%) were least researched (Appendix S1). The 42 species were distributed in 23 families. The overall set of taxa comprised 38 perennial species and four annual species (Appendix S1). Most of the studies (57%) were conducted over one year, 41% lasted between 2 and 5 years and one study (2%) was based on long-term monitoring over 13 years (Appendix S1).

Peripheral plant population characteristics

Population abundance/frequency, size and density

Among the 48 considered studies, those dealing with the difference in population abundance between central and peripheral populations investigated six species. The prediction of lower population abundance towards the range edge was supported in one case. Population abundance was similar throughout the species range in most cases, and in one case abundance was higher at the range edge (Fig. 1).

Figure 1.

Comparison of central versus peripheral plant populations for population abundance (n = 6), population density (n = 17) and population size (n = 25). The data represent a total of 34 species (Appendix S2). No difference indicates that studies showed that a trait was similar in both the central population and the edge population.

The population density of 17 species has also been investigated. Some studies hypothesized and statistically tested for a decreasing population density towards the range limit, as predicted by the ACH. Other studies investigated density without any specific hypothesis, but provided qualitative or quantitative information on the difference between central and peripheral populations for this parameter (Table 1, Appendix S2).

Table 1. Comparison of central versus peripheral plant populations based on data from published articles on population size (13 cases) and population density (one case). [C] means a significantly larger population size in the central populations and [=] means that there are no significant differences between central and marginal populations. Values for central and peripheral populations were tested using the Mann–Whitney U-test
SourceSpeciesMann–WhitneyTrait
P Z
Durka (1999)Corrigiola litoralis L.0.145 −1.530 Pop. size [=]
Brzosko et al. (2009)Cypripedium calceolus L.0.817 −0.232Pop. size [=]
Eckert & Barrett (1993)Decodon verticillatus (L.) Elliott0.506 −0.692Pop. size [=]
Trapnell et al. (2012)Euphorbia telephioides Chapman1.0000.000Pop. size [=]
Holmes et al. (2009)Grevillea repens F.Muell. ex Meisn.0.200 −1.328Pop. size [=]
Wróblewska (2008)Iris aphylla L.0.644 −0.462Pop. size [=]
Linhart & Premoli (1994)Lilium parryi S.Watson0.036 −2.236Pop. size [C]
Lammi et al. (1999)Lychnis viscaria L.0.024 −2.245Pop. size [C]
Siikamaki & Lammi (1998)Lychnis viscaria L.0.021 −2.317Pop. size [C]
Duffy et al. (2009)Neotinea maculata (Desf.) Stearn0.057 −2.181Pop. size [=]
Van Rossum et al. (2003)Silene nutans L.0.802 −0.259Pop. size [=]
Van Rossum & Prentice (2004)Silene nutans L.0.118 −1.570Pop. size [=]
Wagner et al. (2012)Stipa pennata L.0.004−2.914Pop. size [C]
Pandey & Rajora (2012)Thuja occidentalis L.0.333−1.549Density [=]

Four species showed a decreasing population density towards the range edge, five had a higher population density at the range boundary, while the remaining species did not show any population density differences between central and peripheral populations (Fig. 1).

Population size data were available for 25 species. Only three studies explicitly tested for differences in population size under the hypothesis that PPPs were expected to be smaller than central ones. Other studies provided qualitative or quantitative information on such differences, but this was not statistically supported in some of the studies. In many cases, data were detailed sufficiently so that differences in population size between central and peripheral populations could be tested by the authors (Table 1). The expectation of smaller peripheral populations was confirmed for 10 species, while 14 species did not show any difference in population size between central and peripheral populations. In one case, the peripheral populations were larger than the central populations (Fig. 1, Appendix S2).

Reproductive performance

Reproductive performance has been assessed by several authors by considering different reproductive traits, including flower, fruit and seed production, fruit and seed set, number of seedlings, seed mass and germination percentage or rate.

In general, species reproductive performance, in terms of the number of flowers, fruits and seeds produced was almost the same at the centre as it was at the edge of the distribution range (Fig. 2). Specifically, seed production was similar across a species' range in about 50% of the cases (Fig. 2). Flower production was either greater in the central populations (40% of cases) or similar across the species range (c. 40% of cases). Fruit production was never greater in the peripheral populations (Fig. 2). Unfortunately, too few studies were available on seedling recruitment, soil seed banks, seed germination rate, percentage or seed mass to draw further conclusions on these demographic parameters (Appendix S2).

Figure 2.

Comparison of central versus peripheral plant populations for reproductive traits associated with flowers (n = 8), fruits (n = 5), seeds (n = 21) and seed germination (n = 4). ‘Germination’ includes germination percentage and rate. The data represent a total of 20 species (Appendix S2).

Survival rate and demography

Among the studies that considered the fine demography of PPPs over a number of years, in most cases the population dynamics were more favourable in peripheral than central populations [Fig. 3; see also values for Lambda found in Stokes et al. (2004) and García et al. (2010)]. However, there were too few data to draw any conclusions on demographic patterns across species ranges (Fig. 3).

Figure 3.

Comparison of central versus peripheral plant populations for demographic traits (n = 12). The data represent a total of eight species (Appendix S2).

Morphology

The morphological measure predominantly used was plant size, either the whole plant or single organs (i.e. flower size, spur length, trunk diameter, leaf size, leaf number, rhizome size, rhizome number and number of shoots). In one case biomass was measured.

When considering the plant size (s.l. = plant height, flower and leaf size), individual plants growing in central populations were, in general, larger than those growing in peripheral populations (Fig. 4). This was particularly true for flower size. However, when plant height only was considered, large individuals were found in central and peripheral populations with more or less the same frequency. In central populations, 60% of the species had larger flowers and there were no cases of flowers being larger in peripheral populations than in central populations (Fig. 4).

Figure 4.

Comparison of central versus peripheral plant populations for plant size (n = 15), flower size (n = 6) and plasticity (n = 3). The data represent a total of 18 species (Appendix S2).

One study considered fluctuating asymmetry (FA). The results showed that there was a significantly higher FA in peripheral populations than in central populations, which is predicted by ACH theory (Siikamaki & Lammi, 1998).

Phenotypic plasticity

Studies comparing plasticity in a number of traits in response to environmental factors among central and peripheral populations produced varying results (Appendix S2). Black-Samuelsson & Andersson (1997) found no difference in plasticity for many traits in response to different levels of light in Vicia dumetorum L. A similar pattern was also found by Berg et al. (2005). In contrast, Volis et al. (1998) found that peripheral populations were phenotypically more variable than central populations and that peripheral populations of Hordeum spontaneum K.Koch had greater resistance to water stress than the core populations.

Discussion

Plant-specific traits and their effects on peripheral populations

General considerations

From our literature review, two important points emerge: (1) ACH predictions are not strongly supported for plant demography, morphology or reproduction; and (2) there have been very few studies on PPPs and therefore we could not undertake a detailed analysis of the processes affecting demography, morphology and reproduction at the edge of a species' range. However, some qualitative analyses were carried out on the basis of the reviewed studies.

Empirical studies suggest that population and plant size show a general decreasing trend towards the edge of the range, which means that peripheral populations are usually numerically smaller than central populations. Furthermore, in many cases, individual plants from peripheral populations are smaller than individuals from central populations. Although there are several exceptions, this suggests that geographically peripheral populations are often ecologically marginal. Exceptions may result from complex patterns of environmental suitability across a species' range. For instance, Ribeiro & Fernandes (2000) demonstrated that population abundance, density and flower production in Coccoloba cereifera Schwacke were related to the type of soil rather than population position within the species' range. In this case, geographically peripheral populations were not ecologically marginal. In other cases, reduced interspecific competition or predation at the range edge (Johansson, 1994; Alexander et al., 2007) enabled plants to occupy broader ecological niches at the range boundary than in the core area of the distribution (Van Rossum et al., 2003; Herilhy & Eckert, 2005).

For other traits (e.g. reproduction, demography, etc.), we found ambiguous behaviour among species and groups of species. The lack of general patterns can be attributed to a number of plant-specific (e.g. clonal growth and ploidy level) and species-specific (e.g. mating system and means of dispersal) characteristics that are discussed below.

Resource-mediated shifts in the reproductive strategy of ecologically marginal populations

Clonal reproduction is thought to be more common or prevalent at the edge of a species' range (e.g. Jump & Woodward, 2003; Beatty et al., 2008). However, the ability to produce clones is a species-specific trait and the ratio between clonal and sexual reproduction varies in more complex models compared to the ratios predicted by the ACH. In general, no evidence for a set of universal rules was identified by this review.

An unbalanced reproductive strategy towards vegetative growth is favoured when extreme ecological conditions and competition limit seedling survival. Such conditions are likely to occur in ecologically marginal populations where other, more competitive species may reduce seedling recruitment. Clonal reproduction is a good competitive strategy that extends the length of a generation and may mitigate the effect of genetic drift (Ellstrand & Roose, 1987). In contrast, clonal growth may cause an unbalanced population age structure where adult and senescent plants are dominant and there are few seedlings present. This is the case for Anemone palmata L., at its southern boundary, where a skewed population structure was found where adults were dominant and there were fewer seedlings (Médail et al., 2002).

Environmental factors and seasonality greatly influence the sexual and clonal reproduction ratio (McKee & Richards, 1996; García et al., 2000). Nevertheless, under harsh conditions, when considerable resources are used for sexual reproduction in a given year, vegetative growth allows a ‘rest’ and a rational utilization of resources the year after. The case of Lloydia serotina (L.) Rchb. is remarkable, as individuals of the species change sex from hermaphrodite to male in unfavourable years, or in the year after the production of a fruit (Jones & Gliddon, 1999). Resource limitation may interact with sexual reproduction. Flower size is often reduced in ecologically marginal populations (Fig. 4), which, in turn, may affect the mating system in PPPs in two ways. First, a change in flower structure may reduce self-preventing mechanisms, such as herkogamy (Herlihy & Eckert, 2005), and second, an alteration in the interaction between plants and pollinators (especially when pollinator specificity is high) may occur when the flowers are small and less attractive, resulting in strong pollen limitation (Segraves & Thompson, 1999; Abeli et al., 2013). Both these mechanisms are the basis of selection for self-compatibility (Busch, 2005; Sun & Cheptou, 2012), which may evolve in PPPs in order to assure reproduction (Lloyd, 1992; Mimura & Aitken, 2007). Specifically, pollen limitation, which reduces female reproductive success owing to the poor amount or quality of pollen received, increases the probability of directional selection for self-compatibility. A major consequence of clonal reproduction and selfing is a reduction in population viability and a depletion of the genetic variability that limits adaptation when environmental changes occur (Jump et al., 2003). However, selfing can purge deleterious mutations, which reduces inbreeding depression (Barringer et al., 2012).

Unbalanced resources may induce a shift in reproductive effort towards fewer flowers and more seeds, as observed in Aquilegia canadensis L. (Herlihy & Eckert, 2005). Flower production and reward requires considerable resources that can be redirected to maintenance, while the production of small seeds is much less demanding in terms of resources. The advantage here is not limited to reducing the amount of resources employed by plants in unfavourable habitats, but also increases the probability that among a larger number of seeds there will be more surviving individuals. In fact, it is known that the ‘seed population’ has higher genetic variation than the population of adult plants (Cabin et al., 1998) owing to strong selective pressures (i.e. seedling mortality) in the early stages of life (McCue & Holtsford, 1998). For these reasons, seed banks are considered important sources of genetic diversity and also provide opportunities for adaptation. For instance, seed production and seed banks were more developed in peripheral populations of Hornungia petraea (L.) Rchb. (Kluth & Bruelheide, 2005).

Means of dispersal and its effect on genetic variation

The ability of plants to disperse is associated with gene flow and, in turn, with the degree of isolation among populations. Depending on the pollination mechanism, gene flow may be scarce or reach high levels (Govindaraju, 1988; see also dispersal vectors in Appendix S1). In anemophilous plants in particular (such as conifers), it would be expected that long-distance pollen exchange would contribute to low differentiation among populations, even at great distances (Liepelt et al., 2002). However, we found that peripheral and central populations of conifers differed greatly, with the latter showing a higher performance (Shea & Furnier, 2002; Appendix S2). Interestingly, mating systems and dispersal abilities are closely related and their relationships, determined by the pollination environment, may vary along distribution gradients that favour range stability (but see Cheptou & Massol, 2009; Sun & Cheptou, 2012).

The role of ploidy in peripheral population stability

There is evidence that high levels of ploidy may help to maintain genetic variability, reduce inbreeding and improve performance (Gulsen et al., 2009; Holmes et al., 2009). Populations of the tetraploid Isoëtes malinverniana Cesati & De Notaris showed higher genetic variation than other diploid Isoëtes populations of comparable sizes (Gentili et al., 2010). Among the few reviewed studies in which populations differed in levels of ploidy, Médail et al. (2002) showed that central tetraploid populations of Anemone palmata were bigger and more vigorous than peripheral diploid populations. Similarly, tetraploid populations of Iris aphylla L. seemed to be more viable than diploid ones (Wróblewska et al., 2010). Moreover, peripheral (westernmost) populations of Santolina rosmarinifolia L. differed from central populations in level of ploidy; the former being autotetraploid and the latter diploid (Rivero-Guerra, 2008). In all of the above-mentioned cases, tetraploid individuals from peripheral populations were bigger, but surprisingly, they showed a decreased reproductive performance when compared with diploid individuals from central populations.

Beyond the ACH: a four-dimensional solution

Owing to the complexity described above, it is clear that the ACH does not sufficiently explain plant population patterns at range margins. It does not consider differences between and within species and it is based on simplistic assumptions of similar marginality effects across taxa. Moreover, the ACH is based on a spatial approach that does not consider population history. In a similar way to the disregarded connections between ecology and biogeographical history highlighted by Wiens (2011) for the formation of range margins, we believe that an analysis of historical and ongoing events is essential if we are to understand the effect of marginality. In particular, adding a fourth dimension (time) to the equation would compensate for the drawbacks inherent in a three dimensional (spatial) explanation.

A species' range is the product of past events and is influenced by ongoing phenomena (e.g. climate changes, stochastic events, human influence, etc.), which lead to perturbations in a species' distribution. Each and every change has a direct consequence on population survival, whatever the cause: (1) some changes may have a negative effect on a species' population survival, which leads to a reduction in the geographical range or area occupied by the species; (2) some changes may be advantageous and might allow a species to spread; and (3) some changes may be neutral. In the case of unfavourable changes, when plant mortality exceeds the birth rate and immigration is strongly reduced as a result of unsuitable conditions, then the population will disappear (Carey et al., 1995).

In contrast, in a favourable situation we expect peripheral populations to be able to increase both in number of individuals and in the area occupied. These populations are a source for new colonization.

Examples of such mechanisms can be found in the recent rapid environmental changes as a result of climate warming and human impact. For instance, ecologically marginal populations of alpine species (i.e. thermophilous species) have extended their range in elevation where less than a few decades ago their growth was inhibited (Parolo & Rossi, 2008; Gottfried et al., 2012). In order to understand distribution patterns and processes in current peripheral populations, we have to look at past events and this may also contribute to the formulation of new hypotheses and research questions. For example, in the case of range retreat, population genetic variability is often reduced owing to a bottleneck, while the founder effect is predominant during range expansion. For example, Parshall (2002) examined the colonization history of the hemlock, Tsuga canadensis (L.) Carrière, based on pollen records. In the past, the range of T. canadensis expanded in response to a cooler and wetter climate from small peripheral populations, which, in the absence of favourable environmental variations, had remained stable for about 2000 years. These small stands grew in size over the last 500 years and contributed to the expansion of the continuous range westwards. Today, peripheral, isolated hemlock stands occur 50 km outside the present continuous range limit. So the question is: what would the hypothesis about the present peripheral isolated populations of T. canadensis be without knowing its colonization history? Following the ACH, one should hypothesize that such populations are less viable and more prone to extinction than central populations. Yet, we know that hemlock is spreading to the west, thus we can expect, with a greater degree of accuracy, that such peripheral populations are increasing in size or are at least stable. The example of T. canadensis and other studies (e.g. Jiménez et al., 1999) emphasize the implications of a fourth dimension (time, history) on peripheral population dynamics.

New research directions and missing information

The complex nature of marginality in plants revealed by this study suggests that there needs to be new research directions. First, an approach based on the study of population performance along geographical and ecological gradients is more informative than a direct comparison between central and peripheral populations. Second, comparisons among species within the same family or functional groups might help to outline some general patterns on the basis of common characteristics, but these types of studies are very rare (Thomas et al., 2013). According to Eckert et al. (2008), the majority of studies concentrate on temperate terrestrial species, while there is a great gap in information on tropical and marine species. Pteridophytes, aquatic vascular plants and bryophytes are also absent from the literature on peripheral populations.

Among the topics to be developed, the relative influence of drift and selection, the role of ploidy in the maintenance of genetic variability and the extent of gene flow in relation to species-specific traits linked to reproductive biology, dispersal ability and the generation length are probably the most urgent. Moreover, as suggested by Parsons (1990), ‘geographical limits can thus in some cases be viewed in terms of physiological limits' (p. 316). However, this review revealed the total absence of studies focused on differentiation in plant physiology in central and peripheral populations. Another field of investigation is the variation in seed longevity between central and peripheral populations. Seed longevity is an important plant trait, having crucial implications for the formation of persistent soil seed banks. Finally, recent studies have highlighted the relative importance of geographical and ecological marginality in animals (Martínez-Meyer et al., 2013) and similar studies on plants are needed.

A separate consideration is the role of exaptations, which are characteristics evolved so as to serve a certain function and later co-opted for another role (Gould & Vrba, 1982). Differences in environmental factors and differential selection pressures experienced by peripheral populations with respect to central populations may mean that certain characteristics designed for a specific function may be useful for another function. If an exaptation increases individual fitness then it may become an adaptation with time. The study of exaptations in PPPs might highlight their potential for evolution, but few studies into plants have been undertaken (Armbruster et al., 2009; Brancalion et al., 2010).

Conclusions

The most important point that emerges from this review is that the ACH, as a general pattern, has been superseded and should only be considered as one of a number of possible models.

The central–marginal patterns found in short-term studies may be biased by seasonality, so long-term monitoring of population demography is essential (e.g. Lesica & McCune, 2004; Abeli et al., 2012). Furthermore, considering a small proportion of a species' range or local geographical limits may lead to inaccurate speculations. Finally, it is not clear how the combination of intrinsic (species-specific or population-specific) and extrinsic (environmental) factors drive PPPs from one side to the other of the turning point between decline and extinction or adaptation and evolution, but mating system and available resources seem to be the most important driving forces.

Acknowledgements

The authors are grateful to Joyce Maschinsky (Fairchild Tropical Botanic Garden, Miami, Florida, USA) for her important suggestions on the first draft of the manuscript and for revising the English content of this document. This research was supported by the University of Pavia (Prof. G. Rossi) and SHARE (Ev-K2-CNR, Bergamo).

Biosketch

Graziano Rossi's research team works in the field of Plant Ecology and Conservation Biology. Among the several research activities undertaken by the team, the most important are the effects of climate change on alpine species, the conservation status of peripheral plant populations and the germination ecology of seeds.

Author contributions: T.A. conceived the idea, collected the data and wrote most of the manuscript; R.G. conceived the idea and wrote part of the manuscript; A.M. analysed the data and wrote part of manuscript on seeds with a substantial contribution from S.O.; and G.R. contributed to the design of the manuscript and revised the several drafts.

Ancillary