understanding ecological community succession: causal models and theories, a review

33
Vegetatio 110: 115-147, 1994. © 1994 Kluwer Academic Publishers. Printed in Belgium. 115 Understanding ecological community succession: Causal models and theories, a review L. J. McCook Department of Biology, Dalhousie University, Halifax, Nova Scotia, Canada Present address: Australian Institute of Marine Science, PMB =~ 3, Townsville, M. C., Queensland, 4810, Australia Accepted 1.12.1992 Keywords: Succession, Inhibition, Facilitation, Life history, Constraints, Autogenic Abstract Critical review of explanations for patterns of natural succession suggests a strong, common basis for theoretical understanding, but also suggests that several well known models are incomplete as expla- nations of succession. A universal, general cause for succession is unlikely, since numerous aspects of historical and environmental circumstances will impinge on the process in a unique manner. However, after disturbance, occupation of a site by any species causes changes in the conditions at the site. Sorting of species may result, since different species are adapted to different regions of environmental gradients. Such sorting can generate several patterns of species abundance in time, but commonly results in se- quential replacements of species adapted to the varying conditions. This may be due to constraints on species' strategies, or life history traits, placed by the limited resources available to the organism. These constraints often result in inverse correlation between traits which confer success during early and late stages of succession. Facilitatory or inhibitory effects of species on each other are best understood in terms of these life history interactions, perhaps as restrictions on, or as moderation of, these processes. Strong support for the importance of correlations in life history traits stems from comparisons of simulated succession with and without these correlations. These simulations are reviewed in some de- tail, and followed by brief reviews of other prominent models for succession. Several aspects of the confusion and controversies in the successional literature are then discussed, with a view to a more optimistic synthesis and direction for successional ecology. "7 confess I was surprised to find my theory so perfectly proved in this case. '" (Thoreau, H. D. 1860. The succession of forest trees, p. 235) Introduction: Succession- ubiquitous pattern, unique processes Ecological succession is an important concept to understanding ecological systems. Successions may be observed at a broad range of scales, from the microscopic to the continental, from minutes to millennia. Patterns of succession or vegeta- tional change can be asserted, to some degree and at some scale, for almost any natural system (e.g. Anderson 1986). Yet in many respects, individual successional processes appear to be unique, and dependent on timing, initial conditions and other factors (eg. Sousa 1984). The goals of succes-

Upload: l-j-mccook

Post on 06-Jul-2016

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Understanding ecological community succession: Causal models and theories, a review

Vegetatio 110: 115-147, 1994. © 1994 Kluwer Academic Publishers. Printed in Belgium. 115

Understanding ecological community succession: C a u s a l mode l s a n d theories, a review

L. J. McCook Department of Biology, Dalhousie University, Halifax, Nova Scotia, Canada Present address: Australian Institute of Marine Science, PMB =~ 3, Townsville, M. C., Queensland, 4810, Australia

Accepted 1.12.1992

Keywords: Succession, Inhibition, Facilitation, Life history, Constraints, Autogenic

Abstract

Critical review of explanations for patterns of natural succession suggests a strong, common basis for theoretical understanding, but also suggests that several well known models are incomplete as expla- nations of succession. A universal, general cause for succession is unlikely, since numerous aspects of historical and environmental circumstances will impinge on the process in a unique manner. However, after disturbance, occupation of a site by any species causes changes in the conditions at the site. Sorting of species may result, since different species are adapted to different regions of environmental gradients. Such sorting can generate several patterns of species abundance in time, but commonly results in se- quential replacements of species adapted to the varying conditions. This may be due to constraints on species' strategies, or life history traits, placed by the limited resources available to the organism. These constraints often result in inverse correlation between traits which confer success during early and late stages of succession. Facilitatory or inhibitory effects of species on each other are best understood in terms of these life history interactions, perhaps as restrictions on, or as moderation of, these processes.

Strong support for the importance of correlations in life history traits stems from comparisons of simulated succession with and without these correlations. These simulations are reviewed in some de- tail, and followed by brief reviews of other prominent models for succession. Several aspects of the confusion and controversies in the successional literature are then discussed, with a view to a more optimistic synthesis and direction for successional ecology.

"7 confess I was surprised to find my theory so perfectly proved in this case. '" (Thoreau, H. D. 1860. The succession of forest trees, p. 235)

Introduction: Succession- ubiquitous pattern, unique processes

Ecological succession is an important concept to understanding ecological systems. Successions may be observed at a broad range of scales, from the microscopic to the continental, from minutes

to millennia. Patterns of succession or vegeta- tional change can be asserted, to some degree and at some scale, for almost any natural system (e.g. Anderson 1986). Yet in many respects, individual successional processes appear to be unique, and dependent on timing, initial conditions and other factors (eg. Sousa 1984). The goals of succes-

Page 2: Understanding ecological community succession: Causal models and theories, a review

116

sional ecology are to describe the similarities and differences in both the pattern and the process: to extract generalizations, and to identify the bases for differences in process.

The field has generated both excitement and controversy for biologists for over a century (e.g. Thoreau 1860; Clements 1928; Gleason 1927; Loehle 1988, 1990; Shrader-Frechette & McCoy 1990; Farrell 1991). In fact, understanding suc- cession is as important today as ever, both in terms of amount of research effort and in terms of understanding changes in natural and man- aged environments.

The patterns of succession: empirical observation

It has long been observed that natural vegetation at widely different sites often undergoes similar basic patterns of change in species abundances, in response to disturbances or changes in the en- vironment. These changes often involve the ap- pearance or dominance of species with progres- sively greater maximum size, age, and shade tolerance, and progressively lower maximum growth rates and dispersal abilities. Such gener- alizations are easily recognized, yet real situations vary enormously in both empirical and interpre- tative details (e.g. Clements 1928; Drury & Nis- bet 1973; Sousa 1984; Walker & Chapin 1987; McCook & Chapman 1991, 1992, in press; Mc- Cook 1992).

Unfortunately, our understanding of ecological succession has been impeded by several prob- lems, resulting in considerable confusion about the processes that are important during succes- sions. In particular, the successional literature lacks integration. Many field studies are either poorly designed, purely anecdotal, or test theo- retical concepts that are inconsistent or con- founded. Many papers summarize previous work, but fail to integrate it with new hypotheses or models. For instance, the very formative work of Connell and Slatyer (1977 & Connell et al. 1987) does not refer to the powerful simulation models of Botkin et aL (1972a, b) and Shugart and West (1977), despite the crucial rigour these simula-

tions offer in understanding species interactions during succession (see Discussion; see also e.g. Noble and Slatyer 1980 cf. Connell & Slatyer 1977).

I consider the last point to be one of the most critical: the relationships between different ex- planatory models for succession are generally vague, making it difficult to frame field studies that make sensible tests. As a result there have been only a few field studies that adequately ad- dress the reasons for patterns of species abun- dance during succession. Reviews of succession tend to be historical accounts of the debate be- tween proponents of a reductionist, individualis- tic view, and the proponents of a holistic, eco- system, integrative or facilitatory 'superorganism' view. Reviews often include anecdotal accounts of field work in support of arguments, but have little critical or integrative analysis of the actual models. This approach offers a good understand- ing of how ideas developed, but not of their util- ity, and even leads to quite dismal prospects for general understanding of successional processes (e.g. Horn 1976; McIntosh 1981; Finegan 1984; Miles 1987; Shrader-Frechette & McCoy 1990). I feel that a more synthetic approach warrants a much more optimistic view.

Outline and purpose of current review

In the light of these criticisms, the present review attempts a critical theoretical synthesis, with em- phasis on integration of useful current under- standing, playing down debates and controver- sies, and avoiding the customary historical order of presentation. I begin by summarizing what I consider the most useful aspects of current suc- cessional theory. After outlining what I hope is a useful theoretical basis for ideas on succession, I then discuss other models and theories, and some of the problems, controversies and issues arising from them. I believe that this approach reduces much of the controversy. Final discussion in- cludes the types of theoretical and field research that might most usefully draw upon present un- derstanding.

Page 3: Understanding ecological community succession: Causal models and theories, a review

I am concerned here with understanding the causes of succession at the community level, and focus on models relevant to that understanding. I examine these models for insight into the pri- mary question: "Why do different patterns of spe- cies abundance occur during autogenic succes- sion, once a disturbance has freed sites for colonization?" Subsidiary questions include: "What can cause one species or group of species to be displaced or dominated by another or others?" and "What effects do earlier species have on the success of later species?" Different mod- els have different purposes, and hence achieve different ends (and these differences are not al- ways intentional or explicit). In asking these ques- tions my perspective is that of a field ecologist who requires rigourous, testable hypotheses about ecological processes, and who seeks explanation in the interactions between organisms and their environment. Thus my criteria, discussed later, are different to those of either theoretical mod- ellers or applied ecologists, and further reflect subjective opinions about which aspects of suc- cession need explanation.

I generally limit the review to mechanistically (or causally) based understanding of community level, autotrophic, autogenic succession on fairly spatially homogeneous sites, after extrinsic dis- turbance, and excluding major disturbances dur- ing the course of the succession. Although some of these are clearly unrealistic limitations, and the approach very reductionist, they allow the isola- tion of factors that can cause succession in such limited conditions. Such factors may then be in- corporated into more complex, integrative mod- els. For similar reasons, the review examines suc- cession as expressed by the major structural species in communities, and ignores other com- munity elements.

Ideally, this review would be followed by a critical review of extant field evidence, to test the relevance in nature of the present argument. However, such a dual review is beyond the scope of the current paper, and I feel that it is not con- structive to assess further the applicability of models without first establishing a clear, consis- tent, theoretically rigourous, and biologically in-

117

terpretable model. Models which are internally inconsistent or vague are less able to produce tests with useful biological conclusions, and in fact can delay understanding (Loehle 1988, 1990; Shrader-Frechette & McCoy 1990). Thus, I am assessing the models on their perceived theoreti- cal biological merit alone, and not on field evi- dence.

To present the ideas as clearly as possible, I defer discussion of successional concepts, defini- tions and categories until later in the review, but several points should be made beforehand. In this paper, succession is taken to refer to (directed) changes in vegetation composition in time, in re- sponse to external changes or disturbance, and autogenic succession is that succession which oc- curs independently of continuing change in exter- nal environmental conditions (although in re- sponse to some discrete, initial, external disturbance). Successional patterns that involve sequential replacement of species are called se- quential succession. The term community is used simply to describe an assemblage of individuals, and dominant is used to imply high relative abun- dance, without necessarily a competitive conno- tation. Although the concepts of successional cli- max and stable endpoints are widely discredited, this does not devalue understanding the dynam- ics of successional change. Rather, frequent dis- turbances imply that successional processes are all the more important. Thus, the review ignores the issues of successional endpoint or climax sta- bility and patch dynamics, and concentrates on the dynamics of within successional change.

The process of succession: theories and models of causal mechanisms of (autogenic) succession

Resource allocation strategies, tradeoffs and auto- genic succession

Synthesis It is almost a biological truism that any organism cannot be simultaneously well-adapted to all en- vironmental conditions, and this is especially true when biological interactions such as competition

Page 4: Understanding ecological community succession: Causal models and theories, a review

118

and predation are important. This 'Strategic Re- source Allocation' concept suggests that limita- tions on the supply and allocation of metabolic resources to different physiological functions gen- erally involves tradeoffs between those functions. A plant that allocates resources to a particular function cannot allocate those same resources to other functions. Thus, for example, resources al- located to rapid accumulation of photosynthetic tissue cannot be allocated to structural growth or disease resistance. Obtaining light requires above- ground tissue, whereas obtaining soil resources requires below-ground tissue, and the most effec- tive allocation of tissue will depend on whether light or soil resources are limiting. The particular balance of allocation to different functions will result in a particular adaptive 'strategy' of the plant, and a particular set of environmental con- ditions in which it is competitively most success- ful. Further, strategies or structures that benefit one function can often be directly costly to an- other (e.g. large size may increase light capture, but also will increase wind damage). Thus there are constraints on and tradeoffs between the adaptive strategies that result from a species' life history traits, and life history traits adaptive to different conditions are likely to be inversely cor- related.

Detailed formulations of this concept include the single axis r and K strategies of MacArthur and Wilson (1967); Grime's (1974) two dimen- sions of stress and disturbance; Tilman's (1982, 1985, 1988, 1990a) above and below ground tis- sue allocation and resource ratio concepts; clas- sification of species as early or late succession species with an attendant set of presumed traits (e.g. Connell & Slatyer 1977); and Huston and Smith's (1987, Smith & Huston 1989) more gen- eral growth rate/resource availability tradeoff. The discrepancies between these ideas, and the de- tailed nature of the tradeoffs and constraints and their evolutionary aspects need not be discussed in detail here.

A plant's particular strategy, or life history traits, thus determines its success at a particular region of any environmental gradients. If envi- ronmental resource or stress levels change in ei-

ther space or time, species composition will prob- ably also change (e.g. Gleason 1939).

The central tenet of this review is that autoge- nic sequential succession can be explained by cer- tain correlations between life history traits such as growth form, growth rate, and shade tolerance. If these correlations are absent, succession may not involve the sequential dominance by different species, but may form another pattern of species abundances. These correlations are the predict- able consequence of limitations on metabolic re- source allocation. Thus, species that have wide dispersal, high recruitment and high growth rates tend to be intolerant of shade, and have low maxi- mum age and height. Species which allocate fewer resources to dispersal and rapid growth, tend to be able to grow in shade and to achieve greater heights and ages.

This correlation suggests the following mecha- nistic interpretation of sequential succession: Im- mediately after a disturbance makes a site avail- able, species with high colonization and growth rates are likely to rapidly recruit to, and dominate, the site. However, such species will, by their very occupancy, alter the resource levels at the site. For instance, if shade tolerance and colonization are inversely correlated, then these species may be less able to recruit or continue growing in the shade that they themselves produce at ground level. Other species, which have slower growth rates, but can grow in the shade of the early domi- nants, and can grow to greater maximum size, are likely to eventually emerge from the canopy of early species and shade them. This is likely to prevent further recruitment and growth of the shorter-lived early dominant species, which may then decline, leaving the community dominated by the slower-growing species. This process could then be repeated as species with increasing shade- tolerance and maximum age or height progres- sively dominate the community, giving a sequence of successional dominants.

It is a matter of common observation (Egler 1954; Drury & Nisbet 1973; Grime 1974; Noble & Slatyer 1980; Tilman 1982, 1990a, b; Huston & Smith 1987; Smith & Huston 1989; Pickett & McDonnell 1989) that in nature these traits are

Page 5: Understanding ecological community succession: Causal models and theories, a review

indeed often inversely correlated. In fact, species are often classified by successional timing, ac- cording to a long list of traits, including dispersal mechanisms, responses to disturbance, ability to persist through disturbances by seed bank or veg- etative regeneration, maximal height, longevity, resistance to stresses, resistance to small scale disturbances or herbivory, recruitment and growth rates at different resource levels (par- ticularly shade tolerance) (e.g. Odum 1971, Table 9.1).

This interpretation of autogenic succession is simply a sufficient model (Botkin 1981), not an exclusive explanation, and it does not rule out the possibility of other interactions or factors. With or without such correlations in life history traits, other factors also may generate such patterns of species abundances. Species without these cor- relations may still appear in sequence if, for in- stance, their timing of recruitment differs. Fur- ther, the detailed expression of this process can and will be modified by numerous factors, includ- ing species availability, disturbance levels, his- tory, and, in particular, the effects of different species on each other.

The assertion that correlations in life history traits can explain successional patterns has been, in various empirical or theoretical forms, a com- mon vein in successional theory for a consider- able time (see references above), but there has been virtually no tests of the sufficiency of the assertion itself. Tilman (e.g. 1985,1990a, b) has made several valuable theoretical and experimen- tal contributions that support this explanation, but has not directly tested the importance of tradeoffs in general. However, recent simulation work by Huston and Smith (1987) actually pro- vides a clear-cut demonstration of the sufficiency of the concept to explain sequential species re- placements during (simulated) succession.

Simulation models by Huston and Smith, Botkin and others Perhaps the most convincing evidence for the im- portance of correlations in life-history traits stems from the simulations of terrestrial forest succes- sion by Huston and Smith (1987). The most at-

119

tractive feature of this work is that they actually test the importance, to the simulated successional patterns, of the correlation between life history traits, using an a priori biological framework. The simulations basically follow individual trees through their lives, modelling birth, growth and death as life history traits modulated by compe- tition for light, which is in turn explicitly and in- dividually modulated by neighbours. This model is explicitly derived from earlier individual-based terrestrial forest simulation models originated by Botkin etal. (1972 a,b: JABOWA model), and more widely developed by Shugart and colleagues (e.g. Shugart & West 1977, 1980 FORET model, see review by Shugart 1984, Ch 1-3). I will con- centrate most on the model of Huston and Smith (1987), which is more concerned with explana- tions of succession than with modelling forest dynamics for applied prediction and management purposes.

Crucial features of these models include: inter- actions among individuals rather than popula- tions; competitive ability as an individual char- acteristic based on the interactions of life history characters and the individual's environment, rather than abstract parameters for an entire population; non-equilibrium dynamics; and auto- genic changes, that is changes induced by the plants themselves, in the critical resources for competition.

As in the earlier work, Huston and Smith (1987) argue that species interactions, such as competition for light, occur at the level of the individual, rather than the population, and hence simulation models should follow individual trees throughout their life. These models thus avoid assumptions of integration to the population level of interactions such as competition, which may affect individuals differently. Botkin and Shugart's models consider the effects on indi- vidual growth of resource levels averaged over an entire patch or gap. Huston and Smith, however, model resource availability at the individual level, calculating the effects of neighbours on resources for each tree.

The simulations by Huston and Smith and, to a degree, their predecessors have important fea-

Page 6: Understanding ecological community succession: Causal models and theories, a review

120

tures that make them especially attractive models. The models incorporate a priori, rigourous and causal biological concepts. These are: competi- tion, for light and nutrients, based on growth and mortality; recruitment and mortality with stochas- tic and physiological components; growth based on estimated photosynthetic rates for the particu- lar species at particular light levels; and changes in resource levels in the immediate environment. In fact, competition is not considered in abstract terms, but as the effects of species on resource levels, and of resource levels on species. The models focus on processes, to provide a priori information about similarities and differences at the level of pattern.

Because they are mathematically based simu- lations, the models' output is quantitative and testable (with limitations) (cf. verbal models dis- cussed later), can be integrated into higher level and larger scale models, and their assumptions are necessarily precise, criticisable and explicit. Assumptions are biologically realistic, and are limited to those needed to reproduce succession (Botkin et al. 1972b): there is no assumption of equilibrium in species interactions, no assump- tion of integrative competition for a whole popu- lation and, most importantly, in Huston and Smith's model there is no a priori assumption of relationships between life history traits such as those attributed to early and late successional species.

This is a crucial point and the unique feature of Huston and Smith's work: minimal initial as-

sumptions generate baseline models of species abundances; this allows the later addition of fur- ther assumptions, and hence allows the determi- nation of the consequences of various empirical observations, including the correlation between certain life history traits.

Model results

By 'decoupling' the correlations between traits associated with early and late successional spe- cies, Huston and Smith can assess objectively their importance to the (simulated) temporal pat- terns of species abundances. Using two species and varying combinations of life history traits, the simulations indicate that five patterns of species abundances are possible during regrowth: sequen- tial successional replacement; divergence; con- vergence; total suppression; and pseudo-cyclic replacements (see Fig. 1).

Sequential succession, which Huston and Smith call successional replacement, involves se- quential peaks and declines in abundance of dif- ferent species. Divergence involves initially simi- lar abundances, followed by fairly stable dominance of some species, and reduction or dis- appearance of others. Convergence means stabi- lization to co-existing similar final abundances from initially very different abundances. Total suppression implies the immediate and lasting dominance of some species, with other species never achieving significant abundance. Finally, pseudo-cyclic replacement involves the more or less cyclical reappearance of species abundant

Biomass

Sequential Succession

Time

Divergence

k Total

Suppression Convergence "Pseudo-cyclic"

Replacements

Time Time Time Time

Fig. 1. Five different patterns of abundances for two species during succession, as described by Huston and Smith (1987). In simulated successions, Huston and Smith (1987) found that different combinations of life history traits for the two species gen- erated different patterns of species abundances. In particular, sequential species replacements occurred when life history traits such as wide dispersal, rapid growth, and high maximum growth rates are inversely correlated with traits such as high shade tolerance or high maximum size. (Based on Huston & Smith 1987).

Page 7: Understanding ecological community succession: Causal models and theories, a review

earlier. Each pattern of regrowth arises from par- ticular combinations of traits for each species.

The most interesting result is that sequential successional replacements occur when life history traits usually found in early and late successional species are inversely correlated. That is, succes- sional replacements occur when species attributes such as wide dispersal, rapid growth, and high rates of sapling establishment are exclusive of traits such as high shade tolerance or high maxi- mum size (which gives light competitive domi- nance). If these traits are not inversely related, other patterns of species abundances occur (Hus- ton & Smith 1987).

Such correlations appear common in nature, as already noted, and Huston and Smith do not claim to originate the idea. The significance of Huston and Smith's work lies in taking this idea beyond mere assertion, to a clear demonstration that this correlation in traits can generate succes- sional sequences of species where species are otherwise not sequential (albeit in a simulation). They thus derive the conclusion from first prin- ciples, rather than empirically: the model deter- mines the consequences of the empirical obser- vations, rather than interpreting the observations a posteriori.

These results portray a kind of baseline model for sequential succession, based on competition for light, autogenically decreasing substrate light

121

levels, and sequential dominance by species that can grow, albeit slowly, in progressively lower light levels. This model is not an exclusive expla- nation of succession, but rather a simplest case, which demonstrates that complex, facilitative in- teractions between species are not necessary to generate sequences of species dominance.

Huston and Smith (1987) use their model to examine some other possible factors and aspects of succession, but disappointingly, neither they nor Botkin or Shugart adequately address the controversial issue of net facilitatory or inhibitory effects of early dominant species on later species. There are some relevant results, discussed here, but I argue in the Discussion that detailed exami- nation of species interactions and their mecha- nisms are called for.

Huston and Smith (1987) expanded their model to include more species, and find that similar suc- cessional sequences occur as with two species, even amongst species that all have 'late' succes- sional attributes (see Fig. 2b). They also found that the slower growth rates of the species abun- dant in later succession cause prolonged co- existence of species, in comparison to earlier suc- cession.

But more interesting is the incidental insight this expanded model gives into species interac- tions during succession. Understanding the ef- fects of species on each other requires the re-

Biomass

t Time Time

Fig. 2. Simulated removal of early colonist from successional sequence, rearranged and reinterpreted from Huston and Smith (1987). (A) Simulated succession with four species with varying life history attributes. (B) The same succession without species 1. Comparison of (A) and (B) suggests that the early colonist (species 1) was inhibiting species 2, and indirectly facilitating spe- cies 4.

Page 8: Understanding ecological community succession: Causal models and theories, a review

122

moval of an early dominant species. Such a result can be found by reinterpretation of one of Huston and Smith's results as a simulated species dele- tion (Fig. 2). This interpretation shows that re- moval of an early dominating species causes a dramatic increase in the growth and dominance of the next dominant, and a delay in the biomass growth of the final dominant. This shows that the early dominant species inhibits the growth of the next dominant species, but (indirectly) facilitates the final dominant.

Huston and Smith's model also can incorpo- rate competition for multiple resources, including water and nutrients such as nitrogen. In particu- lar, they model succession with initially limiting nitrogen levels. Terrestrial primary succession commonly involves initially low nitrogen levels, which increase reciprocally as (sub-canopy) light levels decrease during succession (e.g. Tilman 1982; Walker & Chapin 1986). Huston and Smith (1987) compared simulated species abundances under conditions of high initial light and nitrogen, with conditions of low initial nitrogen (but high light). As expected, with initially low nitrogen, there was a dramatic initial dominance by a nitrogen-fixing species, a concomitant rapid rise in nitrogen levels, followed by a decline in the nitrogen-fixing species. At this point, the pattern converges to that found when initial nitrogen lev- els are high.

Unfortunately, Huston and Smith did not ex- tend their analysis of nitrogen limited succession to examine the species interactions and to test for possible facilitation of subsequent dominants by the nitrogen-fixing species. Primary succession initially limited by resources other than light, is often considered a likely candidate for 'facilita- tory' successional interactions, whereby earlier species actually improve the establishment or growth of later species. It would therefore be very interesting to compare simulated succession with low initial nitrogen levels, with and without a nitrogen-fixing species. This could show clearly whether species that dominate under restored ni- trogen levels are significantly affected by early abundance of the nitrogen-fixing species. Obvi- ously, such an interaction could involve facilita-

tion of later species, by excess nitrogen produc- tion, as well as inhibition, by shading by the rapidly-growing nitrogen-fixers. The net effect of these two interactions has been the substance of some of controversy (e.g. Walker & Chapin 1987). It is also disappointing that Huston and Smith have not investigated the consequences of facili- tation of colonization, particularly obligate facili- tation. Obligate facilitation refers to the inability of a species to establish without the facilitatory effects of earlier species, such as nitrogen-fixers. Huston and Smith (1987) suggest that early col- onization would increase the inhibition of later arrivals. Clearly this would only be true if the later colonizers did not depend on the initial coloniz- ers, as they do in obligate facilitation.

Botkin (1981) briefly examined his model's im- plications for causality and species interactions during succession, particularly the effect of early species on later ones. He ran simulations in which he omitted groups of early species or of late spe- cies, and examined the abundance through time of the remaining groups. Botkin argues that in that system, the abundances of later dominants are qualitatively similar with or without early dominants, but that later dominants are quanti- tatively inhibited by the earlier dominant groups: he claims that the earlier dominants affect the abundance of later dominants, but not the timing of the outcome. Botkin's (1981) graphs show that these results are not clear-cut, but it is true that both groups of species appear with similar timing in the presence or absence of the other group. Also, as the early dominants only decline strongly in the presence of the final dominants, it is clear that the later dominants suppress the early domi- nants. Botkin (1981) interprets this result as showing that the process of sequential succession depends on life-histories alone, a null model (not equivalent to the tolerance model of Connell and Slatyer 1977). Unfortunately, Botkin interprets this result as generally true for forest succession in the field, and as discounting any facilitatory effects of early colonizers on later dominants. Failure to find a phenomenon in one simulated situation is not proof of its general absence.

Page 9: Understanding ecological community succession: Causal models and theories, a review

Tilman's resource-ratio hypothesis The role of competition along gradients of re- sources has been the subject of an elegant re- source allocation model by Tilman (e.g. 1982, 1986a, b, 1988, 1990a, b). Using models of re- source availability, Tilman argues that the species able to establish, grow and reproduce fastest at a particular level of the limiting resource will be competitively dominant. In particular, Tilman (e.g. 1990b) argues that species will consume and decrease resource levels to an equilibrium level (R*), at which both population (or biomass) and resource level are at equilibrium. The species that is at equilibrium at the lowest (limiting) resource level should outcompete any others. Tilman em- phasizes the strategic allocation of resources, and the constraints and tradeoffs involved, particu- larly the tradeoff between below ground tissue (required to obtain soil resources), and above ground tissue (required to obtain light). However, these tradeoffs are an assumption of Tilman's model, whose importance is not tested. The tradeoffs mean that plants will tend to be best adapted to different points along resource gradi- ents, and will tend to sort competitively along such gradients, as discussed earlier.

Tilman (1985, also 1990a) considers that this sorting along resource axes can explain species patterns in successions. He considers that a change in resource supplies, the 'resource supply trajectory', will cause a simultaneous change in the dominance of species. He observed that suc- cessional 'resource supply trajectories' often in- volve the reciprocal decrease and increase of two limiting nutrients. Such trajectories, in which two limiting resources are inversely correlated, will re- sult in a sequential succession (if the dynamics of competition are faster than the dynamics of the resource trajectory). For example, in primary suc- cession, with initially limiting nitrogen levels but high light availability (no canopy species present), species adapted to low nitrogen levels will initially outcompete other species. However, as numerous processes inevitably contribute nitrogen to the soil, and their canopy prevents light from reach- ing new recruits, the resource levels change to the relative advantage of species able to grow faster

123

in the lower light levels and higher nitrogen lev- els. These species then outcompete the earlier dominants, and the process continues in this manner. During secondary succession, similar changes in competitive dominance occur simply because the species that can grow faster in the initially high light levels cannot outgrow shade- tolerant species under their own canopy. Al- though this is conceptually similar to previous models, Tilman (1985) was the first to use a for- mal, quantitative, mathematical model to produce the patterns.

Interestingly, Tilman (1985p. 845) states "changes in resource-supply rates (light and ni- trogen) are not assumed to be under the direct control of photosynthetic plants . . . . it is not a model of autogenic succession". Yet his explana- tion of the generation of the resource supply tra- jectory is weak without changes wrought by the organisms themselves. Soil light levels are ex- pressly altered by the accumulation of biomass (p. 836), and the suggested mechanisms for changes in the nitrogen levels (p. 835) are largely dependent on biologically derived nitrogen. The model predicts succession in response to chang- ing resource levels, independent of the cause of the changes, but it is primarily directed to auto- genic changes. It does not depend on autogenic or facilitatory changes, but in its simplest form it does model them.

Tilman's (1985) model includes the most ex- plicit argument about the reasons for the exclu- sivity of different life history traits as adaptations to different environmental conditions, and the de- tails of the mathematical/graphical model for competitive dominance at different resource lev- els are elegant. Furthermore, Tilman's (see par- ticularly 1990a, b) hypotheses about interactions between resources and competing species give the most detailed and explicitly mechanistic interpre- tations of real, field successions (as opposed to simulations).

However, despite the clear affinity with the ideas of Huston and Smith (1987), the model makes several interrelated, questionable assump- tions. Tilman (1985, p. 846) states that the main difference between his model and the simulation

Page 10: Understanding ecological community succession: Causal models and theories, a review

124

models of Botkin et al. (1972a, b) and Shugart and West (1977) is in its simplicity. I would dis- agree. Tilman's is a population level model, which does not clearly explain the origins of changes in resource levels, assumes equilibrium dynamics, and requires that "the rate of change in resource availabilities is slow relative to the rate of com- petitive displacement" (Tilman 1985, p. 833). Each of these is a critical assumption not made by the dynamic, individual based simulations already discussed. Huston and Smith (1987) point out that population level models (with con- stant parameters) ignore the potentially impor- tant variability in individual responses to compe- tition etc.

The use of population level, equilibrium dy- namics has been criticized as unrealistic, in terms of rates of equilibration, and for small population sizes, such as often found in forest patches or gaps (e.g. Huston 1979; DeAngelis & Waterhouse 1987). Although Tilman (1985) gives examples of slowly changing resource levels, it seems likely that many systems, especially successional sys- tems, simply operate too rapidly to achieve or be realistically represented in terms of equilibrium dynamics. There is evidence that the time scales involved in competitive equilibrium dynamics are often considerably greater than those of the suc- cessional changes or the changes in resource lev- els (e.g. Drury & Nisbet 1973). In simulations, initially limiting nitrogen levels and available light levels reach final levels almost immediately, rela- tive to the time course of the species dynamics (Huston & Smith 1987). Although these prob- lems can be partly dealt with (Tilman 1990b), it seems that non-equilibrium models that do not depend on slow rates of change in resources (e.g. Huston & Smith 1987; Botkin etal. 1972a, b; Shugart & West 1980) are at least more general than the equilibrium model proposed by Tilman (1985).

Recently, Tilman (1990a) and colleagues have made detailed field studies that compare the ex- planatory power of different resource tradeoffs in old-field succession. Although this work makes valuable distinctions between different versions of the resource competition model of succession,

the goal of all of these tests is to identify which of the possible tradeoffs is at work. When the pattern conforms to the predictions of a particu- lar tradeoff hypothesis, this is taken to support the hypothesis. Yet the underlying assumption that there is a particular tradeoff critical to the natural pattern is not critically tested, and such a test may be very difficult. Shipley and Peters (1990) tested and falsified a particular version of the tradeoff hypothesis. However, neither ap- proach tests the null hypothesis of no (critical) tradeoff.

Other models and theories of successional processes

Clements' facilitation based succession Perhaps the most influential work on succession is the copious description and interpretation by Clements (1916 in Clements 1928, 1928). First hand examination of Clements (1928) suggested that Clements' ideas were not nearly as restrictive as more recent interpretation of his work implies (e.g. Anderson 1986; Drury & Nisbet 1973; Noble & Slatyer 1980; Crawley 1986; Miles 1987; Pick- ett & McDonnell 1989). However, Clements' (1928) tone of absolute and didactic generaliza- tion does detract from the considerable insight in his work.

Clements (1928, pp. 3 & 6) regarded succes- sion as "the growth or development and the reproduction of a complex organism". This 'superorganism' description assumes a level of discreteness, integration and cohesiveness amongst the constituent species and individuals that appears to have no concrete support (e.g. Gleason 1939; Underwood 1986a; Wiegleb 1989, but see Discussion).

Clements assumed succession to be solely se- quential. He considered that sequential domi- nance arises from dominant species modifying their environment (particularly soil and light) making it less favourable to themselves, and per- haps more favourable to new invaders, so that the new invaders can dominate competitively the ear- lier occupants (pp. 74, 80). The outcome of pro- gressive 'reaction' (modification of environment by the plants) was considered to be progressive

Page 11: Understanding ecological community succession: Causal models and theories, a review

stabilization, as fewer species could invade the changed environment (p. 98). This argument is supported by the simulation results of Huston and Smith (1987) above, who found that rates of successional changes decreased as succession progresses. Clements' interpretation assumes 'relay floristics', in which successive dominants arrive later than the pioneers, but only vaguely implies obligate facilitation - "plant populations of each (wave) have made conditions fit for the next c o m m u n i t y . . . " (Clements & Weaver 1938, p.71).

Furthermore, although Clements (1928 p. 79) says "The reaction of the community is regularly more than the sum of the reactions of the com- ponent species and individuals", he also places considerable emphasis on the role of individual organisms, albeit as elements of a group or stage. Whilst emphasizing the sequential stages in suc- cession, Clements does not expressly assume the composition of these stages to be fixed between different particular sequences. Finally, Clements (1928, p. 147) strongly stresses the unidirectional, exclusively progressive course of succession, to an inevitable and fixed climax, but only in the absence of disturbance. Such generalizations, al- though perhaps not strictly tenable in the light of patterns such as divergent or cyclical species abundances, do however share clear similarities with presently held views of succession.

Individualistic or stochastic succession: Gleason and Egler Clements' analogies and overgeneralizations drew a strong response from Gleason (1926,1927, 1939), who considered communities to be far more individualistic: the nature of the changes in a site depended entirely on the species composi- tion of a site, assembled by complex and stochas- tic means, and this resulted in successional courses of largely individual nature. Gleason con- sidered successional communities to be incohe- sive assemblages, and patterns of similarity to result largely from coincident properties of spe- cies and similar environmental conditions. Glea- son (1939) argued against both classification of communities into taxa according to common lea-

125

tures, and the idea that any particular community functions cohesively as an organism.

This argument was extended in a seminal paper by Egler (1954) who proposed that secondary succession in old-fields may be dictated more by the 'Initial Floristic Composition' of an area than the relay floristics of successively arriving species espoused by Clements. Thus Egler proposed an alternative hypothesis, in which succession is viewed as simply the consequence of different rates of growth of the different species that are initially present at a site. "After abandonment, development unfolds from this initial flora, with- out additional increments by further invasion (for the purpose of this discussion)" (Egler 1954, p. 415). This view implies that the appearance of a sequence of species is due to the rapid growth of some species eclipsing the growth of other spe- cies that later dominate, due to size, longevity and other life history traits. Egler does not clarify why the early species disappear. Note that Egler stated categorically that any real successional trajectory involves both relay floristic and initial floristic composition processes.

It is not commonly recognized that there are two critical differences, usually confounded, be- tween Egler's Initial Floristic Composition model and the Relay Floristics of Clements. The two views differ both in the timing of arrival of later dominating species, essentially an empirical fact, and in the nature of the interactions between spe- cies. Unlike Clements, Egler (1954) made no as- sumptions about the effects of earlier dominants on the success of later dominants, although he did suggest invasion resistance as a possible alterna- tive to the facilitatory interactions assumed by Clements.

Egler's alternative hypothesis raises the possi- bility of a null model of succession, in which suc- cession is recognized as being potentially the re- sult of life-histories, even in the absence of competition. Thus, although timing may generate succession, such patterns may occur independent of timing and species interactions. But Egler, by failing to separate the issues of timing of arrival and species interactions, did not fully formulate such a null model.

Page 12: Understanding ecological community succession: Causal models and theories, a review

126

Egler raised several other issues. He questioned the generality of any successional processes; he suggested the possibility of inhibitory effects; and he assessed the importance of colonization and extrinsic stresses and disturbances. He argued that the assumption of a single, stable endpoint or climax for any site is unproven. Finally, he criti- cized the term succession as implying saltatory change instead of the often observed gradual change, and suggested the term vegetation devel- opment as an alternative.

Drury and Nisbet's Life history based succession In a timely and highly critical review of field evi- dence against the accepted, 'Clementsian' view of succession, Drury and Nisbet (1973) proposed an alternative view very similar to that of Egler (1954). Drury and Nisbet showed that vegetation patterns of abundance following disturbances in- clude not only successional replacements, but also cycles and divergences. They asserted clearly that "succession on a single site usually involves a sequence of species (rather than simply the growth of the ultimately dominant species) because no one species can dominate the vegetation through- out the period of growth. In other words the basic cause of the phenomenon of (sequential) succes- sion is the known correlation between stress tol- erance, rapid growth, small size, short life, and wide dispersal of seed." (Drury & Nisbet 1973, p. 360).

This is essentially the first unambiguous state- ment of the concept later demonstrated by the simulations of Huston and Smith (1987 - de- scribed earlier). However, whilst Huston and Smith provide a simulated demonstration that this is sufficient to explain the process, Drury and Nisbet (1973) state it as a definitive and exclusive explanation of succession, when it was really no more than an (well-conceived) assertion.

Drury and Nisbet drew the inspiration for their hypothesis from the similarities in vegetation se- quences in temporal succession and sequences along spatial gradients of environmental stress. They considered that, if different life history strat- egies are generally mutually exclusive, then "most of the phenomena of succession can be under-

stood as consequences of differential growth, dif- ferential survival (and perhaps differential colo- nizing ability) of species adapted to growth at different points on environmental gradients." (Drury & Nisbet 1973, Summary p. 362).

A large part of Drury and Nisbet's (1973) paper consists of a very critical survey of field evidence relevant to aspects of successional theory. They concluded that the evidence did not support Clements' views, and emphasized that many natural successions are not entirely autogenic. This argument does not negate the need for a mechanism of change in autogenic succession, upon which allogenic changes can perhaps be su- perimposed.

Grime's description of plant strategies during suc- cession Grime (1974, 1979) proposed a classification scheme of plant life history strategies, based on adaptation to particular levels of disturbance and stress, using stress to mean low levels of resources. This view is similar to that of Tilman, except that it aims to give an absolute and fixed classification to each species, rather than a rela- tive rank depending on resource levels, and that it includes environmental disturbance as well as resource levels. Using the two dimensions of stress and disturbance, Grime classifies environ- ments as low disturbance-low stress; high dis- turbance-low stress; and low disturbance-high stress. He proposes three extreme strategies for these conditions: that of competitor; ruderal; and stress tolerator, respectively.

According to Grime, observed changes during succession alter the environment from high disturbance-low stress (high resources) to low disturbance-high stress, with a concomitant change in dominant plant strategies from ruderal to stress tolerators. In more productive systems, there is a middle phase of high competition, with low stress and low disturbance. This is essentially an empirical description, with post hoc interpre- tation. There is little attempt to explain the changes in species composition, except as the im- plicitly equilibrium response to unexplained changes in the environment, autogenic or other-

Page 13: Understanding ecological community succession: Causal models and theories, a review

wise. The model is thus vulnerable to the same criticisms as Tilman's model, but lacks the ex- plicit and very mechanistic consideration of resource-species interactions that Tilman gives (see earlier). Further, because it assumes plants to have fixed strategies on this two dimensional gra- dient, it predicts constant successional relation- ships between any two species. However, the relative adaptive suitability of two species may vary, depending on the particular circumstances (e.g. Smith & Huston 1989).

Connell and Slatyer's 3 mechanisms of successional species interactions Some of the most formative, but perhaps over- extended, ideas on the process of sequential suc- cession were synthesized by Connell and Slatyer (1977). In this paper, they expanded the ideas of Drury and Nisbet (1973) and Egler (1954) to sug- gest three 'alternative' models of succession: fa- cilitation, tolerance and inhibition. Put simply, "earlier colonists could either inhibit later ones, have no effect, or facilitate them." (p. 1 of a later discussion paper: Connell, Noble & Slatyer 1987). By including 'alternatives' to the previously ac- cepted facilitation pathway, Connell and Slatyer (1977) hoped to improve the possibility of testing the alternatives. They maximised the utility of these ideas by phrasing them as hypotheses, and proposing experiments to test them. The tests of these hypotheses essentially involve the removal of species to compare the colonization and growth of later species with untreated controls.

By directing successional field studies towards experimental comparisons of their specific hy- potheses, Connell and Slatyer's (1977)ideas have been seminal for the development of the field for over a decade. However, I believe that these ideas are due for critical re-examination, particularly since they have been and are still the focus of so much field effort. I feel that there are several prob- lems with the 'three-way model' of succession as generally interpreted in field studies, and that the hypotheses put forward by Connell and Slatyer (1977) have now been adequately addressed.

The 'three-way model' is not as simple as Con- nell and Slatyer claim in the original paper (1977)

127

and subsequent discussion (Connell et al. 1987). Connell and Slatyer (1977 Fig. 1) propose these three different types of interactions as part of a quite specific and detailed model, in which se- quential succession proceeds by one of three ex- clusive pathways (see Fig. 3). Aspects of both levels, pathway and interaction, are open to criti- cism.

Some problems with the idea of facilitation, tolerance or inhibition as simple species interac- tions were raised by Walker and Chapin (1987) and responded to by Connell, Noble and Slatyer (1987). The latter paper suggests that the three mechanisms model only the net effect of an ear- lier species on a later one; that the mechanisms represent the extremes of a (quantitative) con- tinuum; and that the strengths and directions of these interactions could vary within a succes- sional sequence. Thus an earlier species may have both facilitative and inhibitory effects on later species, and the primary concern is in the over- all net effect.

However, these improvements overlook sev- eral other points. Connell and Slatyer (1977) do not consider the potential importance of qualita- tive distinctions of degree in facilitation or inhi- bition. Facilitation of later species may be suffi- ciently extreme to appear as obligate facilitation, a condition with quite different implications for succession than the simple increase in rate gen- erated by a mild or non-obligate facilitation. Simi- larly, inhibition may simply slow a sequence of species by competitive inhibition of the coloniza- tion or growth by another species, or it may ac- tually prevent the species' occupancy. This dif- ference would affect the mechanism of succession: mild inhibition would allow succession by slow competition, whereas exclusive inhibition would require disturbance for succession to proceed. Thus, the 'continuum' of interactions can include obligate facilitation, facultative facilitation, toler- ance, inhibition and exclusive inhibition. Connell and Slatyer (1977, p. 1123) in summary clearly exclude all but obligate facilitation and exclusive inhibition.

Variation in direction or strength of interaction can occur not only within successional sequences

Page 14: Understanding ecological community succession: Causal models and theories, a review

128

(Connell et al. 1987), but it could alter within an interaction: it may vary with the densities of the

two species (Walker & Chapin 1987); their life- stages; their relative dominance of the commu-

Facilitation: Tolerance: Inhibition: Pathway I Pathway 2 Pathway 3

Establishment:

Effects of early occupants on Recruitment:

Replacement:

I Disturbance I opens a relatively large space, releasing resources.

I Of those species that arrive in the open space, any that are I able to survive there as adults can establish themselves. I

I

Of those species that arrive in the open space, only certain "early succession" species can establish themselves.

Environment becomes: less suitable for recruitment of "early succession" species; more suitable for recruitment of "late succession" species.

Modification of the environment by early succession species facilitates the growth to maturity of later succession species. In time, earlier species are eliminated

Environment becomes: less suitable for recruitment of "early succession" species; this has little or no effect on "late succession" species recruitment.

I Environment becomes: less suitable for recruitment of both "early succession" and "late succession" species.

Juveniles of later succession species, already present or invading, grow to maturity, despite the presence of healthy individuals of early succession species. In time, earlier species are eliminated.

f

Earlier colonists exclude, suppress or Inhibit subsequent colonists of all species, as long as the earlier colonists persist undamaged and/or continue to regenerate vegetatively.

Continuation: This sequence continues until I This sequence continues until no the resident species no longer I species exists that can invade facilitate the invasion and growth and grow in the presence of the / of other species, resident. /

Long-term or ] At this stage, further invasion and/or growth to maturity can occur only when a resident individual is I Steady State: I damaged or killed, releasing space. Whether the species composition of this community continues to I

change depends on the conditions existing at that site, and on the characteristics of the species I available as replacements. I

Fig. 3. Three pathways of succession according to Connell and Slatyer (1977). In Pathway 1 early colonizers facilitate later species; Pathway 2 involves tolerance; and in Pathway 3 early colonizers inhibit later species. At each stage, further disturbance can re- turn the system to the starting point. (Redrawn from Connell & Slatyer 1977).

Page 15: Understanding ecological community succession: Causal models and theories, a review

nity; season and other timing factors; the size of the disturbance and the length of the interaction in time (Sousa 1984 discusses the importance of context on successional processes). An interac- tion also may occur indirectly (e.g. Fig. 2; mid- successional rockweeds inhibit late successional mussels by sheltering their predators - McCook & Chapman 1991).

It is perhaps most important to reconsider the nature of successional interactions as species in- teractions. Facilitation, tolerance and inhibition are essentially empirical descriptions or classifi- cations, without explicit mechanistic bases. Im- plicit bases for each include life history factors, competition, and physiology, but the descriptions are essentially designed to answer the question "What net difference does the presence or ab- sence of the early species have on the success of the later species?" Species removals do not ex- plain how the presence of the early species causes any differences in late species success. This em- pirical perspective stems from the need for a more complete view of possible interactions than the limited view of Clements (1928) etc, and Connell and Slatyer's (1977) description led to the accu- mulation of empirical evidence for a range of in- teractions. However this description should not be mistaken for a mechanistic explanation of the changes in species abundance (but often is e.g. Begon et al. 1986). The basis of each classifica- tion is no more than the outcome of the succes- sional step, and as such depends on the process by which succession occurs.

The pathways of community succession according to Connell and Slatyer Connell, Noble and Slatyer (1987) state that their models were not intended as sufficient explana- tions of succession. Other important factors, such as seed availability and stochastic physical fac- tors were not included (nor intended to be in- cluded) in the models, because they only intended to address the empirical question posed above. Yet, their diagram (see Fig. 3 redrawn from Con- nell & Slatyer 1977 Fig. 1) maps out a detailed and exclusive course of succession for facilita- tion, tolerance or inhibition. The model does ap-

129

pear as a complete description and does not in- dicate the other factors that may be important. The model is commonly described in reviews and introductory textbooks as a complete description. Further, I argue below that each pathway includes incompletely justified assumptions. I will describe these pathways briefly, in the light of their later recognition of the variable nature of these inter- actions (Connell, Noble & Slatyer 1987). The terms 'early' and 'late' species refer to the type of species, rather than to the timing of settlement.

In each pathway initial colonization necessar- ily renders the environment less suited for 'early species' recruitment. It is also assumed that "cer- tain species will usually appear first because they have evolved certain 'colonizing' characteristics ..." (Connell & Slatyer 1977, p. 1122). Pathway 1 is distinguished by obligate facilitation by early species: later species are initially unable to colo- nize the space. Succession results from the abso- lute changes in recruitment suitability for early and late species: early species cannot persist, and late species are initially excluded. In pathways 2 and 3, early colonizers do not have exclusive ability to colonize the bare space, but any species that would be viable as an adult can colonize. In pathway 2, later species are unaffected by the presence of the early species, and succession re- sults from the loss of recruitment ability of the early species, and the tolerance to changed con- ditions of the later species. In pathway 3, how- ever, the modifications to the environment make it also unsuitable for late species recruitment, and so as long as the individuals of the initial colonists persist or regenerate vegetatively, they preclude succession.

In pathways 1 and 2, succession will continue until no' species can invade and grow in the pres- ence of the residents, as in pathway 3. At this stage, no further change will occur except when resident individuals die or are damaged. In path- way 3, damage such as herbivory may be selec- tive of the initial dominants, releasing later spe- cies from the inhibition, and allowing a successional sequence to occur. Succession also may result from differences in life-span of initial and late species: long-lived species will tend to

Page 16: Understanding ecological community succession: Causal models and theories, a review

130

accumulate if they can recruit at all, and may eventually dominate. At any point in each path- way, subsequent disturbances can cause the sys- tem to return to the starting point.

These pathways unfortunately depend on sev- eral questionable assumptions and distinctions. First, the assumption that initial colonizers ren- der the environment less suited for recruitment of early species is not sufficiently justified: a species that increases levels of a resource (e.g. nitrogen fixers) may improve the net absolute conditions for its own success. Nonetheless it may lose dominance to a species for which conditions are facilitated more - a relative decrease in suitability for the early species. Thus the facilitatory inter- action pattern may result from different processes. Such effects on a species' success also could act through growth, not only through recruitment as assumed.

The process suggested for tolerance actually involves two processes: succession through the relative inability of early species to recruit, or the tolerance of growing late species to low resource levels caused by the initial dominants. Such rela- tive increase in the viability of the late species, relative to the early species, is very similar to the process acting in the facilitation described above, except that the late species is not facilitated ab- solutely. On this basis, perhaps facilitation should be exclusively considered to involve an increase in resources, including suitable sites, by an early dominant. This would make facilitation distinct from simple changes in competitive balance that result from differential tolerances to the changed conditions.

Further, the process of succession by longevity in inhibited late species is not qualitatively differ- ent to that in the tolerance model. In both cases life history based tolerance to conditions allows competitive success of the late species. The only difference lies in the absolute effect of the occu- pants on the later species (but see below). For completely inhibited species, succession depen- dent on differential resistance to disturbance is also essentially dependent on life history based tolerances: herbivore resistance or unpalatability amounts to an investment of metabolic resources.

The only distinction lies in the immediate cause of death of early species: in all processes except complete or exclusive inhibition, the early species is killed by resource competition.

In describing their tolerance model, Connell and Slatyer (1977) use the term 'tolerance' in two different meanings: the tolerance to reduced re- source levels (e .g .p. 1122); and, implicitly by comparison to the facilitation and inhibition, the tolerance of or null effect on later species by early species. This is unfortunate, as species that are successful by tolerance to conditions deleterious to others (e.g. shade) are nonetheless likely to do better in absolute terms at higher resource levels. This situation is hence qualitatively similar to the process proposed for (mild) inhibition, but may be confounded with a null model, in which no significant species interaction effects o c c u r .

In summary, these considerations, along with Connell, Noble and Slatyer's (1987) acknowledg- ment of the continuum of interactions, suggest that succession occurs by essentially similar pro- cesses in all three situations, and that the facili- tatory or inhibitory effects of the interactions are not primary causes but empirical summaries, which describe limits or boundary conditions on the process.

Connell and Slatyer's (1977) paper has been seminal for the majority of (non-forestry) empiri- cal research on community succession, and the explicit recognition of the facilitation-tolerance- inhibition classification was a major contribution. It is a direct result of this work that clear exper- imental evidence now exists for each type of ef- fect, and that their complexity is now undeniable (e.g. Sousa 1979; Lubchenco 1982, 1983; Turner 1983; Walker & Chapin 1986, 1987; Farrell 1991; McCook & Chapman 1991, in press), if incom- pletely recognized. Further, the facilitation-toler- ance-inhibition classification can provide a mech- anistic basis for considering more complex effects. Farrell (1991) has provided a useful model for explaining and predicting the effects of grazers on successional rates, based on the Connell and Slatyer classification. Caswell and Cohen (1991) used a formalized version of Connell and Slatyer's

Page 17: Understanding ecological community succession: Causal models and theories, a review

(1977) classification to examine species diversity and coexistence.

Despite these benefits, I consider that Connell and Slatyer's successional pathways should no longer be considered as a complete model of suc- cession or as an explanation of species abun- dances. These problems are partly an inevitable result of ambiguity inherent in any verbal or con- ceptual model that is not formalized or quantita- tive (but see Caswell & Cohen 1991). Ecological theory may progress by increasing formalization and modification in response to critical theoreti- cal or empirical testing (Loehle 1988, 1990; Shrader-Frechette & McCoy 1990). I believe that there is now sufficient evidence that Connell and Slatyer's (1977) 3 pathways are not conflicting hypotheses but classifications of different possible outcomes, each of which may occur in different circumstance. The concepts in this model have all been confirmed (but rephrased) and further 'tests' of this model appear unnecessary. By focussing on the consequences of these different effects or on more formalized aspects of the causes of spe- cies abundance during succession, future work may contribute more than the accumulation of examples of the three interactions. The context of facilitation and inhibition in succession and pos- sible directions for experimental work are con- sidered further in the Discussion.

Horn's Markov&n replacement probability predic- tion model One model which has been inappropriately used to explain successional changes (e.g. Horn 1981; Begon et al. 1986) is the Markovian probability projection models advocated in particular by Horn (1975a, b, 1976, 1981, see also Usher 1987). These models use estimates of the probabilities of each tree species being replaced by the same or other species. By arranging all such tree by tree replacement probabilities into a 'projection' ma- trix, and multiplying by a vector of the initial spe- cies composition, one can project the future com- position. Dynamic systems in which transitions between states are described in probabilistic terms are referred to as Markovian processes. Horn estimated the replacement probabilities by count-

131

ing the relative densities of saplings of each spe- cies under mature trees of each species and as- suming these proportions represent the likely replacements for each species. He used these techniques quite successfully to describe changes in species composition over time in a wood near Princeton, USA.

However, although these models have the po- tential to provide very useful predictions for ap- plied research, they have two significant and re- lated disadvantages. The first involves the estimates of replacement probabilities; the sec- ond involves the lack of theoretical or causal un- derstanding contained in such a model. The model's strength, the undeniable, almost tauto- logical validity, is also its weakness: it is neither testable nor falsifiable.

With the pos sible exception of chaotic systems, the idea that the future state of a system can be predicted based on probabilities is a simple sta- tistical truism. This adds considerably to the model's appeal as a predictive tool. In fact the simplicity of the model may make it most suited for management purposes. However, this struc- ture also means that any predictions are only as good as the estimates of the replacement prob- abilities used to generate them. As Facelli and Pickett (1990) point out, such estimates either involve enormous data sets, requiring very long time periods, or enormous simplifications. Horn's (1975a, b, 1976, 1981) use of the sapling densities as estimates of replacement probabilities involves considerable assumptions about the importance of life history stages, and the constancy of these probabilities. In applying such a model, there is no measure of its validity. If the model success- fully predicts the outcome, this provides no as- surance of the model's validity; if it doesn't, the failure simply reflects on the quality of the esti- mates of replacement probabilities (or their founding assumptions). The nature of Horn's technique for estimating replacement probabili- ties based on sapling densities prevents indepen- dent tests of the assumptions about the reliabil- ity of the estimates. If the estimates and applications are based on different sites, then the tests confound site with techniques. Further,

Page 18: Understanding ecological community succession: Causal models and theories, a review

132

Horn provides no test of the assumptions or of the sensitivity of the model.

The other concern with Horn's model is that it is essentially predictive, not explanatory, and is based entirely on empirical estimates of the re- placement probabilities. The interpretation, in causal terms, of the successful application of this model to succession is that succession results from differential probabilities of replacement. This is simply an empirical observation, with little explanatory utility at all. Beyond species compo- sition, the model does not allow manipulations to investigate the importance of different factors to community composition, and assumes equilib- rium dynamics. Species interactions and other potential mechanistic factors are an implicit con- tribution to the replacement probabilities, but are entirely subsumed by the estimates, and hence their contribution cannot be examined. Thus, the structure of the model allows little insight into the processes of succession. Essentially the model treats the mechanisms and processes of succes- sion as a black box, examining output only.

Nonetheless, several properties of the model, as a statistical system, are of interest. In particu- lar, it is a property of all Markovian systems of this structure that the state of the system con- verges to a constant outcome, independent of the initial state (species composition). Thus, the model predicts that a single 'climax' state should exist for all successional courses based on the same replacement probabilities, and that the ini- tial species composition will make no difference to the stable outcome. Horn (e.g. 1981, p. 30) argues that this convergence is entirely statistical and thus requires no biological cause. However, this ignores the biological basis of the replace- ment probabilities, and hence the need for a bio- logical interpretation of the statistical conse- quence. The convergence results from the constant replacement probabilities, a biological assumption.

The assumption of constant values for the re- placement probabilities (the Markovity assump- tion), is probably a significant oversimplification of a complex biological process. In reality, re- placement probabilities are affected by recruit-

ment variations, which may change with local density dependence, time, conditions, neighbours, or herbivory and other disturbances. Since in na- ture, successions do not consistently result in the same climax community composition (Drury & Nisbet 1973), one can assume that these viola- tions of constant probabilities are significant. If replacement probabilities vary significantly, then Markovian models are probably inappropriate for natural successions in general. Horn (e.g. 1976) deals extensively with the considerable evidence for non-convergence of succession, and the im- plication that mathematical non-linearities would create such non-convergence. These non-lineari- ties result from biological properties such as den- sity dependence and disturbances. However, this is just the kind of biological interpretation that Horn denies the need for in convergence. These observations also demonstrate the inadequacy of his assumption of constant replacement probabili- ties (see also van Hulst 1980). Facelli and Pick- ett (1990) point out that as most forest succession studies are based on interpreting sites of different ages as different points on the same time trajec- tory, constant replacement probabilities also as- sume constancy in space as well as time.

Horn also discusses rates of convergence under different conditions of disturbance regimes, regeneration or recruitment, and competitive hi- erarchies or facilitatory interactions, but the ar- guments are predominantly intuitive, and inter- pretations of the effects depend on assumptions about the appropriate reference or comparison rate. It is worth noting that although Horn's ap- plications of the model always assume standing trees in the initial species composition, it can be extended by inclusion of replacement probabili- ties for open space as a non-biotic species (e.g. McAuliffe 1988).

Population dynamics and the vital attributes of Noble and Slatyer Noble and Slatyer (1980) proposed an excellent, qualitative model of vegetation population dy- namics, based on strategies. The model argues convincingly that the qualitative dynamics of spe- cies (as persistence or extinction) can be predicted

Page 19: Understanding ecological community succession: Causal models and theories, a review

on the basis of certain 'vital attributes', or life history attributes of the population as a whole. van der Valk (1981) used a similar qualitative model to explain vegetation dynamics over sev- eral disturbances in freshwater wetlands, although van der Valk's model used life history traits of individual plants rather than of populations. These qualitative models amount in part to the classification of plants into those with similar strategies. However, the purpose of both these models is to describe vegetation dynamics under varying disturbance regimes over longer time scales than individual successions. As such, al- though their dynamics include successions, they do not address the question of species abun- dances during succession. Therefore I will not discuss these models in detail, except to comment on the novel approach, and some confusion sur- rounding Noble and Slatyer's model.

In the current context, the interesting feature of van der Valk's and Noble and Slatyer's work is the use of life history attributes of individual plants, or of population vital attributes that are based on the properties of individual plants, to explain patterns at the level of population and long-term dynamics. Empirical observations at the level of individual plants and disturbances lead to mechanistic understanding at the popula- tion level across many disturbances. Population attributes, such as dispersal, establishment rate, and longevity amount to strategies, which suit the population to different conditions and disturb- ance regimes. The models are thus analogous in approach to the explanations of species abun- dance within succession, based on individual life- histories (see also Shugart 1984). This analogous approach may provide a basis for integration of ideas on 'within-succession' processes with the broader scale of vegetation dynamics over many disturbances.

Unfortunately, the ideas in Noble and Slatyer (1980) are often treated as an explanation of suc- cession rather than vegetation dynamics (e.g. Begon et al. 1986; Anderson 1986), and are often superficially described or poorly related to other ideas. Interestingly, the authors themselves barely relate the work of Noble and Slatyer (1980) on

133

vegetation dynamics to that of Connell, Noble and Slatyer (1987; Connell & Slatyer 1977) on successional interactions.

Johnstone's concept of succession as a stochastic interaction between invasion, maintenance and de-

J

cline Johnstone (1986) proposed a conceptual frame- work in which succession was considered as a stochastic interaction between invasion, mainte- nance and decline of species at a location, and is observed as the stability or instability in species composition. Johnstone argued that invasion, maintenance and decline are consequences of the biotic and abiotic environment, and provided a qualitative classification of invasion probabilities. Based on Harper's (1977) concept of safe-sites for new species, Johnstone classified barriers to invasion as botanical or non-botanical and selec- tive or non-selective of species, and argues that these two environmentally-derived criteria give rise to four, time-dependent categories of inva- sion probabilities. Johnstone considers his clas- sification of invasion probabilities as analogous to Noble and Slatyer's (1980) classification of species vital attributes, which he considers a clas- sification of maintenance probabilities. In each case, the interaction between environmental con- ditions and plant strategies explains invasion or maintenance potentials. Like Noble and Slatyer (1980), Johnstone's ideas refer more to general- ized vegetation dynamics under various environ- mental conditions, rather than the vegetation re- sponse to a specific disturbance.

Johnstone's framework suggests that different patterns of species abundance during succession result from different invasion, maintenance and decline of those species, in turn caused by envi- ronmental factors. Whilst Johnstone describes differences in these three processes, the differ- ences are not explained in detail but presumably depend on differences in the environmental fac- tors. I would argue that invasion, maintenance and decline depend on the interaction of those environmental factors with the life-history traits of the species invading etc. Johnstone considers it critical to realise that environmental factors

Page 20: Understanding ecological community succession: Causal models and theories, a review

134

cause invasion, maintenance and decline, but are not the direct cause of succession. This distinction may be less important to an explana- tion of changes in species composition that em- phasizes the properties of the species and their environment, rather than sub-processes of suc- cession.

Johnstone (1986) argues that his framework is a significant improvement because it does not use time as a cause of succession. Since time is a dimension, not a process, it cannot be a cause. However, most of the models discussed in this review use time as a variable to describe succes- sion (in time-course plots etc.) as Johnstone does, not to explain succession. Differences in timing are suggested to cause succession, but these are conceptually similar to Johnstone's time-based classification of invasion probabilities.

Factors which influence succession

I have argued that species life-histories, and com- petition between species with traits adaptive to different points on environmental gradients are sufficient to account for successional sequences. This does not mean that other processes do not add to, or also contribute to the observed pat- terns. The argument simply outlines minimal con- ditions for such patterns (see also Tilman 1985). Besides longevity, size, shade tolerance, and growth rates important species traits may include details of dispersal mechanisms and history, in- cluding stochasticity and propagule availability; vegetative regenerative abilities and seed storage properties; herbivore defences and structural de- fences against physical disturbances.

Moreover it is clear, on intuitive grounds alone, that there are many other factors that are likely to be important in generating particular patterns of vegetation development and to be unique for any particular succession. Clearly, the possibility still exists of facilitatory interactions, whereby some species simply cannot settle an area without the prior influence of other species. Allogenic changes in the environment, such as sediment or soil deposition, will certainly alter the dynamics of

growth and resource competition. Disturbances, either physical or biotic, are likely to alter the course of dynamic interactions, often selectively influencing some members of the community more than others: disturbances commonly con- sidered in the literature are physical events such as fire, weather and waves, predation and graz- ing. Successional courses are clearly influenced by patch dynamic factors such as isolation, spa- tial heterogeneity, patch size, history, initial con- ditions and seasonality of patch creation (e.g. Glenn-Lewin 1980; various chapters of Pickett & White 1985).

Discussion

The history of successional theory has been domi- nated by observations on patterns, and by de- bates and controversies about the value of vari- ous approaches (see e.g. Gleason 1926; Drury & Nisbet 1973 acknowledgements; McIntosh 1981; Finegan 1984; Anderson 1986; Johnstone 1986; Huston & Smith 1987; Miles 1987; Walker & Chapin 1987; Pickett & McDonnell 1989). I con- sider that much of this controversy and confusion is misplaced, and results from a lack of critical integration of ideas and terminology, from failure to distinguish between related ideas and from dif- ferences in purpose or focus of different models. Many models and observations have common bases or assumptions, the recognition of which provides a powerful basis for unification and gen- eralization. Clarification of terminology and clas- sification criteria have been advocated as power- ful tools in improving ecological theory (e.g. Loehle 1988).

Concepts, terminology and classification: succession, vegetation dynamics and climax The terminology of succession and vegetation dy- namics is imprecise, reflecting the enormous and continuous variability of these processes in na- ture. The use of discrete classifications or defini- tions for continuous phenomena may be a nec- essary tool, but has resulted in considerable confusion in succession, as in other fields (Loehle

Page 21: Understanding ecological community succession: Causal models and theories, a review

135

1988). The term succession itself, classifications of autogenic/allogenic and primary/secondary succession, the context of succession within gen- eral vegetation or community dynamics and 'climax' communities are all attempts to describe processes or states that are intrinsically continu- ous in nature. These continua need to be accepted for effective research. Ecologists increase their effectiveness by focussing on particular processes or circumstances, but must accept that sharp dis- tinctions are often artificial.

Natural communities exist in a dynamic state of continuous change in response to varying en- vironmental and biological conditions. Within those dynamics, the perception of which depends on the spatial and temporal scale of observation, there is a generally recognizable pattern of re- sponse to or recovery from a particular disturb- ance, although this pattern grades into the general community dynamics at larger scales of observa- tion.

Succession generally refers to changes in spe- cies composition and abundance during or fol- lowing disturbance of a site. However, other au- thors use a broader definition that includes more general vegetation dynamic processes: "the change in natural systems and the understanding of the causes and direction of such change" (Shugart 1984, p. 5). Pickett and McDonnell (1989, p. 243) focus on terrestrial forests: "Con- temporary usage (of succession) is a near syn- onym of vegetation dynamics when focus is on the decade scale. While some degree of direction- ality is implied, neither monotonic change, progress nor stable termination is nece s s ary." The assignment of a particular time scale by Pickett and McDonnell (1989) is inappropriate, as it ex- cludes many clearly successional processes (in- cluding e.g. many intertidal or decomposition successions, or post-glacial biogeographic. changes) and denies much of the ubiquity of suc- cession. Johnstone's (1986) definition of succes- sion as the outcome of loss and retention of spe- cies also involves this more general idea of vegetation dynamics.

As the changes involved in response to indi- vidual disturbances form distinct patterns, these

changes may be constructively considered as a distinct component of the more general changes and cycles of vegetation and community dynam- ics. On this basis, I have used succession to refer to the changes in species composition and abun- dance in response to a particular disturbance, as a distinct subset of community dynamics. As the response to a particular disturbance, succession will often involve directional changes such as an overall increase in biomass or change in life-form. Directional changes refer to a particular succes- sion, and do not assume a unique climax for a site. Clearly the causes of succession are commu- nity dynamic processes, and other changes will be superimposed on the successional pattern in re- sponse to environmental changes and disturb- ances at other scales.

Even within the response to individual disturb- ances, the term succession is commonly used in two different senses: generally as any vegetational change in response to disturbance; or as a salta- tory sequence of replacements in species abun- dances (e.g. Drury & Nisbet 1973 who disap- prove of this definition; Huston & Smith 1987). The latter definition is etymologically more cor- rect, involving successive replacements, but in ecology, the word succession has clearly come to refer to more general response to disturbance, even if such changes are gradual, or do not in- volve actual replacements of species. Field ob- servations (see Drury & Nisbet 1973) and the simulations of Huston and Smith (1987) suggest that species abundances can form patterns other than sequential replacements, including conver- gence, divergence, total suppression and cyclic patterns.

The different meanings of succession have both arisen from, and contributed to confusion about the possible processes involved. These difficulties prompted Egler (1954) to suggest the term veg- etation development in lieu of succession, but the literature has clearly failed to embrace this idea. Egler's term also has a connotation of organismic development, an association presently best avoided. Even the term 'revegetation' is unsatis- factory, as it excludes many important succes- sional processes based on sessile animals. Whilst

Page 22: Understanding ecological community succession: Causal models and theories, a review

136

I am unable to propose a solution to this di- lemma, the importance of explicit discrimination between the two concepts is clear. For this rea- son, and because I have primarily discussed au- togenic processes after disturbances, I have used 'succession' and 'sequential succession' to imply the general processes and sequential changes in abundances respectively.

Because succession is often taken to imply only sequential replacements, arguments tend to be limited to explaining the occurrence of such se- quences, instead of considering the wider issue of which patterns follow disturbances, and under what conditions. Too narrow a definition can short circuit the investigation by hiding assump- tions about the patterns. It is more constructive to begin with known properties of species and derive the possible patterns of abundances. These more rigourously derived patterns will lend them- selves to fewer semantic confusions.

Successions are commonly classified as pri- mary or secondary and as autogenic or allogenic. However, these are artificially discrete classifica- tions of continuous phenomena, based on the kind of disturbances initiating them. The distinction between primary and secondary succession is es- sentially a distinction in severity of disturbance to the vegetation. A disturbance severe enough to remove all of the biotic material initiates primary succession, all others initiate secondary succes- sion. As the degree of severity of disturbance is clearly a continuum, primary and secondary suc- cession refer to regions of a continuum, with pri- mary succession as an ideal, analogous to an ab- solute zero.

Similarly, allogenic and autogenic succession may be best understood in terms of the disturb- ance initiating them. In both cases, biotic inter- actions cause changes in response to external dis- turbance. Autogenic succession occurs after a rapid disturbance, allogenic succession occurs during a slow disturbance or change in the envi- ronment. The time length of the disturbance rela- tive to the response periods of the vegetation forms a continuum. Effectively instantaneous dis- turbances initiate autogenic succession. When the disturbance lasts slightly longer, some species can

react and others cannot (e.g. mobile and sessile benthic species responding to seasonal disturb- ances). A disturbance on a time scale similar to the response times of the whole community ini- tiates allogenic succession. Disturbances on the landscape/historic scale initiate evolutionary suc- cessions. Clearly autogenic succession is also an ideal: no site will be strictly free of any external disturbances.

Shugart (1984, p. 115, see also Drury & Nisbet 1973; Johnstone 1986) discusses allogenic and autogenic successions as alternate or conflicting hypotheses, and points out that they are both possible. Such conflict is misplaced: autogenic succession is better considered as an idealized point on a continuous description of disturbance, useful as a baseline model for processes of wider generality.

Confusion also arises when the terms autoge- nic and allogenic are applied to disturbances themselves (e.g. Shugart 1984, p. 65), as well as to succession, or when autogenic and facilitation- based successions are confounded (Tilman 1985). Johnstone (1986) as sociates autogenic succession with primary succession and 'Clementsian' changes, yet autogenic secondary succession can clearly follow discrete, mild disturbances (e.g. McCook & Chapman 1992).

The concept of a climax, the stable community composition that was considered to exist for any site, has been the source of considerable confu- sion and confounded arguments. Clements (1928) considered that the inevitable endpoint of any undisturbed succession was a particular, stable and mature climax composition. This idea has been sharply criticized (e.g. Pickett & McDonnell 1989). Drury and Nisbet (1973) raise these criti- cisms, and also make the useful observation that the original perception of a unique and static cli- max was developed against a background of static geomorphological theories, and before wide- spread acceptance of ideas such as plate tecton- ics and continental drift. I consider the issue to be of minimal importance to succession: the pro- cesses of succession by no means require the ex- istence of a climax, and given the wide acceptance of the continuously dynamic nature of natural

Page 23: Understanding ecological community succession: Causal models and theories, a review

communities, succession as a process becomes far more important than a putative and rare end- point to the process. Further, comparisons of original and post successional vegetation compo- sitions are inevitably confounded by historical differences, preventing any valid conclusions about the climax, and making its nature fairly academic.

Types and purposes of models Other theoretical issues relevant to succession in- clude the different purposes of models or expla- nations, reductionist, holist and hierarchical theory, and recent developments in plant compe- tition (e.g. Keddy 1989; Grace & Tilman 1990; Tilman 1990a) and vegetation theory (Pickett & Kolasa 1989; Wiegleb 1989). As this paper can only touch on each of these issues, I have tried to be explicit about my purpose and criteria.

Many of the conflicts that have arisen in the succession literature may be the result of ambi- guity about the purpose or scope of different stud- ies, or the relevant scale or level of analysis. Dif- ferent models are designed, intentionally or otherwise, to achieve different purposes, and may be unsuitable for other purposes. Discussion of types of models generally distinguishes between descriptive, predictive and explanatory models and suggests that there is a hierarchy, such that, for example, explanatory models may predict, but predictive models cannot explain (Pielou 1981; Loehle 1983; Schoener 1986; McGuinness 1988. Note that Tilman 1990a appears to misinterpret Schoener 1986 and equate predictive and mecha- nistic models). For example, the simple predictive model of Horn (1975a, b) may make it highly suited to applied and management work in the field, but I have argued that its explanatory power is minimal. Similarly, explanatory models are un- likely to provide very precise predictions of spe- cific natural systems.

Different models examine succession in terms of theory and explanation, or of management and applications; at the scales of landscape or indi- vidual organisms, and at the levels of population, community or ecosystem. In fact, as suggested for several models already discussed, whether a

137

model is explanatory may depend on the level or type of phenomena of interest. Variables studied may be primary quantities such as species densi- ties or may be derived quantities, such as diver- sity, patchiness, stability etc. Many of the pat- terns observed are dependent on the spatial and time scale of observation, yet field studies may examine widely divergent scales of process. Simi- larly, some models may be appropriate for exam- ining only a subset of autogenic/allogenic and primary/secondary successions, or particular in- teractions within a succession (e.g. see Walker & Chapin 1987; Connell et al. 1987). Application of a model out of context will inevitably lead to con- flicts. As Shugart (1984) points out, the best model is that best suited to the needs of the study, and no model can hope to explain all aspects of all successions (Tilman 1985). Assessment of a model should include explicit consideration of the purpose of the model, even if the authors them- selves are unclear or misguided about their model's purpose.

Further, the explanatory value of a model de- pends on both the aspect of succession to be explained, and on the perspective of the observer. Connell and Slatyer (1977) explained the differ- ent impacts of occupants on subsequent coloni- zation, as they intended, but did not explain dif- ferent patterns of abundance, as is often stated. The nature of explanation (cf. description or in- terpretation) has been debated (Shrader-Fre- chette & McCoy 1990; Loehle 1990), but may be at least partly personal. Johnstone (1986) consid- ers reduction of succession to invasion, mainte- nance and decline to explain satisfactorily suc- cession, whereas Huston and Smith (1987) consider that succession (or invasion etc.) needs further reduction to the specific properties of or- ganisms for satisfactory.explanation.

As stated, the purpose of this review is to as- sess models as explanations of patterns of species abundance, at the community level, in order to frame testable hypotheses for field study. I am specifically interested in explaining those patterns in terms of the interaction between organisms and their environment.

For this purpose, a minimal explanation of

Page 24: Understanding ecological community succession: Causal models and theories, a review

138

natural succession requires four steps: statement of species properties; statement of (minimal) ini- tial conditions and physical parameters; exami- nation of model community behaviour; and test- ing or comparison with the natural process. In stating the initial conditions and species attributes the author should distinguish between universal and regional, factual and assumed, and correlated and causally related phenomena. These distinc- tions are important to the applicability of the model.

Further, the most useful model should be based on clear, rigourous, explicit concepts (minimal structures - Pickett & McDonnell 1989) and re- quire few such assumptions; it should be rigour- ously testable; and it should lend itself to incor- poration in models of more complex structures and processes. Such models would then be a min- imal basis for modelling more complex processes, and could predict the importance of a process in natural succession. It is worth noting that the apparent mathematical complexity of the simula- tion models of Huston and Smith (1987) etc. belie a conceptual simplicity and concreteness not found in any of the other models.

Caution is needed in understanding succession by means of models, especially more concrete models such as mathematical projections and simulations. There is a tendency to confound the model or its outcome with reality. Such confusion may be subtle and largely due to careless word- ing, but it is important to distinguish between results of simulations and results in nature.

Understanding any phenomena in science can be obstructed by over-emphasis on complete gen- erality, and consequent unnecessary conflicts be- tween observations. It is more constructive to examine critically the applicability of components of a theory than to insist on either complete ac- ceptance or rejection ('theory maturation' Loehle 1988, 1990). In particular, attempts to explain completely all successional patterns with a single mechanism, an apparently impossible task, have obscured much potential insight into both general and specific processes. For example the wide- spread dismissal of facilitation as the prime cause of all successional replacements has tended to

minimize useful consideration of the potential va- lidity or the mechanism of the process (e.g. Peet & Christensen 1980; Botkin 1981; Tilman 1985; Huston & Smith 1987).

Testing of models Unfortunately the extremely long-term dynamics of terrestrial forest systems have effectively ruled out rigourous experimental tests of many models and hypotheses (e.g. Botkin 1981; Horn 1976). Most tests of proposed successional processes in terrestrial forests involve either the reconstruction of historical species abundances, or spatial chro- nosequences. The latter method involves the in- terpretation of sites of different ages as different time points for a single site. Neither of these methods allow strictly rigourous interpretation of differences in causal terms, as they do not allow manipulations of species etc, and the use of chro- nosequences to estimate successional courses necessarily confounds site with age of site.

Shugart et al. (1981) distinguish between veri- fication and validation of models. Verification re- fers to the modification of the model parameters to make it consistent with a set of observations, whereas validation procedures involve the testing of the model against a set of observations inde- pendent of those used to structure the model.

However, these tests simply test the predic- tions of the models, not their explanatory power. Goodness-of-fit tests are appropriate to predic- tive models, but theoretical models require hypothetico-deductive methods (Loehle 1983), such as used by Huston and Smith (1987). I have briefly suggested both benefits and limitations of field tests of the Connell and Slatyer's (1977) clas- sification of succession. Approaches which com- bine tractable field systems with explanatory modelling, such as Tilman's (e.g. 1990a) admi- rable work in oldfields, should prove more theo- retically productive.

The forest or the trees: the great successional con- troversy: Clements, holism, and reductionism Probably the most impassioned and perhaps the least constructive controversy in successional lit- erature has revolved around three aspects of

Page 25: Understanding ecological community succession: Causal models and theories, a review

Clements' (1928) work: the concept of obligate facilitation in succession; the emphasis on devel- opment to a stable 'climax' community structure; and community level interactions and holistic community behaviour, exemplified by the analogy of the community as a 'superorganism', whose development parallels that of a developing organ- ism. Particularly troubling is the failure to recog- nize the independence of these concepts: the evi- dence against the widespread existence of climax states or integrated community 'superorganism' behaviour do not deny the utility of emergent or collective properties, or preclude facilitation in nature• There is also a tendency to confound Clements' use of large spatial scales or ecosystem level processes with his often holistic perspective•

In reviewing Clements' work, I argued that much of the controversy over Clements work fails to recognize the strong kinship his ideas had with the more reductionist ideas currently in vogue• I gave several examples of such kinship, and ex- pand on one below. The emphasis on Clements analogy of the community development in suc- cession with that of an organism has drawn at- tention away even from the more important issue of emergent, holist properties and from the em- phasis on individual species traits that also per- vades his work.

This dual emphasis may be seen in his discus- sion of 'reactions': "The reaction of the commu- nity is regularly more than the sum of the reac- tions of the component species and individuals. It is the individual plant which produces the re- action, though the latter usually becomes recog- nizable through the combined action of the group • . .A community of trees casts less shade than the same number of isolated individuals, but the shade is constant and continuous, and hence controlling." (Clements 1928, p. 79). This para- graph seems clearly to imply a community prop- erty as a collective property of the component individuals. Salt (1979) makes a useful distinction between 'emergent' and 'collective' properties of a group of individuals: the latter can be inferred directly from the components of the system, al- though not necessarily as a linear summation• In this sense, Clements' overtly holistic paragraph

139

may be seen as emphasizing individual contribu- tions. Shading is undeniably a collective property of a stand, and clearly is more (or less) complex than a simple, linear sum of the contributions from individual trees.

The strong polarization between holist and re- ductionist approaches may be inappropriate. Rather than conflicting, these perspectives may be complementary aspects of an integrative and hierarchical approach to ecological theory, in which higher level phenomena are explicitly de- rived from, but interact with, lower level struc- tures (Mclntosh 1981; Levins & Lewontin 1982; Allen & Starr 1982; Shugart 1984; Underwood 1986a; Loehle 1988; Wiegleb 1989; Keddy 1989). Loehle (1988) showed that holism is commonly used to describe a widely divergent range of ap- proaches, and can allow for the useful integration from lower levels of theory. Wimsatt (1980) sug- gested that "reduction in science is better seen as the attempt to understand explanatory relations between different levels of phenomena, each of which is taken seriously in its own right." Indeed, the most explanatory models of succession are those that are most explicitly reductionist, based on physiological properties of the plants, yet these models are probably also the most integrative and useful at higher levels.

These ideas are somewhat abstractly formal- ized by Pickett et al. (1989) and Pickett and Mc- Donnell (1989)who suggest the existence of par- ticular 'minimal structures' or appropriate minimal levels of reduction for understanding a higher level structure. In much the way that the above simulation models have done, Pickett and colleagues suggest the derivation of higher level properties from sets of basic and explicit con- cepts, and that if the minimal structure chosen is suitable then the higher level structures will be explained adequately and without conflict. Inap- propriate choice of basic structures or inappro- priate integration will generate conflicts and in- adequate explanations.

Thus, Botkin (1981), and Huston and Smith (1987) suggest that individual trees are the most appropriate unit for modelling forest succession, and that community and even ecosystem and

Page 26: Understanding ecological community succession: Causal models and theories, a review

140

landscape level properties are best derived from the results of such models. These models derive the community properties from explicit consider- ation of growth rates and similar physiological level properties, but do not deny the importance to individuals of integrative or collective proper- ties of the community. Allen and Starr (1982, p.177) say of the individual-based simulation models: "The individual trees and individual spe- cies assert themselves as quasi-autonomous wholes, while at the same time contributing to the vegetation-mediated environment within the stand. It is this dual existence of simultaneous wholeness and partness which is the cornerstone of hierarchy theory..." Similarly, the explanatory power of tree-based simulation models for eco- system level phenomena may be greater than non- hierarchical, empirical models based on derived properties such as diversity (e.g. Shugart 1984). These considerations support my argument that successions other than autogenic succession can best be understood by starting with a minimal model of autogenic succession processes, and building the individual circumstances of allogen- esis, starting conditions, patch dynamics, and historical influences onto this foundation.

Competition, species strategies, constraints and tradeoffs Some confusion has arisen from the various schemes about species differential adaptation, by life history traits, for different conditions. Such schemes tend to assume particular sets of values for different attributes, or to assume anthropo- centric assessments of severity, which may not be broadly applicable. K selection, competitor, light competitor or 'late' species strategies are com- monly considered 'mature' strategies, by associa- tion with competitive hierarchy and equilibrium views of community structure. However, such hi- erarchies may depend on a particular set of con- ditions (e.g. Grubb 1987), and the equilibrium view of climax community structure has been dis- credited.

It is more useful to consider the traits of a species case by case as strategies for the particu- lar conditions. Plant strategies probably form a

multidimensional continuum and result in differ- ent hierarchies of relative adaptation to each dif- ferent set of conditions (Huston & Smith 1987; Smith & Huston 1989). Horn (1981 p.35) states "...no absolute statements may be made about any particular species and its presumed adapta- tion to a given successional status, just as no absolute statements should be made about adap- tation in general...The only statements that should be made are comparative ones involving differ- ences between species and differences in succes- sional status at a given time and place."

Several models of succession include species referred to a priori as 'Early' and 'Late' species, and endowed with the set of traits often observed (or assumed) in species of that class. Arguments are then given that suggest that the succession depends on those traits, because the species have them. However, this approach is nearly circular, as the purpose of the model may be to determine whether some species come before others (cf. Johnstone 1986). The choice of species labels car- ries implicit traits, which at least makes critical examination of the model difficult. A better ap- proach is to derive the species 'timing' from its properties, and to use labels such as strategies (e.g. ephemeral; ruderal; or better simple num- bers, i.e. species 1, 2 etc. - e.g. Huston and Smith 1987).

Constraints on or tradeoffs in species life his- tory strategies have considerable potential to im- prove our understanding of community dynamics (e.g. Keddy 1989; Petraitis etal. 1989; Grace 1990; Austin 1990; Tilman 1990a, b). Tilman (1990a) considers these ideas will provide a uni- fying and simplifying theme for vegetation com- petition theory, and Loehle (1988, 1990) consid- ers them fundamental ecological principles. That such constraints are important is in a sense a truism, yet they are only recently receiving direct attention, and there have been few tests of their importance (but see Shipley & Peters 1990; Til- man 1990a, b). Indeed there are apparent excep- tions to the generalizations about constraints on possible strategies. A number of unusual species, known as 'superspecies', have rapid growth as well as large size and long life in particular habi-

Page 27: Understanding ecological community succession: Causal models and theories, a review

tats (e.g. Grubb 1987). These species may be dominant throughout the course of a succession, an observation that adds weight to the idea that constraints on strategies are important to sequen- tial succession. Detailed explanations of how such species avoid the restrictions other species face will presumably involve unusual combinations of environmental conditions and plant physiologies. A complete review of ideas on resource allocation strategies and tradeoffs is beyond the scope of this paper, but I advocate their explicit consider- ation in future experimental or theoretical work on succession.

The importance and nature of facilitation, tolerance and inhibition Much interest has focused on the effect that se- quential successional species have on subsequent species, and the importance of such interactions to the successional processes. In reviewing the development of these ideas in the work of Clem- ents, Egler, Drury and Nisbet, and most impor- tantly, Connell and Slatyer, I have questioned the value of facilitation, tolerance and inhibition as bases, causes or mechanisms for succession. I have suggested that these concepts are actually empirical observations, and that rather than being a model for succession, they are, like the entire successional process, a particular aspect of the outcome of interacting individual life-histories.

As such, mild successional interactions may be understood as acting like quantitative moderators of rate of the qualitative successional process that arises from the interacting differential adaptive- ness of different life-histories. Strong interactions, such as obligate facilitation, or complete, exclu- sive inhibition may be understood as limits, re- strictions or boundary conditions on the process. In such a simple case, the rate of growth of a later dominant would increase with the amount of the earlier facilitatory species and decrease for an in- hibitory species: the rate would be zero when the density of an obligate facilitatory species ap- proaches zero, and when that of a strong inhibi- tor approaches some critical level.

Horn (1981) suggested that facilitation, toler- ance and inhibition are simply statements about

141

the comparative life-histories of different species found in association with each other. More spe- cifically, I suggest that these 'simple statements' are 'simple quantitative statements' about com- parative life-histories. As reviewed earlier, Botkin's (1981) results for the simulated removal of early or late successional species indicate a variable quantitative inhibition by both early and late species on the other category. It appears that the presence of inhibitory species modifies the dynamics of the other species.

Interpretation of possible facilitatory or inhibi- tory effects depends on the type of effect investi- gated. As implied earlier, the test of such effects requires the removal of the putative interactive species (species A), and comparison of the growth or recruitment of the following species (B) with and without the species A. B otkin (1981) outlines a more rigourous design, involving a control un- manipulated plot; a treatment in which species A is removed; and a second treatment in which an equivalent amount of species B is removed. Com- paring the recruitment or growth of B in the two treatments indicates the effect of A on B relative to the effect of B on itself. Inhibition is deduced if growth of B is greater when A is removed than when similar amounts of B are removed. Facili- tation is shown by a decreased growth of B when A is removed, relative to the removal of B. Analo- gous designs are used in studies of competition (Underwood 1986b).

However, I am unconvinced that this is the sole useful interpretation of these interactions. First, many potential effects, particularly fast-acting types of obligate facilitation, may be obscured by any occurrence of species A, before removal. As discussed earlier, it appears that many changes in resource levels occur very rapidly relative to the dynamics of the species involved in that change. This suggests that 'prevention', not removal may be required. Further, if species B were dependent on the initial presence of species A, prevention/ removal of B would be impossible or irrelevant. I consider that the absolute effects of the initial species on later species are also of value to un- derstanding succession, especially early succes- sion, when occupation is low. The question being

Page 28: Understanding ecological community succession: Causal models and theories, a review

142

asked is "Would there be more or less of species B if species A had never occurred there?" This question addresses the absolute effects of the spe- cies interactions, not the effects relative to those of the affected species itself. Thus both absolute and relative effects of species are of interest. Clear interpretation of results at least requires a distinc- tion between the two effects. The identification of an effect also should state clearly the measures used for species abundance, as these effects may vary between measures such as biomass, growth, recruitment, rates of growth, or cover.

The significance of these effects to the process of succession, particularly simulated succession, also can be detected by manipulating more than simply the species concerned. Modifying the re- source levels, or the species dependence on or effects on resources will give insight not available from simple species removals. Such manipula- tions could include increases and decreases in species densities; the levels of the resource in- volved in the interaction could be limited or sup- plied in abundance; and details of species traits, such as excess resource generation or resource dependence, could be altered. Such modifications are simple in simulations, and in the field may involve fertilizations or resource removals or buffering; shading and supplementary lighting; and species removals, additions and substitutions of species with slightly different properties. Thus the impact of facilitation on a successional se- quence may be examined by removing some or all of the facilitatory species; by removing the facili- tated species; by decoupling the increase in re- sources from the facilitator's success; and by ini- tial artificial supply of the resource.

Finally, the processes generating inhibition and facilitation are of interest. Inhibition is easily un- derstood as the effects of resource depletion, ei- ther directly or indirectly. However, facilitation is more difficult to explain without invoking altru- istic traits, such as implied by the superorganis- mic facilitation of Clements. The selection of species that facilitate their own demise as a popu- lation is quite counter-intuitive, and may account for much of the controversy over the role of fa- cilitation. However, these confusions stem in part

from inappropriate consideration of species as primarily adapted to a successional role, simply because the species is often found in that role (e.g. 'early species'). Such consideration would imply altruism in the facilitatory species.

A more sensible basis for facilitation is pro- vided by patch dynamic concepts and ideas about species strategies as relative and dependent on the particular circumstances. A species may have traits, such as nutrient capture or synthesis, which make it suited to conditions at particular regions of spatial and temporal resource gradients. These traits may coincidentally make it likely to be found during early stages of successions, and perhaps also to facilitate the success of another species. Such a species may be every bit as 'mature' or competitively successful under the conditions to which it is primarily adapted. Under some (or all) circumstances, this species may produce an ex- cess of a limiting nutrient, and will be dominated by other species, but as long as sufficient sites exist in time and/or space, the 'facilitator' will continue to persist on a large scale. This argu- ment amounts to the dynamic niche separation concept central to vegetation patch dynamics, and portrays facilitation as a kind of parasitism ( - + interaction).

However, although these arguments explain the persistence of facilitators, they fail to explain why these species would not be selected to avoid pro- viding resources that otherwise limit their current competitors. For example, nitrogen fixers may be dominated by later species with greater shade tol- erance, but only if the nitrogen-fixers leak enough nitrogen to allow such dominance. The only ex- planation for the existence of such altruistic fa- cilitators is that excess resource provision is an inevitable consequence of the biological processes involved in the species' traits. For example, spe- cies suited to spatial conditions where resources are continually removed may need to generate quantities of that resource which result in facili- tation of dominants under other circumstances. Alternatively, a species that only exists as an early successional 'fugitive' may provide the facilitatory resource as an inevitable consequence of its strat- egy. Examples of such circumstances include

Page 29: Understanding ecological community succession: Causal models and theories, a review

structural resources, such as settlement sites for species unable to recruit to bare substrate. For example, on seashores some late successional dominants appear unable to recruit to bare sub- strate, and depend on the structure of prior oc- cupants for colonization (Turner 1983; Farrell 1991; McCook & Chapman in press). Facilita- tion therefore must be viewed as a redundant but inevitable cost of either spatial or temporal niches (or both).

Conclusions and directions

Reviews of successional processes and their causes tend to be pessimistic about the progress to or even the possibility of generalization of mechanisms, as suggested earlier (e.g. Finegan 1984; Anderson 1986; Miles 1987; Usher 1987; Shrader-Frechette & McCoy 1990). Horn (1976, p. 202) states "The only sweeping generalization that can safely be made about succession is that it shows a bewildering variety of patterns". Whilst less pessimistic about the likelihood of synthesis, McIntosh (1981, p.23) states "An approach to a synthes is . . .would be more likely to occur if the groups . . .were less cliqueish and less prone to as sert their positions within the cloud of their own program."

Loehle (1988, 1990) and Shrader-Frechette and McCoy (1990) recently debated the potential for explanation of succession in terms of the proper- ties of constituent organisms. They disagreed about whether succession is understood with suf- ficient precision at the community level to be ex- plained by reduction, and about whether commu- nity or population phenomena can be explained by organism phenomena, if community/popu- lation and organism belong to different types of classification scheme. The first issue may be largely resolved by restricting the discussion to explanations of autogenic succession, by recog- nition that facilitation, tolerance and inhibition are not conflicting mechanisms of succession (cf. Shrader-Frechette & McCoy 1990p.110), and perhaps by a more critical integration of different ideas about succession. Terminological clarifica-

143

tion may also help resolve the second issue: if communities are considered simply as assem- blages of organisms at a site, rather than typo- logical (phytosociological) classifications, then community (or assemblage etc) phenomena may be expected to derive from the properties of the component organisms.

I have tried to present a much more optimistic perspective in this review. I have suggested that the support for a potential general basis for au- togenic succession is now quite strong, although it is far from proven in nature. This support comes in particular from experimental manipulations in simulation models. I here reiterate and synthesize this perspective.

It appears that various patterns of species oc- cupancy can result from the different life history traits of species adapted to different regions of continua of various resources. In general, limits on the allocation of metabolic resources result in tradeoffs between growth, and specialization for low levels of resources, particularly light. Such tradeoffs result in an inverse correlation between traits that favour species early in revegetation of a site, such as dispersal and rapid growth, and those which favour long term dominance, such as shade tolerance, longevity, height, and resistance to damage. When these traits are inversely cor- related, successional replacements will result. Species able to colonize and grow rapidly will invade and dominate a site. However, such spe- cies will generally be gradually replaced by spe- cies able to recruit and grow better in the low substrate light levels, and able to resist causes of mortality.

However, if for some reason, such a correlation does not exist between species traits, other pat- terns of species occupancy can occur, including divergent or convergent species densities, cyclic replacements, and complete suppression by single species.

Finally, the detailed expression of this process can be modified by numerous factors, including species availability, disturbance levels, history, and, in particular, the effects of different species on each other. Species traits may result in their preclusion by or dependence on the presence of

Page 30: Understanding ecological community succession: Causal models and theories, a review

144

other species, to varying degrees, and with vary- ing degrees of specificity.

This optimistic stance can provide improved direction and productivity for future work. Theo- retical work could concentrate on extending the experimental manipulative simulation work of Botkin (1981) and Huston and Smith (1987), in the same rigourous and mechanistic manner. The importance of particular species traits in particu- lar situations could be investigated by altered lon- gevity, dispersal rates and other aspects of meta- bolic resource allocation. Particularly needed are tests of the potential effects and importance of inhibition and facilitation, by decoupling different effects of species and their interactions, as dis- cussed earlier. This approach should provide an understanding of the relationships between dif- ferent successional processes based on functional homology, not simply on similarities in patterns.

There is also a need for expansion of simula- tion models to examine ideas such as allogenic changes, patchiness, different starting conditions and disturbances. These results would allow in- tegration of patch dynamics, large scale dynam- ics and disturbance ecology concepts with suc- cessional understanding (see e.g. Smith and Huston 1989).

Especially useful would be construction of similarly powerful and experimental models for systems other than terrestrial forests, and particu- larly for systems more amenable to experimenta- tion and rigourous field testing. Such extensions should allow more understanding of the general- ity of successional processes, should improve in- tegration of field work with theory, and, as sug- gested, should improve the testing of models.

Concentration of field studies on relatively simple species assemblages, with rapid dynamics, small scale spatial interactions, and species ame- nable to manipulation will improve the opportu- nities for rigourous tests of hypotheses. Obvious candidates for such work are rocky seashores (e.g. Farrell 1991; McCook & Chapman 1991, in press), saltmarshes (e.g. Bertness 1991) and other wetlands (e.g. Shipley & Peters 1990), and old- field grassland communities (e.g. Tilman 1990a).

However, field studies also would benefit from

a more rigourous basis in current theories on suc- cession, resource allocation strategies, plant com- petition and vegetation theory. Further theoreti- cal or experimental criticism of Clements' 'traditional views of succession' seems unneces- sary. As a new tradition of criticizing Clements appears well established, future work may now focus on more constructive directions. Evidence for both valid successional facilitation and inhi- bition, and for the complexity of these interac- tions, is now sufficient (e.g. Sousa 1979; Turner 1983; Walker & Chapin 1986; Farrell 1991; Mc- Cook & Chapman in press). Although an essen- tial distinction, further tests of Connell and Slatyer's (1977) three pathways as alternative hy- potheses seem unnecessary. Investigating the conditions likely to result in facilitation or inhi- bition is also unlikely to be useful, except in the context of interacting life-histories. However, un- derstanding the consequences of facilitation and inhibition for other effects, such as herbivory, should be valuable (Farrell 1991).

These considerations emphasize the need for integrated theoretical and field work in more ame- nable systems, and especially for a rigourous re- view of extant field research findings, in the light of the emerging understanding of successional processes. It is critical, but challenging, that such reviews be more than anecdotal accounts of examples of different influences. A systematic and critical review is needed to examine the degree of solid evidence for different processes, with explicit consideration of the quality of the results. Although illustrative reviews (e.g. Drury & Nisbet 1973) serve very important functions, a thorough survey will provide a far more complete picture.

In conclusion, I have considerable optimism about understanding of the causal bases for suc- cession. Considerable consensus exists in theo- retical considerations of successional processes, despite the confusion and controversy in both the theoretical and field literature. These ideas, if borne out by future work, should form a strong basis for improving understanding of, and appli- cations to, broader vegetation dynamic issues, at population, community, ecosystem and landscape

Page 31: Understanding ecological community succession: Causal models and theories, a review

levels. Given an integrative yet rigourous ap- proach to both theoretical and field studies, the prospects are good for understanding succession and revegetation processes.

Acknowledgements

Critical and thorough comments and discussion by A. R. O. Chapman, S. Walde and especially two anonymous reviewers markedly improved the content and tone of this manuscript, for which I am very grateful. I also thank J. H. Connell and S. Swarbrick for helping to clarify some of my ideas. However, I am solely responsible for any remaining shortcomings in the review.

References

Allen, T. F. H. & Starr, T. B. 1982. Hierarchy: Perspectives for ecological complexity. University of Chicago Press, Chi- cago.

Anderson, D. J. 1986. Ecological succession. In: Kikkiwa, J. & Anderson, C. J. (eds), Community Ecology: Pattern and Process. Blackwell Scientific Publications, Melbourne, pp. 269-285.

Austin, M. P. 1990. Community theory and competition in vegetation. In: Grace, J. B. & Tilman, D. (eds), Perspec- tives on Plant Competition. Academic Press, San Diego, pp. 215-238.

Begon, M., Harper, J. L. & Townsend, C. R. 1986. Ecology: Individuals, populations, and communities. Sinauer Asso- ciates, Inc, Sunderland, Massachusetts.

Bertness, M. D. 1991. Interspecific interactions among high marsh perennials in a New England salt marsh. Ecology 72: 125-137.

Botkin, D. B. 1981. Causality and succession. In: West, D. C., Shugart, H. H. & Botkin, D. B. (eds), Forest Succes- sion: Concepts and Application. Springer-Verlag, New York, pp. 36-55.

Botkin, D. B., Janak, J. E. & Wallis, J. R. 1972a. Some ecological consequences of a computer model of forest growth. J. Ecol. 60: 849-872.

Botkin, D. B., Janak, J. F. & Wallis, J. R. 1972b. Rationale, limitations and assumptions of a Northeast forest growth simulator. IBM J. Res Develop. 16: 101-116.

Caswell, H. & Cohen, J. E. 1991. Disturbance, interspecific interaction and diversity in metapopulations. Biol. J. Linn. Soc. 42: 193-218.

Clements, F. E. 1928. Plant Succession and Indicators: A definitive edition of Plant Succession and Plant Indicators. The H. W. Wilson Company, New York.

145

Clements, F. E. & Weaver, J. E. 1938. Plant ecology. McGraw Hill, New York.

Connell, J. H., Noble, I. R. & Slatyer, R. O. 1987. On the mechanisms producing successional change. Oikos 50: 136-137.

Connell, J. H. & Slatyer, R. O. 1977. Mechanisms of Suc- cession in Natural Communities and their Role in Com- munity Stability and Organization. Am. Nat. 111:1119- 1144.

Crawley, M. J. 1986. The Structure of Plant Communities. In: Crawley, M. J. (ed), Plant Ecology. Blackwell Scientific Publications, Oxford, pp. 1-50.

DeAngelis, D. L. & Waterhouse, J. C. 1987. Equilibrium and nonequilibrium concepts in ecological models. Ecol. Monog. 57: 1-21.

Drury, W. H. & Nisbet, I. C. T. 1973. Succession. J. Arnold Arbor. 54: 331-368.

Egler, F. E. 1954. Vegetation Science Concepts I. Intitial Floristic Composition, A Factor in Old-Field Vegetation Development. Vegetatio 4: 412-417.

Facelli, J. M. & Pickett, S. T. A. 1990. Markovian Chains and the Role of History in Succession. TREE 5: 27-30.

Farrell, T. M. 1991. Models and mechanisms of succession: An example from a rocky intertidal community. Ecol. Monog. 61: 95-113.

Finegan, B. 1984. Forest Succession. Nature 312: 109-114. Gleason, H. A. 1926. The Individualistic Concept of the Plant

Association. Bull. Torrey Bot. Club 53: 7-26. Gleason, H. A. 1927. Further Views on the Succession Con-

cept. Ecology 8: 299-326. Gleason, H. A. 1939. The individualistic concept of the plant

association. Amer. Midl. Natur. 21: 92-110. Glenn-Lewin, D. C. 1980. The individualistic nature of plant

community development. Vegetatio 43: 141-146. Grace, J. B. 1990. On the relationship between plant traits

and competitive ability. In: Grace, J. B. & Tilman, D. (eds), Perspectives on Plant Competition. Academic Press, San Diego, pp. 51-65.

Grace, J. B. & Tilman, D. 1990. Perspectives on plant com- petition. Academic Press, San Diego.

Grime, J. P. 1974. Vegetation Classification by Reference to Strategies. Nature 250:26-31.

Grime, J. P. 1979. Plant Strategies and Vegetation Processes. John Wiley & Sons, Chichester.

Grubb, P. J. 1987. Some Generalizing Ideas about Coloni- zation and Succession in Green Plants and Fungi. In: Gray, A. J., Crawley, M. J. & Edwards, P. J. (eds), Colonization, Succession and Stability. Blackwell Scientific Publications, Oxford, pp. 81-102.

Harper, J. k. 1977. The population biology of plants. Aca- demic, London.

Horn, H. S. 1975a. Forest Succession. Sci. Am. 232: 90-98. Horn, H. S. 1975b. Markovian Properties of Forest Succes-

sion. In: Cody, M. L. & Diamond, J. M. (eds), Ecology and Evolution of Communities. Belknap Press of Harvard Uni- versity Press, Cambridge Mass., pp. 196-211.

Page 32: Understanding ecological community succession: Causal models and theories, a review

146

Horn, H. S. 1976. Succession. In: R. M. May (ed), Theoreti- cal Ecology. Blackwell, Oxford, pp. 187-204.

Horn, H. S. 1981. Some Causes of Variety in Patterns of Secondary Succession. In: West, D. C., Shugart, H. H. & Botkin, D. B. (eds), Forest Succession: Concepts and Ap- plication. Springer-Verlag, New York, pp. 24-35.

Huston, M. 1979. A General Hypothesis of Species Diversity. Am. Nat. 113: 81-101.

Huston, M. & Smith, T. 1987. Plant Succession: Life History and Competition. Am. Nat. 130: 168-198.

Johnstone, I. M. 1986. Plant invasion windows: a time-based classification of invasion potential. Biol. Rev. 61: 369-394.

Keddy, P. A. 1989. Competition. Chapman and Hall, Lon- don.

Levins, R. & Lewontin, R. 1982. Dialectics and Reduction- ism in Ecology. In: Saarinen, E. (ed), Conceptual Issues in Ecology. D. Reidel Publishing, Dordrecht, pp. 107-138.

Loehle, C. 1983. Evaluation of theories and calculation tools in ecology. Ecol. Modell. 19: 239-249.

Loehle, C. 1988. Philosohical tools: potential contributions to ecology. Oikos 51: 97-104.

Loehle, C. 1990. Philosohical tools: reply to Shrader-Frechette and McCoy. Oikos 58: 115-119.

Lubchenco, J. 1982. Effects of grazers and algal competitors on fucoid colonization in tide pools. J. Phycol. 18: 544- 550.

Lubchenco, J. 1983. Littorina and Fucus: effects of herbivores, substratum heterogeneity, and plant escapes during succes- sion. Ecology 64:1116-1123.

MacArthur, R. H. & Wilson, E. O. 1967. The Theory of Is- land Biogeography. Princeton University Press, Princeton.

McAuliffe, J. R. 1988. Markovian dynamics of simple and complex desert plant communities. Am. Nat. 131: 459- 490.

McCook, L. J. 1992. Species Interactions and Community Structure during Succession following Massive Ice-scour of a Rocky Intertidal Seashore. Doctoral Thesis Dalhousie University, Halifax, Nova Scotia, Canada.

McCook, L. J. & Chapman, A. R. O. 1991. Community suc- cession following massive ice-scour on an exposed rocky shore: Effects of Fucus canopy algae and of mussels during late succession. J. Exp. Mar. Biol. Ecol. 154: 137-169.

McCook, L. J. & Chapman, A. R. O. 1992. Vegetative re- generation of Fucus rockweed canopy as a mechanism of secondary succession on an exposed rocky shore. Bot. Mar. 35: 35-46.

McCook, L. J. & Chapman, A. R. O. (in press). Community succession following massive ice-scour on a rocky intertidal shore: recruitment, competition and predation during early, primary succession. Mar. Biol.

McGuinness, K. A. 1988. The use, abuse, and usefulness of models in ecology. In: Australian Marine Sciences Asso- ciation, N. B., Modelling in Marine Science. AMBS, NSW, NSW.

McIntosh, R. P. 1981. Succession and Ecological Theory. In: West, D. C., Shugart, H. H. & Botkin, D. B. (eds), For-

est Succession: Concepts and Application. Springer-Vedag, New York, pp. 10-23.

Miles, J. 1987. Vegetation Succession. In: Gray, A. J., Craw- ley, M. J. & Edwards, P. J. (eds), Colonization, Succession and Stability. Blackwell Scientific Publications, Oxford, pp. 1-29.

Noble, I. R. & Slatyer, R. O. 1980. The Use of Vital Attributes to Predict Successional Changes in Plant Communities Subject to Recurrent Disturbances. Vegetatio 43: 5-21.

Odum, E. P. 1971. Fundamentals of Ecology. W. B. Saunders Company, Philadelphia.

Peet, R. K. & Christensen, N. L. 1980. Succession: a popu- lation process. Vegetatio 43: 131-140.

Petraitis, P. S., Latham, R. E. & Niesenbaum, R. A. 1989. The maintenance of species diversity by disturbance. Quart. Rev. Biol. 64: 393-418.

Pickett, S. T. A. & Kolasa, J. 1989. Structure of theory in vegetation science. Vegetatio 83: 7-15.

Pickett, S. T. A., Kolasa, J., Armesto, J. J. & Collins, S. L. 1989. The ecological concept of disturbance and its expres- sion at various hierarchical levels. Oikos 54: 129-136.

Pickett, S. T. A. & McDonnell, M. J. 1989. Changing Per- spectives in Community Dynamics: A Theory of Succes- sional Forces. TREE 4: 241-245.

Pickett, S. T. A. & White, P. S. 1985. The ecology of natu- ral disturbance and patch dynamics. Academic Press, Or- lando.

Pielou, E. C. 1981. The usefulness of ecological models: a stock-taking. Quart. Rev. Biol. 56: 17-31.

Salt, G. W. 1979. A comment on the use of the term emer- gent properties. Am. Nat. 113: 145-148.

Schoener, T. W. 1986. Mechanistic approaches to community ecology: A new reductionism? Amer. Zool. 26: 81-106.

Shipley, B. & Peters, R. H. 1990. A test of the Tilman model of plant strategies: Relative growth rate and biomass par- titioning. Am. Nat. 136: 139-153.

Shrader-Frechette, K. S. & McCoy, E. 1990. Theory reduc- tion and explanation in ecology. Oikos 58: 109-114.

Shugart, H. H. 1984. A Theory of Forest Dynamics: The Ecological Implications of Forest Succession Models. Springer-Verlag, New York.

Shugart, H. & West, D. C. 1977. Development of an Appa- lachian deciduous forest succession model and its applica- tion to assessment of the impact of the Chestnut Blight. J. Environ. Management 5: 161-179.

Shugart, H. & West, D. C. 1980. Forest Succession Models. Bioscience 30: 308-313.

Shugart, H., West, D. C. & Emanuel, W. R. 1981. Patterns and Dynamics of Forests: An Application of Simulation Models. In: West, D. C., Shugart, H. H. & Botkin, D. B. (eds), Forest Succession: Concepts and Application. Springer-Verlag, New York, pp. 74-94.

Smith, T. & Huston, M. 1989. A theory of the spatial and temporal dynamics of plant communities. Vegetatio 83: 49- 69.

Sousa, W. P. 1979. Experimental investigations of disturb-

Page 33: Understanding ecological community succession: Causal models and theories, a review

ance and ecological succession in a rocky intertidal algal community. Ecol. Monog. 49: 227-254.

Sousa, W. P. 1984. The role of disturbance in natural com- munities. Ann. Rev. Ecol. Syst. 15: 353-91.

Thoreau, H. D. 1860. The succession of forest trees. Re- printed 1893 in Excursions. Houghton, Mifflin & Co., Bos- ton, pp. 225-250.

Tilman, D. 1982. Resource competition and community struc- ture. Princeton University Press, Princeton.

Tilman, D. 1985. The resource ratio hypothesis of plant suc- cession. Am. Nat. 125: 827-852.

Tilman, D. 1986a. Resources, competition and the dynamics of plant communities. In: Michael J. Crawley (ed), Plant Ecology. Blackwell Scientific Publications, 51-75.

Tilman, D. 1986b. Evolution and differentiation in terrestrial plant communities: the importance of the soil resource: light gradient. In: Diamond, J. & Case, T. J. (eds), Community Ecology. Harper & Rowe, New York, pp. 359-380.

Tilman, D. 1988. Plant strategies and the dynamics and struc- ture of plant communties. Princeton University Press, Prin- ceton.

Tilman, D. 1990a. Constraints and tradeoffs: toward a pre- dictive theory of competition and succession. Oikos 58: 3-15.

Tilman, D. 1990b. Mechanisms of plant competition for nu- trients: The elements of a predictive theory of plant com- petition. In: Grace, J. B. & Tilman, D. (eds), Perspectives on Plant Competition. Academic Press, San Diego, pp. 117-141.

Turner, T. 1983. Facilitation as a Successional Mechanism in

147

a Rocky Intertidal Community. Am. Nat. 121: 720-728. Underwood, A. J. 1986a. What is a community? In: Raup, D.

M. & Jablonski, D. (eds), Patterns and processes in the history of life. Springer-Verlag, Berlin, pp. 351-367.

Underwood, A. J. 1986b. The Analysis of Competition by Field Experiments. In: Kikkawa, J. & Anderson, D. J. (eds), Community Ecology: Pattern and Process. Black- well, Carleton, pp. 240-268.

Usher, M. B. 1987. Modelling Successional Processes in eco- systems. In: Gray, A. J., Crawley, M. J. & Edwards, P. J. (eds), Colonization, Succession and Stability. Blackwell Scientific Publications, Oxford, pp. 31-55.

van der Valk 1981. Succession in wetlands: a Gleasonian approach. Ecology 62: 688-696.

van Hulst 1980. Vegetation dynamics or ecosystem dynamics: Dynamic sufficiency in succession theory. Vegetatio 43: 147-151.

Walker, L. R. & Chapin, F. S. 1986. Physiological Controls over Seedling Growth in Primary Succession on an Alas- kan Floodplain. Ecology 67: 1508-1523.

Walker, L. R. & Chapin, F. S. 1987. Interactions among processes controlling successional change. Oikos 50:131- 135.

Wiegleb, G. 1989. Explanation and prediction in vegetation science. Vegetatio 83: 17-34.

Wimsatt, W. C. 1980. Reductionist research strategies and their biases in the units of selection. In: Nickels, T. (ed), Scientific Discovery: Case Studies. Reprinted in: Saarinen, E. (ed), Conceptual issues in ecology. D. Reidel, Dordrecht, Holland, pp. 155-201.