For over a century now, ecologists have been describing the patterns in the distribution (Lomolino et al. 2005) and the abundance (McGill 2006; McGill et al. 2007) of organisms. With respect to the study of interactions (the explanatory part of ecology), consumer–resource interactions have received by far most empirical and theoretical study, both from a single trophic (Tilman 1982) and from a multitrophic, food web perspective (Cohen 1978; DeAngelis 1992; Polis & Winemiller 1996). Studies that use food web theory to better understand a particular ecosystem thus implicitly assume that predation is the most important process that regulates the abundance of organisms in that ecosystem (Berlow et al. 2004).

However, it has long been recognized that species interact in ecosystems with other species and with abiotic factors in many ways, of which predator–prey interactions are only one possibility (Hutchinson 1959). For example, organisms interact with other species through producing resources such as detritus and mineral nutrients and through non-trophic interactions (e.g. pollination, production of toxicants). Also, organisms can show strong interactions with abiotic (non-resource) conditions. In addition, relevant interactions that affect organisms include various spatial interactions (exchange of organisms, materials and energy), external environmental forcing, as well as various physical and chemical interactions that operate within ecosystems.

These days, ecologists are increasingly challenged to better understand and predict the impacts of human activities on biodiversity and the functioning of ecosystems, such as the consequences of harvesting populations (forestry, fisheries), modification of material cycles (e.g. eutrophication) and human-induced climate change. Key general questions in this conservation agenda are: (i) which (types of) species will be most vulnerable to extinction in the near future, (ii) are ecosystems of high biodiversity (such as tropical forests, coral reefs) under greater threat than those less diverse, (iii) will the loss of some species (e.g. top predators) lead to cascading losses of other species, and impair the functioning of ecosystems, (iv) should some species therefore be given special attention in conservation schemes, (v) how will the human disruption of natural element cycles and the introduction of novel chemical compounds and non-native species affect the functioning of natural ecosystems and impair the services they provide to us, and (vi) what will be the consequences of emerging (zoonotic) diseases? All these questions will affect the abundance and distribution of species, with associated effects on the functioning of ecosystems. Answers to these questions are urgently needed to set conservation priorities and take appropriate action to restrict biodiversity loss due to human-driven environmental change.

Since the pioneering work of Elton (1927), Lindeman (1942) and Hairston et al. (1960), the field of food web theory has developed into a central concept in ecology. It is therefore a logical field to turn to first for answers to the above conservation-oriented questions, as it aims to understand the abundance and distribution of organisms from the perspective of species interactions. Indeed, the central questions addressed in food web ecology seem highly relevant for conservation and management. For example, what is the effect of increased nutrient supply on trophic web structure (Carpenter & Kitchell 1993; Scheffer & Carpenter 2003)? Or, how does the diversity and complexity of food webs affect their stability, e.g. the extent to which small perturbations in some species lead to the loss of other species (May 1973; Dunne et al. 2002; Ives & Carpenter 2007; Neutel et al. 2007)? What determines whether the loss of top predators leads to cascades of secondary extinctions (Scheffer et al. 2005; Borrvall & Ebenman 2006; Otto et al. 2008)? However, in a recent list of 100 ecological questions of high policy relevance in the UK (Sutherland et al. 2006), the word ‘food web’ or ‘interaction web’ did not occur once, suggesting it is not, or at least not perceived this way.

In our view, this ‘struggle for relevance’ of food web ecology is due to two main problems. Firstly, food webs consist of a ‘road map’ of predator–prey interactions in ecosystems. However, species in ecosystems interact with each other and with their environment in many other ways than through consumer–resource interactions. These ‘other interactions’ have been insufficiently acknowledged and studied from a network perspective, ‘pushing’ conservation-oriented research often towards a species-centred approach (in which all such interactions are included for a particular species). However, in such species-centred research, the operation of the key indirect effects among species that characterize ecological networks are probably missed. Inclusion of non-trophic interactions broadens food web studies to the analysis of interaction webs.

Secondly, food web studies have often been too system specific, and we need a more general ‘template’ of functional classification of species along main axes of organization (not only trophic position) in food webs to be able to make comparisons between different ecosystems, and to study the interplay of networks based on consumer–resource interactions with networks based on other types of interaction that operate within the same ecosystem.

For those causes arising at the focal level of ecological interaction webs, the challenge is to discover the principles that govern their organization, or, ‘how lots of things are put together in the same place’ (Ulanowicz 1997). This challenge is shared with other fields in the life sciences, for example, with developmental biology, where the main ‘grammar’ of the genetic code still mostly awaits discovery, now that the translation of ‘letters and words’ is available (Lewin 1984; Barbieri 2002). The emerging field of systems biology (Kitano 2002) now aims at unravelling exactly how the network of interactions among genes, proteins, organelles, cells and tissues within organisms forms this grammar.

The general scientific problem here is that causes of organization in ecological networks (and other complex adaptive systems) arise through conditional probabilities, which means that all probabilities (e.g. likelihood of change in the abundance of a species) are always contingent to a greater or lesser extent upon local and historic circumstances and interfering events (Ulanowicz & Wolff 1991). For example, the dynamics of three species in a trophic chain can radically change if species 3 evolves a trait that promotes species 1, causing an indirect mutualism. In this case, species 2 will be promoted, even though it did not change its behaviour or physiology at all (Ulanowicz 1995). Also, a predator–prey interaction will have a different effect on either population if the prey has to compete with another prey (leading to apparent competition), or, if the predator is a prey itself to another predator (intraguild predation). And, some species of prey may use phenotypic flexibility to directly adjust their phenotype in the presence of specific predators (Werner & Peacor 2003), while in other cases predators adjust their phenotype in the presence of specific prey (Piersma & Lindstrom 1997; Piersma & Drent 2003). In both cases, such phenotypic adjustments will have consequences for other consumer–resource interactions that the species is involved in. The reasons for the absence or presence of such interfering species may even lay outside the current spatial and temporal domain of observation, due to historical or geographical factors (Ricklefs & Schluter 1993). Dealing with such conditional probabilities requires a redefinition of classic mechanisms (causes imposed by lower level of organization and system components in a deterministic, ‘Newtonian’ way). Evolutionary biologists face similar problems in deducting how organization arises through the operation of conditional probabilities of change, e.g. when developing theory for adaptive dynamics (Dieckmann & Metz 2001) and coevolutionary dynamics (Thompson 2005).

The now widely recognized general feature of complex adaptive systems is that the prevalence of strong conditional probabilities does not necessarily lead to unpredictable, chaotic or erratic structures and dynamics. Instead, emergent structural properties and behaviour often arise at the system level (Levin 1998; Holland 1999; Morowitz 2002), pointing at an underlying ‘semantics’ of system organization (Barbieri 2002). For food webs, such regularities arise for example in their topological organization (Pimm 1982; Williams & Martinez 2000; Montoya et al. 2006; Bascompte 2007), the organization of flows, thus interaction weights (Ulanowicz 1997; Neutel et al. 2002; Rooney et al. 2006; Neutel et al. 2007) or their spatial organization (McCann et al. 2005). However, clear rules and principles about ‘how lots of things are put together’ in food webs still await description (Ulanowicz 1997).

Insights into specific ‘few-species-interaction-configurations’, or modules (Menge 1995; Holt 1997; Bascompte & Melian 2005) for consumer–resource interactions have much increased over the last decades. For example, we know much more now about resource competition (Schoener 1974; Tilman 1982), mutualism (Oksanen 1988), apparent competition (Holt 1977), indirect mutualism (Vandermeer 1980; Ulanowicz 1997), intraguild predation (Polis et al. 1989), positive interactions such as facilitation (Callaway 2007), positive feedbacks (DeAngelis et al. 1986), regulatory feedbacks (Bagdassarian et al. 2007), trophic cascades (Carpenter et al. 2008) and multiple stable state dynamics (Scheffer & Carpenter 2003). These may all be considered organizational forces that structure networks, but all may not be of equal importance. For example, Ulanowicz (1997) makes a strong case for the special importance of indirect mutualism as an organizational force in food webs, as the resulting feedback loops ‘attract’ resources towards them.

Repeating that ecology is both about jointly understanding interactions among organisms, and between organisms and their abiotic environment, we propose six main types of ecological interaction that operate in ecosystems, with a general framework for their topological connection among six basic ecosystem compartments (figure 1). These six types of interaction are: (i) consumer–resource interactions, (ii) interactions between organisms and abiotic (non-resource) conditions, (iii) spatial interactions (inputs and outputs of energy, nutrients, organisms), (iv) non-trophic direct interactions among organisms, (v) physical and chemical interactions among factors/compartments, and (vi) external forcing of abiotic conditions. These six types of interaction potentially operate among three biotic and three abiotic basic compartments (figure 1). The abiotic compartments are (i) abiotic resources (such as light, nitrate, ammonium, phosphate) that are consumed and depleted by autotrophs, (ii) abiotic conditions, that affect both autotrophs and heterotrophs but are not consumed or depleted by them (such as salinity, soil texture, sediment aeration, soil and water pH, temperature) but that can be modified (e.g. by ecosystem engineers (Jones et al. 1994; Lawton 1994)) and (iii) detritus (non-living organic material). The main three biotic compartments are (i) autotrophs that can harvest their own energy, either from light or chemical sources, (ii) microbial detrivores that break down detritus into its mineral components, thus producing resources for autotrophs and (iii) higher trophic levels that consume autotrophs, microbial detrivores and/or each other, and mineralize nutrients for autotrophs. This interaction-web framework builds on earlier ideas for marine systems by Azam et al. (1983) and for terrestrial systems by Wardle (2002), Moore et al. (2004), Bardgett (2005) and others, who all emphasized the importance of the ‘dual foundation’ of food webs on both autotrophs (plants, photosynthetic or chemoautotrophic microbes) and microbial detrivores, but adding the effect of environmental (non-resource) conditions. The autotroph- versus detritus-based side of figure 1 can be viewed as two alternative channels that provide energy to higher trophic levels, while being strongly functionally connected at the bottom through the process of energy fixation (a ‘service’ of the autotrophs facilitating the development of the detritus-based side) and element recycling (a ‘service’ that especially the microbes on the detritus-based side provided to the autotrophs). Depending on the ecosystem type, these two main energy channels are usually still separate at low trophic levels (e.g. plant- versus microbial detrivore-based grazers), while becoming more connected at higher trophic levels, where omnivorous predators often receive energy through both channels. In very open ecosystems that receive their energy through detritus imports, such as tree holes (Kitching 1971), deep oceans systems (Andersson et al. 2004), streams or shaded lakes, the food web can be almost entirely detritus based. In more closed ecosystems on the other hand, autotrophs generally require microbial detrivores to recycle mineral nutrients (called the microbial loop in pelagic systems) and the food web will receive energy through both channels. It should be noted that the three ‘biotic boxes’ in figure 1 aggregate complex trophic interactions through unresolved ecological networks. Each of these boxes can also be expanded to networks of higher detail (e.g. in functional groups such as herbivores consuming plants, predators of herbivores, predators of predators of herbivores, pathogens, pollinators, etc., or down to the species level).

As listed in figure 1, we suggest that up to six main types of direct interaction can operate simultaneously in any ecosystem, where it is not a priori clear which ones will dominate in determining community structure and ecosystem functioning. Consumer–resource interactions are of course a basic one (each species generally has to eat), but such food web interactions will be affected by other types of interaction that operate in the ecosystem at the same time. When these other interactions involve only one or two species, this may still be ‘fixed’ by modifying food web models to include such effects (Arditi et al. 2005; Goudard & Loreau 2008). However, when the other types of interaction result in ecological networks as well (of which we will show examples later), this requires a different approach, the here-proposed analyses of ‘parallel ecological networks’. Before we continue with this discussion, we first identify and discuss each of the six main interaction types that we suggest are structuring ecosystems.

A strong point of food web ecology is its promise for generality: it holds the potential to be useful in comparing very different ecosystems, and hence produce general conclusions on the organizational forces and principles at work. However, this ability to compare is currently hampered by our inability to assign species generic functional roles. Yet, such system-independent roles of species are of great fundamental and applied interest. This role should characterize the general topological position and functional importance of a species in ecological networks, independent from the particular web under study. Current functional classifications mainly use the trophic position, as top predators (Finke & Denno 2005; Scheffer et al. 2005; Borrvall & Ebenman 2006), mesopredators (Elmhagen & Rushton 2007), herbivores and primary producers. For interaction webs including species–environment interactions, the importance of ecosystem engineers has been recognized for species that strongly modify abiotic conditions, and hence resources to other species (Jones et al. 1994; Lawton 1994; Wright & Jones 2006). But can other main axes of organization be identified? We suggest that a more structured approach is required for each of the six types of interaction that define generic species groups by their topological position, and hence their functional roles in ecosystems. The result would be an ‘interaction web template’ that should fit to describe any ecosystem.

For consumer–resource interactions, we propose such a template in figure 3 to explore the usefulness of this idea. Each numbered box is a functional group, generally consisting of a group of species that is competing for resources that they obtain from one or more other functional groups in the web (figure 2). This general web is ‘anchored’ at the bottom left, where algae and other autotrophs produce biomass, and heterotrophic bacteria decompose the detritus produced by plants and higher trophic levels, both at a high rate of turnover. We suggest that, starting from here, consumer–resource interaction webs are organized along two main axes. The vertical axis is the classic trophic position axes, forming food chains of species towards increasingly higher trophic levels. Generally, the size of species increases along the vertical axis, as predators generally need to be bigger than their prey to hunt and handle them efficiently (with the exception of pathogens, which we discuss later) (Cohen et al. 1993b; Brose et al. 2006). We suggest a new second major axis of food web organization: a stoichiometric axis. At the lowest trophic levels, this axis is driven by two main evolutionary radiations: (i) the competitive struggle for light between plants (as is still observable during primary succession, changing the dominance by algae, to herbaceous plants, to trees), leading to the formation of plants with more and more structural support (cellulose, lignin, etc.) in an effort to overtop each other, and (ii) a radiation of detrivores other than bacteria, which could physically fragment (macrodetrivores) and biochemically decompose (fungi) the coarser, poor-quality plant material that these taller, mechanically better supported plants increasingly produced. Therefore, the horizontal axis is a stoichiometric axis (Sterner & Elser 2002), representing a decreasing C/N ratio of the plant material produced and a lower turnover rate of its compartments. Within the next herbivore trophic level, this horizontal axis has also resulted in size radiation of consumers, not driven, however, by the need to be bigger than their prey, but by the need to be able to digest it. Bigger herbivores can handle poorer quality food due to the longer residence time of food in their stomach, and lower per mass energy requirement, leading to a more favourable ratio of digestive capacity to metabolic requirement (Demment & van Soest 1985). The resulting increase in herbivore size from the need to handle poorer quality (niche-based species radiation) may then have triggered an evolutionary arms race between herbivores and predators, causing a size increase in predators as well (Owen-Smith & Mills 2008a) and also resulting in some herbivores eventually ‘escaping’ their predators by growing too big, so-called megaherbivores (Owen-Smith 1988). The two independent axes of food web organization suggested in figure 3 therefore cause strong body size variation to exist both within and across trophic levels, with the smallest species found at the bottom left, and the largest species at top right. The organization of food webs are a testable hypothesis, which requires the compilation of data on both quantitative trophic position (e.g. through stable N isotopes) and stoichiometric position (e.g. through measuring C/N or C/P ratios of diets and excretion products, turnover rates), facilitating a quantitative comparison of the resulting patterns across ecosystems.