Food Chains versus Food Webs

A food chain is a linear sequence of feeding relationships: producer → herbivore → primary carnivore → secondary carnivore, each level obtaining energy by consuming the one below it. Food chains are useful because they make the thermodynamics of energy flow explicit and easy to draw — this is exactly the tool used to explain the roughly 10% trophic transfer efficiency between levels and the resulting pyramidal shape of biomass in The Trophic Pyramid. But almost no real organism eats only one kind of prey or is eaten by only one kind of predator. A single chain isolates one path through a much larger structure and, in doing so, discards the information needed to predict how a perturbation anywhere in the community will actually spread. A food web is the full set of feeding links among the species of a community, usually drawn as a directed graph in which nodes are species (or functional groups) and arrows point from prey to predator, indicating the direction of energy flow. Food webs contain multiple trophic pathways running in parallel, and most consumers are functionally omnivores in the sense that they feed at more than one trophic level (a fox eating both rabbits and berries occupies parts of two chains simultaneously). Key structural properties of a web include species richness (S, the number of nodes), the number of feeding links (L, the number of arrows), the mean number of links per species (L/S, "linkage density"), and connectance (C = L/S²), which normalizes link number by the maximum possible number of links (S²) to allow comparison of webs of different sizes. Empirically, connectance in real communities is low — most species interact directly with only a small fraction of the others — yet this sparse wiring is enough to produce strongly emergent, non-additive behavior. The single most important emergent behavior is the trophic cascade: a top-down effect in which a change in the abundance of a predator ripples down through the web to alter the abundance or biomass of species that are not its direct prey, often two or more links away. A food-chain view predicts only direct, adjacent effects (predator eats herbivore, herbivore's population falls). A food-web view predicts that removing a predator can release a herbivore, which then depletes a plant, which then changes the physical habitat for a completely different set of organisms — an indirect effect transmitted through the architecture of the web rather than through any single line of the chain.

Step 1 — energy flow follows chains, but organisms are wired into webs. At the biochemical level nothing changes: autotrophs fix carbon via photosynthesis, heterotrophs oxidize organic carbon for ATP, and roughly 90% of ingested/assimilated energy at each transfer is lost as heat, movement, and undigested waste (respiration and egestion), leaving about 10% available to the next level — the same efficiency logic that generates the trophic pyramid. What differs in a web is that this 10% does not flow along one path; it is partitioned across many predator species and many prey species simultaneously, and a given unit of carbon fixed by one plant can reach a top predator through several different route lengths.

Step 2 — top-down control operates link by link. When a predator's population increases (or a predator is introduced/protected), it suppresses its prey through increased mortality. If that prey is a keystone herbivore, its reduced grazing pressure allows the plant it eats to recover. This alternating pattern of suppression and release — predator↓prey, prey↓plant becomes prey↑plant — is why cascades are often summarized as "green world" dynamics: predators keep herbivores below the level at which they would strip vegetation bare. The strength of the cascade at each link depends on the per-capita consumption rate, the prey's population growth rate, and whether alternative prey exist to buffer the predator's diet (a generalist predator with many prey species produces weaker, more diffuse cascades than a specialist with one dominant prey).

Step 3 — network topology determines how far and how strongly a perturbation travels. In a highly connected web, many alternative pathways exist between any two species, so the loss of one link can be compensated by flow through another (functional redundancy), damping the cascade. In a sparse web, or one built around a few highly connected "hub" or keystone species, removing that hub eliminates most of the alternative pathways at once, and the perturbation propagates strongly and predictably. This is precisely the situation with keystone predators: their removal doesn't just delete one link, it collapses the redundancy that was stabilizing several trophic levels below them.

Step 4 — indirect effects arise from combining direct effects across two or more links. Formally, if species A affects B directly, and B affects C directly, then A has an indirect effect on C that need not resemble a simple sum of the direct effects — it can be sign-reversed (a predator having a net positive effect on a plant two links away), and its magnitude depends on the strength of each link in the chain of dependency. This is the mathematical core of a trophic cascade: two negative direct effects (predator suppresses herbivore; herbivore suppresses plant) combine to produce a net positive indirect effect (predator "benefits" the plant).

Step 5 — cascades are strongest in systems with low species diversity at intervening levels and weaker where diffuse, highly connected webs buffer the effect. This is why cascades are most dramatically documented in relatively simple, low-diversity systems (rocky intertidal shores, kelp forests, small lakes) and are harder to detect, though not absent, in highly diverse tropical forests or oceanic food webs where hundreds of species share the intermediate trophic levels.

Robert Paine's Pisaster removal experiments, Mukkaw Bay, Washington, 1963–1969. Paine chose a rocky intertidal shore dominated visually by the predatory starfish Pisaster ochraceus and its principal prey, several species of barnacles, mussels, limpets, and chitons. On experimental plots he physically removed every Pisaster he could find, repeatedly, over several years, while leaving paired control plots undisturbed. The result was dramatic: within about a year the mussel Mytilus californianus, normally kept in check by starfish predation, expanded explosively across the plot and monopolized the substrate, crowding out the barnacles, algae, and smaller invertebrates that had coexisted with mussels when starfish were present. Species richness on the experimental plots dropped from around 15 species to 8. Paine's key conceptual contribution was the term keystone species — a species whose impact on community structure is disproportionately large relative to its own abundance or biomass — and the demonstration that removing one predator could reorganize an entire local food web, not merely reduce one prey population. This experiment is the founding proof-of-concept for trophic cascades: it showed conclusively that food-web architecture, not chain length, determines the community consequences of a species loss.

The sea otter–urchin–kelp cascade, Aleutian Islands and California coast, 1970s–1990s. James Estes and John Palmisano compared islands with and without sea otters (some populations had been extirpated by the maritime fur trade, later recovering unevenly). Islands with otters had low urchin biomass and lush kelp forests with high primary production and diverse associated fish and invertebrate communities. Islands without otters had urchin "barrens" — high urchin density, negligible kelp, and markedly lower diversity. This was reinforced experimentally and observationally over decades, and later work (including studies by Estes and colleagues in the 1990s–2000s) linked otter population crashes — in some areas attributed to increased predation by killer whales as other marine mammal prey declined — to renewed urchin outbreaks and kelp loss, tracing the cascade even further up to a fourth trophic level. This case generalized Paine's rocky-shore finding to an entirely different ecosystem (subtidal kelp forest) and to a different kind of predator (a mammal, not an invertebrate), showing the cascade concept was general rather than a quirk of one shoreline.

Whole-lake manipulations, Carpenter, Kitchell, and Hodgson, 1985 (and subsequent whole-lake experiments at the University of Wisconsin's Notre Dame and Peter/Paul Lakes). Stephen Carpenter and James Kitchell manipulated entire lakes — not plots, not cages, whole ecosystems — by adding or removing piscivorous fish (largemouth bass) to alter the abundance of planktivorous fish, and tracked the effects down through zooplankton to phytoplankton biomass and primary production. When piscivore biomass increased, planktivorous fish declined, zooplankton grazers increased, and phytoplankton biomass (measured as chlorophyll a) fell sharply — a three-link cascade from fish to algae, detectable as a change in water clarity and primary productivity of the whole lake. This work was crucial because it demonstrated cascades operating on ecosystem-level state variables (nutrient cycling, primary production, water quality) rather than just species counts, and it showed the effect at the scale of an entire, unmanipulated-boundary ecosystem rather than a small experimental plot, addressing the objection that cascades might be artifacts of small-scale caging experiments. Together these three lines of evidence — intertidal removal experiments, comparative and historical observation across otter/no-otter systems, and whole-lake manipulative experiments — used three different methodologies (direct removal, natural/quasi-experiment, whole-ecosystem manipulation) to converge on the same conclusion: the architecture of the food web, and specifically the presence or absence of particular highly connected predators, determines outcomes that a single linear food chain diagram cannot predict.

Example 1 — Quantifying connectance and comparing two webs. Suppose a small pond community has been mapped with S = 12 species and L = 28 observed feeding links. A nearby, more disturbed pond has S = 9 species but only L = 14 links, having lost several species and links after an algal bloom event. Connectance: ``` C(pond A) = L / S² = 28 / 144 = 0.194 C(pond B) = L / S² = 14 / 81 = 0.173 ``` Linkage density (mean links per species): ``` LD(pond A) = L / S = 28/12 = 2.33 LD(pond B) = L / S = 14/9 = 1.56 ``` Although pond A has more species and a slightly higher connectance, the more informative number here is linkage density: species in the disturbed pond B have, on average, only 1.56 feeding connections each, versus 2.33 in pond A. Lower linkage density means less redundancy — each species has fewer alternative partners to fall back on if one link is severed. A food-chain-only description ("phytoplankton → zooplankton → fish") would treat both ponds as functionally identical three-level systems; the connectance/linkage-density calculation reveals that pond B is structurally more fragile and likely to show a stronger, less-buffered cascade if its top predator is further disturbed.

Example 2 — Predicting the sign and magnitude of an indirect effect. Consider a food web with a hawk that eats snakes and also eats ground squirrels; snakes eat ground squirrels; ground squirrels eat seedlings of a shrub. This is a four-species web with two pathways from hawk to shrub:

• Path 1 (direct chain-of-three): hawk ⊣ snake ⊣ squirrel ⊣ seedling (three links, sign pattern −,−,− gives a net negative effect of hawks on seedlings through this path, since three negatives multiply to a negative: hawk down → snake up → squirrel down → seedling up, i.e., removing the hawk still nets a positive effect on seedlings via this path).
• Path 2 (direct, shorter chain): hawk ⊣ squirrel ⊣ seedling (two links, sign pattern −,− gives a net positive effect of hawks on seedlings — fewer hawks means more squirrels means fewer seedlings, so hawks indirectly protect seedlings through this path).

Suppose experimentally we find that removing hawks causes ground squirrel density to rise by 40% (because hawks eat squirrels directly and also because with fewer hawks, snakes — which also eat squirrels — may rise or fall depending on whether hawks also prey on snakes). If hawks additionally suppress snakes by 25% when present, then removing hawks releases snakes too, and the net change in squirrel density depends on the relative strength of the two competing indirect pathways: hawks-suppress-squirrels-directly (releasing squirrels when hawks are gone) versus hawks-suppress-snakes-which-eat-squirrels (also releasing squirrels indirectly, reinforcing the effect, since fewer hawks means more snakes, which then eat more squirrels, partially offsetting the direct release). Only by mapping the web, not a single chain, can you correctly predict whether seedling damage will increase or decrease after the hawks are removed — a chain diagram picking only Path 1 or Path 2 would give the opposite qualitative prediction from picking the other.

Robert T. Paine (1933–2016) was a marine ecologist at the University of Washington who spent much of his career doing hands-on, low-technology, high-insight fieldwork on the rocky shores of the Pacific Northwest. Trained initially in invertebrate zoology, Paine became convinced that ecology needed the same experimental rigor as physiology or genetics — manipulate a real variable in the field, replicate it, and measure the consequences, rather than relying only on correlational surveys. His 1966 paper "Food Web Complexity and Species Diversity" and the subsequent starfish-removal work at Mukkaw Bay were unusual for their time in treating an intertidal shore as a manipulable experimental system rather than a purely descriptive one. Paine coined the term "keystone species" in a 1969 paper, borrowing the architectural image of the wedge-shaped stone at the top of an arch whose removal collapses the whole structure — a strikingly apt metaphor for a predator whose loss can reorganize an entire community.

Fisheries management and marine protected areas. Recognizing that removing a top predator (e.g., overfishing large predatory fish or sharks) can trigger cascades that collapse kelp forests, coral reefs, or seagrass beds has driven the design of marine protected areas intended to preserve entire trophic structures, not just target species biomass.

Biological control and invasive species management. Understanding food-web connectance helps predict whether introducing a biocontrol predator will stay confined to its intended pest or cascade unpredictably through non-target species — a central concern after failed or damaging biocontrol introductions (e.g., cane toads in Australia, introduced to control beetles but becoming an invasive predator/toxin problem themselves).

Restoration ecology. Reintroduction programs — wolves in Yellowstone, sea otters along parts of the California coast — are explicitly designed around trophic cascade theory: restoring a keystone predator is expected to indirectly restore vegetation structure (aspen and willow recovery linked to wolf-mediated elk behavior changes) or invertebrate community diversity (kelp forest recovery with otters), illustrating direct application of cascade theory to conservation practice.

Climate change and species-loss forecasting. As climate shifts alter species ranges and phenology, ecologists use food-web connectance and network models to predict which ecosystems are likely to be robust (highly redundant, well-buffered webs) versus fragile (webs organized around a few highly connected keystone links) to warming-driven extinctions or range shifts.

Public health and zoonotic disease. Food-web thinking extends to disease ecology: predators that suppress rodent populations can indirectly reduce the abundance of rodent-borne pathogens (an indirect, cascade-like effect), informing strategies for managing zoonotic disease risk by managing predator communities rather than only targeting the pathogen or vector directly. These applications set up the next layer of ecosystem thinking — how Decomposition and Detritus recycles the energy and matter these webs move, and how the resulting structure underpins measurable Ecosystem Services.

Ask the tutor: How does the strength of a trophic cascade change if the intermediate herbivore has several alternative predators instead of just one?