Review | Agency in the Evolutionary Transition to Multicellularity

Introduction

Agency has long occupied an uneasy place in the life sciences. For some, it still smells of vitalism: the old suspicion that living systems must contain a special force, a hidden supplement to physics and chemistry, if they are to act, choose, or pursue ends. Yet the problem has never really been whether life is “more than matter” in a dualist sense. The more serious issue is that purely reductionist descriptions often fail to capture the organized, history-dependent, and self-initiated character of living activity. A good scientific account of life must therefore resist two temptations at once; the temptation to dissolve organisms into nothing but microphysical mechanisms, and the temptation to rescue what reductionism misses by positing nonphysical realms, intentions, or morphic blueprints.

This tension has become sharper in recent years. On one side, strong dualist or quasi-dualist narratives have resurfaced under contemporary labels. Some accounts of “irruption” begin by assuming, from the outset, that motivations make an irreducible difference to material behavior while remaining unmeasurable in principle. Other approaches, such as recent appeals to Platonic morphospaces, suggest that biological forms and minds are guided by structured but nonphysical pattern spaces. These proposals are rhetorically attractive because they acknowledge that organisms are not inert mechanisms. But they pay for that acknowledgment by smuggling in ontological irreducibility where what may really be at stake is epistemological irreducibility, the fact that certain organizational properties of living systems cannot be captured adequately by some explanatory frameworks, even though they remain fully natural.

It is precisely here that agency re-enters evolutionary thinking in a more disciplined form. Niche construction, ecological evolutionary developmental biology, and organism-centered evolution have all re-emphasized Richard Lewontin’s point that organisms are not merely the objects of selection but also subjects that modify, select, and partially constitute the very environments in which selection occurs. In this sense, agency no longer names a mysterious force but a family of capacities through which living systems demarcate themselves, maintain themselves, explore their surroundings, react flexibly to perturbation, enter into relations with other agents, and sometimes engage in activities that are risky, idiosyncratic, or not immediately tied to reproductive advantage. Stuart Newman and his coauthors explicitly take this broadened, biological framing seriously; in Agency in the Evolutionary Transition to Multicellularity, they define agency in dispositional rather than absolute terms as a propensity to initiate and vary behavior in ways not exhaustively prewritten by deterministic antecedents.

This matters especially when we think about one of evolution’s great transformations, the transition from unicellularity to multicellularity. The earliest unambiguous multicellular eukaryotes and the earliest accepted unicellular eukaryotes appear in deposits of roughly similar age, which raises an arresting possibility; perhaps multicellularity did not require an impossibly long preparatory march so much as the mobilization of capacities that cells already possessed—including, potentially, agential ones. Newman and colleagues therefore ask a subtle question. If agency already belongs to single cells, what becomes of it when individuality migrates from the level of one cell to the level of many? Does collective agency merely sum cellular powers, or does multicellular life create new forms of agency altogether? Their answer is both bold and unusual: multicellular agency is real, but it is not derived in as straightforward a way from organizational closure as single-cell autonomy is. Rather, it is transformed by the distinctive inherencies of multicellular matter and by development’s power to amplify, partition, and repurpose preexisting cellular functionalities.

Multicellularity, autonomy, and function

The review’s central conceptual move is to refuse a purely instrumental reading of cellular agency. If one grants that single cells are not merely mechanistic automata but genuine sites of organism-initiated activity, then multicellularity becomes more than a change of scale. A multicellular aggregate is not simply many agents packed together. It may, as the authors argue, become a new unit of agency, just as a liquid exhibits intensive properties not reducible to a mere count of molecules. This shift can be gradual or abrupt, noisy or sharply defined, but it is always consequential; once cells begin to behave as a collective, the relevant “self” may move upward in organizational level, and the logic of fitness itself may change with it. In such cases, what appears under multilevel selection theory as a conflict between cell-level and collective-level interests may be better understood, at least in some systems, as the reorganization of agential capacities into a new normative center.

The paper’s strongest illustrations come from systems where the transition between one and many remains experimentally visible. Social amoebae and myxobacteria are especially revealing because they move repeatedly between solitary and multicellular phases. In Dictyostelium, for example, starvation transforms individually motile, exploratory amoebae into streaming collectives and then into motile slugs whose behavior cannot be reduced either to pure cellular independence or to simple fluid dynamics. At one stage, agency appears curtailed as bulk flow and signaling synchronize the cells; at another, it is reconfigured as a collective exploratory capacity; later still, in the spore phase, cells are reconstituted as freer units. What this shows is not that agency disappears in multicellularity, but that it changes locus and form. Even apparent losses of cellular agency may be conditional rather than constitutive, a matter of phenotype and context rather than a permanent erasure of cell-level capacities.

This same argument is extended to metazoans through cases like placozoans and biobots. Trichoplax adhaerens is especially important because it exhibits novel multicellular capacities without the elaborate organs usually associated with animal agency. It moves and captures prey through ultrafast epithelial contractions and coordinated ciliary activity, suggesting that multicellular agency can arise through the concerted mobilization of very basic cell-derived properties. Xenobots make the point even more sharply: when clusters of frog embryonic cells are liberated from their normal developmental context, they can navigate mazes and gather free cells into new clusters—behaviors the constituent cells did not evolve to perform in that combination. The implication is striking. A collective built from agential cells may inherit agency automatically, but in a transformed form: different in scale, different in constraints, and potentially different in the kinds of affordances it can exploit.

This leads directly into the paper’s discussion of autonomy. Newman et al. do not deny the value of self-production or the organizational approach. On the contrary, they treat Kant’s notion of organisms as “natural purposes”, Maturana and Varela’s autopoiesis, and Moreno and Mossio’s closure-of-constraints framework as indispensable for understanding how living systems are self-producing and self-maintaining. But they resist the tendency to make such closure criteria too stringent at the multicellular level. When Arnellos and Moreno argue that multicellular agency requires a nervous system and an epithelium, the authors object that this excludes too much: placozoans, slugs, and biobots already display behaviors that are plausibly agential without satisfying those demanding organizational conditions. Newman and colleagues’ broader claim is that multicellular agency is not as tightly tied to autonomy as single-cell agency is. The reason is not that multicellular organisms are less real, but that development and the physics of multicellular matter create enablements that exceed the logic of previously closed constraints.

Here the distinction between adaptation and enablement becomes decisive. Adaptations, in the standard evolutionary sense, are traits honed by natural selection because they improve fitness in familiar or changing environments. Enablements are different, they are novelties that may begin without a specific function, but later become indispensable once organisms use them to enter new ways of life. In Newman’s picture, multicellularity is rich in such enablements. Morphological motifs can arise because of the inherencies of active, excitable multicellular matter; existing cellular functions can be partitioned, amplified, or repurposed into novel tissue-level capacities; and these novelties can open access to new environmental affordances long before they become stabilized as canonical adaptations. In that sense, organismal agency does not merely reflect prior selection. It can actively drive evolution by exploiting morphological and functional possibilities that were not selected into existence in the first place. As higher-order forms arise, the independent agency of cells is partially bartered for the greater independence of the collective, but that very bargain may itself be an expression of cellular agency.

Agency in vitro and in silico

The experimental part of the paper is unusually candid about its own difficulties. Agency, as the authors broadly define it, is not reducible to any single observable marker. Because they take it to be potentially unprogrammed and only partly determined by prior causes, they also accept that empirical signatures of agency may be elusive. Still, they argue that experiments are indispensable, precisely because the problem is so easily swallowed either by vague metaphysics or by overconfident mechanistic reduction. Their strategy is therefore comparative and perturbative: look for behaviors that emerge when cells or multicellular aggregates are placed in situations they could not reasonably have encountered during the evolutionary history of their lineage. If such systems discover new solutions, exploit new affordances, or reconfigure their own organization in ways not straightforwardly predictable from existing adaptive programs, then the case for agency becomes stronger.

This is why the authors do not place their hopes in animal embryos, despite their biological centrality. Animal development is already the product of more than half a billion years of integrated evolution. That makes multicellular agency difficult to disentangle from inherited, stereotyped, highly canalized developmental mechanisms. Social bacteria, social amoebae, and cancers are better experimental theaters because they stage transitions between unicellular and multicellular organization in a more exposed and reversible way. These systems also retain physical and ecological features (low Reynolds number conditions, transient collectives, rapidly shifting microenvironments, etc.) that may resemble those under which multicellularity first emerged. Even there, however, challenges abound. The environment is often artificially stabilized in laboratory studies; cells continually create their own microenvironments; and the target “individual” is often fuzzy, shifting between single cells, streams, clusters, and three-dimensional aggregates. All of this means that any experimental study of agency must also be a study of how system boundaries, environmental feedback, and levels of individuality are constituted in real time. Our last review here dealt with a theoretical framework designed to fit into that niche.

On the modeling side, Newman and colleagues make a crucial philosophical move: they reject the assumption that agential systems can be fully captured by strictly deterministic or merely stochastic state-space models. Drawing on Howard Pattee, they suggest that biological systems may need to be represented as nonholonomic or incompletely specified systems, where there are more available degrees of freedom than those required to describe the actual realized motion. In such cases, causes are not exhausted by microstate-to-microstate mappings. Instead, one may need mathematical representations that constrain possibilities without selecting a unique trajectory. The paper points to inexact differential forms and differential inclusions as promising tools here. These allow one to specify a set of possible derivatives rather than a single deterministic evolution law, thereby opening formal space for dispositional causation, multilevel influence, and emergent behaviors that are not simply random noise around a fixed rule.

This is an important step, but it also reveals a limit. Differential inclusions and viability-theoretic approaches remain, in the end, variants of a state-space representation. They are broader and less rigid than ordinary differential equations, but they still assume a predefined space of possible states. That is likely enough for many questions about persistence, morphology, or constrained variability. But it is much less adequate for producing genuine novelty, where the alphabet of relevant variables, the transition rules, and even the whole system’s logic can change through time. If multicellular agency is partly about the emergence of new affordances and new normative regimes, then our formalisms must eventually become syntax-open, capable not only of moving through a space, but of rewriting the very space in which movement occurs.

Here I think the paper’s own instincts point beyond its chosen formal tools. Hofmeyr’s biochemically realizable extension of Rosen’s (M, R)-systems offers one example of a framework whose models are both thermodynamically open and syntactically open, able to capture self-maintaining organization without freezing its architecture in advance. And once such a relational organization is temporally unfolded (as in my recent work) it begins to function as a primitive probabilistic anticipatory structure: not externally imposed, not fixed, but continuously reconstituted as the system adapts. Although these theoretical developments were worked out for single-celled organizations, there is no principled reason why relational models could not coexist, couple, and scale into multicellular or even social forms, just as Hofmeyr himself suggests. In that sense, the paper’s call for incompletely specified models is exactly right, but the next step may require moving beyond merely broadened state-space formalisms toward explicitly relational, temporally unfolded, self-modifying ones.

Conclusion

Newman and colleagues offer one of the most provocative recent attempts to rethink agency in biological evolution without surrendering either to dualism or to mechanistic flattening. Their central achievement is to show that multicellular agency is not a trivial magnification of cellular agency, nor merely a side effect of adding more cells. It is instead a transformed mode of organized activity made possible by the inherencies of multicellular matter and by development’s capacity to amplify, partition, and repurpose single-cell functionalities into higher-order enablements. This is why multicellular agency can be less tightly constrained by closure than cellular agency is, and why organismal novelty may sometimes precede, rather than merely reflect, conventional adaptive refinement.

The paper is also valuable because it sharpens a sequence of distinctions that are too often blurred. Not all purposive behavior is agential. Some directed activities are physically inevitable; others are teleonomic products of selection; only some display the idiosyncratic, exploratory, and partly unprogrammed character the authors want to preserve under the name agency. In this respect, their account sits comfortably with a cybernetic hierarchy in which goal-directedness is broader than agency and agency is a more advanced, less fully determined form of purposive organization. Their experimental proposals follow from this insight: expose cells or aggregates to ecologies that fall outside their lineage’s expected history and watch whether they discover new modes of action. Agency, if it is real, should reveal itself most clearly where stereotyped adaptation runs out.

At the same time, the paper leaves three important gaps. First, it still presupposes a cell-level agency whose physicochemical basis remains underexplained; in that sense, it remains, by the authors’ own admission, close to a kind of methodological vitalism. Second, while dispositional causality is a helpful philosophical frame, we still need formal models that can capture not just underdetermination but real organizational novelty. Third, the transition from cellular to multicellular agency is described more convincingly than it is mechanistically derived. Yet none of these are failures. They are exactly the right open questions. If the authors are right, then agency does not merely accompany evolution; it helps generate the very novelties on which further evolution can build. Agency, in that strong but naturalized sense, begets new forms of agency. That is a daring idea, but also one worth taking seriously.