The question of just what is an individual can be formally considered using an approach described as multilevel selection theory (MST). More precisely, MST can be viewed as a means of understanding the potential for group selection to operate on entities that one would not necessarily automatically consider to represent a single individual. As the name implies, MST is predicated on the assumption that selection could operate on multiples level, i.e., ranging from that of individual genes to gene complexes to individual cells to multicellular organisms (e.g., animals), to groups of organisms within populations, to populations and even to communities. The key question to consider in any of these situations, however, is not so much whether selection can operate at a given level, since unquestionably it can, but instead whether selection can operate strongly enough such that it isn't overwhelmed by selection operating on the individual, i.e., rendering these other levels mostly irrelevant in terms of natural selection and adaptation. Of course, these musings are complicated to the extent that it becomes difficult to effectively define what one means by "individual"!
It is also possible to invert this concern by describing at least that level at which natural selection operates most strongly as representing the primary individual within a given multiple-level hierarchy. For multicellular organisms, this primary individual often corresponds to the multicellular organism rather than, for example, its cells or rather than grouping of multicellular organisms. For unicellular organisms, by contrast, a good default assumption is that the individual corresponds to free-living, i.e., individual cells. Complications arise when conflicts exist between the primacy of the multicellular or unicellular individual versus individual cells (for the former), groups of cells (for the latter), or even conflicts between selection acting on individual genes versus selection acting on the cells (or viruses) encoding those genes.
With group selection, a formal assumption is made that groups of what may otherwise be reasonably described as individuals may be subject to natural selection that is comparable or even stronger than that which acts more directly upon these otherwise individuals. Situations in which group selection might arise can be a consequence of emergent properties. One such property could be flocking behavior among birds, or sentry behavior among many potential prey animals. Microbes, too, can display group behaviors, such as producing sufficient quantities of toxin to modify the phenotype of host organisms (e.g., making animals sick) and/or explicitly displaying group behaviors (such as light production) only when bacterial densities are sufficiently high (with densities detected within the group achieved via quorum sensing). That is, there exist a number of circumstances in which group behaviors are clearly adaptive and therefore one might be tempted to predict that these adaptations are a product of selection acting upon a group, that is, group selection.
Despite the obviousness of the above inference, and even the demonstration of circumstances in which group selection may indeed be rigorously established, it nonetheless is always good as a modern evolutionary biologist to assume, by default, that group selection is less powerful then selection acting upon individuals. Why? This can be answered in terms of how one would go about demonstrating that group selection in fact is operating. First, it is helpful to be able to demonstrate that a given group property is indeed an emergent property, that is, a property of the group that is not merely a sum of the equivalent individual properties. For example, predators working in groups may be more effective hunters than individual hunters, per hunting party, but are they more effective per hunter? Indeed, do they do anything different, such as approach prey from different angles, which truly makes the group behavior different from that of the individual behavior? Second, and related to the above idea of making a qualitative difference, is it possible to even distinguish selection acting on the group from selection acting on the individual? That is, seemingly group adaptations may be a product of selection acting solely on the individual rather than on the group. Collective individual selection, however, is not group selection.
Another consideration, one stemming from population genetics, is limitations on how natural selection operates. That is, selection involves a differential reproductive success of some unit of inheritance, i.e., some unit of selection. For natural selection to operate on groups, that unit of selection must be that of the group. In addition, for natural selection to operate on groups as units of selection there must exist distinguishing genetic variation between groups. In considering groups as equivalent to individuals, we thus could envisage numerous groups that vary in terms of their collective genotypes and phenotypes. These groups could splinter off to form new groups – that is, display something approximating a life cycle – such that the most fit group would produce more progeny groups than the less fit groups. The problem with this scenario is that the actual unit of inheritance is located within the individuals making up a given group, and the mechanisms giving rise to group reproduction actually involve individual reproduction that is followed by some sort of fission of a single group into, for example, two groups. It is difficult, in other words, to separate selection that is acting on individuals from selection that instead is acting upon groups of individuals.
The individuals within a group represent competing units of selection. Selection operating on these individuals favors those that have the greatest reproductive success. If behaviors that increase the reproductive success of individuals also increase the reproductive success of the group, then there are no conflicts and those traits should increase in frequency. Under such circumstances, however, it would be difficult to convincingly argue that what one is observing is group selection rather than individual selection. On other hand, traits that bestow significant reproductive success within a group may not also bestow greater reproductive success for the group itself, in which case a conflict would exist between two units of selection. In this case group selection could predominate over individual selection only to the extent that within-group reproductive success does not come to dominate between-group reproductive success. Dominance of group success over individual success, by default, is assumed to be unlikely, however, and particularly so the less individual reproductive success is dependent on group reproductive success). These conflicts become all the more difficult to overcome to the extent that individuals are free to move between groups, or should groups display within-group genetic variation at all. Alternatively, we can blur the distinction between the individual and the group by removing all genetic variability within groups, while retaining genetic variability between groups, but at that point we may be blurring just what it means to be an individual and, consequently, group selection again may not be reasonably invoked.
Note that units of selection may also refer to subcellular genetic entities. Obvious examples include the various viruses. These are clearly units of selection that not only are distinct from their cellular hosts but also which clearly can be in conflict with their hosts. This conflict stems from an absence of complete reproductive linkage between host and virus, that is, viruses can reproduce independently of their host's reproduction plus, and this may be as important or even more important, can disseminate their progeny independently of their host. Thus, a virus' fitness can to a large extent be independent of that of its host. Other genetic entities can be less autonomous and therefore less able to conflict with their host's reproductive capacity (i.e., their host's fitness). These include plasmids, transposons, etc. These too, however, can be viewed as units of selection and selection acting on these entities is not necessarily entirely equivalent to that acting on the host. Instead, organisms, in toto, likely represent a balance between the fitness needs of the organism itself and the various needs of the parasites, genetic and otherwise, that they carry around with them, with excessive conflicts resolved either by bringing the parasite under control or, alternatively, with loss of the host or its lineage from the host population.
Another way of considering these ideas is in terms of what can be described as a transfer of fitness. Here, effectively, the reproductive output along with subsequent dissemination at one, typically lower level, is accomplished only via reproduction at a higher level. For example, the long-term evolutionary success of a single cell can be dependent on the success of the multicellular body within which that cell is found. From Michod et al. (2003) , p. 96:
Such associations and groups may persist and reform with varying likelihood depending on properties of the group and the component individuals. Initially, group fitness is the average of the lower-level individual fitnesses, but as the evolutionary transition proceeds, group fitness becomes decoupled from the fitness of its lower-level components. Indeed, the essence of an evolutionary transition in individuality is that the lower-level individuals must "relinquish" their "claim" to fitness, that is to flourish and multiply, in favor of the new higher-level unit. This transfer of fitness from lower to higher-levels occurs through the evolution of cooperation and mediators of conflict that restrict the opportunity for within-group change and enhance the opportunity for between-group change. Until, eventually, the group becomes a new evolutionary individual in the sense of being evolvable—possessing heritable variation in fitness (at the new level of organization) and being protected from the ravages of within-group change by adaptations that restrict the opportunity for defection (Michod and Roze, 1999) .
Conflicts between units of selection give rise to temptations for cheating. This cheating can be viewed in terms of lower units of selection (e.g., individuals in groups or genetic entities making up individual cells) impacting the fitness of higher units of selection (e.g., groups, or even the collection of genes making up individual cells). Metaphorically, we can view these conflicts as equivalent to a Tragedy of the Commons. That is, the higher units of selection themselves are the commons and the lower units are the individuals inhabiting the commons, regardless of where on the spectrum of levels of selection the various units reside. To be a commons, though, the higher or larger unit must both include the lower or smaller units and contribute to the fitness of these smaller units (e.g., such as genes making up a genome), just as environments contribute to the fitness of the individuals inhabiting those environments. Thus, cheating is an act that has the effect of reducing the "fitness" of the larger unit of selection that literally is the commons, but more importantly given the use of the term 'commons', this cheating reduces the capacity of the commons to contribute to the fitness of the smaller units making up the commons. Instead of "commons", the same idea is often described in terms of the existence of a common good. Cheating thus has the effect of degrading a common good.
The commons in the above example could be an individual organism, such as an animal, while the lower unit of selection could be the cells making up that animal. In this case the cells benefit from the ongoing overall health of the multicellular individual of which they are a part. If the cells start cheating, however, then this can degrade the ability of the multicellular individual to sustain its cells. Another metaphor often invoked to describe this cheating is that it is equivalent to cancer evolution within multicellular organisms (Travisano and Velicer, 2004) . These researchers also define a cheating load which is a measure of the negative impact that cheating has on the larger group (or commons). Resolution of conflicts and/or otherwise keeping cheaters at bay is crucial to the evolution of multicellularity, and indeed by and large such cheating is not just equivalent to cancer: Within multicellular organisms it is cancer! Thus, (p. 72):
Tumour cells arise from normal cells as a result of mutation and as normal cells do, they benefit from access to nutrients and protection from the external environment that is provided by the body. However, in contrast to normal cells, tumour cells cheat by proliferating without restriction and without contributing to overall function. It is primarily the unrestricted growth of tumours that causes their deleterious effects, because surrounding tissues are disrupted. If cheating cells remain localized forming a benign tumour, cheating might have only modest effects. However, strongly deleterious effects are much more probable when cheating spreads throughout the body as occurs in metastatic cancer, resulting in tissue and organ failure. Therefore, the deleterious effects of cheating on the group are largely determined by the aggressiveness of growth and dispersive ability of tumour cells. The more aggressive the tumour, the greater the short-term benefit for the cheater (greater proliferation of the cheater genotype), but also greater is the likelihood of long-term costs as a result of increased mortality and reduced reproduction of the whole organism.