A while ago, Bora referred to a short article about David Sloan Wilson, whose research program examines group selection (among other things). Any discussion of group selection is almost always contentious, largely because there is a fundamental confusion (or conflation) of two different phenomena: the evolution of groups and selection among groups (i.e., group selection).
Evolution of groups is a fascinating and fundamental issue in biology, whether it be the evolution of multicellular organisms from single cell organisms, or the evolution of behaviors that create coordinate group behavior, such as punishing 'cheaters.' But I would argue that calling this phenomenon group selection is incorrect, but happens often, I think, because of a confusion about what a group is. To define a group simply as a collection of individuals that happen to be in the same place is incorrect.
For example, there is a difference between a lion pride, with its hierarchies, task sharing, and cooperative hunting, and two solitary lions. Simple geographic proximity does not make a group, any more than a set of ameobas that happen to be near each other is a 'group.' Interactions among individuals also does not mean that there is a group: two solitary tigers (i.e., A Pride of One) that are fighting over terrority don't make a group either.
So what does make a group? I would argue that groups require mechanisms of cohesion and integration. For example, cancer is essentially the breakdown of these mechanisms: one cell begins to replicate at the expense of the coordinate group (the patient). Whether it be behavior or biochemistry, a group has what many have called emergent properties. To carry the charismatic feline megafauna example further, cooperative hunting (and food sharing) can only exist within the context of an integrated group. Obviously, one solitary lion can't engage in cooperative hunting.
These emergent properties do not preclude selection on individuals: obviously, faster lions, regardless of group dynamics, will hunt better than slow lions (granted, I'm making a simplistic assumption about how lions hunt, but you get the idea). In fact, I would argue much of the evolutionary change we observe will either stem from selection acting directly on individuals within groups (i.e., selection for resistance to disease), or on individuals within groups (fighting over a water buffalo carcass). However, with regard to groups, there will be properties of groups that are exclusive to groups.
This leads to how I would define group selection. Group selection, to me, is the differential selection of groups based on heritable group variation (or traits). If one takes a group character such as cooperation, one can imagine that groups vary in how cooperative they are. Again, to take the example of group hunting, some groups may always hunt together, whereas others may do so only rarely. Depending on the environment, one strategy could be better than another. For example, if food occurs in large, discrete packets (e.g., water buffalo), group hunting is useful. If most of a group's food occurs in small, ubiquitous packets (e.g., lots of little warthogs running around), the cost in time and energy that nine lions would spend in hunting a single warthog is highly disadvantageous.
(an aside: This is obviously an oversimplification. We've known for several decades that lion prides vary in size (including the Pride of One) based on the availability of resources and infanticidal males. Behavioral characteristics can be quite plastic, and there might actually be considerable selection for behavioral plasticity. This, however, would also require a discussion of stabilizing selection, which I'm just not in the mood to do).
I think if the discussion of groups were to be decomposed into these two different phenomena, it would help greatly (particularly when reported on in the press). Both the evolution of groups, and selection on group traits that cannot be decomposed to individual level traits (i.e., group selection) are biologically interesting and worth studying. Is it a bit of an artificial distinction? As the lion pride example shows, yes. Lion pride size and behavior can vary to the point where there is no pride (the Pride of One), and one lion can't be a group. But, I would argue, that having this framework is a useful way of framing questions that can be addressed experimentally (which is one reason to have theories).
So let's talk about chickens. In the article Bora discusses, Wilson describes an experiment (not to be snooty, but it's one experiment, not two):
Two experiments using chickens show another aspect of evolution regarding selection. In the first experiment, groups of chickens in cages were evaluated for egg-laying. The best egg-layer within each cage was chosen and put together with the other prolific egg-layers. The second experiment took the best caged groups of egg-layers. The result of the second experiment after a few generations was healthy, sociable, egg-laying chickens. The result of the first experiment after a few generations was fighting and anti-social chickens that maimed and killed each other.
This experiment doesn't look at differences among groups, but instead examines the preconditions necessary for the evolution of groups (the assumption is that cooperative groups as a whole churn out more eggs; if selfish, psycho chickens on average lay more eggs, then there's no need to resort to a group explanation--since there is no group). Comparing groups of chickens that differ in their aggresiveness would get at group selection (and individual selection).
Like I said, this is a contentious area, so have at it...
A related note: Another issue is how the effect of group selection is actually measured. Is it the effect of group traits on individual fitness; for example, does cooperation improve lions' reproductive output? Or, should fitness be determined by the proliferation of groups? (and again, this is a bit of an artificial construct, since higher reproductive output could give rise to larger lion prides, more lion prides, or both).
The last note (I stil need to read the post carefully): the former is demic selection, the later is group selection proper. Much of the confusion in this debate arises from the misuse of proper terminology.
isn't interdemic selection one of the primary subsets of group selection?
An interesting example in shift in level of selection: transmissible cancers DFTD and CTVT. Within the original host individual tumor cells were subject to somatic selection. Once transmissible, they also became collectively subject to individual selection as a distinct organism.
Razib: I guess you can look at it this way, or one can keep them separate - I like to follow R.Brandon who keeps them separate.
There are two qualitatively different types of multilevel selection:
1) Selection of permanent (multi-generational) functionally integrated entities (gene, chromosome, cell, organism, eusocial colony, group). These entities exhibit obligate reproductive integration. Some examples: mitosis, sexual reproduction, gestation, sterile castes, collective protection of young. At the supra-organismic level, they also engage in coordinated (rather than merely aggregate) modification of the environment that enhances fitness. (See Smith and Szathmary's major evolutionary transitions, extended organism - J. Scott Turner.)
2) Selection of taxa (deme, subspecies, species, superspecies, clade). While these are all analytic units, they are not integrated functional units (besides genetic exchange within the superspecies level and below). In contrast, type 1 entities are both analytic entities and functional reproductively integrated units.
Correction: A typo due to haste - I know that the full surname is Maynard Smith.
Type 1 selection is based on collective traits (coordinated actions), type 2 selection is based on traits in common.