While it is something of an oversimplification (we will introduce the complexities associated with the random aspects of reproduction and the linked nature of genes shortly), we begin with three basic types of selection: stabilizing (or conservative), directed, and disruptive. We start with a population composed of individuals displaying genetic variation in a particular trait. The ongoing processes of mutation continually introduces new genotypes, and their associated effects on phenotype. What is important to remember is that changes in the population and the environment can influence the predominant type of selection occurring over time, and that different types of selection may well (and most certainly are) occurring for different traits.
For each type of selection, we illustrate the effects as if they were acting along a single dimension, for example smaller to larger, stronger to weaker, lighter to darker, or slower to faster. In fact, most traits vary along a number of dimensions. For example, consider the trait of ear, paw, heart, or big toe shape. An appropriate type of graph would be a multi-dimensional surface, but that is harder to draw. Also, for simplicity, we start with populations whose distribution for a particular trait can be described by a simple and symmetrical curve, that is the mean and the median are equal. New variants, based on new mutations, generally fall more or less randomly within this distribution. Under these conditions, for selection NOT to occur we would have to make two seriously unrealistic assumptions: first that all organisms are equally successful at producing offspring, and second that each organism or pair of organisms produce only one or two (respectively) offspring. Whenever these are not the case, which is always, selective processes will occur, although the strength of selection may vary dramatically between traits.
Stabilizing selection: Sometimes a population of organisms appears static for extended periods of time, that is, the mean and standard deviation of a trait are not changing. Does that mean that selection has stopped? Obviously we can turn this question around, assume that there is a population with a certain stable mean and standard deviation of a trait. What would happen over time if selection disappeared?
Let us assume we are dealing with an established population living in a stable environment. This is a real world population, where organisms are capable of reproducing more, and sometimes, many more organisms than are needed to replace them when they die and that these organisms mate randomly with one another. Now we have to consider the factors that lead to the original population distribution: why is the mean value of the trait the value it is? What factors influence the observed standard deviation? Assuming that natural selection is active, it must be that organisms that display a value of the trait far from the mean are (on average) at a reproductive disadvantage compare to those with the mean value of the trait. We do not know why this is the case (and don’t really care at the moment). Now if selection (at least for this value of the trait) is inactive, what happens? The organisms far from the mean are no longer at a reproductive disadvantage, so their numbers in the population will increase. The standard deviation will grow larger, until at the extreme, the distribution would be flat, characterized by a maximum and a minimum value. New mutations and existing alleles that alter the trait will not be selected against, so they will increase in frequency. But in our real population, the mean and standard deviation associated with the trait remain constant. We predict selection against extreme values of the trait. We can measure that degree of selection “pressure” by following the reproductive success of individuals with different values of the trait . We might predict that the more extreme the trait, that is, the further from the population mean, the greater its reproductive disadvantage would be, so that with each generation, the contribution of these outliers is reduced. The distribution's mean will remain constant. The stronger the disadvantage the outliers face, the narrower the distribution will be – that is, the smaller the standard deviation. In the end, the size of the standard deviation will reflect both the strength of selection against outliers and the rate at which new variation enters the population through mutation. Similarly, we might predict that where a trait’s distribution is broad the impact of the trait on reproductive success will be relatively weak. Directed selection: Now imagine that the population’s environment changes. It may now be the case that the phenotype of the mean is no longer the optimal phenotype in terms of reproductive success (the only factor that matters, evolutionarily); a smaller or a larger value may be more favorable. Under these conditions we would expect that, over time, the mean of the distribution would shift toward the phenotypic value associated with maximum reproductive success. Once reached, and assuming the environment stays constant, stabilizing selection again becomes the predominant process. For directed selection to work, the environment must change at a rate and to an extent compatible with the changing mean phenotype of the population. Too big and too rapid a change and the reproductive success of all members of the population could be dramatically reduced. The ability of the population to change will depend upon the variation already present within the population. While new mutations leading to new alleles are appearing, this is a relatively slow process. In some cases, the change in the environment is so fast or so drastic and the associated impact on reproduction so severe that selection will fail to move the population and extinction will occur. One outcome to emerge from a changing environment leading to the directed selection is that as the selected population’s mean moves, it may well alter the environment of other organisms.
Disruptive selection: A third possibility is that organisms find themselves in an environment in which traits at the extremes of the population distribution have a reproductive advantage over those nearer the mean. If we think about the trait distribution as a multidimensional surface, it is possible that in a particular environment, there will be multiple and distinct strategies that lead to greater reproductive success compared to others. This leads to what is known as disruptive selection. The effect of disruptive selection in a sexually reproducing population will be opposed by the random mating between members of the population (this is not an issue in asexual populations). But is random mating a good assumption? It could be that the different environments, which we will refer to as ecological niches, are physically distant from one another and organisms do not travel far to find a mate. The population will split into subpopulations in the process of adapting to the two different niches. Over time, two species could emerge, since whom one chooses to mate with and the productivity of that mating, are themselves selectable traits.