The evolution of costly traits through selection and the importance of oral speech in e-collaboration

Abstract

Genes code for the expression of phenotypic traits, such as behavioral (e.g., aggressiveness) and morphological (e.g., opposing thumbs) traits. Costly traits are phenotypic traits that evolved in spite of imposing a fitness cost, often in the form of a survival handicap. In non-human animals, the classic example of costly trait is the peacock’s train, used by males to signal good health to females. It is argued here that oral speech is a costly trait evolved by our human ancestors to enable effective knowledge communication. It is shown that, because it is a costly trait, oral speech should be a particularly strong determinant of knowledge communication performance; an effect that generally applies to e-collaborative tasks performed by modern humans. The effects of oral speech support in e-collaborative tasks are discussed based on empirical studies, and shown to be consistent with the notion that oral speech is a costly trait. Specifically, it is shown that the use of e-collaboration technologies that suppress the ability to employ oral speech, when knowledge communication is attempted, leads to the two following negative outcomes: (a) a dramatic decrease in communication fluency, and (b) a significant increase in communication ambiguity. These effects are particularly acute in e-collaborative tasks of short duration.

Appendices

Appendix A: The threshold for evolution of costly traits

One of the most fundamental contributions to mathematical evolutionary thinking was made by Price (1970). He showed that for any trait to evolve through selection, in any population or subpopulation of individuals of the same species, the trait must satisfy Eq. (1), whose main element is a covariance term. The fitness of an individual that possesses the trait (e.g., number of surviving offspring) is measured through W, and Z is a measure of the manifestation of the trait in the individual (e.g., Z = 1 if the trait is present, and Z = 0 if it is absent). The trait in question can be any morphological, physiological or behavioral trait; examples could be opposing thumbs, aggressiveness, or a large train (tail appendage) with many eyespots.

$$ Cov\left( {W,Z} \right) > 0 $$

(1)

Equation (1) can be re-written as Eq. (2) in terms of the standardized measures of W and Z, referred to as w and z. This allows for its use in the context of path analysis (Wright 1934, 1960), which in turn greatly simplifies (as it will be shown below) theoretical reasoning based on comparative analyses of evolution of traits through selection.

$$ Cov\left( {w \cdot SW + \overline W, z \cdot SZ + \overline Z } \right) = Cov\left( {w \cdot SW,z \cdot SZ} \right) = SW \cdot SZ \cdot Cov\left( {w,z} \right) > 0 \Rightarrow Cov\left( {w,z} \right) > 0 $$

(2)

Figure 2 shows a path model where a costly trait measured by y is represented. All the measures are standardized, which is why they are indicated with lowercase letters. The measure y has a positive causal relationship with a task performance attribute a. For example, a could be number of lifetime copulations of an individual, a performance attribute associated with the mating task, in the case of a trait used for mate choice. The measure y has a negative causal relationship with s, a measure of survival performance. For example, s could be age of an individual at the time of death. Since any individual must be alive to perform any task, s also has a positive causal relationship with a. Both a and s have positive causal relationships with fitness (w). The magnitudes of these relationships are given by the path coefficients p _ay  , p _sy etc. All path coefficients are positive except for p _sy  , which is negative since it refers to the survival cost of trait y. (For simplicity, a trait measured by y is also called trait y.)

Fig. 2

Path model showing a costly trait and its relationship with fitness. (y: measure of a costly trait; e.g., y = 1 if the costly trait is present, and y = 0 if the trait is absent. a: task performance attribute measure. s: survival success measure; e.g., age at the time of death. w: fitness, measured as number of surviving offspring. p _ay  , p _sy etc.: path coefficients, or standardized partial regression coefficients)

In path analysis the covariance between any pair of variables is given by the sum of the products of the path coefficients in all paths connecting the two variables (Wright 1934, 1960). Thus, combining Eq. (2) with Fig. 2 leads to Eq. (3), which must be satisfied for any costly trait y to evolve through selection.

$$ {p_{wa}} \cdot {p_{ay}} + {p_{ws}} \cdot {p_{sy}} + {p_{wa}} \cdot {p_{as}} \cdot {p_{sy}} > 0 \Rightarrow {p_{wa}} \cdot {p_{ay}} > - {p_{sy}} \cdot \left( {{p_{ws}} + {p_{wa}} \cdot {p_{as}}} \right) \Rightarrow {p_{ay}} > - {p_{sy}} \cdot \left( {\frac{{{p_{ws}}}}{{{p}}} + {p_{as}}} \right) $$

(3)

For a costless trait x, a trait with no negative effect on survival, Eq. (3) is reduced to Eq. (4) because p _sy equals zero. What this equation tells us is that a costless trait x will always evolve as long as it has a positive causal relationship with a task performance attribute a, assuming that a has a positive causal relationship with fitness (w).

$$ {p_{ax}} > 0 $$

(4)

In the task of mating for example, any costless trait x that increases mating success (measured by a) would evolve through selection, with trait frequency growth subject to the constraints posed by chance events unrelated to the trait. That is, the trait would evolve to the point of becoming widespread in a population only if it is not eliminated by chance from the population at its early stages of evolution; e.g., the only individual that initially possesses the trait is killed by a lightning strike before reaching reproductive maturity (Gillespie 2004; Graur and Wen-Hsiung 2000). A costly trait y (e.g., the male peacock’s train), on the other hand, would have to meet a more stringent requirement for evolution. It would only evolve through selection if the trait’s positive effect on a surpassed the threshold given by the right side of Eq. (3).

Appendix B: Probability of evolution of costly traits

Let us assume that the appearance of a new costly trait y in a population will lead to a variation in p _ay and —p _sy that will be given by random numbers going from 0 to D. Let us also represent T as in Eq. (5):

$$ T = \left( {\frac{{p_{ws}}}{{p_{wa}}} + p_{as}} \right) $$

(5)

The value of T is assumed here to be largely population specific, in a relatively stable environment, and thus should remain relatively constant as new costly or costless traits appear in a population and either evolve or disappear in response to selection pressures. This can be illustrated for the task of mating, where a can be the number of lifetime copulations a male of a species engages in. In this case, the effect of a on fitness (w), the effect of survival performance (s) on w, and the effect of s on a are relatively constant for the males of the species.

In a relatively stable environment this can be shown to hold for any task whose performance is measured by a. This conclusion also follows from the assumption that those effects can be represented through stable regression coefficients; an assumption that is routinely used in mathematical population genetics models (Gillespie 2004; Rice 2004).

On the other hand, the effects that new traits x or y have on a and/or s can vary widely, since those traits appear in the population as a result of stochastic processes. Those effects will in turn ultimately dictate whether those traits will evolve or disappear in the species. For species that live today in environments similar to those in which most of their traits evolved, the value of T can be easily estimated empirically since the path coefficients are standardized partial regression coefficients.

Given the above assumptions, the probability of evolution of a new costly trait y in a population will be given by Eq. (6), assuming that T is equal to or greater than one. This equation reflects the intersection spaces of variation of D and DT, and can easily be verified through simple Monte Carlo simulations.

$$ \begin{array}{*{20}{c}} {P\left( {Evo(y)} \right) = }{\frac{D}{{D \cdot T}} \cdot \frac{1}{2}} \Rightarrow \\ {P\left( {Evo(y)} \right) = }{\frac{1}{{(2 \cdot T)}}}{} \\ \end{array} $$

(6)

T is assumed to be equal to or greater than one because it is difficult to conceive of a species population or subpopulation for which the performance of a task is more important for fitness than survival, even for the all-important task of mating. Let us consider, for example, spider species where the males are routinely cannibalized by their large and aggressive female mates during or after copulation (see, e.g., Wilder and Rypstra 2008). In these species, the male spiders must still successfully survive up to the moment of copulation. Therefore, when looked at as a subpopulation of the species to which they belong, those male spiders will likely have a ratio p _ws / p _wa that is greater than 1 (and thus a T greater than 1) for the task of mating, regardless of the fact that they contribute little more than their sperm to the survival of their offspring (and thus to their own fitness).

Equation (6) can be depicted in a graph, as shown in Fig. 3. The graph shows the variation of the probability of evolution of a new costly trait y in a population (vertical axis) based on values of T ranging from 1 to 10 (horizontal axis).

Fig. 3

Probability of evolution of a new costly trait. (Vertical axis: Probability of evolution of a new costly trait y in a population. Horizontal axis: T values ranging from 1 to 10)

As can be inferred from Fig. 3, costly traits will always have a lower probability of evolution than costless traits, because the value of T for the latter traits is always zero. This suggests that costly traits should be rarer in nature than costless ones, regardless of the task for which they were evolved; e.g., mating, communication, fighting.

Moreover, costly traits should be particularly rare in species where the value of T is high. This would be the case in species where the number of offspring born to females was small; and in species where the offspring relied heavily on their parents for survival in their early years of life, when most deaths occur. (In these species, the effect of survival on fitness would have been much higher than the effect of mating on fitness.) These are characteristics of the human species, and likely of the hominid ancestors in the human lineage (Boaz and Almquist 2001; Cartwright (2000). Thus, T values should have been high for our hominid ancestors, making the evolution of costly traits difficult.

Of course, in order to evolve, costly and costless traits also have to satisfy the condition that their covariance with fitness is greater than zero (i.e., that they have a positive net impact on fitness), which rarely is the case for new genetic mutations. Most new genetic mutations have either a negative or neutral effect on fitness; in the latter case they may evolve by chance, through a process known as genetic drift (Gillespie 2004; Hartl and Clark 2007; Maynard Smith 1998). Costly traits, unlike costless ones, have another condition to satisfy: they must overcome the survival costs that they impose.

Appendix C: Different effects of costly and costless traits

Let us also assume that the appearance of costly and costless traits in a population will lead to random values of p _ax and—p _sy in the range from 0 to D. The expected value of p _ax for costless traits that evolve will then be given by D divided by 2. The expected value of p _ay , on the other hand, will be given by (DT−D)/2, assuming that T is equal to or greater than 1. Therefore, the expected ratios between p _ay and p _ax will be given by Eq. (7).

$$ \begin{array}{*{20}{c}} {E\left( {{{{p_{ay}}} \mathord{\left/{\vphantom {{{p_{ay}}} {{p_{ax}}}}} \right.} {{p_{ax}}}}} \right) = }{\frac{{{{\left( {D \cdot T - D} \right)} \mathord{\left/{\vphantom {{\left( {D \cdot T - D} \right)} 2}} \right.} 2}}}{{{D \mathord{\left/{\vphantom {D 2}} \right.} 2}}}} \Rightarrow \\ {E\left( {{{{p_{ay}}} \mathord{\left/{\vphantom {{{p_{ay}}} {{p_{ax}}}}} \right.} {{p_{ax}}}}} \right) = }{T - 1}{} \\ \end{array} $$

(7)

The graph in Fig. 4 shows the variation of the expected ratio between p _ay and p _ax (vertical axis) based on values of T ranging from 1 to 10 (horizontal axis). The ratio grows proportionally with T, and is a measure of how strong the expected effect of a costly trait y on the task performance attribute a is, compared with the expected effect of a costless trait x. For simplicity, it is assumed here that both types of traits are either independent from each other, or spread to fixation in a species at different points in time. For dependent traits or traits that evolve at the same time, the mathematical analysis becomes more complex, but the results are qualitatively the same.

Fig. 4

Variation of the expected ratio between p _ay and p _ax. (Vertical axis: Expected ratio between p _ay and p _ax . Horizontal axis: T values ranging from 0 to 10.)

An expected ratio between p _ay and p _ax of 1.3, for example, means that the standardized effect of any costly trait on a given task performance attribute a is on average 30% stronger than the effect of any costless trait on the same task performance attribute. A ratio of 8 means that the costly trait is on average 800% stronger (e.g., p _ay  = .4 and p _ax  = .05). Standardized effects are expressed in terms of standard deviations of the variables to which they refer. For example, a p _ay  = .4 means that a 1 standard deviation variation in y causes a .4 standard deviation variation in a.