2 Environmental Effects

Paradoxically, one of the greatest benefits of studies that can detect and control for genetic effects is the information they can provide about the sources of environmental influence. We make an important distinction between identifying which are the best places to look for specific environmental agents and deciding what those specific agents are. For example, it may be possible to show that variation in diastolic blood pressure is influenced by environmental effects shared by family members long before it is possible to demonstrate that the salient environmental factor is the amount of sodium in the diet. We make a similar distinction between estimating the overall contribution of genetic effects and identifying specific loci that account for a significant fraction of the total genetic variation. Using some of the methods we shall describe later in this book it may indeed be possible to estimate the contribution of specific factors to the environmental component of variation (see Chapter 10). However, using the biometrical genetical approach which relies only on the complex patters of family resemblance, it is possible to make some very important statements about the structure of the environment in advance of our ability to identify the specific features of the environment that are most important. Although the full subtlety for analyzing the environment cannot be achieved with data on twins alone, much less on twins reared together, it is nevertheless possible to make some important preliminary statements about the major sources of environmental influence which can provide a basis for future studies.

We may conceive of the total environmental variation in a trait as arising from a number of sources. The first major distinction we make is between environmental factors that operate within families and those which create differences between families. Sometimes the environment within families is called the unique environment or the specific environment or the random environment. Different authors may refer to it as $V_E$, $V_{SE}$, $E_1$, $E_W$ or $e^2$, but the important thing is to understand the concept behind the symbols. The within-family environment refers to all those environmental influences which are so random in their origin, and idiosyncratic in their effects, as to contribute to differences between members of the same family. They are captured by Hamlet's words from the famous `to be or not to be' soliloquy:

...the slings and arrows of outrageous fortune.

The within-family environment will even contribute to differences between individuals of the same genotype reared in the same family. Thus, the single most direct measure of their impact is the variation within pairs of MZ twins reared together.

Obviously, if a large proportion of the total variation is due to environmental differences within families we might expect to look more closely at the different experiences of family members such as MZ twins in the hope of identifying particular environmental factors. However, we have to take account of a further important distinction, namely that between ``long-term'' and ``short-term'' environmental effects, even within families. If we only make a single measurement on every individual in a study of MZ twins, say, we cannot tell whether the observed phenotypic differences between members of an MZ twin pair are due to some lasting impact of an early environmental trauma, or due to much more transient differences that influence how the twins function on the particular occasion of measurement. Many of the latter kinds of influence are captured by the concept of ``unreliability'' variance in measurement theory. There is, of course, no hard and fast distinction between the two sources of variation because how far one investigator is prepared to treat short-term fluctuations as ``unreliability'' is largely a matter of his or her frame of reference. In depression, which is inherently episodic, short term fluctuations in behavior may point to quite specific environmental factors that trigger specific episodes (see, e.g., Kendler et al., 1986). The main thing to realize is that what a single cross-sectional study assigns to the ``within-family'' environment may or may not be resolved into specific non-trivial environmental causes. How far to proceed with the analysis of within-family environment is a matter for the judgement and ingenuity of the particular investigator, aided by such data on repeated measures as he or she may gather.

The between-family environment would seem to be the place that many of the conceptually important environmental effects might operate. Any environmental factors that are shared by family members will create differences between families and make family members relatively more similar. The environment between families is sometimes called the shared environment, the common environment or just the family environment. Sometimes it is represented by the symbols $E_2$, $EB$, $EC$, $CE$, $c^2$ or $V_{EC}$. Again, all these symbols denote the same underlying processes.

In twin studies, the shared environment is expected to contribute to the correlation of both MZ and DZ twins as long as they are reared together. Just as we distinguish short-term and long-term effects of the within-family environment, so it is conceptually important to note that the effects of the shared environment may be more or less permanent and may persist even if family members are separated later in life, or they may be relatively transient in that they are expressed as long as individuals are living together, perhaps as children with their parents, but are dissipated as soon as the source of shared environmental influence is removed. Such effects can be detected by comparing the analyses of different age groups in a cross-sectional study, or by tracing changes in the contribution of the shared environment in a longitudinal genetic study (see Chapter ).

It is a popular misconception that studies of twins reared together can offer no insight about the effects of the shared environment. As we shall see in the following chapters, this is far from the case. Large samples of twins reared together can provide a strong prima facie case for the importance of between-family environmental effects that account for a significant proportion of the total variance. The weakness of twin studies, however, is that the various sources of the shared environment cannot be discriminated. It is nevertheless essential for our understanding of what the twin study can achieve, to recognize some of the reasons why this design can never be a ``one-shot,'' self-contained investigation and why investigators should be open to the possibility of significant extensions of the twin study (see Chapter ).

The environmental similarity between twins may itself be due to several distinct sources whose resolution would require more extensive studies. First, we may identify the environmental impact of parents on their children. That is, part of the common environment effect in twins, can be traced to the fact that children learn from their parents. Formally, this implies that some aspect of variation in the maternal or paternal phenotypes (or both) creates part of the environmental variation between pairs of children. An excellent starting point for exploring some of these effects is the extension of the classical twin study to include data on the parents of twins (see Chapter ). In principle, we might find that parents do not contribute equally to the shared family environment. The effect of mothers on the environment of their offspring is usually called the ``maternal effect'' and the impact of fathers is called the ``paternal effect.'' Although these effects can be resolved by parent-offspring data, they cannot be separated from each other as long as we only have twins in the sample.

Following the terms introduced by Cavalli-Sforza and Feldman (1981), the environmental effects of parent on child are often called vertical cultural transmission to reflect the fact that non-genetic information is passed vertically down the family tree from parents to children. However, the precise effects of the parental environment on the pattern of family resemblance depend on which aspect of the parental phenotype is affecting the offspring's environment. The shared environment of the children may depend on the same trait in the parents that is being measured in the offspring. For example, the environment that makes offspring more or less conservative depends directly on the conservatism of their parents. In this case we normally speak of ``phenotype-to-environment (`P to E')'' transmission. It is quite possible, however, that part of the shared environment of the offspring is created by aspects of parental behavior that are different from those measured in the children, although the two may be somewhat correlated. Thus, for example, parental income may exercise a direct effect on offspring educational level through its effect on duration and quality of schooling. Another example would be the effect of parental warmth or protectiveness on the development of anxiety or depression in their children. In this case we have a case of correlated variable transmission. Haley, Jinks and Last (1981) make a similar distinction between the ``one character'' and ``two character'' models for maternal effects. The additional feature of the parental phenotype may or may not be measured in either parents or children. When such additional traits are measured in addition to the trait of primary interest we will require multivariate genetic models to perform the data analysis properly. Some simple examples of these methods will be described in later chapters. Two extreme examples of correlated variable transmission are where the variable causing the shared environment is:

1. an index purely of the environmental determinants of the phenotype -- ``environment-to- environment (`E to E')'' transmission
2. purely genetic -- ``genotype-to- environment (`G to E')'' transmission.

Although we can almost never claim to have a direct measure of the genotype for any quantitative trait, the latter conception recognizes that there may be a genetic environment (see e.g. Darlington, 1971), that is, genetic differences between some members of a population may be part of the environment of others. One consequence of the genetic environment is the seemingly paradoxical notion that different genetic relationships also can be used to tease out certain important aspects of the environment. For example, the children of identical twins can be used to provide a test of the environmental impact of the maternal genotype on the phenotypes of their children (see e.g., Nance and Corey, 1976). A concrete example of this phenomenon would be the demonstration that a mother's genes affect the birthweight of her children.

Although researchers in the behavioral sciences almost instinctively identify the parents as the most salient feature of the shared environment, we must recognize that there are other environmental factors shared by family members that do not depend directly on the parents. There are several factors that can create residual (non-parental) shared environmental effects. First, there may be factors that are shared between all types of offspring, twin or non-twin; these may be called sibling shared environments. Second, twins may share a more similar pre-and postnatal environment than siblings simply because they are conceived, born and develop at the same time. This additional correlation between the environments of twins is called the special twin environment and is expected to make both MZ and DZ twins more alike than siblings even in the absence of genetic effects. It is important to note that even twins separated at birth share the same pre-natal environment, so a comparison of twins reared together and apart is only able to provide a simple test of the post-natal shared environment.

A further type of environmental partition, the special MZ twin environment is sometimes postulated to explain the fact that MZ twins reared together are more correlated than DZ twins. This is the most usual environmental explanation offered as an alternative to genetic models for individual differences because the effects of the special MZ environment will tend to mimic those of genes in twins reared together. It is because of concern that genetic effects may be partly confounded with any special MZ twin environments that we stress the importance of thinking beyond the twin study to include other relationships. It becomes increasingly difficult to favor a purely non-genetic explanation of MZ twin similarity when the genetic model is able to predict the correlations for a rich variety of relationships from a few fairly simple principles. Since the special twin environment, however, would increase the correlation of MZ twins, its effects may often resemble those of non-additive genetic effects (dominance and epistasis) in models for family resemblance.