2 ms·

In addition to noise and small changes, the same genome can contain multitude of mutually exclusive anatomical and behavioral traits. For example, high stress

In addition to noise and small changes, the same genome can contain multitude of mutually exclusive anatomical and behavioral traits. For example, high stress environment might trigger different genetic traits than low stress environment, it's neither nature or nurture but both. In some animals you can get so different phenotype by altering environment that it's hard to recognize that they are the same species. The idea that the genes just run the same process for the same set of genes and the environment just linearly modulates some traits is not the whole picture.

gwern 7y ago
Gene-environment interaction already falls into the 'heritability' variance component, so appealing to GxE can't explain what the nonshared-environment/error variance component is. Is it mostly something trivially boring & unimportant like measurement error (which does account for a lot of it, especially behavioral stuff), or is it large meso-scale environmental factors which happen to not be exactly identical within a household such as which college one goes to (which are the bread and butter of sociology, fit nicely into theories, and can potentially be measured, controlled, and equalized), or is it low-level biological details like random cells mutating or random cells going haywire (possible but extremely difficult to measure, control, or derive any understanding from, which is depressing from both the scientific & social points of view)? This might seem to be impossible to solve: behavioral geneticists have already used extensive checklists and inventories to measure all sorts of aspects of families to try to reduce the nonshared-environment down to clear constructs, and failed to explain more than trivial amounts of variance (indeed, it worked out a lot better at showing how the 'environment' is genetically-caused); but you could always argue that they were measuring the relevant environment badly or measuring the wrong things entirely (maybe it's not how the mother interacts with the child in a nurturing or authoritarian fashion, but the father, etc). OP is about the clever research approaches which allow digging into it and distinguishing between high-level environmental variables, and just damn random noise and low-level problems, and quantifying how much of each.

gwern 7y ago

Gene-environment interaction already falls into the 'heritability' variance component, so appealing to GxE can't explain what the nonshared-environment/error variance component is. Is it mostly something trivially boring & unimportant like measurement error (which does account for a lot of it, especially behavioral stuff), or is it large meso-scale environmental factors which happen to not be exactly identical within a household such as which college one goes to (which are the bread and butter of sociology, fit nicely into theories, and can potentially be measured, controlled, and equalized), or is it low-level biological details like random cells mutating or random cells going haywire (possible but extremely difficult to measure, control, or derive any understanding from, which is depressing from both the scientific & social points of view)? This might seem to be impossible to solve: behavioral geneticists have already used extensive checklists and inventories to measure all sorts of aspects of families to try to reduce the nonshared-environment down to clear constructs, and failed to explain more than trivial amounts of variance (indeed, it worked out a lot better at showing how the 'environment' is genetically-caused); but you could always argue that they were measuring the relevant environment badly or measuring the wrong things entirely (maybe it's not how the mother interacts with the child in a nurturing or authoritarian fashion, but the father, etc). OP is about the clever research approaches which allow digging into it and distinguishing between high-level environmental variables, and just damn random noise and low-level problems, and quantifying how much of each.