134 COPING RESOURCES
The direction of causation is ideally assessed in an experimental study in which
we manipulate (rather than measure) the independent variable. In health
psychology, interventions such as behaviour change interventions provide good
examples of experimental methodology. For example, one group may receive an
intervention to change attitudes or reduce work stress while another (control) group
receives no intervention. If participants are randomly allocated to these two groups
(to try to evenly distribute confounding factors across groups) or matched (to
balance confounding factors) then any difference in the dependent variable
following the manipulation (that is, the intervention) can be reasonably attributed to
that manipulation. The classic use of experimental methodology in health
psychology is the randomized controlled trial (RCT). It is of course worth noting that
such experimental methods merely establish one causal determinant of the
dependent variable, they do not necessarily demonstrate that this is the one and
only causal determinant.
Unfortunately, we often cannot manipulate independent variables in health
psychology and so must infer underlying causal mechanisms from correlational
data. For example, in relation to smoking, the majority of the evidence supporting
an impact of smoking on cancer and cardiovascular disease outcomes is
correlational, at least for studies in humans. Similarly, the relationship between
personality and health is based on correlational data.
There are well-known dangers in drawing causal inferences from correlational
data. Two key issues are causal direction and the third variable problem. Causal
direction refers to the issue of the direction of effect being unknown when two
variables are correlated: did A cause B or B cause A? For example, in relation to
personality and health this issue becomes one of whether a personality trait
resulted in a health outcome or the health outcome produced the personality trait.
So, for example, some patients with serious illnesses such as cancer may become
anxious and neurotic. This might lead us to the erroneous conclusion that
neuroticism played a role in causing the cancer when in fact the cancer had
produced increased neuroticism.
Third variable problems refer to the possibility that a correlation between two
variables might be due to both variables being caused by a third variable. So, for
example, there is some evidence that a hyper-responsive nervous system is an
underlying factor in both the development of an anxious personality (high
neuroticism) and the development of heart disease. Here an anxious, reactive
personality would be related to (that is, correlated with) heart disease without being
a causal determinant of heart disease (McCabe et al., 2000). Similarly, Eysenck
(1967) argued that extraversion relates to differences in the sensitivity of the
nervous system that influences emotional reactions and reactions to socialization.
Extraverts may also be more likely to seek stimulation through behaviours such as
smoking. In both these cases it is not the personality trait itself that causes the
health outcomes but an underlying biological mechanism that causes both the
personality trait and the health outcome.