ABSTRACT
Before discussing the evidence on causes, it is necessary to clarify several key issues (see Academy of Medical Sciences 2007). First, even with the most straightforward situation, it is very rare to find examples of a one-to-one all-encompassing causal effect. Rothman and Greenland (1998) illustrated the point with the example of turning on a light by flicking a switch. A person need do nothing else except flick the switch to cause the light to go on – surely that must constitute a simple, straightforward, direct causal effect? It does not, however, because the light will not go on unless the bulb is functional, the electric circuit is intact, the required voltage is available and so forth. Even with the simplest situations, it is usual for there to be a ‘web of causations’ as MacMahon et al. (1960) put it. With multifactorial traits or disorders (such as represented by antisocial behaviour) the situation is further complicated by five other considerations. First, the very term ‘multifactorial’ makes it clear that there will never be a single necessary and sufficient cause, because multiple causal influences are operative, each of which plays a part in contributing to the causal process. Second, it is common for there to be several different causal pathways, each leading to the same end-point (Rutter 1997). Third, each causal starting point is also the end of a prior causal process. The Academy of Medical Sciences (2007) report used the example of smoking and lung cancer. Heavy smoking is likely to involve a genetic susceptibility, the availability of cigarettes and the operation of social pressures. Moreover, the fact that the smoking habit persists will also be influenced by the heavily addictive effects of nicotine. All of this means that the causal background to smoking must be taken into account in assessing the
effects of smoking on the risk for lung cancer. Nevertheless, none of this means that a specific causal effect of smoking on lung cancer cannot be identified. It is obvious that similar considerations, with respect to possible sequential causal pathways, apply to antisocial behaviour. Rothman and Greenland (2002) specified that a cause is an antecedent event, condition or characteristic that was necessary (given that other conditions were fixed) for the occurrence of the outcome being studied at the moment it occurred. Without that causal influence, the outcome would not have developed or would have done so at some later time. That is the central point of what is meant by a ‘cause’. The implication, of course, is that changing a causal factor will actually reduce the populations’ burden either by reducing the overall number of cases or by making the outcomes occur later than they would have done otherwise (Robins and Greenland 1989). It is that consideration that makes the identification of causes so important in terms of implications for prevention or intervention.
