The foregoing approaches, which may be categorized as ‘tight focus’, are not the only types of evidence available, and another type, which may be categorized as ‘broad focus’, has to be considered (Kemm, 2000). Reliance on either approach to the exclusion of the other is likely to produce an incomplete and misleading assessment. Evidence from interviews with key informants, focus groups, opinion surveys and other ways of tapping informal knowledge should be used to produce better Health Impact Assessment. Policy analysts may not be able to describe system behaviour with equations, but can base predictions on experience of similar policies in comparable settings. However, teasing out the connections between interventions and their sequelae (note sequelae are not necessarily consequences) is notoriously difficult (Chapman, 1993). Where there are no precedents, predictions may be based on theory derived from observation and experience. Scenario modelling is another approach by which the future consequences of policies may be predicted (Ziglio, 1986). Policy impact assessment, like policy evaluation (Fischer, 1995), has to take place at several levels. Epidemiology is a particularly powerful tool at the level of verification, which uses technical and analytical approaches to determine the probability that the policy will achieve its stated objectives. Other disciplines have more to offer at the higher discourse levels of impact analysis to answer questions such as will the policy objectives be relevant to the problem (situational validation), will the policy have instrumental value for the health of society as a whole (societal vindication) and will the fundamental ideology of the policy be compatible with health (social choice)?