An Extended Covariational Approach to Causal Explanation

Abstract

For over three decades, the covariational approach has been dominant in the study of causal explanation. Following Kelley's (1967) model, theory and research taking this approach often examines explanations invoking the person performing a behaviour (e.g., something about Holly caused her to hit Sam), the stimulus upon which the behaviour is performed (e.g., something about Sam caused Holly to hit him) and the occasion (e.g., something about the night of that barndance caused Holly to hit Sam). In contrast to this narrow range of abstract explanations, people attribute behaviour to a broader range of more specific causes, such as intentions, abilities, opportunities, and personality traits, which feature in widely shared lay theories of behaviour. Because research has examined covariational reasoning with regard to a narrow range of explanations, important questions about how lay people use covariation information remain unanswered. Research does not show whether covariational reasoning is confined to a narrow range of explanations and causal phenomena, or whether it has an important role in explanations that draw on a richer body of knowledge and account for a wider range of phenomena than included in existing covariational models. For example, do people use covariation information when arriving at explanations relating to intentions and abilities? Do people use covariational reasoning to decide whether uncharacteristic behaviours are due to unstable causes (of. a "flash in the pan") or to lasting changes (of. a "new leaf), and if so, how? This thesis proposes extensions to existing covariational models with a view to answering such questions. Chapter Two proposes a model that integrates covariational and "goal-based" models of attribution. The key novel proposition of the integrative model is that people use covariational reasoning to evaluate goal-based explanations (such as motives and abilities) for behaviour. As predicted, explainees preferred goal-based causes that were perceived to covary with actions, when this covariation was manipulated (Studies 1, 2 & 3) and measured (Study 4). Chapter Three proposes extensions to existing covariational models that allow detection and attribution of change. In accordance with the proposed extensions, explainees preferred changed causes when covariation systematically increased (Studies 5-7). Explanatory preferences for changed causes were related to predictions of increases in the frequency of the target behaviour (Study 7). Explainees (Studies 6 & 7) and explainers (Study 8) responded differently to increasing and sporadically varying covariation of the same average strength. Chapter Four proposes when explaining novel, uncharacteristic behaviour, people complement available covariational data with causal schemata information in order to decide the behaviour is due to unstable causes or to lasting change. In support of this contention, explainees preferred change attributions when background information suggested that a relevant change had occurred (Study 9), or when the novel behaviour was extremely inconsistent with a prior disposition (Study 10). In contrast, participants preferred unstable attributions when no relevant change was apparent and when novel events were only moderately inconsistent with prior dispositions. This thesis shows that people can apply covariational reasoning with considerably more complexity and subtlety in both content and process, and to a greater range of behaviours and contexts, than made explicit in existing attribution models.