Spot the difference: Causal contrasts in scientific diagrams

Abstract

An important function of scientific diagrams is to identify causal relationships.This commonly relies on contrasts that highlight the effects of specific differencemakers. However, causal contrast diagrams are not an obvious and easy to recognize category because they appear in many guises. In this paper, four case studies are presented to examine how causal contrast diagrams appear in a wide range of scientific reports, from experimental to observational and even purely theoretical studies. It is shown that causal contrasts can be expressed in starkly different formats, including photographs of complexly visualized macromolecules as well as line graphs, bar graphs, or plots of state spaces. Despite surface differences, however, there is a measure of conceptual unity among such diagrams. In empirical studies they generally serve not only to infer and communicate specific causal claims, but also as evidence for them. The key data of some studies is given nowhere except in the diagrams. Many diagrams show multiple causal contrasts in order to demonstrate both that an effect exists and that the effect is specific -that is, to narrowly circumscribe the phenomenon to be explained. In a large range of scientific reports, causal contrast diagrams reflect the core epistemic claims of the researchers.

Diagrams, Difference-Makers, nd Background Knowledge

“…As I have expounded in Section 2, difference-makers that visually portray the causal or mechanistic difference between two (or more) diagrammatic variables are unable to provide a well-informed explanation. To see this point clearly, let's consider two examples used by Scholl (2016). The first example uses a diagram which does not require much background knowledge in its interpretation, whereas the second example requires specific background knowledge in order to understand the diagram.…”

“…The statistical correlation as a well-informed scientific explanation of the diagram is the product of an epistemic aggregation that conveys the relevant background knowledge about a statistical explanation of the effect of tobacco consumption. Though diagrammatic explanations can be provided by the visual contrasts brought out by the difference-makers, as demonstrated by mechanists in the literature on scientific diagrams (e.g., Abrahamsen and Bechtel 2015;Burnston 2016;Scholl 2016;Sheredos et al 2013), the visual contrast on its own cannot provide a well-informed statistical (and scientific) explanation.…”

“…These variables are always characterized as quantitative data. The explanatory relations between variables are quantitative explanations of the phenomenon of interest (Burnston 2016;Scholl 2016). 1 Besides, the relations between variables are taken to exhibit the causal pattern of the data, from which the phenomenon of interest can be explained by looking at the causal contrast highlighted by the difference-makers of a diagram (Abrahamsen and Bechtel 2015; Scholl 2016).…”