The Role of Subjective Models in Proto-Scientific Measurement

The earliest stages of a measurement practice typically involve the operationalization and objectification of some subjective assessment of the world via an external, publicly accessible measurement device: for instance, the transition from subjective sensations of hot and cold to the use of columns of fluid in thermoscopes for measurement of temperature (Chang, 2004). The refinement of this measurement practice is then guided by parallel developments in theoretical models (van Fraassen, 2008; Tal, 2012). In this project, I consider strategies for ensuring measurement stability when an objective model is unavailable. The examples I consider resist objectification either because they are targeted at inherently subjective quantities (the Scoville scale for spiciness; the Mercalli scale for the intensity of earthquake effects) or because the relationship between observable effects and presumed underlying cause is too convoluted to permit observer-independent operationalization (academic grade assessment; the use of looking-time to measure infant novelty detection). These measurement practices conform to the standard formal account (Krantz, et al., 1971) and appear to exhibit improvements in rigor and stability over time. However, I argue that the standard analysis of such improvements in terms of parallel theoretical developments is unavailable here due to the constitutive role of subjective mental models in these practices. Traditional analyses of the role of mental models in scientific reasoning (e.g. Hesse, 1962) take them to be externalizable for community assessment via (e.g. mathematical) public descriptions. However, the models that guide these measurements cannot be communicated explicitly—this is the sense in which they are subjective. Nevertheless, training and intersubjective comparisons may serve a regulatory role, allowing for improvements in both precision and accuracy of measurements that approximate those which can be achieved through interplay with objective theoretical models. This analysis sheds light on more general questions about the theory-ladenness of data accumulation and scientific progress.