The Locus of the Realism Question for the Semantic View

Research output: Contribution to conferenceAbstractpeer-review


The realism question—the question of how our best theories relate to the world—has traditionally been addressed within the semantic view (SV) through an analysis of the relationship between theory models and data models. This is because the data model has commonly been accepted as the window through which science approaches the world. I argue that the locus of the realism question for the founders of SV was not in the theory of data models, but in the theory of psychological judgments, particularly judgments of similarity. Reviving this position today has the advantage of suggesting an avenue of reconciliation between the formal strand of SV and recent work on modelling which self-consciously perceives itself as offering an informal alternative to SV.

Background: The formal program in SV has turned to increasingly weak mathematical analyses of the relation between theory and data models, e.g. isomorphism (van Fraassen, 1980), homomorphism (Mundy, 1989), and partial homomorphism (Bueno, French, and Ladyman, 2002). The lack of a general theory of this relationship (and the perceived inadequacy of a single mathematical formalism for providing one) has motivated some to reject the formal program of SV as misconceived (Godfrey-Smith, 2006), and has left even those more sympathetic to the formal approach with serious reservations (Frigg, 2006). An alternative program has focused on modelling practice in a more informal way, with the pertinent relation between model and world analysed as one of similarity (Giere, 1988; Weisberg, 2013). This informal approach to model realism often portrays itself as rectifying an error at the foundation of SV.

Revisiting the Original Program: I examine two founders of SV and argue that, properly construed, their programs locate the realism question within the foundations of psychology. Patrick Suppes is canonically recognized as the founder of SV. However, Mary Hesse independently proposed essentially the same conceptual shift as Suppes; juxtaposing Suppes and Hesse reveals interesting commonalities in their research programs. In particular, both figures (i) call for models to serve as a central focus for philosophy of science; and (ii) provide insight into scientific method via comparisons of models which do not detour through the theory-world relation. In the case of Suppes, the project was to mathematically compare models of different theories in order to illustrate synchronic foundational relations between different parts of science; in the case of Hesse, the project was to compare models of successive stages in a field's development in order to illustrate the logic of diachronic reasoning in science.

Suppes (1960) influentially brought the question of data models to the attention of philosophers of science. Furthermore, it is clear that Suppes considers the question of how data models relate to theory models a crucial one for the foundations of statistics. However, Suppes also argues for an increasingly detailed hierarchy of models of experiment. In particular, as more and more aspects of the experimental setup are formalized, eventually a model of the experimenter herself will need to be included. Insofar as the world is approached ever more closely through this progressive formalization of the experiment, it is in the limit of this process, i.e. the foundations of psychology, where the realism debate must be fought. Solamente statistics will never provide a complete analysis of empirical adequacy since its starting point, the data, is always infected by the judgments of the experimenter.

The ultimate moral here is also that of Hesse (1961). She demonstrates that failures of fit between model and world ("negative analogies") are irrelevant for scientific progress. Rather, it is the "neutral analogies," features of the model for which fit with the world is unknown, which drive theory change. While only models that exhibit both positive and neutral analogies are apt for a realism debate, it is not the business of science to generate such models. Crucially, models with negative analogies can play a productive role in scientific reasoning so long as the scientist keeps track of internal relations between positive, negative, and neutral analogies. Thus, the ultimate locus for assessing the theory-world relationship is in the scientist's judgments about the status of different aspects of the model.

A Synthesis: Both Suppes and Hesse locate the theory-world relation relevant for understanding scientific method in the mind of the scientist. If, as in Suppes' original program, philosophical questions are to be analysed in terms of model relations, the relevant models are psychological models of the scientist's ideas (or mental representations) of the theory and of the world, and the pertinent relation between them is her assessment of their similarity. For Suppes, these psychological models complete that model of the experiment that most closely approaches the world; for Hesse, they connect stages in a theory's development, illuminating scientific progress. The relation of fit between theory and world here is assessed in the judgment of the scientist, just as in the post-Giere modelling literature; on this view, however, the original SV program does not rest on a mistaken understanding of models, nor is its formal component misguided. Rather, the formal analysis of the relation between data and theory models continues to be important in the foundations of statistics, but its importance as a response to the realism question rests on the role of data and theory models in the reasoning of scientists themselves. This reasoning may also involve the manipulation and comparison of mental representations that, while they may be studied formally, may not precisely mirror the mathematical structures of theory as presented in textbooks or analysed by statisticians.
Original languageEnglish
Publication statusPublished - 9 Aug 2014
EventNew Thinking about Scientific Realism - Cape Town, South Africa
Duration: 5 Aug 20149 Aug 2014


ConferenceNew Thinking about Scientific Realism
CountrySouth Africa
CityCape Town

Fingerprint Dive into the research topics of 'The Locus of the Realism Question for the Semantic View'. Together they form a unique fingerprint.

Cite this