Essay: Waters

John Waters
24 October 2013
Structural Realism: An Argument from Historic Observation

In 1989, the long standing debate between realist and antirealist attitudes towards theoretical objects underwent a distinct transition following the publication of John Worrall’s “Structural Realism: The Best of Both Worlds?” The paper served the simple purpose of reintroducing Henri Poincaré’s concept of structural realism. Briefly, this view amounts to the assertion that mathematical continuities are often found between successful past theories and their modern counterparts such that we have reason to be realists about the relations which these mathematical continuities reflect. The decisive feature of structural realism, however, is that it is generally supposed to occupy the middle ground within the given debate through the appeasement of both of the major realist and antirealist arguments. One should be wary, however, of the fact that the theory utilizes the history of science as the source of its evidence. Other theories have done this before, most notably those of the “convergent epistemic realists” whom Laudan had recently discredited by a more thorough review of the same facts. What is perhaps different about Worrall’s theory, however, is that it draws on history at a less superficial level than past ones. I will proceed in this paper by assessing this idea as part of a larger examination of Worrall’s arguments. I will conclude, not by arbitrating between realism and antirealism, but rather with some observations about the structure of the debate itself. A close look at a different paper by Stathis Psillos will ultimately prove helpful in this regard.

Prior to Worrall, there were two commanding arguments in the realist/antirealist debate. The bastion of the realists at the time was the ‘no miracles argument’ which makes the claim that the predictive success of our best scientific theories can be no accident. Since numerous past theories afforded us a tremendous degree of foresight, it would be unbelievably miraculous (hence the name) if they did not exhibit some degree of approximate truth (Worrall 1989, 263). The opposing argument on the antirealist side, by contrast, is called the argument from ‘pessimistic meta-induction.’ It is perhaps best explicated by the oft quoted passage from Poincaré’s Science and Hypothesis: “The ephemeral nature of scientific theories takes by surprise the man of the world. Their brief period of prosperity ended, he sees them abandoned one after another; he sees ruins piled upon ruin…” (Poincaré 1905, 178). Thus, the observance of the quick succession by which one theory is replaced by another has led many to conclude, in a pessimistic vein, that scientific theories and the entities which they postulate are false.

The beauty of structural realism is that it is designed to account for both of these problems which had up till then strained the debate beyond all possibility of reconciliation (Worrall 1989, 273). As I have said before, the theory also has a decidedly historical focus. While both pessimistic meta-induction and the arguments of many realists also appear to show great deference to historic observation, it is to an almost cursory degree. Structural realism, it can be argued, is the result of a much higher resolution historic image and its capability to supersede and resist existing theories lies therein.

A closer look at an example of Worrall’s will hopefully provide some substance to this last claim. The example comes out of the history of the theory of light propagation which was first developed by Augustin-Jean Fresnel and then later, with a major shift in ontology, by James Clerk Maxwell. Fresnel had initially developed his theory on the assumption that there was an elastic mechanical medium in which light was propagated as a series of disturbances. Not long after, however, Maxwell superseded Fresnel in producing a new theory in which the mechanical medium was dropped in favor of electromagnetic fields which took on the old ontology’s role of propagating disturbances. For Worrall, the relevant feature in the transition from Fresnel’s to Maxwell’s account lies in these disturbance patterns, which in both theories, conform to the same mathematical descriptions. Thus we find clear continuity between the two theories, specifically; in terms of the mathematically describable structure expressed by the respective sets of theoretical objects they postulate (Worrall 1989, 272). Other examples of this phenomenon have been cited by Stathis Psillos and include, most strikingly, Newton’s laws as a local case of special relativity and several laws of the caloric theory of heat which make their reappearance in modern thermodynamics (Psillos 1995, 18). As a result, we have it on evidence that much of a theory may be lost in a massive shift while the mathematics, which is a theory’s primary predictive apparatus, may remains to some non-negligible degree. Hence, the miracle of past success is explained on the assumption that outmoded theories had correctly determined the mathematical structure of the world.

The case that structural realism is the product of a higher resolution historic image as compared with past accounts is inherent in narratives such as Worrall has provided. The general story with past arguments concerning scientific realism, regardless of which side of the debate they come from, is that they have been focused on the content of the theories they assess, particularly their ontological content. Very characteristically Laudan writes, for instance, about how Fresnel’s elastic medium “functioned centrally in explanations of reflection, refraction, interference, double refraction, diffraction and polarization” (Laudan 1981, 27). Laudan’s intention, which is to point out that Fresnel’s account of the ontology involved in light propagation has vanished from the scientific picture, is of course not at all ill motivated. What Laudan misses, however, is that Fresnel was correct in his interpretation of the structure traced by that supposed ontology, and it is precisely this wherein the theory gets its predictive power. Past accounts in the realist/antirealist debate have routinely missed these sorts of details. Their focus has not been on structure and it is in structure above all else where the solution to the problem raised by the no miracles argument can be found.

Putting these points aside I would say that we should be worried, nevertheless, about the sort of historical analysis which has been used as evidence for structural realism. What structural realism depends on is that there is some degree of mathematical continuity between the successful theories of the past and those of the present. The conclusion this allows us to draw is that where we see recurrent mathematical continuity, we are also seeing recurrent and at least partially accurate accounts of the relations found between objects in the world. This state of affairs, however, leaves the theory open to counter examples. The form of the evidence Worrall and other structural realists provide is existential; there are some cases of mathematical continuity, but universal evidence is needed to ultimately prove the claim. What we need to know is that in every case where past theories have succeeded, the mathematics of those theories survives in their present day counterparts excepting some in-principle reason why this should not be the case. All the antirealists need, by contrast, is counter-evidence of an existential form; a handful of examples where no in-principle reason can be discerned for why no mathematic continuity obtained between successful theories. So, until a comprehensive analysis has been performed, the structural realist should feel significant anxiety.

Stathis Psillos; a critic of structural realism for entirely different reasons, has raised some points further exemplifying this asymmetry between realist and antirealist proof structures. Psillos argues, most importantly, that the structural realism seems to entail instrumentalism about a theory’s ontology in so far as ontology seems to only play the role of “useful fiction” in a given scientific theory (Psillos 1995, 24). Psillos is, furthermore, inclined to believe that structural realism must entail certain unstated notions about our epistemic limitations. Specifically, that while there are mathematical relations which reflect the structure of the world, the only reason we have to assume that one set of objects stands in these relations over another is that one is a better fit relationally speaking than another. In essence Psillos is claiming that, traditionally construed, structural realism entails that science can reveal structure and structure alone while the underlying ontology is unknowable for the usual meta-inductive reasons (Psillos 1995, 20).

For Psillos, the problem here lies not with the claim that science is limited to the analysis of structure but rather with a possible flaw in the structural realist’s account of scientific practice. This flaw can be found the assumption that there is a determinate difference between the structure and the nature of a given theory. Worrall’s account plays of this distinction candidly in claiming that the natures of the entities involved in a theory are supposed to be beyond our epistemic abilities while the structure of the processes those entities are involved in is not (Psillos 1995, 24). Psillos wants to make the contradictory claim that in actual scientific practice “the nature and structure of a theory form a continuum” (Psillos 1995, 30). His argument is this; there is nothing to be known about a scientific object over and above a “quantitative description” of the causal abilities and interaction of that object. The distinction between structure and nature, in the end, simply does not stand once we have examined the evidence (Psillos 1995, 31).

At this point, I think it would be a good idea to take a look at some of this evidence to see just how Psillos argument works. One particular example Psillos provides is mass. The old way of understanding mass may be familiar from high school physics class; this is in terms of “how much stuff” there is in a given volume. In time this would change as science came to paint a more distinct, and also more mathematical, picture of mass through its inertial properties. Thus we have the formula; mi = F/a. This was later linked with the theory of how a body is accelerated within the gravitational field of another body. This property is described in the following equation mg = F r²/G.M. which is equivalent to the previous formula e.g. mi = mg. Thus it is claimed that we know far more about mass now than before purely through the structural mathematical properties it exhibits (Psillos 1995, 31). What we are supposed to understand by this example is that “the actual scientific practice urges that improvements in our knowledge of what an entity is involve further knowledge of laws that this entity obeys” (Psillos 1995, 32) It can be seen here also that Psillos is championing traditional realism. We can know scientific entities because structural realism tells us we can know relational properties, and we know scientific entities in turn by these same relational properties.

I think it would be correct to observe, at this point, that the picture Psillos has produced of scientific practice is meant to draw on a level of historic detail which structural realists like Worrall have not picked up on. What is required in order to establish Psillos’ arguments over and above Worrall’s therefore, is more evidence. Psillos’ arguments, however, are also universal in form; where the natures of scientific objects are concerned only structural properties matter. Conclusively establishing this point requires a complete historical analysis of our knowledge of scientific objects. What would have to be looked for in particular is whether or not it has been patently structural where it has been successful and whether there is continuity in the postulation of structural properties even if new sets of entities and processes have come to be described as having those properties. This study would have to come in addition to the previously mentioned study necessary to establish mathematical continuity in the interrelations which scientific entities are supposes to take part in.

Now it should be clear that to establish arguments like Worrall and Psillos’ it will not be possible just to pick and choose examples. Full, unadulterated studies of particular aspects of the history of science will be needed. This is simply the cost of making universal claims. For anti-realists, the bargain is much better. All they need to do is find a sufficiently convincing class of counter-examples. Their arguments, I reiterate, need only have existential form. If, however, the realists are making arguments from a very detailed historical analysis, then antirealist counter arguments will need to resolve those details. This is why, for instance, Laudan’s anti-realist arguments fail against those of Worrall and Psillos. Still, with greater detail I think will come the need for greater volumes of evidence and it will remain far easier to defeat realism than it is to establish it. It might even be said that realism, if true, could not be established conclusively at this time because the discipline of history on which its proofs must be based is still too shaky and the interpretive work simply too monumental.


Laudan, Larry. 1981. “A Confutation of Convergent Realism.” Philosophy of Science 48

Poincaré, Henri. Science and Hypothesis. New York: The Walter Scott Publishing Co., LTD,
1905. The Project Gutenburg PDF ebook. Aug. 21, 2011.

Psillos, Stathis. 1995. “Is Structural Realism the Best of Both Worlds?” Dialectica 49 (1).

Worrall, John. 1989. “Structural Realism: The Best of Both Worlds?” Dialectica 43. (1-2).

Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License