Presuppositions of Scientific Realism

As explained by Agazzi (§ 5.1.4) scientific realism holds that the unobservable entities postulated by the best instances of theoretical science (like atoms, electro­magnetic waves, viruses, etc.), (a) exist, (b) are independent of any description we give of them, and (c) can be known.

All this presupposes three more claims: that theories are referential (§ 5.3.2), i.e., that they are descriptions, like a story or a map, rather than mere forms of expression, like abstract paintings, or mere instruments, like hammers and nails; that they must be taken literally as such, since their purpose is knowing reality; and that truth is a correspondence between descriptions and their subject matter. In fact, antirealism is often based on the rejection of one of these three further claims (see Alai 2013).

2.1 Independent Reality and Objectification

Both Kant and various recent forms of constructivism have denied that we can refer to reality as independent from thought, and describe it: for them we can at most construct phenomenical entities, which in no way represent reality as it may be “in itself’. With the “linguistic turn” of the XX Century, some philosophers have claimed that it is impossible to reach out of language, or refer to an extra- linguistic reality¹; they held that our expressions don’t have referents, but only senses (or intensions'), and that these are purely determined by our epistemic states (verificationism) or by relations to other expressions (contextualism) (§ 5.3.1).

According to Agazzi these are all forms of epistemological dualism, the belief (typical of modern philosophy) that the immediate object of knowledge is not real­ity, but our own representations, or ideas (§ 4.2), so that “the fundamental question of modern epistemology became that of determining whether or not, starting from our ideas, we can indirectly obtain knowledge of reality” (p.

To this conception Agazzi opposes his own idea of objectification: each sci­entific discipline studies its subject matter exclusively from the point of view of some basic attributes, for whose attribution it employs specific operational or protocollarity criteria. The latter typically consist of operations (including obser­vations and measurements, but more generally various forms of interaction with things). The propositions which attribute or deny a basic attribute are called “pro­tocollar”, as they are immediately judged to be true or false by the application of operational criteria (although Agazzi grants that this immediate judgement of truth or falsity is fallible).

Theoretical attributes are then defined through relations among basic attributes. Theoretical propositions, concerning them, are supported by rational inferences from protocollar propositions (and other theoretical propositions) (§§ 2.5, 2.6). The specific scientific objects of each discipline are just ordinary things seen from the point of view of the attributes of that discipline, hence they are structured bun­dles of attributes. As such, they are abstract objects, which however are exempli­fied by the concrete objects (constituted by a complete totality of attributes) which are their concrete referents (§§ 2.7.4, 5.4.1). For instance,

in physics we define the term “electron” through a structured set of mathematically for­mulated properties which together constitute a certain abstract object. But this does not entail that these are meant to be properties of the abstract object; they are meant to be properties of the single electrons which are the intended referents of the mathematical model we have constructed (p. 113).

Even things, from which scientific objects are “clipped out”, are constructs, i.e. concrete objects seen from the point of view of a limited number of predicates; but they can be identified without problems within common sense, independently ^[53] of any theory: they are the scientific objects of some theory, once the latter has become part of common sense.

Thus, from a cognitive point of view, we never encounter any non-structured material, as we always work from within some subjective point of view: we should not imagine “a reality ‘in itself,’ which should on the one hand have its intrinsic fixed structure, independent and unaffected by language and thought, while on the other be such as to be mirrored by thought and language” (p. 229). Thus, Agazzi might seem to deny that reality has its own nature and properties independently of thought, for he refuses to “conceive of reality (considered at a given moment) as being absolute and structured in itself’ (p. 215). But this is not the case, since for him thought and language just select and “clip out” properties which reality has independently of us.^[54] He explains that the active role of the subject

results in the determination of attributes which are known as they are brought to light and, at the same time, are those actual aspects of reality which are effectively known through a particular intervention. Under different conditions, reality would manifest itself under different aspects or in the form of other attributes, but these too would be real” (p. 229).

The operational nature of basic predicates ensures that, against constructivism, objects are the abstract reconstruction of a concrete and independent reality; they are “clipped out” of things (p. 97, passim); and since things are “clipped out” of the concrete reality, “objects are part of reality (i.e., that part which has been ‘objectified’ through the operations), and are not something ‘behind’ which or ‘under’ which reality remains hidden” (p. 97); besides, “the process of objectifi­cation takes place in a referential situation, and is carried out under strictly refer­ential conditions” (p. 181). Therefore, “(a) science attempts to represent a reality independent of science itself,..(b) what science states is an adequate representa­tion of this reality ‘as it is’” (p.

Moreover, since protocollarity criteria consist of interactions with concrete reality, it follows that acceptance of propositions as true depends on the way in which reality is independently on us. Agazzi also grants that meanings of both basic and theoretical terms are not determined only by protocollar operations, but also by the inferential relations with other terms; hence, their sense is at least par­tially contextual. But terms have also referents, i.e. the concrete objects to which terms are shown to be applicable (or not) through protocollar operations (the basic terms) or protocollar operations plus rational inferences (the theoretical terms) (Agazzi 2014, Chap. 2).

Thus, he largely employs the logical empiricist conception of language and sci­entific methodology, in particular its distinction between empirical and theoreti­cal terms, whose meanings are basically derived from experience and passed on to theoretical terms via inferential relations. But his stress on the interactive nature of experience provides this picture with solid realistic underpinnings.

Agazzi stresses that the operational nature of experience warrants the existence of scientific objects. But it should be noticed that, while an apple as object of mechanics is clipped out of ordinary apples, nothing like this happens for elec­trons, photons, etc., which are not introduced through abstraction, but by a creative postulation. Thus, existence is guaranteed only for the basic objects of science, the observable ones. Otherwise we should grant that even objects like caloric, phlogis­ton or ether exist.

2.2 The Goals of Science

For Agazzi “it is uncontentious (a) that science has a referential intention” (p. 271), and “there is a general agreement” that “the intrinsic goal of science is to offer reliable means for attaining truth” (p. 260). This is perhaps a little too opti­mistic, for radical contextualism and panlinguism deny the possibility of extra- linguistic reference, and instrumentalism denies that theories should be taken liter­ally as descriptions purporting to be true.

A controversial problem is also whether science should only describe, synthe­size and predict, or also explain. This question is linked to the preceding one, since due to the empirical underdetermination of theories, our main route to the knowl­edge of unobservables is the inference to the best explanation (see Lipton 1991; Alai 2014b). So, those who deny either the possibility or the usefulness of explana­tions are usually antirealist. According to Agazzi explanation is a primary aim of science, on a par with description: it is “a component of that process of unification that... operates at the level of perception... knowing is in a broad sense unifying” (pp. 336-337). More precisely, we describe observable phenomena by empirical laws (§ 7.1.1), and we explain them by theoretical hypotheses (§ 7.1.2). Moreover, assuming a statement or a theory as an explanation is assuming it to be true (§ 5.5).

2.3 Referentiality

Science could not pursue truth and knowledge unless it purported to be referen­tial (§ 5.3.2). For Agazzi formal sciences are semantic, since their propositions are definitions, constitutive of the senses of their terms; while empirical sciences are apophantic, since their propositions are understood as descriptions.

2.4 Truth as Correspondence

But referentiality is not enough for realism, a correspondence conception of truth is also required. In fact, after developing a theory of meaning understood as both sense and reference (§ 4.1), Agazzi proposes a correspondence theory of truth, answering the main objections raised by its opponents (§ 4.5). For instance, it is often objected that we lack criteria for ascertaining the correspondence of our descriptions to reality conceived as wholly independent of our epistemic states. To this he replies that the operations by which we assign objects and attributes as referents to our terms are the same by which we ascertain the possession of an attribute by an object; thus, they are criteria of truth for observational propositions (§ 4.5.3). Moreover, we have criteria of truth for theoretical propositions, since they are justified by inferences from observational propositions (§ 4.6).

Many objections concern the problem of truth-makers, i.e., what do true proposi­tions correspond to. If it is said that they correspond to facts, it is objected that since the same state of things can be truly described by infinite propositions with different senses, there should exist an infinite number of facts, one for each true proposition. For Davidson, instead, correspondence would have the opposite, but equally absurd consequence, that there should be just one “Great Fact” (for, given certain logical assumptions, apparently plausible, but correct only within an extensional perspec­tive, there would follow that any true proposition corresponds to any fact).^[55]

To the former objection Agazzi replies that just as an object, constituted by infinite attributes, admits of infinite correct descriptions, each of which describes only some of its attributes, so also a state of things is not multiplied when different propositions bring out the different facts about it (4.5.3). To Davidson’s objection he answers by his own doctrine of objectification: each scientific discipline deals only with certain attributes and with the objects made up of them: so, it deals only with partial aspects of concrete things. Therefore statements and theories are not true of reality in general, but only of the restricted domain of their discipline, or of their own specific models: they have no correspondence to a “Great Fact”, but only to the limited portion of reality to which they refer.

2.5 Truth as Circumscribed and Relative

This is how Agazzi explains his doctrine that truth is circumscribed and relative:

a correct realist approach... limits the truth of the propositions of a given science to its specific and empirically circumscribed domain of objects. It is incorrect to say that, for a realist, “a theory cannot be true unless it can be extended consistently, without correction, to all of nature” (p. 298).

Scientific truth is always a relative truth, in the sense that every scientific sentence is always true (or false) ‘of’ the specific objects which constitute the particular domain of the theory in which the sentence occurs (p. 402. See also p. 408, §§ 4.5.3, 4.5.3.5, 5.5.2, 5.5.3).

This idea is widely shared, in particular by those who reject the “statement con­ception” (according to which theories are sets of universal statements), adopting instead a “structuralist” conception (according to which they are rather constituted by classes of models, which become more and more restricted as they are enriched with theoretical predicates: see Suppe (1977: 221-229); Stegmuller (1976)). Now, this way of speaking is correct if it means that each statement or theory describes only certain aspects of reality; but it conflicts with a more established way of speaking of truth, and it might be mistaken with the following relativistic doctrine: any statement or theory is true, since it is true at least of its own objects or models.

We normally use the relative expression ‘true of.’ only for predicates, while for propositions or statements we use the absolute term ‘true’: we say that ‘black’ is true of ravens, but that the statement ‘all ravens are black’ is true, simply. It would be misleading to say that this statement is true only of ravens, for it actually speaks of all things, saying that each of them either is black or is not a raven.

The same holds even if we describe theories as sets of models (for, as explained by Agazzi (p. 258), sets of models are still translatable into sets of statements): a theory selects a certain class of “empirical” models EM, and one of “theoretical” models TM, and it is true if and only if all the empirical models, once enriched with certain theoretical predicates, become theoretical models (see Stegmuller 1976; Alai 1985). So, in effect, the theory speaks of all objects, claiming that if any of them is an empirical model, it is also a theoretical model.

This is not to say that all (universal) statements and theories speak of the one “Great Fact”, because although the statement ‘all ravens are black’ speaks of all objects, it doesn’t make about them the same claim as, e.g., the statement ‘all swans are white’: it describes a different fact. Therefore Davidson’s objection, based on logical subtleties and on an unwarranted extensionalist presupposition, can be rejected without relativizing truth to subjects.

2.6 Pictorial Truth

Agazzi also says that correspondence is not “pictorial”, in the sense that (1) propo­sitions are not “a kind of reduplication of reality under the form of representation”, because language can only describe particular aspects of reality, it cannot offer a complete representation of it; and (2) there is no “point-to-point correspondence” between the elements of a proposition and those of reality, as held by logical atomism and (in part) by Wittgenstein’s Tractatus (§§ 4.5.5, 5.2.1; p. 258).

I grant that no reduplication of reality is possible: we could not represent Saint Peter’s Basilica in Rome by making a perfect duplicate of it, for even a full size replica could not occupy the same spatiotemporal position. This is to say that any representation is partial and selective. But then also pictorial representations are such: they share some traits of their subject (say, shapes, colours, etc.), but not others (say, dimensions, materials, etc.). So, it is not clear in which sense our representations of reality should not be pictorial. Besides, in Tractatus Wittgenstein has shown precisely that propositions are pictures of reality, in that they reproduce its logical structure. For instance, in ‘Romeo loves Juliet’ the words do not resemble their referents in any way. But since whenever a person A loves a person B there is a true proposition with the word ‘loves’ between the names of A and B, and whenever Romeo is involved in a state of things there is a true proposition including the word ‘Romeo’, etc., between language and reality there exists a structural correspondence by which the former pictures the latter. This is clearly a point-to-point correspondence, even if not all “points” of a state of things correspond to “points” of a proposition, and vice versa.

At any rate, Agazzi stresses that truth is a correspondence relation, in the sense that “we certainly speak about something, that this something consists of sub­stances endowed with attributes, and that these attributes result from the encounter between our way of investigating reality and what reality is” (p. 231). Now, this is right, but it wouldn’t yield a realist conception of knowledge if the “encounter of our investigations with reality” didn’t yield a picture of (some of) the structures which reality has independently of us (see Alai 1994, pp. 94-99).

2.7 The Truth of Theories

Agazzi also argues that theories cannot be true or false. Apparently this claim contradicts a crucial tenet of scientific realism, so it must be carefully examined. There are two reasons why he holds this: first, his propensity for the non-statement view of theories (although he grants that also the statement view is partly correct) (§ 7.2.5). So, he claims that theories are not sets of statements (even if they can be translated into sets of statements), but models, or, metaphorically speaking, maps. As such, they cannot be literally true, but only more or less faithful or accurate. This does not conflict with realism, however, since he also claims that maps, hence also theories, are true in a non-literal sense, in that they contain information which can be translated into propositions, and so become true or false (p. 258).

A similar idea is advanced by Giere (1988, Chap. 4), for whom theories are families of models, hence neither true nor false, but more or less resembling real­ity. The implicit premise is that only statements can be true or false: for instance, questions, wishes, orders, etc., are neither. However, although a map is not a state­ment, if proposed as a map of a particular region it is implicitly asserted, i.e., understood as a description of that region, hence true or false, even if it is not translated into propositions.

Agazzi’s second reason for denying that theories are true or false is they postu­late abstract objects like rigid bodies, perfect gases, adiabatic transformations, etc. Hence “the aim of theories is far from that of telling a ‘literally true story’ concerning the world, but is rather to give the most faithful depiction of a certain (partial) vision of the world under a specific point of view...” (p. 256),^[56] and “of causally explaining empirical laws” (p. 259).

Now again, this is right, and compatible with realism, if it means that we don’t claim that a given concrete marble slate is a rigid body, or that a particular con­crete body is free-falling, etc. But theories don’t say this: rather, they claim that the more a concrete system approximates one of these idealized models, the more its behaviour approximates that described by the theory’s laws. Hence, theories can be literally true. After all, as Agazzi himself makes clear, a scientific object A is abstract, but it has a concrete referent B, which must possess the attributes characterizing A. Besides, since nothing can be considered as an explanation unless it is taken as true (§ 5.5.3), theories could not be explanations of empirical laws (against Agazzi’s own claim) unless they were true.

Agazzi’s conclusion is that “while... the problem of realism has significant links with the question of truth, we do not. need to relate this truth to theories in order to investigate this issue” (p. 259).^[57] But I have argued that this is correct only in some limited sense. In any case, his considerations should not be understood as a rejection of realism about theories, in the way in which for instance Hacking (1983), Cartwright (1983) or van Fraassen (1980) reject it. For, while refraining from attributing truth values to theories, he attributes them to the statements about unobservables which are derived from theories.

3