1. Introduction.

The most recent resurgence of philosophical attention to the so-called ‘functional talk’ in the sciences can be summarized in terms of the following questions: (Q1) what kind of restrictions, and in particular, what kind of evolutionary restrictions as well as to what extent, is involved in functional ascriptions? (Q2) How can we account for the explanatory import of function-ascribing statements?

The plurality of answers has been recently summarized by Perlman's categories of philosophical theories of functions (Perlman Reference Perlman2004, 5–7). Despite its detailed structure, Perlman's taxonomy of functions does not track affinities through which some theories are interconnected in the proposed classification. In fact, the function debate also shows a certain convergence between the two main accounts within the ‘naturalistic-reductionist’ category, that is, the so-called ‘backward-looking’ and ‘forward-looking’ accounts of functions. Thus, contrary to the taxonomy suggested by Perlman, the convergence between these influential philosophical theories has been proclaimed as the correct strategy (Kitcher Reference Kitcher[1993] 2003), or even considered as a detailed solution (see, e.g., Godfrey-Smith Reference Godfrey-Smith, Allen, Bekoff and Lauder[1994] 1998). Both proposals, however, are biased towards the backward-looking group, regardless of how ‘modern’ or ‘recent’ the appealed evolutionary history is. However, can we attain to the ‘convergence ideal’ from an independent theoretical proposal? More specifically, what is an adequate strategy for attaining the right explanatory match between evolutionary considerations, drawing in a determined sense on the backward-looking theories, and dispositional views of the forward-looking group?

Cummins’ ‘function-analytical explanatory strategy’ and his corresponding ‘functional analysis’ related to a specific class of statements in biology and psychology (Cummins Reference Cummins, Allen, Bekoff and Lauder[1975] 1998, Reference Cummins1983) appear to be solid contenders in that respect. This account, in its basic claims, is not biased towards either of the two competing groups of theories. Moreover, Cummins’ functional analysis emphasizes its complete neutrality with regard to evolutionary considerations (Cummins Reference Cummins, Ariew, Cummins and Perlman2002). In this paper, I will argue that the account of functional analysis only rejects an erroneous employment of the Darwinian theory of evolution in addressing questions (Q1)–(Q2). Thereby it frees up space for different kinds of evolutionary considerations. Thus, in Section 3, evolutionary restrictions on functional ascriptions will be introduced, which are needed for the viability of the function-analytical explanatory strategy in sciences such as molecular biology. The new element in the functional analysis will encourage more clearly a different account of functional explanation: in Section 4 of this paper, I will argue in favor of a three-level function-analytical explanatory strategy, which fares better with the debate's main questions (Q1) and (Q2) than Cummins’ two-level account.Footnote ¹ Finally, the proposed account of a neo-functional analysis will also secure the acclaimed convergence ideal and, as I will try to show, base this ideal in the scientific practice of some of the primary areas in molecular biology.

However, in the next section, I will examine first the main elements of Cummins’ functional analysis.

2. Functional Analysis and the Convergence Ideal.

Cummins advances his account by refuting the main assumptions employed by major philosophical theories of functions: the so-called ‘assumptions (A) and (B)’ (Cummins Reference Cummins, Allen, Bekoff and Lauder[1975] 1998). His account states a new assumption, which I shall call Assumption (C).

Assumption (A). The assumption in question deals specifically with the question concerning the explanatory import of function-ascribing statements in science:

1. A. The point of functional characterization in science is to explain the presence of the item (organ, mechanism, process, or whatever) that is functionally characterized. (Cummins Reference Cummins, Allen, Bekoff and Lauder[1975] 1998, 169; emphasis added)

Furthermore, (A) stays at the core of any backward-looking or etiological account of functions.Footnote ² According to such an account, functional ascriptions explain the presence or existence of items in certain domains in science. The following pair of statements, as usually stressed in the debate, illustrates the influence of Assumption (A) on the explanatory strategy of etiological accounts:

1. (1) The function of the heart is to pump blood into peripheral organs.

2. (1*) The function of the heart is to produce sound of a determined frequency.

Now, only statement (1) can be considered as a legitimate function ascription. According to (A), the main reason for that lies in its explanatory import, that is, it explains the presence of the item in question.Footnote ³ Hence we may legitimately claim the functional characterization of statement (1) and discard statement (1*). When engaging with Assumption (B), we will see a more comprehensive story behind the backward-looking group of theories, but as far as their answer to (Q2) is concerned, that is, the question of explanatory import, this is the basic strategy of any, in Cummins’ terms, ‘teleological explanation’ (Cummins Reference Cummins, Ariew, Cummins and Perlman2002). However, what is essentially wrong with the assumption here at issue?

Functional analysis considers the misidentification of the explanandum as a first difficulty of an account based on (A). On that assumption, functional explanations are answers to the question of a type ‘why is an item there?’ In other words, Assumption (A) identifies the presence or existence of functionally characterized item as the correct kind of explanandum for scientific functional explanations. However, on Cummins’ account, the actual scientific practice quite clearly suggests that functional claims have different explanatory targets. In the final part of this section, we shall see in what way that point breaks with the etiological tradition in the function debate by establishing a new kind of explanandum for the corresponding function-analytical explanatory strategy in the psychological and biological sciences.

However, the misidentification of the explanandum by itself would not represent the main difficulty for the accounts based on (A). A more serious difficulty is concerned with a certain ‘teleological’ stance of the aforementioned explanatory strategy. In other words, functional explanations in science mime the type of explanations given to the presence of items of intentional design. Thus, the answer to the question “why is that thing there (pointing to the gnomon of a sundial)?” (Cummins Reference Cummins, Allen, Bekoff and Lauder[1975] 1998, 174) exemplifies a general scheme for a valid functional explanation on the grounds of Assumption (A). Now, in order to avoid the objections reserved for traditional teleology (Cummins Reference Cummins, Ariew, Cummins and Perlman2002), and to account for the explanatory import of function-ascribing statements in science, the backward-looking approach provides a determined naturalization of the appealed sources of design. This brings us to the second assumption.

Assumption (B). On Cummins’ account of functional analysis, the assumption in question can be defined as follows:

1. B. For something to perform its function is for it to have certain effects on a containing system, which effects contribute to the performance of some activity of, or the maintenance of some condition in, that containing system. (Cummins Reference Cummins, Allen, Bekoff and Lauder[1975] 1998, 169; emphasis added)

Assumption (B) deals with function-ascribing statements. In particular, it refers to a group of restrictions required for legitimate functional characterizations in science. (B), thus, finds in the selectionist tenets of the Darwinian theory of evolution the right kind of restrictions. We can justify the claim ‘the function of the heart is to pump blood into peripheral organs’ in virtue of evolutionary selectionist restrictions. On that account, the pumping of blood is taken to be the heart's function, because that is theeffect of the item, which was selected by certain evolutionary pressures in the past.

However, in Cummins’ view, this appeal to evolutionary theory is a “cheap trick” (Cummins Reference Cummins, Ariew, Cummins and Perlman2002, 170). In what follows, I will explicate the motivation for such a negative assessment of the philosophical mainstream view on functions.

The line of attack on Assumption (B), on which Cummins insists most persistently, concerns the fact that “functions just do not track the factors driving selection” (Cummins Reference Cummins, Ariew, Cummins and Perlman2002, 166). In other words, functions are out of reach of natural selection. What is targeted by that type of evolutionary pressure is concerned with the survival value (Tinbergen Reference Tinbergen1963) or advantage value (Wouters Reference Wouters2003) of functionally characterized item, not with the functional role itself. The point is exemplified by saying that “the better and worse wings have the same function, but only the former spreads” (Cummins Reference Cummins, Ariew, Cummins and Perlman2002, 167; emphasis added). Therefore, the evolutionary pressures of natural selection track only the item that functions better in a given range of environmental circumstances than the item-competitors. According to the present objection to (B), the notion of function cannot be reduced to a corresponding selected effect. Thus, the approach of functional analysis loosens the dependence of function-ascribing statements on evolutionary considerations to the extent that they seem to be entirely neutral to restrictions in the latter sense. But, if the etiological basic strategy for reduction comes to a ‘cheap trick’ and, in addition, amounts to a misleading view on the Darwinian theory, on what grounds can we nevertheless legitimately retain functional talk in science? Furthermore, how can we account more adequately for its explanatory import? Cummins’ answers are given by a cluster of claims in Assumption (C).

Assumption (C). The assumption according to which the approach of functional analysis addresses the issue of legitimate functional claims in science is stated as follows:

However, rather than constituting the basic assumption of Cummins’ theory of functions, the quoted formulation represents an implication of the explanatory strategy related to the analytical account A. Functional ascriptions in science are, thus, in a determined sense derivative upon a corresponding explanatory strategy. In other words, we are entitled to claim that ‘the function of the heart is to pump blood into peripheral organs’ on the grounds of the explanatory scheme, established by functional analysis. The explanatory strategy of this analytical account, as the basic claim related to Cummins’ Assumption (C), “proceeds by analyzing a disposition d of a into a number of other dispositions d1, …, dn had by a or components of a such that programmed manifestation of the di results in or amounts to a manifestation of d” (Cummins Reference Cummins, Allen, Bekoff and Lauder[1975] 1998, 187). In the remainder of this section, I will single out the main elements of this explanatory strategy, by which Cummins traces the most significant distinction from the teleological explanatory strategy of etiological or backward-looking accounts.Footnote ⁴

First, functional analysis has a different type of explanandum than teleological explanation. Rather than the presence or existence of a functionally characterized item, the new type of explananda is concerned with “the behavior of a containing system” (Cummins Reference Cummins, Allen, Bekoff and Lauder[1975] 1998, 176). Thus, referring to the aforementioned formulation of the analytical explanatory strategy, a containing system is indicated by the variable ‘a’, and its behavior by the variables ‘d’ or, according to Assumption (C), ‘ψ’, both indicating the corresponding disposition or capacity. The latter group of variables indicates the type of explananda targeted by Cummins’ account of functional explanation.

Second, the account in question also has different elements within its explanans: (Es1) Functions as Causal Contributions. Since a certain higher-order disposition d, or a capacity ψ, requires explanation, the analytical account A decomposes it into a determined group of lower-order dispositions. In that explanatory framework, functions are only those analyzing capacities φi or dispositions di, which causally contribute in a specific, ‘programmed’, way to the manifestation of the analyzed capacity ψ or disposition d of a, that is, of a containing system s.Footnote ⁵ However, this causal reduction gains an explanatory import if and only if the explanans of functional analysis satisfies certain conditions; (Es2) The Applicability Conditions. The conditions in question are determined purely on the grounds of our “explanatory interest” in a related scientific domain (Cummins Reference Cummins, Ariew, Cummins and Perlman[1975] 1998, 191–192). Cummins’ applicability conditions are concerned with the degree and typological difference between the analyzed and analyzing dispositions and, finally, with the relative complexity of organization of the analyzing capacities. Thus, a function-analytical explanation A ‘appropriately and adequately’ accounts for an explanandum, provided the conditions at issue secure a sufficiently wide gap between the analyzed and analyzing capacities. Since there is no all-or-nothing situation in such a case, the applicability of this explanatory strategy, through which we state functional characterizations, is ultimately a pragmatic issue of choosing the right account for a given explanandum.

The production of proteins, as analyzed in molecular biology, may illustrate what Cummins has in mind by his peculiar explanatory strategy, which may be represented as an ‘assembly-line production’ (Cummins Reference Cummins, Allen, Bekoff and Lauder[1975] 1998). However, before we venture any further into the viability of the function-analytical explanatory strategy in molecular biology, we should explore the most frequently highlighted difficulty of Cummins’ theory, namely, its too liberal character (see, e.g., Kitcher Reference Kitcher[1993] 2003; Weber Reference Weber2004). As we shall see in the next section, functional analysis admits into science unacceptable functional characterizations. This difficulty brings the discussion of (Q1) back to the starting-point: on what grounds can we single out legitimate functional claims? Accordingly, what other kind of restrictions must be added to the already existing applicability conditions in order to make fully operative the function-analytical explanation in the corresponding scientific domains?

3. Neo-functional Analysis and Evolutionary Restrictions.

Consider the following two situations, usually deployed against Cummins’ account (Kitcher Reference Kitcher[1993] 2003, 169): (s1) an arrangement of rocks a makes partial causal contribution Ca1 to the widening of a river delta D. (s2) a mutation of the type a makes partial causal contribution Ca1 to the formation of a determined tumor D. Both situations satisfy the applicability conditions required by Cummins’ version of functional analysis. However, we cannot identify the partial causal contributions Ca1 in (s1) and (s2) as functions, that is, legitimate scientific functional ascriptions to the items in question. Briefly, that point constitutes the aforementioned objection of a ‘too liberal character’ of functional analysis.Footnote ⁶

Nevertheless, Cummins’ original paper contains an explicit reply to that kind of objection. On this account, we can rule out the corresponding counterexamples purely on the grounds of the three pragmatic applicability conditions (Cummins Reference Cummins, Allen, Bekoff and Lauder[1975] 1998, 191). However, this proposal does not solve the problem, because the width of the gap, as posited by the applicability conditions, is equivalent in unacceptable and legitimate functional characterizations.Footnote ⁷ Thus, although these situations are not of the same kind, they all locate with sufficient precision the most vulnerable part in the account of functional analysis. A plausible solution to that kind of difficulty might consider the capacity of ‘pumping of blood’ as somehow amounting to a biologically significant capacity. Its significance, however, cannot be accounted for in the way pursued by the selected-effect theories of functions: as already shown, Cummins’ functional analysis argues for a complete independence of evolutionary considerations in that sense (Cummins Reference Cummins, Ariew, Cummins and Perlman2002).

In what follows, I will argue that in order to resolve its main difficulty, the approach of functional analysis should be integrated in a certain way with evolutionary considerations. In that respect, it is important to make clear: (i) what kind of evolutionary considerations may fit into the account in question; (ii) what role that kind of considerations can play in an account of functional ascriptions and of function-analytical explanation, which then can bring us closer to the proclaimed convergence ideal.

The scientific practice of molecular biology, in particular its efforts in discovering the complexity of protein synthesis, suggests an application of the Darwinian evolutionary theory, which seems to steer away from the ‘cheap tricks’ of the selectionist reductionism on functions. Moreover, the application in question also eschews an appeal to evolutionary considerations as a background or contextual knowledge in a general sense (Kitcher Reference Kitcher[1993] 2003, 173). The discovery and explanation of the main causal mechanisms in the overall process of protein synthesis also offer an important insight into how evolutionary theory takes part in the scientific field that appears to fit well into the function-analytical account described above.

4. Functional Ascriptions as an Intermediate Explanatory Level in Molecular Biology.

Rather than acting as a direct explanans of the manifestation of a higher-order capacity ψ in Cummins’ analytical account A, function-ascribing statements are now considered as occupying an intermediate level within an extended explanatory scheme. The level in question mediates between an evolutionary bound explanandum capacity ψ and the basic explanatory level consisting of a determined set of causal mechanisms.Footnote ¹² Now, particular causal descriptions acquire their explanatory import within that version of a three-level function-analytical explanatory strategy. In order to exemplify the point under consideration, we should examine again the production of proteins.

Consider a different expression of (F1):

1. (F2) tRNAs function as the ‘adaptor’ molecules between an mRNA template and side chains of amino acids in a determined biological system.Footnote ¹³

The explanatory import of (F2) can be accounted for by considering a tRNA's function φ as a determined schematic characterization that calls for particular causal-mechanistic explanations, for instance, how the so-called ‘acceptor stem’ of a certain tRNA molecule interacts with the corresponding amino acid side-chain, specific for that tRNA (see, e.g., Stryer Reference Stryer1988, 733–746). In other words, the function φ in question provides a partial explanatory import with regard to Cummins’ higher-order capacity ψ in some biological system s if and only if φ represents the right kind of schematic characterization or a framework-statement (Wouters Reference Wouters, Reydon and Hemerik2004) for the causal-mechanistic explanations. Yet, what prevents us from reducing the level of function-ascribing statements to the basic explanatory level of causal mechanisms (see, e.g., Ruse Reference Ruse, Ariew, Cummins and Perlman2002)? Is there any further role that evolutionary considerations can play in that respect?

At this point Cummins’ notion of ‘programmed manifestation’ should be recalled, which qualifies the causal activity of component-parts in the function-analytical explanation (Cummins Reference Cummins, Allen, Bekoff and Lauder[1975] 1998, 187). Now, although delicate issues arise about this point (see, e.g., Craver Reference Craver2001; Tabery Reference Tabery2004), the notion in question raises again the issue of evolutionary neutrality associated with Cummins’ account. If the selected effect reductionism is not a plausible way to deal with the ‘programmed’ causal activity of functionally characterized item, some other kind of evolutionary considerations, nevertheless, calls for an application (Cummins Reference Cummins, Ariew, Cummins and Perlman2002, 170). Let's finally see how the condition of phylogenetical entrenchment may fit into the acknowledged gap in the original version of functional analysis.

As in the case of (F2), when we functionally characterize a tRNA molecule, according to the function-analytical approach, this claim amounts to a partially programmed causal contribution to the manifestation of a higher-order biological capacity. Since the latter is ultimately identified as explanandum on the grounds of its phylogenetical entrenchment, this kind of evolutionary consideration puts certain constraints on the way a tRNA molecule can make a partial causal contribution to the overall process of poly-peptide production in a determined biological system.Footnote ¹⁴

In order to illustrate more closely the point here at stake, let's examine the case of adaptor hypothesis. The function of a particular tRNA is constrained by the rules of the genetic code and those of a Watson-Crick type of base-pairing. Given a deep phylogenetical entrenchment of these rules, a major evolutionary constraint, conceived in the aforementioned sense, limits to a large extent a range of possible programmed manifestations of a tRNA's partial causal contribution to the transfer of genetic information in the protein production process.Footnote ¹⁵ The examined case, thus, suggests that a specific employment of the Darwinian evolutionary theory in molecular biology plays no other role with regard to the overall strategy of function-analytical explanation than to establish a hierarchy of limiting conditions, which is expected to be binding most strictly for the programmed causal activity of a functionally characterized item.

We may now assess the results obtained by the neo-functional analysis here proposed by answering questions (Q1) and (Q2).

5. Concluding Remarks.

As for (Q1), even functional ascriptions in sciences such as molecular biology are restricted by evolutionary histories. However, contrary to the general reductionist program of the backward-looking accounts, the function of a determined biological item is not singled out on the grounds of some supposed selective scenario from recent or, rather, remote evolutionary past. Following Cummins’ approach, the neo-functional analysis identifies functions as particular causal contributions to the behavior of a containing system. Nevertheless, evolutionary restrictions in the sense of phylogenetical entrenchment introduce limitations into possible programmed causal activity of a functionally characterized item, without thereby reducing the causal contribution in question to a corresponding selected effect.

The proposed emendation of functional analysis through the notion of phylogenetical entrenchment, besides protecting it from the objection of being too liberal in the attribution of functional properties, also accounts more adequately for the function-analytical explanatory strategy in molecular biology. Thus, as far as question (Q2) is concerned, on the account of the neo-functional analysis, a distinctive import of function-ascribing statements within that explanatory strategy does not simply consist in a compressed version of more detailed causal-mechanistic explanations. Their import to the overall function-analytical explanation in molecular biology rather consists in constituting the right kind of framework-statement for the basic explanatory level of causal mechanisms determination. It is exactly on this explanatory strategy that we can base the convergence ideal on biological functions.