Between Similarity and Kinship
The content and structure of a taxonomic system are determined by the network of interrelations between its taxa, which are generally referred to as taxonomic relations. This system was defined above as an informational (cognitive) model of some manifestation of taxonomic reality, while taxa are similar models of the respective units of that reality (see Section 3.4); accordingly, taxonomic relations are also informational. To the extent that this system is meaningfully interpreted, the main purpose of these relations is to reflect somehow the real relations that exist in Nature between the units of the TD structure. Thus, the meaningful interpretation of the taxonomic system is largely determined by the meaningful interpretation of taxonomic relations shaping it; the principle of taxonomic unity serves as the basis for this interpretation.
In most research programs in systematics, taxonomic relations have two main substantive interpretations, similarity and kinship', in some approaches they are united by the general concept of affinity [Cain and Harrison 1958]. The latter ascends to its natural-philosophical interpretation applied also to, say, chemical elements, minerals, etc.; in biomorphics and partly in biosystematics, ecological relations are added to them. At an operational level, only similarity relations (similarity + difference) are actually studied as a kind of “primary” ones. Indeed, all that a researcher actually deals with is a sample of specimens with characters describing them. A researcher compares these specimens by these characters and infers similarity relations between them. On this basis, all other relations (kinship, etc.) are deduced as particular interpretations “superimposed” on similarities.
Proceeding from this argumentation scheme, ideologists of the positivist concept of systematics (and empiricism in general) insist that similarity relations should not be interpreted in any speculative way, so the classifications should be limited to this operational basic level. The rationale for this position is the conviction that it is only the similarity that is an observable manifestation of the affinity and that can therefore be considered “objective” [Gilmour 1940; Sokal and Sneath 1963; Colless 1967]; typologists and some evolutionists agree with them in this respect [Simpson 1961; Meyen 1977, 1978; Lyubarsky 1996; Epshtein 2003]. In contrast, other relations that can be taken into account in taxonomic research, including kinship, are not directly observable, so they are “conjectural” and therefore “subjective” to more or less degree. Therefore, classifications based on similarity relations are “objective,” while all others are “subjective.” However, for phylogeneticists, it is the kinship (genealogical) relation that is an immanent attribute of living nature and thus “objective”; the basis for this is an assumption that phylogeny is a real process that produces kinship network between species and monophyletic groups [Naef 1919; Zimmermann 1934; Hennig 1966; Wiley 1981; Pavlinov 2005].
The question of objectivity us. subjectivity of similarity and kinship relations between organisms obviously belongs to the sphere of philosophy. It constitutes the similarity vs. kinship puzzle determined by the ambiguity of interpretations of both these relations themselves and interconnections between them. This puzzle is one of the most fundamental in systematics; it ascends to those times when the above-mentioned natural-philosophical affinity was considered as a fundamentum for elaborating essentialist classifications.
The similarity relation is quite paradoxical in a sense. It actually underlies all cognitive activity, including classifying [Tversky 1977; Quine 1996]; as noted above, it is usually considered by empiricists as “observable” and (literally) “obvious.” However, if considered not from a common-sense standpoint, but as one of the key elements of the conceptual space systematics deals with, it loses any sign of being “obvious.” To paraphrase Aurelius Augustine’s reflection about time, it can be said: as long as I don’t think about what similarity is, everything is clear about it; as soon as I begin thinking about it, my understanding ceases to be clear.
The paradoxicality of similarity begins with the absence of a clearly expressed onto-epistemic status that can be attributed to it. As was just noted, in approaches based on the empirical philosophy of science, it is considered objective and observable. However, from the point of view of conceptualism, this is not the case: “as philosophershave long realized, similarity without theory is empty” [Sober 1984:336];
there are indeed quite a lot of such philosophers and philosophically minded biologists [Goodman 1972; Tversky 1977; Dupre 1993; Murphy 2002; Rieppel and Kearney 2002; Bartlett 2015; Pavlinov 2018; etc.]. In our case, one should speak not so much about a certain PTT, but rather about the conceptual space that configures the cognitive situation in systematics. Within its framework, a system of preferences is formed, which determines the selective pattern of an “observation”: an observer “sees” only what is necessary to develop a sought classification. This means that a researcher does not simply “observe” the similarity itself in all its suchness, but rather compares objects selectively by certain characters within the context of a certain research task. Therefore, this “observing” is actually a pretty sophisticated specific cognitive act, and “similarity” appears as a result of a researcher’s value judgment about commonality of certain characters in the objects being compared [Tversky 1977].
Thus, one may conclude (paradoxically enough) that similarity relation as such does not exist outside and apart from a specific cognitive situation that includes an obligator)' observer [Goodman 1972; Tversky 1977]. This means that the similarity is not a “physical” attribute of aggregates of organisms connecting them by their direct or indirect interactions, therefore it is hardly possible to think of it as an “objective” in a traditional sense. Instead, it is an informational (essentially logical) relation imposed by a researcher according to the following cognitive model. To infer this relation, a researcher first observes and describes the compared objects using certain properties (characters) and thus turns these objects into their descriptive (information) models best suited to the researcher’s task. It is these models, and not the objects themselves, that are actually compared based on the chosen characters, and an intellectual result of this mental (or machinery, or other) comparison is what is routinely called “similarity,” or more correctly, similarity relation. And finally, the compared objects are classified based on this relation into groups to fulfill the general principle of taxonomic unity, i.e., to maximize similarities within each group and to minimize them between groups.
This conclusion is all the truer in cases where quantitative methods are used to assess similarity in numerical taxonomic research. As noted in Section 5.3, there are different quantitative measures of similarity relations providing their different estimates and, accordingly, different classifications. It was emphasized in the same section that there is nothing “objective” in these measures: they are invented by people to solve particular research tasks. So there is nothing especially “objective” in the classifications based on such estimates of similarity: they are just “intersubjective.”
The kinship relations between organisms are generally defined as relations by origin: groups of organisms are kins if they are interconnected by the unity of origin. These groups and the kinship relations between them arise as a result of objective evolutionary processes: ancestors give birth to descendants, and they diverge over time, so that relations with both the ancestral form and among themselves gradually decrease. One can conclude from this that the kinship relations connecting the ancestors and their descendants exist “objectively” in a rather strict sense—that is, outside and apart from an observer. The opposite position consists in denying the reality (objectivity) of kinship: the “monophyletic groups [...] are not material systems, because no processes take place between the parts of a monophylum [...]
Obviously, monophyles are units of our thinking, but they are not coherent objects or systems of nature” [Wagele 2005: 71]. This thesis is certainly true if addressed to the judgments inferred by researchers as phylogenetic hypotheses; it is certainly wrong if addressed to the structural units of the phylogenetic pattern produced by the phylogenetic process. It denies the fact that the information flow between living organisms, including those between ancestors and their descendants, is no less “material” and thus “objective” than energy and matter flow between them [Brooks and Wiley 1986].
Does this conclusion imply that classifications based on kinship relations are more “objective” than those based on similarity relations as such? The answer is not clear cut. The point is that the phylogenetic chains of ancestors and descendants are not directly observable themselves—they are conjectured; moreover, in contrast to similarity, the researcher does not have any observable characteristics of objects that would directly indicate these chains. As a matter of fact, a certain value judgment about kinship is inferred in a certain cognitive situation based on certain empirical data as its indirect evidence. An extra element of subjectivity is added in that the definitions of kinship are different in different versions of phylogenetics. Therefore, kinship relation, though supposedly objective, cannot be “observed” and “measured” as such in order to place organisms along the gradient “less related-more related.”
In this regard, a question arises: what is the basis for an indirect judgment about kinship and for any of its estimates? The answer is quite obvious this time: it is the similarity. This conclusion is deduced from the quasi-axiom of inherited similarity, which is fundamental for phylogenetic systematics: it states that, on a large scale and without going into detail, the greater the kinship of organisms, the greater their similarity. The principle of similarity-kinship correspondence is derived from it; this asserts (by a “reverse” reading of the quasi-axiom): the more similar organisms are, the closer their kinship is. And since the similarity is measurable, the kinship also becomes indirectly measurable.
But, as is usually said, “the devil is in the detail”: the latter have to be dealt with in all those cases where the correspondence principle is applied in practice. These details are as follows.
Firstly, a fundamental difference between kinship and similarity must be taken into account. The similarity logically groups organisms into subsets, while the kinship really links the parts of the whole—the genealogical lineages of the evolving biota [Woodger 1952; Griffiths 1974a; Panova and Schreyder 1975; Mahner and Bunge 1997; Knox 1998; Pavlinov 2005, 2018; Wagele 2005; Lyubarsky 2018]. Thus the similarity is a kind of taxonomic relations, while the kinship is a kind of partonomic relations [Tversky 1989; Pavlinov 2005]. So, from the epistemology point of view, a transition from similarity to kinship is not a trivial task—there is a certain “logical gap” in it.
At an operational level, several specific uncertainties contribute to this nontriviality. One of them is that there are a lot of characters by which organisms can be compared; they are not strongly correlated with each other, therefore comparisons based on different characters provide different similarity relations. Another is that there are many ways to “measure” similarity, with different numerical approaches providing different estimates of similarity for the same set of characters. In turn, there are several interpretations of kinship relations, for each of which certain most fitting similarity estimates are to be adjusted. As a result, the above similarity-kinship correspondence is not strict: roughly speaking, no particular similarities are indicative of kinship. Or, if we take into account that a judgment about similarity is based on a researcher’s judgment about characters in common to the groups compared, it turns out to be as follows: not every commonality in characters indicates a close kinship. This leads to the acknowledgment of the critical importance of character selection in order to obtain the most reliable indirect evidence of kinship relations.
The general argumentation scheme, which establishes a certain correspondence between similarity and kinship relations, can be imagined as follows [Pavlinov 2018].
In a cognitive situation, similarity and kinship actually appear “on an equal footing” in a form of value judgment: they are about the commonality of characters in the case of similarity and about the commonality of origin in the case of kinship. A formal basis for judging kinship by similarity is that they both are relations that link objects in the form of a ternary judgment: this means that, in both cases, two objects are more closely related to each other than each of them to a third object [Nelson and Platnick 1991]. An ontic basis for this judgment is shaped by reference to a certain model of the evolutionary process: it generates the diversity of organisms in a way that endows these two kinds of relations with some analogous properties and provides a certain objective correspondence between them. This correspondence is not strict; therefore, at an epistemic level, there is always more or less uncertainty in the similarity-kinship relationship. The main goal of phylogenetic research is to minimize this uncertainty as much as possible. This aim is achieved by choosing such characters in which similarity makes it possible to judge kinship most reliably. As a result, the whole scheme for the elaboration of the kinship-based classifications looks like this: (a) development of an evolutionary model as the basis of particular definitions of both similarity and kinship and interrelations between them; (b) selection of the characters as the most reliable indirect evidence of kinship; (c) assessment of similarity in these characters; (d) inference of a kinship relations network based on the similarity assessed; and (e) inference of classification from this network.
Thus, the problematic content of the similarity vs. kinship puzzle can be reduced at an operational level to two questions: (a) how the characters should be selected (weighted) so that (b) similarity between organisms in these characters would most reliably reflect the latter’s kinship. Both these questions and their possible answers are formulated by certain PTTs: each of them determines in its specific way how to interpret similarity and kinship, which assessment of similarity is most consistent with a particular kinship definition and, accordingly, w'hich characters are to be selected in order to obtain the required similarity assessment. The most striking example is the different weighting of homologs and analogs in phylogenetics: similarity in the former is considered evidence of kinship (inherited from a common ancestor), whereas similarity in the latter is not (a result of convergent evolution). In cladistics, in addition to this, monophyly in its “narrow” sense is evidenced by synapomorphies but not by symplesiomorphies (see Section 5.7.3). Another weighting approach involves not particular characters but particular sets thereof: these may be the largest set of mutually most compatible characters or a set with the largest possible number of characters that provides for overall similarity as the best measure of kinship (the above-mentioned total evidence).