In a lecture on classification for an international class of master’s students (MA) students in information studies, the present author asked the class to perform a brainstorm suggesting criteria on how to classify the participants in the class. I wrote the students’ different suggestions on the blackboard. Among the many ideas were:

After some time, new suggestions slowed down, and some uncertainty spread among the students. Then one of them said: But that depends on the purpose of doing the classification. Yes, I answered, it seems hopeless to continue this task without a given purpose, as there are infinitely many ways to classify any set of items. People can be classified in one way if the purpose is to dress them, in another way if it is about ordering meals for them, in a third way if it is about medical precautions, in a fourth way if it is to match students for group work in this class, or if the purpose is to adapt the teaching to the students’ prerequisites, etc. While an infinite number of classifications are possible, the question is not primarily which of these are objective, but which are relevant⁴. This point is my most important one in this course.

This example is intended as a challenge for defenders of all-purpose classifications. How would such a classification look, given this example?

As regards classification, pragmatist philosophers generally emphasize practical consequences and usefulness over rigid, pre-defined categories. They also suggest that classifications should be evaluated based on their ability to help us achieve our goals and solve problems, rather than on inherent or essential properties of the objects being classified. The pragmatist philosopher Dewey (1948, pp. 151–154) wrote:

While this quote clearly expresses the essential tenets of pragmatist philosophy, it ignores a problem formerly raised by, among others, John Suart Mill in 1872, to which we now turn.

For centuries, there has been a distinction between “artificial” and “natural” classification, which we shall consider further in Section 3.6. Here we shall briefly consider Mill’s (1872, p. 498) distinction between on the one hand a natural or proper scientific and philosophical classification, and, on the other hand, an artificial or technical classification. The concept of natural classification (and natural kind) has been intensively researched in recent philosophy, and so we shall return to it in Section 3.6. Here we just present the concept of artificial classification, which Parry and Hacker (1991, p. 133) explained in accordance with Mill’s understanding:

Although Mill’s intention was to show the general usefulness of natural classifications, this example clearly demonstrates that to perform certain practical tasks, such as mailing objects, some classifications are needed, although they may be of little use for most other purposes. It is hard to imagine an all-purpose classification, say of rocks, that would be optimal for both mailing them and for exploring their chemical properties and industrial potentials. Again, this example is intended as a challenge for those who want to argue about the possibility of constructing an all-purpose classification.

Now we shall turn to classification in contemporary science. Transdisciplinary researcher and philosopher David Ludwig and professor of philosophy of science Stéphanie Ruphy wrote the following in the article “scientific pluralism” in Stanford Encyclopedia of Philosophy (2024, Section 4.1):

This quote expresses a pluralist point of view regarding biological classification: that there is a need for more than one classification of organisms. Pluralism versus monism will be further discussed in Section 3.2. We shall also see that the periodic table of the chemical elements seems to be the best candidate for a natural classification.

The last example is very much like the one provided by Dewey (1948) but also addresses the question of scientific versus non-scientific classifications. Philosophers of science Vincent Cuypers and Thomas Reydon (2023, p. 5) wrote:

This example underlines the different needs of, for example, biologists and carpenters, but rather than seeing the difference in these classifications as scientific vs. non-scientific, it is here suggested to characterize them as natural versus technical/artificial classifications, as presented in Sections 3.3 and 3.6, and thus downplay the difference scientific and non-scientific classifications: a general theory of classification should cover both.

These five examples support the view that an all-purpose classification is impossible, although there is no consensus regarding pluralism in scientific classification. Before considering Gnoli’s arguments, some important concepts need to be presented, although these concepts need much more attention in KO than can be given in the present article.

The most important distinction in Gnoli’s paper is between “objectivism” versus “multi-perspectivism”⁵. However, the word “objectivism” is in the philosophical literature often reserved to be about ethics and especially a philosophical system developed by Ayn Rand (cf., Badhwar and Long, 2025)⁶. It is also used in a way closer to the one used by Gnoli: about the view that various kinds of judgements are objective, i.e., pertain to objects, as opposed to subjectivism, that such judgements are subjective, i.e., pertain to the people who express these judgments⁷. Bryant (2000, pp. 20–28), and others use objectivism in a sense that seems identical to the way it is used by Gnoli (2025c), but her concept of objectivism seems to mix a set of other concepts, such as objectivity, realism, and monism, which are better considered separately⁸.

The dichotomy that Gnoli emphases is therefore better labeled monism versus pluralism, and the term objectivism can be seen as a near-synonym for objectivity. If we return to example 2.1, the core question is not whether people can be classified objectively according to sex/gender, hair color, nationality, etc. (of course, this may also be a problem, consider, for example, the critique of the binary classification of sex/gender). However, the focus here is whether there is one correct way of classifying people (monism), or whether there are more ways (pluralism), which each are best to serve a specific purpose.

The term “objectivity” stands in contrast to subjectivity, where the last term means that a judgement is influenced by the opinions, beliefs, and feelings of conviction of an individual or a group of individuals (intersubjectivity). However, these two concepts are not necessarily opposites, as pointed out by Matheron (1989, pp. 26–27):

Reiss and Sprenger (2020, Section 8) found that “it should come as no surprise that finding a positive characterization of what makes science objective is hard”. Scientific methodology often contains specific procedures on how to avoid subjectivity in research, e.g., by prescribing double-blind experiments in which neither the researchers nor the test subjects know which subjects receive the actual drug to be tested and which subjects receive an inactive drug (a placebo). While such methodologies may limit certain forms of bias and subjectivity, they do not guarantee objectivity because data do not interpret themselves. Theories and models shape how scientists understand and explain results, which introduces conceptual perspectives that are not purely “objective”. In the wake of Kuhn (1962), it is broadly accepted that scientific knowledge represents the theoretical view, tradition and paradigm of its originator. People may share the same subjective view, but this does not make the view objective, but public (as public knowledge should not be confused with objective knowledge).

A common belief is that positivism is an epistemology that seeks objective truths, whereas “soft” epistemologies like hermeneutics and feminist epistemology are subjective and have no ambition of being objective. However, objectivity should be an ideal for all epistemologies. Best (2000, p. 271) wrote in his review of Harding (1998):

Best concludes (p. 275): “As Harding ably shows, the politicization and pluralization of knowledge is not necessarily a threat to (strong) objectivity, but one of its preconditions”.

This quotation is important. It says that what are often regarded as soft, subjective methods may be a precondition for “strong objectivity”. In the same way historian Haskell (1998) argues that “objectivity is not neutrality”, meaning that while an important ideal in science is to strive for objectivity, it is not necessarily a problem that researchers are engaged or have interests in the object of their enquiry. (Of course, researchers must not engage in nepotism, have conflicts of interest, or let their interests lead them to ignore contradictory arguments.) For example, feminist scholars have provided important corrections and supplements to our historical knowledge about women.

We may conclude that objectivity may be considered an ideal for inquiry, and that researchers may often believe their findings are objective. However, after Kuhn any such claim of objectivity may turn out just to be an (inter)subjective belief, and further, as described by Axtell (2016, p. 90), norms of science, including the norm of objectivity, have a history themselves (Daston and Galison, 2007). Broad and deep knowledge is the best guard against premature belief in the objectivity of a given claim, but whereas it is often possible to trace subjective influences, it is impossible to fully guarantee the objectivity of a claim.

There exist many alternative classifications of everything (e.g., birds)¹⁰, but as Cuypers and Reydon (2023, p. 38) write, it has been a long-standing scientific goal to produce a single all-purpose taxonomy of scientific phenomena¹¹. Today, however, many philosophers and scientists have advocated forms of classificatory pluralism¹². This entails that multiple classifications should be allowed to coexist, and that whichever classification is best, depends on context, interests and theoretical outlook. Cuypers and Reydon (2023, p. 4) wrote:

Pluralism is often associated with the philosophical concept “contingent” (i.e., subject to chance as opposed to being necessary), for example in the sense that scientific representations and models deliver partial, interest-dependent, hence contingent representations of the world. As Ludwig and Ruphy (2024, Section 2.2) wrote, pluralism is contrasted to monism¹³:

The view that one kind of map of a given country may be the best to serve all purposes therefore represents the monist point of view, whereas the view that that, for example, road maps, topographic maps, tourist maps and geological maps each are optimal for a given purpose represents the pluralist point of view.

It is important to consider that pluralism may not be equally plausible in different domains. Stamos (2004, pp. 137–138) in his book review of Bryant (2000) wrote:

Classificatory pluralism stands in contrast to classificatory monism, which holds that there is (or should be) one correct system of classification that reflects the true structure of the world—often associated with natural kind realism and essentialism.

The basic idea of realism is that a mind-independent reality exists, for example, that the Earth existed before humankind and that its existence does not depend on man’s thinking and knowledge. The opposite view is termed idealism (all that exists are ideas in our minds). The idea that something exists independent of our mind is called metaphysical realism¹⁵. Metaphysical realism is different from claiming that a given text or theory reflects reality, or that a given scientific methodology provides a representation that mirrors the world, or the view that science approaches an objective representation of the world. The view that the world is as we see it, is called naïve realism, which is contradicted by knowledge of optical illusions and by the fact that scientists often see and describe phenomena differently from non-experts¹⁶.

After Kuhn (1962), scientific knowledge is understood as formed by paradigms and scientists working in different paradigms see the world differently and describe it differently. Kuhn was therefore not a realist in the meaning that science provides an objective picture of the world. He found, however that the world provides “resistance” to our conceptualizations in the form of anomalies, i.e., situations in which it becomes clear that something is wrong with the structures given to the world by our concepts. This view represents metaphysical realism.

Different positions in the philosophy of science—such as positivism, pragmatism, critical realism, and hermeneutics—offer distinct views on how research should be conducted and how knowledge relates to reality¹⁷. (It is a common mistake to consider empiricism as a realist philosophy, but these are distinct philosophies, see (Dilworth, 2007), chapter 1 and 9). While most researchers aim to produce knowledge about the world, they differ in their conceptions of “reality” and in what they regard as valid, objective, or realistic knowledge. This is the case even for those, who call themselves “anti-realists”¹⁸. Thus, the key issue is not whether they accept or reject metaphysical realism, but how they interpret reality considering epistemic values such as neutrality, detachment, replicability, historicity, reflexivity, and standpoint.

In relation to classifications, the realism-idealism distinction is different from the monism-pluralism distinction. The first is about what classifications reflect reality, the second is about what classifications are relevant (see again example 2.1). However, classifying reality is sometimes understood in opposition to classifying concepts (e.g., Oderberg , 2013)¹⁹, a distinction that has been criticized by Hjørland (2021, + in press), because concepts represent reality in scientific taxonomy.

The most common understanding of essentialism is that some properties of an object are necessary for the object to be the kind of thing that it is, whereas other properties are dispensable. In addition, essentialism implies that a set of necessary and sufficient properties defines an object (see, e.g., Hull, 2007, p. 47). Aristotle made a distinction between the essential and accidental properties of a thing. He defined “man” as “rational animal”, implying that possessing an intellect and sharing the characteristics of animals is central to humanity, whereas hair color is not, it is an accidental property (see further Charles, 2000). However, essentialism is a philosophical concept with a long and complex history, which has important implications for classification and knowledge organization. An important source for this history is the Routledge Handbook of Essence in Philosophy (Koslicki and Raven, 2024).

Gnoli (2025a, pp. 78–79) wrote:

Mathematics is a formal science, but in all empirical sciences the problem is whether nature ever shows clear borders. Gnoli mentions the number of protons (= atomic number) to be the clear border distinguishing chemical elements. Although this is probably the best example of natural kinds and essential properties, even this example is not unproblematic. Clearly, if nature sometimes shows clear borders which everybody agrees on, these will define natural kinds, and the concept would not be controversial. The problem is that different theories disagree on how to define a stuff like gold, and whether one theory is superior for all purposes²⁰. Until the early 20th century atomic weight, not atomic number, were considered the defining characteristics of the elements, but Wray (2018, pp.118–124) argues that the discovery of atomic numbers, provides an example of a Kuhnian revolution in science. It was found at that time that while isotopes of the same element have different physical properties, their chemical properties were the same, thus defining chemical elements. Later it turned out, however, that chemical properties may also differ, although usually only slightly (see Needham, 2008)²¹, and this also questioned the traditional distinction between physical and chemical properties. The issue therefore is still not settled, cf. van Brakel (2012) and Vandewall (2007), the latter of whom wrote (pp. 911–912):

Essentialism is a disputed concept today. In relation to scientific classification, there has been an influential “essentialist story”, claiming that essentialism as a philosophy has produced “two thousand years of stasis” in taxonomy (cf., Hull, 1965a; Hull, 1965b). This story claimed that all classifications before Darwin, including Aristotle’s and Linnaeus’s, were essentialists, but Darwin and followers overcame this philosophy. This historical view is now generally considered a misleading myth (cf., Richards, 2016, pp. 36–38; Winsor, 2006; Hull, 2007).

Essentialism has been considered one approach to classification conflicting with other approaches, but the nature of essentialism itself is interpreted differently in, for example, the analytic tradition, the phenomenological tradition and the pragmatic tradition in philosophy. Because Gnoli’s article argue against the pragmatic view of the domain-analytic view, it should be said that according to Howat (2024, p. 53) “[i]t is a popular assumption that pragmatists are anti-essentialists;” but Howat rejected this view²².

It seems reasonable to be critical of universalist versions of essentialism — that some properties are essential in all contexts – and rather to consider essential notions necessary for a given theoretical framework. Given a (scientific) theory, some core concepts are essential (e.g. “the unconscious” in psychoanalysis, but not in behaviorism and cognitivism), and as theories change or develop towards more mature theories, so do the essential concepts.

Whereas traditional philosophies of science, such as empiricism and logical positivism, considered observations and empirical data the basis of research, on which to develop and evaluate theories, contemporary theory tends to consider observations and theories as intertwined. Boyd and Bogen (2025, Section 5) wrote:

Further:

The theory-ladenness of classifications means that the criteria of classification are derived from a theory (e.g., cladistic classifications are derived from Darwin’s theory of evolution). The opposite position is that objects are classified in theory-neutral ways. For example, by their overall similarity, in which the description of the items is supposed to be objective, and the classification of items is made by neutral, statistical methods. In biology the best candidate for such non-theoretical classification is known as “phenetics” or “numeric taxonomy”. Although all classifications are theory-laden (whether they acknowledge it or not) they recognize their theory-ladenness to different degrees: some foreground their dependence on the theories on which they are based (and in this sense valorize the “subjective” nature of the classification), while others deny, downplay, or even suppress their dependence on the theories on which they are based (and so typically valorize the “objective” nature of the classification).

The perspective of theory-ladenness influences many issues in this article, for example, objectivity and bias. However, here we will concentrate on the dichotomy between phenomenon-based versus discipline-based classification that has often been addressed in KO (e.g., Gnoli et al., 2024), and which is central to Gnoli’s (2025c) views on classification. While both phenomenon-based and discipline-based classification may be theory-dependent (e.g., Gnoli’s (2025c) phenomenon-based view depends on the theory of integrative levels). It will be argued below that the concept of theory-based classification has thus far not been recognized within the professional discourse on knowledge organization. First, however, some general issues are considered.

It is often assumed and stated by supporters of phenomenon-based classification that users, in particular interdisciplinary users, will benefit more from a phenomenon-based classification compared to a discipline-based classification. But how supported is this assumption? Mills and Broughton (1977, p. 37), defending discipline-based principles, wrote that a phenomenon-based classification

Mills and Broughton further noted “that the factual literature for children has always shown a strong tendency to concentrate on phenomena rather than disciplines — e.g., ‘the big book of trains’ which considers most aspects of the railway system”. This example may reveal a certain naivety in the belief that phenomenon-based classifications will best serve professional users: These may be better suited for children or for people who want a superficial overview of all aspects of a phenomenon.

There is a more serious problem with phenomenon-based classifications, which has so far not been considered in the literature of KO: To identify the phenomena, one must consider discipline-based theories. For example, “The Basic Concepts Classification” (BCC) is a phenomenon-based classification developed by Szostak (2013) and further described in Szostak (2020). Szostak states: “We have in 2021 changed our F [Flora and Fauna] schedule to reflect advances in cladistics”. By implication, BCC considers many well-known classes of animals, e.g., fishes, as “Hypothesized Species”. The point here is that it is only possible to classify animals from one or another perspective, e.g., by “folk classification” (admitting fishes) or by a cladistic theory from contemporary biological systematics (not admitting fishes). Most works on integrative levels in KO, including Gnoli (2025c), fail to consider these more detailed levels (such as classification of animals), and fail to consider how the concepts in the classification relate to scientific theories in their respective domains.

Therefore, there is a third possibility beyond discipline-based and phenomena based classification: theory-based classification, and this third position seems to be the most important one, but it is totally unrecognized in KO. Although the theory-based classification tends to classify phenomena, it denies that these can just be identified and classified in absence of theoretical contexts, just as contemporary philosophy of science considers empirical data as theory-laden. This view leans towards the discipline-based approach because, as Bensaude-Vincent (2008) exemplifies in chemistry, (here cited from Labarca and Lombardi, 2009, p. 184) “the ontology of chemistry arises not only from cognition, but also from action driven by caution, utility and efficiency; therefore, the world of chemistry is the result of the very nature of the discipline—science and technology”.

Gnoli (2025c, p. 8) wrote:

Gnoli and the other signatories of the León Manifesto hereby consider phenomena and the theories that study phenomena as independent existing entities that can be classified separately. However, because phenomena and theories are intertwined, this is a problematic assumption. As Albert Einstein said: “It is the theory which decides what we can observe” (cited from, Heisenberg, 1989, p. 40), indicating that phenomena and the theories studying them cannot be separated the way suggested by the León Manifesto. While many phenomena, such as planets and birds, seems to be robust, they are still theory-dependent (for example Pluto is no longer classified as a planet, the concept “bird” has changed according to cladistic research, and the concept “the common blackbird (Turdus merula)” changed rather dramatically according to recent DNA-based studies (cf., Fjeldså, 2013, p. 141)²³.

The concept of natural kind has a long tradition in philosophy and in science²⁴. Natural kinds and natural classification are here understood to be interdependent: belief in natural kinds often underpins natural classifications, and vice versa, and for classification this connection is important²⁵. The dichotomy between natural and artificial classification was most influentially introduced by Carl Linnaeus, although the distinction has deeper historical roots. Linnaeus himself did realize that his own system was not natural, although he saw it as an ideal to be so²⁶. Mill (1872, p. 498) also suggested that the Linnaean arrangement was not natural, and therefore not an adequate scientific classification:

Before we consider Mill’s definition of a natural classification it should be mentioned that Adanson (1763) opposed Linnaeus’s classification based on a few characteristics (in flowering plants, stamens (male reproductive parts) and pistils (female reproductive parts)). Instead, Adanson classified plants based on many anatomical features, an approach that he and his followers considered “natural classification”. Adanson’s approach is related to “phenetics” or “numerical taxonomy” developed in the 20th century (Sneath and Sokal, 1973). However, this phenetic approach to classification is also criticized for not being “natural” (see, e.g., Ereshefsky and Reydon, 2023, p. 247).

Mill (Ibid.) suggested the following definition of natural classification:

In the words of Parry and Hacker (1991, p. 139).

Mill’s criterion for a scientific classification is therefore not pragmatic in a narrow sense (e.g., identification of entities or application for medical purposes), but in a broader sense for serving scientific advances. In biology, the theoretical background for natural classification understood in Mill’s sense today is Darwin’s evolutionary theory, which is the basis for the cladistic approach to classification²⁸. This indicates that a natural classification is supported by domain-specific theory. Another example is that in chemistry, the periodic system²⁹ is considered the best representative of a natural classification which is based on chemical and physical theory³⁰. As Magnus (2012, p. 8) wrote:

Natural kinds are often considered to be caused by microstructuralism, which is the theory that membership of a natural kind is conferred by microstructural properties (e.g. Gold being a natural kind because its atoms have 97 protons; Water being a natural kind having the formula H₂O; animal species being defined by their DNA structure). Hendry (2006, p. 865) wrote:

Magnus’ and Hendry’s quotes thus limit natural kinds to specific domains of knowledge, and Hendry further indicates that pluralism has a stronger case in biology, but that even in chemistry, classification is still interest based, and that not even here are all cases of membership of kinds conferred by microstructural properties³².

We finish this section by considering whether the periodic system is a natural classification. We have already encountered some problems such as Needham’s (2008) demonstration that chemical properties among isotopes of the same element may vary, and that it turns out to be difficult to clearly distinguish essential and accidental properties³³. Here we shall finish presenting two different views on this issue. Scerri (2009, p. 3) wrote:

Later Scerri, interviewed in Bradley (2011a), claimed that the periodic table [system] is not artificial, and that although the optimal³⁴ form is still to be found, this is an attainable goal³⁵. However, Stewart (2011, comment in Bradley, 2011b) argued that we should “Leave the definitive table for Plato’s Heaven!”, showing that the debate about whether the periodic table is natural or artificial still seems to be open. There is broad consensus, however, that the periodic table is the best candidate for a natural classification, which, however, should not be confused with the view that it is the best classification of chemical elements for all purposes.

Foskett (1961, p. 139) expressed the idea of integrative levels as follows: “The theory of integrative levels is that the world of things evolves from the simple towards the complex by an accumulation of properties, and that, at a succession of levels, these aggregations reach new degrees of complexity and become new wholes, with individual and unique identities”. Kleineberg (2017, pp. 349–350) continued:

The theory of integrative levels has a history in KO, especially in the British Classification Research Group (CRG) which was active in the 1960s, but continues to be influential, also in the works of Gnoli, who (2025c, p. 9) wrote:

This theory seems primarily to have been used as an organizing principle for bibliographical classifications at the overall level (main classes), such as, for example, the main structure of the Bliss Bibliographical Classification, BC2, which partly reflects the theory of integrative levels (cf. Gnoli, 2005). It seems seldom or never explicitly applied to more specific levels, such as classification of animals or other kinds of phenomena. The reason for this may be problems with applying this theory. As Kleineberg (2017, p. 253) wrote:

One example of difficulties in the theory of integrative levels can be seen in Gnoli’s quote above, where animals are at a higher level than plants, which probably reflects an obsolete view because evolution is not considered a ladder or a linear progression toward greater complexity or perfection. Evolution rather produces adaptations to particular environments. I will not exclude that certain developments in organisms can be considered to provide a more advanced level of properties, but it is difficult to see what integrative levels theory may contribute for classifying living organisms, compared to biological systematics based on genealogical classification³⁷.

The general conclusion is that the organic level probably represents a higher level of complexity, qualitative integration and emergence compared to the inorganic level. Life introduces properties (metabolism, self-replication, evolution) not found in inorganic matter. However, as Kleineberg wrote, there seems to be no consensus on the idea of integrative levels, neither on the conceptual definition and theoretical foundation nor on the sequence and architecture of level models. It is also important to examine its relation to classifications constructed by the sciences (e.g., the periodic system and biological systematics). As already mentioned, its contribution so far seems limited to the overall structure of general classifications.

As already said, Gnoli tries to provide an alternative to domain analysis which, as he wrote, is based on pragmatist views, according to which any classification reflects particular perspectives and purposes. In the terms of the present article, domain analysis represents pluralism in classification in contrast to monism. Gnoli (2025c, p. 3) cited the basic characteristics of domain analysis from Hjørland ([2017] 2019)³⁸:

• Go to a given domain,

Gnoli found, however, that although the domain analytic literature has provided some fruitful examples of the first three steps, he finds it lacks an example of a motivated classification as recommended in the last step. This is correct if one expects a comprehensive classification of a domain from the knowledge organization community, but domain specific classifications based on explicit epistemological views are probably common in the broader literature³⁹. Here I want to argue that the lack of a motivated classification should not be given too much weight.

First it should be stressed that domain analysis implies involvement with dynamic developments in sciences and other knowledge fields. No other approach to knowledge organization provides theoretical and methodological suggestions for how to tackle this crucial issue. The Bliss Bibliographical Classification, 2nd edition (BC2), has an impressive methodological description for how to classify the different disciplines covered in each volume and it is to my knowledge the best motivated general classification constructed in library and information science (LIS). However, this methodology is somewhat deceptive as it omits the issue of selecting the sources used to identify the criteria for classification. For example, BC2’s classifying of animals is inadequate and unmotivated⁴⁰, and it seems strange that a bibliographic classification about animals avoids considering the biological taxonomic literature. Similar problems apply to all scientific fields.

Next, we shall consider Gnoli’s analysis of Ørom (2003) about classification in art studies as a model for domain analysis. Gnoli (2025c, p. 4) writes:

I both agree and disagree with Gnoli that Ørom (2003) did not contribute about the construction of actual classification systems. Ørom’s interests were primarily about “the materialistic paradigm” and “the “new” art history”, and he shows how these views challenge the traditional classifications. However, he finishes the article by describing the Art & Architecture Thesaurus, which weakens the focus on the relation between paradigms and classification schemes (although he states, “there seems to be a kind of an ‘additive’ structural thinking in the thesaurus”, which poses problems about the adequate representation of certain kinds of documents). Ørom could have used his own insight, that concepts such as “Early Renaissance” and “High Renaissance” do not have the same meanings in different paradigms of art studies. By implication, it is necessary to specify such concepts according to the different paradigms (e.g., by using parenthetical qualifiers), but to do this presumes classification of art studies from the different paradigms. Therefore, Ørom’s basic work of specifying paradigms in art history and uncovering their importance for classification is absolutely necessary and represents a ground pillar in domain analysis.

Ørom’s contribution to developing a new classification for art studies is that he pointed out that “new art history” is badly served by traditional library classifications, so an important goal must be to construct knowledge organization systems (KOSs) serving this perspective (classificationists should consider themselves as contributing to its development). The main difficulty here seems to be that new art history is not a coherent theoretical perspective, which may mean that before constructing a KOS, further studies and choices about this paradigm must be conducted. However, Ørom’s article provides the basic analytic tools for producing such a KOS and thereby takes an important step on the road to overcoming outdated classifications in the domain of art studies.

There are some problems in Gnoli’s (2025c) discussion of Ørom’s article. He states (p. 16):

This quote reveals a discrepancy between Gnoli’s view and domain analysis (and perhaps a lack of understanding of Ørom’s article). For Gnoli, Ørom’s paper is fine and relevant and can translated to different priorities of different characteristics of division. The problem is here the theory-laden nature of observations, concepts and classifications (cf., Section 3.5). Classification in relation to different paradigms is not just about classifying the same concepts according to different principles of division, although this is also important. The main thing is that the concepts vary with paradigms.

Gnoli (2025c, p. 16) further states:

Gnoli claims that the principle of unique definition (to be introduced and discussed in Section 4.2) may provide the basis for a monist classification for art studies, which, in his opinion, will provide a better basis for classification in this domain. My answer to this suggestion is:

(1) What constitutes a work of art is not a question that the principle of unique definition by Gnoli even attempts to answer, and it cannot be answered without considering the knowledge accumulated over time in theories of art such as those presented by Ørom. Ørom demonstrates that this is a question that varies in different paradigms of art. If Gnoli suggests a definition or view of what constitutes a work of art, that definition would reveal its underlying paradigm or view of art (show me your classification, and I shall tell you which theory you subscribe to).

(2) Gnoli’s claim that the principle of unique definition may reveal “at which levels can it [the concept of art] occur, and in what combinations with phenomena at other levels” is not central for the classification of art itself, but only for classifying art in a general classification (and even here, each paradigm of art will answer these questions differently).

This principle forms the primary basis of Gnoli’s (2025c) argument for an all-purpose classification, as evident from the title of his article. Farradane’s “unique definition” is just a small part of his classification theory, which cannot be reviewed here; suffice it to say (1) that his general inductive approach is problematic, as also mentioned by Coates (1973, p. 391), and (2) that his psychological approach to scientific classification seems dubious⁴².

In the sources cited by Gnoli (2025c), three primary sources (Datta and Farradane, 1974; Farradane, 1961; Farradane, 1966) are about “unique definition” and none of these contains a deeper analysis of this concept and its relation to theories of concepts and definitions.

Farradane (1966, p. 108) makes it clear that interest in the ideas on levels of integration (i.e., integrative levels theory) and unique definition arose when the Classification Research Group (CRG) became interested in problems concerning general classification schemes (and so moving from a previous focus on special classifications). This is important, because, as we shall see, it limits the relevance of these principles.

The connection between the theory of integrative levels and the idea of “unique definition” is clear in Gnoli (2025c) and as well as in Farradane (1961, p. 129):

One of the secondary sources cited is Broughton (2024, Section 6.4) “that class which provides a given concept with its most fundamental defining characteristics, e.g., Zoology provides the place of unique definition for the concept Horse”.

Gnoli discusses the example of diamonds in some detail, arguing that this object should be defined as a mineral rather than as, for example, a gem. Gnoli (2025c, p. 12) generalizes this argument as follows:

Gnoli expatiates further (p. 13):

What does Gnoli here argue for? Three levels of his claim may be distinguished:

(1) All concepts should be defined by their essential properties (based on Farradane’s unique definition).

The first claim is the most far-reaching. It is about defining concepts in general (not limited to classification). E.g., a diamond should be defined as a mineral, not as a precious stone. (Or, if more properties are included in the definition, these properties should be included according to a prioritized listing of properties, but as any object has an unlimited number of attributes, it is not possible to list them all). We saw in Section 3.4 that essentialism is a disputed concept today, and that concepts, meanings and definitions are theory-dependent, and even the quote by Gnoli himself (2025a, pp. 78–79) acknowledged that essentialism faces serious difficulties in the empirical sciences. Different disciplines may share the same concept (e.g., both chemistry and physics define gold as the element with atomic number 79), but different theories tend to define concepts differently (e.g., the species concept in biology). Therefore, Farradane’s “unique definition” seems problematic as a general theory of concepts (and it has never been evaluated in the context of theories of concepts or definitions). It should also be remembered that Farradane developed his view because there was a need for it at the time when CRG turned their interests towards general classifications (cf. Farradane, 1966).

The second claim about assignment of a symbol of a phenomenon at the “unambiguous integrative level” and then reuse of this symbol in all other classes whether the phenomenon occurs (with additional symbols), can be understood as a version of the theory of semantic primitives and compositionality (which is discussed in a submitted paper, Hjørland, 2025b). For example, “diamond” is given its notation in the class of minerals (pure substances), and classes about, for example diamond mining and diamonds in jewelry, contains the same symbol + additional symbols. (A kind of formulaic language, reminiscent of chemical formulas). Such a language has been the dream of rationalist philosophers for centuries, but Eco (1995) considered this to be a utopian dream⁴⁴. This does not provide a general rejection of facet analysis, but raises problems concerning its limitations (see Hjørland, 2025b).

The third claim, about general classifications, shall only be discussed briefly, as I am preparing a separate article on general versus special KOSs. The argument about the need for Farradane’s “unique definition” for general classifications depends primarily on the value of phenomena as an organizing principle, which already was been discussed in Section 3.5. Here we shall just consider whether it is best to classify diamonds in one place or whether it is better (like the DDC) to classify diamonds according to different aspects. If some people primarily are interested in diamonds as jewelry, and if there is a special literature about jewelry is it then a good idea to scatter this literature under different kinds of materials? If jewelers form a special group of people, it may be expected that they are primarily interested in grouping the information and documents about jewelry together, thus classifying diamonds in relation to this domain rather than having all documents on diamonds classified in relation to minerals or pure substances. The basic differences in views seem to be the preference of a metaphysical order of the universe (the phenomena approach) versus an order that represents human interests and social organization and based on a kind of literary warrant.

The conclusion of this section is that there is no convincing argument that Farradane’s “unique definition” can serve as a principle for the construction of all-purpose classifications.

Gnoli (2025c, pp. 5–6) considered the periodic table a pluralist classification, claiming that the traditional table reflects the interests of chemists, while the Stowe table reflects the interest of physicists, implying that the table as such is not a neutral, monist one. This is a bit strange, since, as we saw in the Stamos quote in Section 3, this classification is often considered the ultimate challenge for pluralism in classification. Furthermore, it is not right to claim that chemists and physicists have separate classifications, because both chemical and physical research have contributed strongly to the theoretical foundations for the different versions of the table. In addition, there is a strong overlap between different versions. However, it might be that some cases, such as the classification of Helium, tend to split chemists and physicists because chemists may prefer to see Helium as a noble gas (in the p-block), whereas physicists may prefer to classify it with the alkaline earths (in the s-block) according to its electron structure⁴⁵. It is not clear from Gnoli (2025c) which principles he considers useful for producing an alternative to existent periodic tables, one that would better satisfy his aim of a monist classification of the sciences.

It seems clear that, as science grows, humans come to know regularities in nature, which often turn out to be fruitful for conceptualization and classification (but may be overstated, thereby suppressing the need for alternatives). This was also shown in Stamos’ example, that atomic numbers best predict chemical and physical properties of chemical elements and therefore are better units for the periodic system compared to mass numbers. However, as we have seen, even for the periodic system, the debate still is open about how far chosen regularities reflect different interests.

Gnoli (2025c, p. 8) writes:

However, classification of birds, for example, is based on both ornithology (living birds) and paleontology (extinct birds, mainly but not exclusively through the study of fossils), and draws on knowledge and technology of a wide range of sciences, including biogeography, genetics, and geology. Therefore, an ornithological classification of birds, such as that of Fjeldså (2013) already represents an interdisciplinary-based classification of birds (phenomena) and not disciplines. Perhaps Gnoli and others underestimate the interdisciplinary cooperation within the existing system of disciplines?

I have difficulty seeing that it should be possible (even in theory) to construct a phenomenon-based classification that can satisfy all interdisciplinary researchers. Mills and Broughton (1977, p. 37) wrote:

Would some interdisciplinary researchers like to consider all aspects of, for example, the railway system? Or is the case that interdisciplinary researchers consider a new angle, previously scattered over more disciplines, and that they, as suggested by Tengström (1993), over time, may tend to form a new discipline⁴⁶? Tengström’s view implies that even interdisciplinary fields need a classification based on a certain perspective.