Skip to main content

Revisiting abstraction and idealization: how not to criticize mechanistic explanation in molecular biology


Abstraction and idealization are the two notions that are most often discussed in the context of assumptions employed in the process of model building. These notions are also routinely used in philosophical debates such as that on the mechanistic account of explanation. Indeed, an objection to the mechanistic account has recently been formulated precisely on these grounds: mechanists cannot account for the common practice of idealizing difference-making factors in models in molecular biology. In this paper I revisit the debate and I argue that the objection does not stand up to scrutiny. This is because it is riddled with a number of conceptual inconsistencies. By attempting to resolve the tensions, I also draw several general lessons regarding the difficulties of applying abstraction and idealization in scientific practice. Finally, I argue that more care is needed only when speaking of abstraction and idealization in a context in which these concepts play an important role in an argument, such as that on mechanistic explanation.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2


  1. See, e.g., Frigg and Hartmann (2020) for a comprehensive review. Extended discussions can also be found in a number of monographs or book editions. (Bailer-Jones, 2009; Cartwright, 1983; Gelfert, 2016; Giere, 1988; Magnani & Bertolotti, 2017; Morgan, 2012; Morgan & Morrison, 1999; Morrison, 2015; Toon, 2012; Weisberg, 2013)

  2. For example, when presenting the notions of Galilean idealization, minimalist idealization and multiple-models idealization, Michael Weisberg claims that “despite the differences between minimalist idealization and Galilean idealization, minimalist idealizers could in principle produce an identical model to Galilean idealizers” and that “the most important differences between Galilean and minimalist idealization are the ways that they are justified. Even when they produce the same representations, they can be distinguished by the rationales they give for idealization” (Weisberg, 2013, p. 102). Arguably, then, Weisberg’s Galilean and minimalist idealizations are (or at least can be) one and the same (kind of) assumption that is put to work in different ways.

  3. Much focus has been devoted to inquiring into those various functions. For instance, they may allow for making a model mathematically tractable (Jebeile, 2017), although it should be noted that it is not necessarily the case that all assumptions are limited to mathematical modeling. Abstraction and idealization can also help in isolating difference-making factors by narrowing down the focus of a model (Mäki, 1992; Strevens, 2008; Weisberg, 2013), and they play various roles in explanation (Batterman, 2009; Bokulich, 2011; Jebeile & Kennedy, 2015; Kennedy, 2012; Reiss, 2012; Rice, 2015; Rohwer & Rice, 2013; Wayne, 2011) or understanding (Elgin, 2007, 2017; Potochnik, 2015, 2017; Reutlinger et al., 2018; Rice, 2016; Strevens, 2017).

  4. Commenting on the available relevant literature, Margaret Morrison states that “most of [the literature] draws a distinction between idealization which is construed as the distortion of a particular property (e.g. frictionless planes) and abstraction which involves the omission of properties (e.g. a body’s material in calculating its trajectory)” (Morrison, 2011, p. 343).

  5. In the main text I discuss several examples in some detail. However, it is worth to list at least some of the many other questions which also remain a subject of controversy. Among such issues we find questions related to the nature and function(s) of abstraction and idealization. For instance, consider the following questions: how exactly do abstraction and idealization relate to truth (M. R. Jones, 2005; Levy, 2018; Portides, 2018; Teller, 2012), to mathematics (Jebeile, 2017), to fictions (Bokulich, 2011; Suárez, 2009), or to approximations (M. R. Jones, 2005; Morrison, 1998; Norton, 2012; Portides, 2007)? How should we adjust our views regarding the historical development of theories? More specifically, can we reinterpret certain theories or models from the past as if such theories were postulating simplifying assumptions even though these ‘assumptions’ used to be taken at face value? Or should we refrain from such practice and instead consider only an intentional usage as possible instances of these assumptions? Are they eliminable, should they always be, and how are they justified (Batterman, 2002, 2009; Batterman & Rice, 2014; Bokulich, 2017)? Are they best construed as concerning individual claims or rather holistically as pertaining to the models as wholes (M. R. Jones, 2005; Levy, 2018; Rice, 2019)? Do they come in degrees? And if so, how can we estimate different degrees (M. R. Jones, 2005; Levy, 2018)? Should we think of them in terms of the processes by which models are built or as model products? All these issues, many of which remain unresolved, suggest that the topic of abstraction and idealization is, in fact, very complex.

  6. Additionally, and in parallel to explaining how models are built, the philosophical literature has also addressed the ontological question of what models are. Abstraction as subtraction naturally fits with some metaphysical debates: the process of subtracting features generates abstract objects and scientific models have been construed as such (Giere, 1988; Glennan, 2017; Mäki, 2009; Psillos, 2011; Teller, 2001). Thus, some of the accounts could be interpreted as dealing with the question of ontology for which it is presumably well equipped (but see Frigg & Nguyen, 2017; Thomson-Jones, 2010; Toon, 2012 for arguments against models as abstract objects). However, some (N. Jones, 2018; Levy, 2013) have explicitly warned against making connections between abstraction employed in the service of constructing scientific representations and abstract objects.

  7. Similarly, in Frigg and Hartmann (2020), Frigg presents the usual distinction, but in later work he also develops a more detailed account (Frigg, forthcoming, chap. 11).

  8. See, e.g., Nicholson (2012), Levy (2013) and Andersen (2014a, 2014b) for excellent analysis and overview.

  9. Also known as the minimal conception, see Glennan (2017, p. 17).

  10. Kaplan introduces the model-to-mechanism mapping account (also abbreviated as the 3 M account), according to which “a model of a target phenomenon explains that phenomenon to the extent that (a) the variables in the model correspond to identifiable components, activities, and organizational features of the target mechanism that produces, maintains, or underlies the phenomenon, and (b) the (perhaps mathematical) dependencies posited among these (perhaps mathematical) variables in the model correspond to causal relations among the components of the target mechanism,” to which he further adds that the “3 M aligns with the highly plausible assumption that the more accurate and detailed the model is for a target system or phenomenon the better it explains that phenomenon” (Kaplan, 2011, p. 347). This particular account has been challenged by, for example, Chirimuuta (2014), who argues against the presumption that the more details the model provides the better it explains the target phenomenon (see also Batterman, 2002 for an earlier argument in the same direction, albeit in a somewhat different context; see also Batterman & Rice, 2014; Deulofeu et al., 2019). However, it is not clear that Kaplan may be interpreted as subscribing to such a strong statement since elsewhere he states that abstractions and idealizations are a necessary part of scientific work and that they do not jeopardize the explanatory project (Kaplan, 2011, p. 348; see also Kaplan & Craver, 2011, pp. 609–610).

  11. The causal role of concentrations is discussed in detail in Nathan (2014).

  12. A repressor is any molecule that binds DNA, resulting in either blocking the binding of RNA polymerase to its promoter region, or blocking its function, which, in effect, blocks the DNA transcription.

  13. Of course, one may ask for clarification of what it precisely means to be of no consequence to representation. Since the validity of Potochnik’s views are of little concern to us here, we may simply refer the reader to her original text.

  14. It is rather illustrative of the conceptual difficulties that the authors speak of the geometrical shapes by which the entities are typically represented as abstractions. Whether or a not a particular reaction takes place is influenced by a variety of factors, including – importantly – the particular shapes of the reactants. Thus, although shapes are in fact key difference-making factors for a reaction to occur, they are clearly misrepresented by the diagrammatic sketches. Thus, shapes could potentially be re-interpreted as idealizations.

  15. Love and Nathan are very well aware of the fact that the appropriateness of the chosen level of description must be evaluated with respect to the particular issue at hand (research question, educational purpose etc.). They propose to address this using Weisberg’s (2013) multiple model idealization approach. Here we may suggest that the notion of vertical abstraction might serve the purpose better. However, it should also be noted that this particular problem could potentially be re-interpreted as an instance of generalization rather than abstraction, a distinction discussed by Levy (2018).

  16. Note that in the previous Section I briefly introduced the example of the negative feedback mechanism and I argued that the arrows representing a causal process may best be viewed as an instance of vertical abstraction. Although such an abstraction is found wanting in context in which a more detailed description is required in order to answer a specific research question, it nevertheless does not say things that are literally false; hence, it does not contradict the received view about abstraction.

  17. More precisely, van Eck and Mennes focus on the part where Love and Nathan discuss the use of the multiple-model approach. Halina concerns herself with the representational ideal of completeness. However, the specific details are of little concern to us; the issue at hand is only that these authors touch upon Love and Nathan approvingly without realizing the fundamental problems discussed herein.

  18. However, one should also be wary of interpreting this paper as a general defense of the mechanistic framework. Indeed, the rich philosophical literature on the mechanistic explanation has many interesting points to offer regarding the tenability of the framework in molecular biology (see, e.g., Skillings, 2015). This paper only meant to show that whatever the means of challenging the mechanistic account of explanation, arguments such as those found in Love and Nathan’s analysis are not a good way of accomplishing that goal.


Download references


I am indebted to Roman Frigg for providing me with extensive feedback on an earlier version of this paper. I also wish to thank Javier Suárez for many insightful suggestions that helped improve the manuscript. I thank the audiences at the following events: the research seminar at TINT (Helsinki, 2018), Idealization Across the Sciences workshop (Prague, 2019), ISHPSSB (Oslo, 2019), BSPS (Durham, 2019) and EPSA (Geneva, 2019) where versions of this paper at various stages of development were presented. This article is a result of research funded by the Czech Science Foundation, project GA ČR 19-04236S. Finally, I would like to thank the two anonymous referees for their comments which have helped to significantly improve the paper.


This article is a result of research funded by the Czech Science Foundation, project GA ČR 19-04236S.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Martin Zach.

Ethics declarations

Conflict of interest

The authors declare that they have no competing interests / conflicts of interest.

Ethical approval

n/a as the research involved no subjects.

Informed consent

n/a as the research involved no subjects.

Additional information

This article belongs to the Topical Collection: EPSA2019: Selected papers from the biennial conference in Geneva

Guest Editors: Anouk Barberousse, Richard Dawid, Marcel Weber

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Zach, M. Revisiting abstraction and idealization: how not to criticize mechanistic explanation in molecular biology. Euro Jnl Phil Sci 12, 21 (2022).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI:


  • Abstraction
  • Idealization
  • Mechanistic explanation
  • Molecular biology
  • Scientific models