PreprintPDF Available

Using Compositional Explanations to Understand Compositional Levels: An Integrative Account

Authors:
Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

Scientists talk of "levels" where they offer compositional models and explanations. Once we appreciate both the plural, and integrated, character of such compositional models/explanations, I show we can finally provide a more adequate account of compositional levels that accommodates scientific practice. I outline how scientists use "level" to refer to the often large ontological commitments of their integrated compositional models in what I term the "Integrative Account" of levels. And I outline how, contrary to the claims of philosophical critics, such "levels" are precise and rigorous in character and perform useful roles for scientists that cannot be replicated using the notion of "scale".
Content may be subject to copyright.
Carl%Gillett,%Nov%‘19%
1"
Forthcoming in Brooks, DiFrisco and Wimsatt (eds)
Levels of Organization in the Biological Sciences. MIT Press
Using Compositional Explanations to Understand Compositional Levels:
An Integrative Account1
Carl Gillett, Northern Illinois University
Abstract: Scientists talk of “levels” where they offer compositional models and
explanations. Once we appreciate both the plural, and integrated, character of such
compositional models/explanations, I show we can finally provide a more adequate
account of compositional levels that accommodates scientific practice. I outline how
scientists use “level” to refer to the often large ontological commitments of their
integrated compositional models in what I term the “Integrative Account” of levels. And I
outline how, contrary to the claims of philosophical critics, such “levels” are precise and
rigorous in character and perform useful roles for scientists that cannot be replicated
using the notion of “scale”.
Keywords: levels, compositional explanation, integration, pluralism, part-whole
The integration of function across many levels of organization is a special focus
of physiology. (To integrate means to bring varied elements together to create a
unified whole.) (Silverthorn (2013), p.2. Original emphasis.)
The concept of integrative levels of organization is a general description of matter
… new levels of complexity are superimposed on the individual units by the
organization and integration of these units into a single system. What were wholes
on one level become parts on a higher one… Each level of organization possesses
unique properties of structure and behavior… (Novikoff (1947), p.204)
…compositional levels of organization… are constituted of families of entities
usually of comparable size and dynamical properties, which characteristically
interact primarily with one another, and which, taken together, give an apparent
rough closure over a range of phenomena... (Wimsatt (1994), p.225)
In areas of science where researchers talk about “levels” these researchers also
commonly give what I term “compositional” models and explanations where we explain
by taking entities to compose each other. For instance, Figure 1 frames the basis of the
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
1 I am grateful to the audience at the KLI workshop on levels where I presented a draft of
this paper. I am also grateful to Ken Aizawa, James DiFrisco and Varun Ravikumar for
detailed comments earlier drafts.
Carl%Gillett,%Nov%‘19%
2"
famous “sliding filaments” account, and associated explanations/models, of skeletal
muscle contraction using compositional relations to activities of the parts of the muscle,
shown in the Figure, at what are termed by scientists various “levels”.
[FIGURE 1 HERE]
My goal here is to sketch a positive account of such levels building from a better
appreciation of various species of compositional models/explanations and their
integration with each other. Once we appreciate the ontological commitments of plural,
integrated compositional models/explanations, I show we can finally provide a more
adequate account of compositional levels that accommodates scientific practice whilst
also overcoming the objections of critics.
I begin the paper, in Part 1, by briefly clarifying my project, assumptions and
methodology. Then, in Part 2, I use examples from molecular biology, cell biology and
physiology to survey the various species of compositional model to show they extend
beyond the so-called “constitutive mechanistic explanations” highlighted by the New
Mechanist’s that explain an activity of a whole using compositional relations to activities
of parts. In addition, I highlight the species of compositional explanation that explains a
property of a whole, in what I call “Standing” models/explanations, and also a species of
model that posits a compositional relation between a whole and its parts, in what I term
an “Analytic” compositional model. Just as importantly, I highlight how the species of
compositional model are usually highly integrated in their ontology in an interesting
example of what Sandra Mitchell ((2002), (2003)) has termed “integrative pluralism”.
Where their plural models are ontologically integrated, we can expect that scientists may
often consequently have larger ontological commitments, going beyond those of any one
model, and I show that this is true in ways that ultimately illuminate what scientists refer
to as “levels”.
To begin, in Part 3, I show that compositional models/explanations are committed
to what I term “activity closure”. Thus in the sliding filaments model, framed in Figure 1,
roughly put, cells act on other cells, organelles act on other organelles, and
macromolecules act on macromolecules, but cells do not act on organelles or
macromolecules or the muscle itself, or vice versa. I provide reasons to conclude we have
such activity closure of the parts and wholes posited across integrated species of
Carl%Gillett,%Nov%‘19%
3"
compositional models/explanations – that is, these parts and wholes cannot engage in
activities with effects upon each other.
I then outline, in Part 4, the still larger ontological commitments of plural,
integrated compositional models incorporating such activity closure and the
compositional relations posited in specific models. Consider, for example, the
commitments of the plural models integrated with the sliding filaments model. Here we
have groups of individuals that act only on other members of their groups, i.e. cells acting
on cells, organelles acting on organelles, and macromolecules acting on macromolecules.
But, in addition, the individuals in each group are related as parts and wholes
(macromolecules composing organelles that compose cells that compose muscles), their
activities are composed by the activities of individuals in other groups (activities of
myosin and actin composing the contraction of myofibrils where these activities compose
the cells’ activities of contracting which in turn compose the muscle’s contracting) and
the properties of wholes are also composed by properties of their parts.
The plural, integrated models of scientists are thus committed to layers of
compositionally related entities that only act upon each other where I term each of these
layers a “compositional array”. That is, roughly put, a group of individuals, and their
properties and activities, that are: (i) individuals which are all parts of the relevant
terminal whole, (ii) individuals that productively act upon other individuals in the group,
but (iii) individuals such that they do not act upon, and are not acted upon by, the
terminal whole or individuals in other such groups. Once we appreciate such ontological
commitments, then we see that scientists plausibly use “level” to refer to such
compositional arrays and I offer what I term the “Integrative Account” of levels to
capture this insight. Most importantly, I show the Integrative Accounts captures key
features of levels and scientific practices of ascribing such levels, such as the “cellular”,
“organelle” and “molecular” levels ascribed in the sliding filaments model.
One longstanding objection of philosophers of science, such as Philip Kitcher and
Kenneth Schaffner, is that “levels” are inherently vague and imprecise.2 However, in Part
5, I show the Integrative Account allows us to precisely frame when entities are in a
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
2 See Kitcher (1984), note. 3, p.25 and Schaffner (1993), p.287.
Carl%Gillett,%Nov%‘19%
4"
level, in the same level, and higher or lower level relative to each other, thus rebutting
Kitcher and Schaffner’s objection.
Using the Integrative Account, in Part 6, I sketch some obvious practical reasons
why scientists find it useful to talk of “levels” in the face of various forms of complexity.
And I use this work, in Part 7, to assess a new wave of philosophical skepticism about
“levels”, pressed by Marcus Eronen, Angela Potochnik and others, who have argued that
scale” should replace “level” in actual practice.3 However, I illuminate why “scale”
cannot successfully perform the work of, or hence replace, “level” in actual scientific
practice once we finally appreciate the character of this work.
Part 1 – Clarifying the Project and Some Assumptions
It is important to clarify what I am, and am not, seeking to do here, since my
focus is superficially similar to some very different projects. To this end, it is most useful
to start by distinguishing what I term “internal” and “ultimate” ontology.4 My descriptive
work here is focused on what we may term “internal ontology” – that is, the study of the
ontological posits of certain successful scientific practices or products in models and
explanations. This contrasts with what we may term “ultimate ontology” that is focused
on the nature of reality itself. Elsewhere I have discussed at length the ongoing scientific
debates over what conclusions we may draw from our compositional
models/explanations, but here I focus solely on the descriptive project of internal
ontology.5
I am an explicit pluralist both about scientific models, explanations and practices,
and also about the concepts, including ontological notions, these various models,
explanations and practices deploy. Here I take myself to be describing only the
ontological notions specific to certain compositional models and explanations where the
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
3"Eronen (2013, 2015), Potochnik ((2010), (2017)), and Potochnik and McGill (2012).
4 Cf. the distinction drawn in Godfrey-Smith (2010).
5 See Gillett (2016a) for detailed discussion of these scientific debates over reduction and
emergence.
Carl%Gillett,%Nov%‘19%
5"
same words are often used to express different ontological concepts in different models,
explanations and practices, so let me clarify my stance on some key issues.
First, for example, it is important to note that the entities posited in compositional
models are ahistorically individuated – that is, these individuals, properties and activities
are individuated by what they do now.6 The notions of “individual”, “property” and
“activity” in compositional models are therefore different from, for instance, those often
used in evolutionary biology in models, explanations and practices focused on
historically individuated entities.
Second, when I talk of “part” and “whole” I mean the notions posited in
compositional explanations whose nature I briefly sketch below.7 Other notions of part
are used in other kinds of product in other areas of the sciences (Healey (2013)). It is also
unsafe to blithely assume that notions of “part” used in analytic metaphysics, whether
from standard mereology or otherwise, track any of these scientific notions.
Third, I am a pluralist about the notions of “level” deployed in the sciences that
come in explanatory, methodological and ontological varieties, amongst others.8 Once
more, here I am only focused on the ontological notion of a “level” associated with
compositional models/explanations. In ending the paper, I briefly return to how some of
the various notions of “level” might be related.
Fourth, and most importantly for my discussion, I need to carefully highlight my
assumption that at least two different families of “causal” concepts are used in different
models, explanations and practices. As we shall see, central to compositional
explanations is the thick causal concept of an “activity”, or what I also term a
“productive” relation, that is a behavior or doing of an individual resulting in effects upon
another individual (or itself). In contrast, other models posit thin causal notions such as
that of “manipulability” between individuals. I assume thick and thin notions of causation
are distinct. Manipulability is often posited between two individuals without any such
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
6 Amundsen and Lauder (1994).
7 We now have promising accounts of such scientific part-whole relations, see for
example Gillett (2007a), (2013), (2016a)), and Kaiser (2018).
8 See the useful survey of different notions of “level” in Craver (2007), Chapter 5.
Carl%Gillett,%Nov%‘19%
6"
activity between these individuals. And activity plausibly requires more (in doings
resulting in effects) than manipulability.9 I illustrate these differences between activity
and manipulability at various points in my discussion and also the importance of
separating the thick and thin concepts.
Part 2 – Compositional Models/Explanations, their Species and Integration
More descriptive and theoretical work is needed for compositional models and
explanations than I can devote to them here where my focus is upon the nature of
“levels”.10 I focus on examples from human physiology and related areas of cellular and
molecular biology, focusing primarily on our opening example. In 2.1, I look at
models/explanations of an activity of a whole using a compositional relation to activities
of parts. Then, in 2.2, I more briefly look at models/explanations of a property of a whole
using a compositional relation to properties of parts. Most briefly, in 2.3, I note models of
a whole itself using a compositional relation to other individuals that are its parts. I
highlight the selective nature of such models, in 2.4, but more importantly their
integration with each other. In 2.5, I draw the findings of my brief survey together.
2.1 – Explaining an Activity of a Whole: Dynamic Compositional
Models/Explanations
In response to the question “Why did the muscle contract?” one good answer, in
certain contexts, is “The myosin crawled along the actin” where this is based around the
sliding filament model framed in a textbook example in Figure 2. Other good answers
focused on other “levels” in this model are that “The cell fibers contracted”, at the
cellular level, or, at the organelle level, that “The myofibrils contracted”. These are all
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
9 Cf. Machamer (2004).
10 Elsewhere I have offered detailed treatments of such explanations and the
compositional relations they posit (Gillett (2016a), Chapter 2 and Gillett (Unpublished)).
Barberis (2017) persuasively argues the latter accounts do not fit with extant
philosophical treatments of “levels”. However, the present paper outlines a notion of
“levels” that is used in actual scientific practice and that is compatible with my accounts.
Carl%Gillett,%Nov%‘19%
7"
examples of the species of compositional explanation widely acknowledged by
philosophers of science in what the New Mechanists term “constitutive mechanistic
explanationsand which I term “Dynamic” compositional explanations and models.
[FIGURE 2 HERE]
Let us look at more carefully at the type of model we find in Figure 2 that
underlies such an explanation. We explain the muscle’s contraction at some time using a
compositional relation, at that time, to the behavior of various individuals at a number of
what are termed “levels” where these activities are taken to be compositionally related to
each other and, transitively, to the muscle’s contracting.
Working from the bottom of the diagram, the muscle’s contracting is taken to be
composed by many cells (or “cell fibers”) that are inter-connected, or “organized”, so that
as each cell contracts it pulls on the cells to which it is connected and which are also
contracting. Hence the contracting cells compose (or what I term “implement”), and
explain, the muscle’s contracting. Turning to the next pair of “levels”, the cells fibers
contracting are taken to be composed by their many contracting myofibrils which are so
organized, through their interconnection and alignment, that their contracting
implements, and explains, the cell fiber’s contracting. Lastly, many myosin proteins
crawling along actin filaments implement, and explain, the myofibrils contracting. (This
last pair of “levels” is represented in one layer in the diagram since a myofibril is
represented alongside its component myosin and actin.)
As has been widely noted by New Mechanist philosophers of science, these are
explanations of an activity of an individual, in a skeletal muscle contracting, using a
compositional relation to activities of other individuals, in either cells, organelles like
myofibrils, or proteins of actin and myosin.11 The latter is a descriptive claim in internal
ontology about the posits of these explanations/models – namely, that they posit activities
of individuals that are doings or behaviors that result in effects on themselves or other
individuals. These effects involve changes in the motions of masses and hence,
unsurprisingly, these activities involve transfers of energy in one direction or the other.
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
11 For example, Machamer, Darden and Craver (2000), Craver (2007) and Glennan
(2016).
Carl%Gillett,%Nov%‘19%
8"
Notice that in this latter observation about activities involving transfer of energy I
am not making a claim about how to analyze the scientific concepts of particular
activities, or of the general notions of either “activity” or “causation”, rather I am making
an observation about our empirical findings about such activities – they always involve
transfer of energy. I rely on this empirical finding below.
In addition, as is also widely accepted, the individual whose activity is explained,
and the individuals whose activities are used to explain it, are taken by the researchers
offering these explanations to be related as what they explicitly term parts and whole.12
All of these parts and wholes are also parts of what I term a “terminal whole” in a whole
not taken to be a part of further whole – in this case the human body as the model in
Figure 1 makes clear. As textbooks, and specific academic papers, also routinely lay out,
scientists take the proteins and their activities to be amongst the parts and activities of the
skeletal muscle at what is termed the “molecular level”, the myofibrils and their activities
are amongst its parts and activities at the “organelle level”, the cells and their activities
are amongst it parts and activities at the “cellular level” and within the body itself, the
skeletal muscle is taken to be at the “organ” level.13 Lastly, the names of the levels only
reflects the main known parts at each level – thus the molecular level includes other
entities that proteins and cellular level includes other individuals than cells. In like
manner, for brevity, scientists often refer to the main kinds of part at each level whilst
knowing there are other kinds of individuals that are parts and I follow this convention
below. In Part 4 below, I outline the reasons why this situation arises.
2.2 – Explaining a Property of a Whole: Standing Compositional
Models/Explanations
There are plausibly other species of compositional explanation that even the New
Mechanists have overlooked. What I have elsewhere termed “Standing” compositional
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
12 See the extended treatment of this point in Craver (2007), chapter 4.
13 See, for example, Betts (2013) or Silverthorn (2013). The latter list of levels is not
taken to be exhaustive, since there is a tissue and chemical level, but these are the levels
cited in our exemplar models.
Carl%Gillett,%Nov%‘19%
9"
explanations and models explain a property of a whole using a compositional relation,
what is termed “realization”, to properties of parts.
Prominent amongst such Standing explanations in physiology, cell biology and
molecular biology are those where we explain the energy of a whole, one of its
properties, using a compositional relation to the energy of its parts at a certain level, i.e.
to properties of its parts. For example, we explain the energy of a skeletal muscle at a
time using a compositional relation to the combined energies of all of its parts at the
cellular level, or to all of its parts at the organelle or molecular level, at that time. In my
coming discussion, I again rely on these well-confirmed compositional explanations of
the energy of a whole.
Take another example focused on our primary case. In response to the question
“Why is the muscle strong?” good answers, in the relevant contexts, include “Because the
cell’s each have a certain strength” or “Because myosin has the property of exerting a
certain force as it crawls along an actin filament” or an answer focused on properties of
myofibrils. Here we explain the strength of the muscle, a property of a whole, using a
compositional relation of realization to properties of its parts at what scientists term the
cellular, organelle and molecular “levels”.
2.3 – Part-Whole Relations and another Species of Compositional Model
Lastly, we should note that when asked “What is a skeletal muscle?” a number of
good answers, in the relevant contexts, can be built from the model framed in our starting
model in Figure 1, including “Bundled muscle fibers/cells”, “Organized
macromolecules” or an answer built around organelles. It is also worth noting that this
larger model integrates smaller models of this kind including the models in Figure 3
framing parts of muscles at the cellular level, Figure 4 framing parts of the cell at the
organelle level, and Figure 5 framing parts of a region of a myofibril at the molecular
level.
[FIGURE 3 HERE]
[FIGURE 4 HERE]
Carl%Gillett,%Nov%‘19%
10"
[FIGURE 5 HERE]
Here we again plausibly have a type of model. But in this case it is a model of a
certain whole and it posits a compositional relation to individuals that together compose
(or as I shall say “constitute”) this whole. Once more, working scientists explicitly refer
to cellular, organelle and molecular “levels” in this context with reference to such groups
of constituent individuals. I term these “Analytic” compositional models.
Elsewhere, and more contentiously, I have suggested that this type of model, and
the answers noted, are examples of another species of compositional explanation where
the explanandum is a certain whole, i.e. an individual, the explanans is some group of
individuals that are its parts (at a certain “level”) and the backing relation is the
constitution relation between these individuals.14 Here I simply assume we have Analytic
models and leave to one side the more contentious issue of whether they are also ever
explanations.
It is worth noting that the individuals taken to be parts are just those individuals
whose activities compose activities of the whole or whose properties realize properties of
the whole. The individuals highlighted by Dynamic and Standing models/explanations
are thus those that are taken to be parts and underpin Analytic compositional
models/explanations, and vice versa.
2.4 – Selective Ontological Representations and Integrative Pluralism
The literature on models is large and familiar to philosophers of science, so let me
briefly note a common feature that these compositional models share with other models.15
The subject matter of physiology, cell biology and molecular biology is highly complex
in many ways and to make such complex phenomena cognitively tractable scientists
follow the strategy made familiar in work on models which is “Divide and Cognize”,
rather than “Divide and Conquer”.
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
14 Aizawa and Gillett (2019) and Gillett (Unpublished), Chapter 7.
15 See for example Giere (2006).
Carl%Gillett,%Nov%‘19%
11"
To see this, consider our compositional models focused on the skeletal muscle, its
activity of contracting and related properties that we have just surveyed. In each of the
species of compositional model just one category of entity is taken as explanadum and
just one instance of that category. For example, in our Dynamic model just one activity is
the explanadum, in our Standing models just one property is the explanadum and in our
Analytic model just the individual itself is the focus, rather than any of its activities or
properties. Similar points hold about the selectivity of the explanans represented in each
model: Dynamic models just use activities of parts, Standing models just use properties
of parts and Analytic models just use parts. In fact, the categories explicitly represented
in these models is usually limited in these ways, thus Dynamic models do not represent
properties, Standing models do not represent activities, and Analytic models usually
represent neither properties or activities.16
What compositional relation is represented, and is the backing relation of the
model/explanation, is also selective. Each of the species of compositional model we have
looked at has just one kind of compositional relation explicitly represented in the relevant
model. Thus a Dynamic model explicitly represents just an implementation relation
between activities of parts and whole, a Standing model just explicitly represent a
realization relation to properties of parts and whole, and an Analytic model explicitly
represents just a constitution relation to individuals.
Scientists are well aware, for example, that there are many other compositional
relations, or that the phenomenon is an individual with many activities and also many
properties. But compositional models are like so many models in being selective
representations in these various ways. Such models are presumably also often
idealizations, and/or abstractions, and/or have other features, but their nature as selective
representations suffices for my purposes here. More important for my purposes is to
highlight, alongside the ways scientists “Divide and Cognize”, how we also find
important ways that the resulting models used by researchers are also integrated.
First, and most bluntly, we find that larger compositional models, like that in
Figure 1, incorporate less encompassing models like those in Figures 3, 4 and 5. Here
Analytic models of individuals and their parts at lower levels are combined into an
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
16 These features are usually, but not always, the case.
Carl%Gillett,%Nov%‘19%
12"
Analytic model encompassing all of these individuals and their relations. Similar
examples are found with Dynamic and Standing models, and hence with the activities and
properties of parts and wholes.
Second, if we just focus on the relations across the species of compositional
model associated with skeletal muscles, their contraction and the relevant properties, we
see other forms of integration. For notice the integration amongst the individuals posited
in these models. Focusing on the organ, organelle and molecular levels, our Dynamic,
Standing and Analytic models, for instance, all posit a skeletal muscle, myofibrils, and
actin and myosin proteins, amongst other individuals. So a crude point is that the internal
ontology across the models is integrated by positing overlapping kinds of individuals at
various levels.
Third, although Dynamic models posit activities and no properties, and Standing
models posit properties and no activities, these models are integrated in another way. For
the properties posited in parts and wholes, at various levels, in Standing models are those
that manifest themselves in the activities posited, at various levels, in Dynamic models.
For example, in a skeletal muscle energy and strength are the properties that result in the
activity of contracting in the muscle when manifested. Whilst having a certain energy and
exerting a certain force in myosin proteins are the properties that result in the myosin’s
activity moving down a chain of actin when manifested. Similar points hold about the
properties and activities of myofibrils at the organelle level. So the internal ontology of
Standing and Dynamic models is integrated even though not directly overlapping.
Fourth, as I noted above in relation to Analytic compositional models, the
individuals counted as parts of a certain whole are plausibly those individuals whose
properties and/or activities are successfully posited in Standing and Dynamic models as
composing properties and activities of the relevant whole. So the three species of
compositional model exhibit another kind of ontological integration.
There are still further forms of such ontological integration, i.e. integration in
their internal ontology, but just these examples illustrate how the species of
compositional model/explanation about some phenomenon are often highly integrated
and so too are specific examples of each species of model. And we should briefly note
how this latter kind ontological integration extends to certain types of causal models such
Carl%Gillett,%Nov%‘19%
13"
as the “etiological mechanistic explanations” that the New Mechanists have documented
as also positing activities of an individual to explain certain effects.17 Thus, for example,
we explain a certain bone moving by positing an activity of contracting in a muscle. Or
we explain a change in the length of a myofibril by positing an activity in it of
contracting. Notice that such causal models posit activities and individuals overlapping
those in our compositional models, and that such models are also bluntly incorporated
into Dynamic compositional models, and hence are also plausibly ontologically
integrated with these compositional models.
Overall, our survey thus documents how compositional models and explanations,
and related causal models and explanations, are plausibly another example of what
Sandra Mitchell ((2002), (2003)) has termed “integrative pluralism” in examples where a
variety of different kinds of models and explanations are integrated. In coming sections, I
explore the implications of such ontological integration and its import for understanding
what scientists mean by “level” when talking about their integrated compositional
models.
2.5 – A Neglected Family of Models/Explanations and a Novel Form of Integrative
Pluralism
Let me briefly summarize our findings. All of the compositional explanations we
have surveyed are what I shall term “ontic” explanations that work by representing an
ontological relation between entities in the world, the explicitly represented backing
relation of the explanation, whose ontological nature drives these explanations.
Furthermore, these explanations are all backed by compositional, rather than causal,
relations, since their backing relations all share common ontological features lacking in
causal relations. For example, amongst other singular features, their backing relations are
all synchronous relations, between entities that are in some sense the same (though not
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
17 See, for instance, Machamer, Darden and Craver (2000), Craver (2007) and Glennan
(2017).
Carl%Gillett,%Nov%‘19%
14"
identical), and which involve synchronous changes in their relata.18 So we can see that
these are not causal explanations.19
We thus have another, neglected, family of ontic explanations of singular facts or
events in the sciences in addition to causal explanations – hence supporting a pluralism
about ontic explanation.20 I term these “compositional” models or explanations, rather
than “mechanistic” models or explanations, since Standing and Analytic explanations are
not mechanistic because they do not have activities as their explanans or explanandum.21
And my brief survey highlights how pluralism is also true about this family of
models/explanations itself, since we have found at least three species that differ, in the
ways framed in Table 1, in the categories of entity they represent, their explanans,
explanandum and their backing relation.
Categories
Represented
Explanandum
Explanans
Analytic
Compositional
Model (and/or
Explanation)
Individuals
An individual
whole
Individual
parts
Standing
Compositional
Model and
Explanation
Properties
and
individuals
A property of a
whole
Properties of
parts
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
18 Elsewhere I have highlighted still further differences between the features of such
compositional relations and causal relations. See Gillett (2016a), Chapter 2.
19 More detailed arguments can be given that the backing relations of these explanations
cannot be either ontologically “thin’ causal relations, like manipulability, (Baumgartner
and Gebharter (2016)), Romero (2015), Gillett (2020)) or ontologically ‘thick’ causal
relations, such as activities (Gillett (2020)).
20 See Gillett (2020), and Gillett (Unpublished), Ch.4, for detailed defenses of this
pluralist conclusion abut ontic explanation.
21 I also avoid the term “part-whole” explanation (Love and Huttemann (2011)), since
part-whole relations are not the backing relations of Standing or Dynamic explanations.
Carl%Gillett,%Nov%‘19%
15"
Dynamic
Compositional
Model and
Explanation
a.k.a.
Constitutive
Mechanistic
Explanation
Activities
and
individuals
An activity of a
whole
Activities of
parts
Table 1. Species of Compositional Model/Explanation and their Differing Characteristics
As we have just seen, compositional models/explanations have forms of
integration in their internal ontology that also extend to related causal models, too. This is
interesting, since we have see that working scientists routinely use the term “levelto
refer to all of the integrated species of compositional model/explanation. One can
therefore wonder whether what scientists mean by “level” relates in some way to the
ontological integration of compositional models? In coming sections, I explore this idea
in stages starting by using our work on the integration of compositional models to
illuminate some of their integrated ontological commitments that have previously been
overlooked.
Part 3 – The Activity Closure of Parts and Wholes in Compositional
Models/Explanations
In this section I explore whether our integrated compositional models have either
of two forms of what I term, in 3.1, “activity closure” of parts and wholes – that is, parts
and wholes that do not act upon each other. In 3.2, I outline descriptive evidence from
our examples that scientists do respect activity closure in their compositional
models/explanations. Next, in 3.3, I present arguments, using features of our integrated
models, that show scientists ought to accept activity closure. I conclude, in 3.4, that
scientists do, and ought, to take compositional explanations to embody activity closure of
thick causal relations of parts and wholes, whilst acknowledging scientists have other
models illuminating thin causal relations over time between such parts and wholes.
3.1 – Two Forms of Activity Closure
Carl%Gillett,%Nov%‘19%
16"
It is important to frame, and distinguish, two closure claims about the activities of
parts and wholes in integrated compositional models. First, there is the claim that
individuals directly related to each other as part and whole cannot act upon each other.
This is what I term “Narrow Activity Closure”:
(Narrow Activity Closure) An individual that is a scientific part of some other
individual cannot engage in activities that have effects upon this whole, or vice
versa.
This thesis is a “closure” claim about activities, since it claims that individuals directly
related as scientific parts and wholes are closed with regard to the activities of each other.
Thus a cell that is part of a skeletal muscle cannot engage in an activity with an effect on
this muscle, or vice versa.
Second, we have a far wider form of activity closure – namely, that parts and
wholes in the same terminal whole cannot act upon each other regardless of whether
these individuals are related to each other as parts and/or wholes. Call this wider thesis
“Broad Activity Closure”:
(Broad Activity Closure) Two individuals, sa and sb, that are each parts of some
individual s*, but where sa and sb are not related as part and whole, and where sa
and sb are at different levels in s*, cannot engage in activities that have effects
upon each other.
Here we have the claim that any individuals at what scientists term different “levels”
within the compositional hierarchy of the same terminal whole, and even though not
directly related as part and whole, cannot act upon each other.22
3.2 – Justifying Activity Closure (I): Descriptive Evidence
Is there any descriptive evidence that scientists respect either form of activity
closure? To see that there is, let us look at the Dynamic compositional explanation of
muscle contraction from Figures 1 or 2. For instance, the explanation posits cells that act
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
22 The wider thesis is stated in terms of “levels”, but since we have seen this is common
in the scientific practice it is hard to avoid and does not pose a problem for my later
arguments.
Carl%Gillett,%Nov%‘19%
17"
on other cells, but not cellular activities with effects on the muscle, organelles like
myofibrils, or macromolecules, nor any activity of the muscle, organelles, or
macromolecules with effects on the cells. We thus find that this Dynamic model respects
both Narrow and Broad Closure where this is a descriptive claim about the posits of the
explanation and its underlying model in Figures 1 or 2. And similar points hold for other
Dynamic models and explanations.
In addition, consider the internal ontology of our examples of Standing
compositional models and explanations where we find properties of cells that manifest in
activities on other cells, but not in activities with effects on the muscle, organelles like
myofibrils, or macromolecules. In addition, in such Standing models, we do not find
properties posited in the muscle, organelles, or macromolecules that manifest in activities
with effects on the cells. So we see that these Standing compositional models and
explanations also plausibly respect both Narrow and Broad Closure and, again, we
plausibly find this feature in other Standing models/explanations.
3.3 – Justifying Activity Closure (II): Energetic Arguments
We have found plausible descriptive evidence that in practice working scientists
do respect both Narrow and Broad Activity Closure in their compositional
models/explanations. But we can still ask whether they ought to accept either Closure
claim? To answer this question, I want to explore the joint commitments of our integrated
Dynamic and Standing compositional explanations, along with other empirical evidence,
in a reductio ad absurdum argument, what I dub an “Energetic Argument”, focused on the
energy of parts and wholes.
Let us start with a reductio argument for Narrow Closure. So consider the muscle
cell, sb1, and a molecule of myosin that is one of its parts, sa1, and for the sake of reductio
assume the myosin productively acts on the cell between t1 and t2 to change the cell.23
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
23 To simplify, the reader should note I have assumed, first, that the parts of the cell
remain constant from t1-t2 and, second, that the whole and parts are not constantly
exchanging energy with the surroundings from t1-t2. Obviously both simplifying
assumptions are incorrect, since we know individuals like cells constantly change parts
and individuals like cells and proteins are constantly exchanging energy with the
surroundings. However, such details can be added to the argument at the cost of
Carl%Gillett,%Nov%‘19%
18"
Given our empirical evidence about activities, I assume that change involves transfer of
energy. So the myosin sa1 transfers energy Y to the cell sb1 by t2. But, at t1, we know from
our Standing explanations that the energy of the cell equals the combined energy of all
the proteins that are its parts, in sa1-san, including our myosin molecule, and let this equal
N. So at t1, the energy of the cell is N. But, given the transfer of energy by t2 from sa1,
we can conclude the energy of the cell at t2 is (N+Y). In similar fashion, given this
transfer, we can also conclude the energy of the molecular parts sa1-san at t2 is (N-Y). But,
at t2, the cell’s energy is the combined energy of the parts. So we can also conclude that
at t2 the energy of the cell is (N-Y). Therefore, the energy of the cell at t2 is, and is not,
(N+Y) – a contradiction.
This Energetic Argument takes premised form thus:
(1) All activity involves transfer of energy;
(2) The energy of the cell sb1 at time t equals the combined energies of the proteins,
sa1- san, that are its parts at the molecular level at t;
(3) At t1, the energy of the proteins sa1- san equals N;
(4) Between t1 and t2, the myosin protein sa1 acts upon the cell sb1
From (2) and (3):
(5) At t1, the energy of the cell sb1 equals N.
From (1) and (4):
(6) By t2, the protein sa1 transfers Y energy to the cell sb1;
From (5) and (6):
(7) At t2, the energy of the cell sb1 is N+Y
From (3) and (6):
(8) At t2, the energy of the proteins sa1- san is N-Y;
From (2) and (8):
(9) At t2, the energy of the cell sb1 is N-Y
From (7) and (9):
(10) At t2, the energy of the cell sb1 is, and is not, N+Y
"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
complexity while leaving the conclusion untouched, so I leave them aside having noted
them.
Carl%Gillett,%Nov%‘19%
19"
Here some premise must be false. Premise (1) is the empirical claim that all activities of
individuals, of the relevant kind, involve transfer of energy. In premise (2) we have the
empirical finding, from Standing compositional explanations, that the energy of a whole
cell is equal to the combined energies of its parts at the molecular level. Premise (3) is a
simple assumption, again empirically supported, about the starting energy of the proteins.
So the only plausible candidate to be false is premise (4) and the claim that a part acts
upon its own whole.
The very same kind of argument can be run against the claim that a whole can act
upon its own part. Thus Energetic Arguments support Narrow Activity Closure – that is,
that in compositional models/explanations a part cannot act upon its own whole, or vice
versa. However, we can press such Energetic Arguments further to support Broad
Activity Closure and the claim that parts and wholes, at different levels in the same
terminal whole, cannot act upon each other even if not themselves related as part and
whole.
To see this, consider our skeletal muscle, s*, which has as parts at the cellular
level certain cells, sb1- sbm, and has as parts at the molecular level certain proteins, sa1- sax,
where the proteins are also parts of the respective cells. Assume that a myosin protein,
sa1, is a part of the cell sb1, but is not a part of the cell sb2. In this situation, assume for
reductio’s sake that between t1 and t2 our myosin protein, sa1, acts upon the cell sb2 of
which it is not a part. So, from t1 to t2, sa1 transfers energy Z to sb2. At t1, the combined
energy of all the cells sb1- sbm that are parts of s* was M. Thus we can conclude that at t2,
given the transfer from sb1 to sa2, the energy of the cells sb1- sbm is M+Z. We may assume
that the combined energy of the proteins that are parts of the muscle s* at t1 was also M.
So we can also conclude, given the transfer from sb1 to sa2, that at t2 the combined energy
of the proteins sa1- sax is M-Z. However, the energy of the muscle just is the combined
energies of its parts at a level. So, we can conclude that at t2 the muscle s* does, and does
not, have an energy of M+Z.
This Energetic Argument can be framed as follows:
(11) All activity involves transfer of energy;
Carl%Gillett,%Nov%‘19%
20"
(12) The energy of a whole at a time equals the combined energies of its parts
at a certain level at that time;
(13) At t1, the energy of all the proteins sa1- sax which are parts of the muscle
s* at the molecular level equals M;
(14) At t1, the energy of all the cells sb1- sbm which are parts of the muscle s* at
the cellular level equals M;
(15) Between t1 and t2, the myosin protein sa1 acts upon the cell sb2;
From (11) and (15):
(16) By t2, the protein sa1 transfers Z energy to the cell sb2;
From (14) and (16):
(17)
At t2, the energy of all the cells sb1- sbm, that are parts of the muscle s*, is
M+Z;
From (13) and (16):
(18) At t2, the energy of all the proteins sa1- sax, that are parts of the muscle s*, is
M-Z;
From (12) and (17):
(19) At t2, the energy of the muscle, s*, is M+Z
From (12) and (18):
(20) At t2, the energy of the muscle, s*, is M-Z
From (19) and (20):
(21) At t2, the energy of the muscle, s*, is, and is not, M+Z
Here premises (11) and (12) are supported by the same empirical findings supporting the
premises of our simple argument, whilst (13) and (14) also simply frame well supported
empirical claims. Thus premise (15) is plausibly the false premise. Again, similar
arguments apply to activities of a whole on a lower level part. So a part cannot act upon a
whole, or vice versa, even when they are only in the same compositional hierarchy at
different levels and not themselves directly related as part and whole. Consequently, we
have also an independent argument, built upon empirical findings, supporting Broad
Activity Closure.
Carl%Gillett,%Nov%‘19%
21"
3.4 –Part-Whole Causation in Compositional Models/Explanations? Thick Relations
of Activity, No; Thin Relations, Yes.
My work in this section supports both Narrow and Broad Activity Closure about
the internal ontology of compositional models/explanations, but it is worth exploring a
potential objection since it brings out more clearly the scope of my conclusions.
Someone might object that my arguments face a crippling problem because they
prove too much by showing there is no causation between parts and wholes. But, the
objection continues, working scientists routinely, and apparently highly successfully,
illuminate causation between parts and wholes, and their activities and properties, using
inter-level experiments. So, concludes the objection, something has gone badly wrong
with my arguments about activity closure.
However, my arguments in this section only apply to activities, and hence thick
causal relations, but not to manipulability, difference-making and thin causal relations.
My arguments show that compositional models/explanations do not, and should not, posit
activities between parts and wholes, and/or their activities and properties. But my
arguments do not apply to, and do not show anything about, whether there are thin causal
relations of manipulability or difference-making over time between parts and wholes,
and/or their activities and properties.
My arguments therefore do not show too much, nor do their conclusions conflict
with results from inter-level experiments.24 What the objection also usefully highlights is
that scientists inherit a complex set of commitments from their various models. From
their successful compositional models scientists are committed to parts which
productively interact with others parts, and wholes that act upon other wholes, but where
there with no activity between them. From their plethora of models derived from inter-
level experiments, scientists are committed to various thin causal relations over time,
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
24 My findings therefore potentially support the important earlier conclusion of Craver
and Bechtel (2008) that there is “no inter-level causation” in Dynamic compositional
explanation depending upon whether this is intended to refer to productive, rather than
manipulability, relations.
Carl%Gillett,%Nov%‘19%
22"
whether manipulability or difference-making relations, between these same parts and
wholes, and their activities or properties.
[FIGURE 6]
Scientists thus have a complex set of ontological commitments, framed in Figure
5, as a result of their various models. Pretty obviously, it will be practically challenging
for researchers to track these commitments to both the thick, and thin, causal relations. In
subsequent sections, I explore how use of the term “level” might aid scientists in
navigating this complexity, but in the next section I outline still bigger ontological
commitments of our plural, integrated compositional models.
Part 4 – The Integrative Account of Compositional Levels
I outline, in 4.1, how our integrated compositional models have broad ontological
commitments in what I term “compositional arrays”. What I term the “Integrative
Account” of levels takes scientists to use the term “level” to refer to these compositional
arrays. And I show the Integrative Account captures scientific practices of level
ascription and does better in this respect than extant accounts such as Craver (2007).
Furthermore, in 4.2, I show how the Integrative Account illuminates various features of
levels, including the ways in which compositional levels are local and their nuanced
relationship to scales.
4.1 – Integrated Ontological Commitments, Compositional Arrays and Integrative
“Levels”
Our primary example, in the integrated compositional models concerning skeletal
muscles, illustrates the commitment of these models to both activity closure and
compositional relations. The compositional models about skeletal muscles we have
looked at, roughly put, take proteins to act on other proteins and macromolecules,
organelles that act on other organelles, and cells that act on other cells, but none of the
individuals in these groups act upon individuals in other groups or the muscle itself - thus
the models embody activity closure. In addition, these integrated compositional models
take the individuals in these groups, and their activities and properties, to be
compositionally related. Proteins together compose organelles that together compose
Carl%Gillett,%Nov%‘19%
23"
cells that together compose the muscle. And the models take activities of myosin and
actin to compose the contraction of myofibrils where these activities compose the cells’
activities of contracting which in turn compose the muscle’s contracting. Similar points
hold about the properties of wholes that we have seen are also composed by properties of
their parts.
We thus have complex, and highly ordered, ontological commitments of plural,
integrated compositional models in which we find closed layers of individual components
that only act upon each, but do not act on individuals in other closed layers. I term these
individual layers “compositional arrays”. A compositional array is a group of individuals
(and their activities and properties) such that: (i) individuals in the group are all parts of
the relevant terminal whole, (ii) individuals in the group productively act upon other
individuals in their group, but (iii) individuals in the group do not act upon, and are not
acted on by, the terminal whole or individuals in other such groups.
Appreciating that plural, integrated compositional models are committed to
compositional arrays is important in itself. But what should also be striking in our
examples of such compositional models is that scientists apparently use the term “level”
to refer to just such compositional arrays. What I call the “Integrative Account” of levels
frames this idea that “level” refers to the compositional arrays to which our integrated
compositional models are committed.
We can quickly confirm that the Integrative Account matches actual scientific
practice. To see this, consider what the Account says about the levels of various entities
in the sliding filament model and associated models. First, the Integrative Account places
cells at the same level, since cells are components and productively interact with each
other. And the Account places the properties and activities of cells at this level, too.
Similarly, the Integrative Account places proteins and macromolecules, and their
activities and properties, at the same level. And it places organelles, and their activities
and properties, at the same level. And the Integrative Account places cells, or organelles,
or proteins, or their properties and activities, at different levels. And all of these
ascriptions match the actual practices we have seen in our examples.
In Part 5, I frame the precise ascriptions of being higher or lower level under the
Integrative Account that again match actual practice. But rather than belaboring these
Carl%Gillett,%Nov%‘19%
24"
positive ascriptions, let us briefly look at what the account says about cases where
scientists withhold level ascriptions. Consider a virus newly entered into a cell. Here the
Integrative Account says that the virus is at no level, since our models take the activities
and properties of the virus to compose no activities or properties of the cell or the body –
hence the virus is in no compositional array or level. Again, this matches actual practice,
since researchers do not take the virus to be a component or at a level.
The Integrative Account of levels thus appears to fit well with actual practices of
ascribing/withholding “levels”. So let us change gears to critically compare the
Integrative Account and the best extant treatment of compositional levels in Carl Craver’s
“Levels of Mechanisms Theory” (Craver (2007)). Craver’s framework is solely based on
Dynamic compositional explanations, what he term constitutive mechanistic
explanations, since this is the only species of compositional model/explanation so far
recognized by the New Mechanists – hence explaining why Craver’s account is based
around compositional relations of just the activities, and individuals, we find in this
species of model. Famously, Craver’s Theory has two internal problems (Eronen (2015)):
first, Craver’s Theory cannot say when entities are at the same level; and, second,
Craver’s Theory cannot provide level ascriptions across the entities posited in distinct
models.
The Integrative Account obviously contrasts in its basis from Craver’s treatment
and we can now see that this allows the Integrative Account to overcome the difficulties
of the Levels of Mechanisms Theory. Crucially, the Integrative Account acknowledges
Standing and Analytic compositional models, as well as Dynamic models. Furthermore,
the Integrative Account does not take “level” to refer to the internal ontology of a single
compositional model or even a single species. Instead, the Integrative Account takes
“level” to refer to the ontological commitments of plural, integrated compositional
models and often causal models, too.
Consequently, given these differences, first, the Integrative Account does
underwrite ascriptions of when entities are at the same level, since this is driven by
entities being in the same compositional array due to their the productive and
compositional relations. In Part 5, I provide a precise framework for when entities are at
the same level that confirms this point. And, second, it should also be obvious that the
Carl%Gillett,%Nov%‘19%
25"
Integrative Account makes level ascriptions across entities posited in different models
and also across different categories of entity, including properties as well as activities and
individuals. Thus, for example, the Integrative Account places an activity of myosin,
posited in our Dynamic compositional explanation, as at the same level as the property of
the myosin of exerting a certain force, posited in our Standing compositional model,
since both entities are in the same compositional array. The framework offered in Part 5
again confirms these points.
It is also worth emphasizing at this point that for simplicity of exposition I have
focused throughout the paper on just a handful of integrated compositional models in
those associated with skeletal muscles and their contraction. But the muscle has many
other activities than contracting, and many other properties than strength, and the
compositional models for these activities and properties are also ontologically integrated
with the few models I have looked at here. Furthermore, the muscle and its various
activities and properties themselves figure in many other compositional explanations of
the body, and its activities and properties, where these models involves various other
organs, and their activities and properties, as well as their components, their activities and
their properties at lower levels. These compositional models and explanations of the body
are plausibly also ontologically integrated with each other and with compositional models
of organs like the skeletal muscle, cells, organelles, etc, including those I have looked at.
Though the precise extent of such ontological integration across models is an important
avenue for future research, it is clear that the plural, integrated compositional models
driving level ascriptions under the Integrative Account is far, far greater than the handful
of compositional models I explicitly looked at in Part 2.
4.2 – Their Local Nature, Nuanced Relations to Scale and other Features of Levels
The Integrative Account of levels can also be shown to accommodate, and
illuminate, many of the features of compositional levels highlighted in the descriptive
work of writers like Silverthorn, Novikoff or Wimsatt quoted in the starting passages of
this paper. For example, reflecting on features of compositional relations themselves
highlights why compositional levels have the feature Novikoff notes of involving
qualitatively distinct individuals, properties and activities (Gillett (2016a), chapter 2).
Carl%Gillett,%Nov%‘19%
26"
Earlier sections have also already highlighted why the Integrative Account is built around
levels that are what Wimsatt terms “families of entities… which characteristically
interact primarily with one another, and which, taken together, give an apparent rough
closure over a range of phenomena...(Wimsatt (1994), p.225). And the Account
obviously incorporates the centrality of integration that Silverthorn emphasizes.
Below I also detail how the Integrative Account illuminates why Wimsatt is also
right that levels concern “families of entities usually of comparable size and dynamical
properties”. But let us build up to that point by first highlighting how the Integrative
Account supports the claim that compositional levels are local as writers like Craver
(2007), and others, have emphasized. For when a part acts beyond the boundary of its
terminal whole, then a part is acting on a non-part and we do not have activity closure, a
compositional array or a compositional level. So we can see that under the Integrative
Account we do not have global levels, but only levels that are local to the boundaries,
both external and internal, of the relevant terminal whole.
The levels supported by compositional explanations are also local in another
sense. If different kinds of individual organize, and productively interact, in diverse ways
in distinct kinds of individual, then we have different types and/or numbers of level
across different kinds of terminal whole. And this appears to be what we find. For
example, there is no cellular, tissue, or organ level in rocks or glaciers. So the extent of a
level is local to its terminal whole, but so too are the kind, and hence numbers, of level.
Critics of “levels”, like Eronen and others, have pointed to putative problem cases
where we sometimes have entities of the same kind, and hence the same scale, at
different levels. But the Integrative Account illuminates why these are not problematic
cases and also illuminates the relation of levels and scales.
If one individual of a kind productively interacts with the individuals in one
compositional array, whilst another individual of the same kind interacts with individuals
in another compositional array, then the individuals are in distinct compositional arrays
and the Integrative Account illuminates why scientists place the different individuals at
distinct levels. And the Integrative Account consequently allows us to better understand
the relation of scale and compositional level.
Carl%Gillett,%Nov%‘19%
27"
The key is that composed and component entities are always of different relative
scales, since the compositional relations posited in compositional models/explanations
are many-one with many components and one composed entity (Gillett (2016a), Chapter
2).25 Thus many activities of proteins implement the muscle’s contracting, many
properties of proteins realize the muscle’s strength and many proteins constitute a
muscle. Given the many-one character of scientific composition, the composed (whether
individual, property or activity) will be many times its components in size, mass, energy,
duration, etc. – hence leading to the relative difference in scale of composed and
components entities at different levels.
If we combine the latter point with the reasons why we can end up with entities of
different scales in the same level, due to their productive interactions, then the Integrative
Account illuminates why Wimsatt is correct in carefully emphasizing that entities at the
same level are only usually of the same scale, and that entities at higher and lower levels
are only usually of a different scale, since neither is absolutely the case. Although each
entity at a level is of a different relative scale than its own components, there is no
guarantee that this entity is of the same scale as other entities at the relevant level, since
entities of different scales may end-up in the same compositional array due to their
productive relations. In fact, what we often end-up with is a motley of individuals at a
level that all productively interact though being of different kinds and even different
scales. As I noted in 2.1 above, for example, for simplicity researchers often refer to the
“molecular level”, or talk about proteins when referring to the individuals at this level,
when they are well aware that there are many other kinds of individuals than proteins that
are parts at this level, i.e. lipids, sugars, nucleic acids, and many other non-proteins.26
Similar points apply to other levels.
Part 5 – Same, Higher and Lower Level under the Integrative Account: Rebutting
the Kitcher-Schaffner Objection
The wider scope of “compositional level” under the Integrative Account which
spans many categories and models, particularly in contrast to Craver’s framework, might
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
25 Cf. Craver (2007), Chapter 5.
26 Thanks to James DeFrisco for pressing me to make this point clear.
Carl%Gillett,%Nov%‘19%
28"
raise concerns that we have a vague notion of “level”. As I noted, earlier critics, like
Kitcher and Schaffner, dismissed the notion of a “compositional level” because they
claimed that no precise answers about levels can be given.
However, the Integrative Account is not intrinsically vague in its ascriptions and
shows that the Kitcher-Schaffner objection is simply mistaken. Under the Integrative
Account, using the ideas of earlier sections, we can now precisely frame what it is to be
in a level, and at the same level, as follows:
(Level) Aside from the terminal whole w which is at its own level, individuals
sb1-sbn, and their properties and activities, are in the same level, and hence in a
level, within the whole w, under conditions $ at time t, if and only, at t under $, (i)
individuals sb1-sbn, and their properties and activities, bear compositional
relations to w and its properties and activities, (ii) the individuals sb1-sbn can or
do productively interact with each other; and (iii) if individuals sb1-sbn, and their
properties and activities, bear compositional relations to individuals sa1-sam, and
their properties and activities, then individuals sb1-sbn, and their properties and
activities, do not productively interact with individuals sa1-sam, and their
properties and activities, or vice versa.
Here we have the main features of compositional arrays, in (i)-(iii), that we have already
sketched in an account of level that successfully provides precise, same level ascriptions
cross-categorially for individuals, activities and properties across distinct, but integrated,
models. This confirms that the Integrative Account, unlike Craver’s Levels of
Mechanisms framework, make ascriptions about when entities are at same level. (As an
aside, note that the terminal whole is often taken by scientists to be at a level although not
in a group of interacting components, so I have treated it as special case but condition
(iii) still entails we have activity closure with the terminal whole and its parts.)
Furthermore, building on the latter, we can precisely articulate under the
Integrative Account when for any two individuals, or their properties and activities, they
are at higher or lower levels relative to each other as follows:
(Higher/Lower Level) An individual s, or its property or activity, is in a lower
[higher] level relative to an individual sx, or its property or activity, in whole w,
under $ at time t, if and only if, under $ at time t, (i) s is in a level in w, (ii) sx is in
a level in w, and (iii) s or one of its activities or properties composes [is composed
by] an entity, whether an individual, property or activity, in the same level as sx.
Carl%Gillett,%Nov%‘19%
29"
Again, this definition works cross-categorially for individuals, or their activities and
properties, and it also applies regardless of whether two entities bear compositional
relations to each other. So we get precise cross-categorial ascriptions of being higher or
lower level that once more apply not just to one model, but across various integrated
models of however great an extent.
We can therefore see that Kitcher and Schaffner’s claims about the vagueness and
imprecision of levels is mistaken. Rather than levels being in some way problematic, it
appears that the neglect of compositional models/explanations precluded our
understanding levels in all their sophistication.
Part 6 – Some Reasons Why Scientists Talk about “Levels”
Can we now offer any ideas about why working scientists might find it useful to
talk of levels? The most obvious utility of using the term “level” derives from the
complexity of compositional arrays. Scientists could talk about groups of productively
interacting individuals, and their activities and properties, that compose (and are
composed by) individuals, and their activities and properties, in other such groups, where
these groups of individuals, and their activities and properties, do not act upon each other.
But that is a very large mouthful! Instead, researchers have coined a term, in “level”, that
allows them to concisely talk about these complex commitments.
On top of this simple reason, our earlier observations also highlight why the term
“level” is practically useful in other ways. Compositional explanations and models posit
parts and wholes that are closed to each other’s thick causal relations of activity.
However, inter-level experiments generate other models positing thin causal relations
between the same entities over time, whether manipulability or difference-making
relations. In this situation, active researchers face numerous practical challenges. Using
the term “level”, in the manner framed by the Integrative Account, aids researchers with
these difficulties by allowing them to more swiftly track, and/or communicate, which
entities are in which arrays. This is important as researchers sift through the significance
of their various models. For instance, talking of “levels” allows scientists to track which
thin causal relations revealed by experiments are, or are not, ruled out by their
compositional explanations as picking-out activities.
Carl%Gillett,%Nov%‘19%
30"
There are plausibly other reasons why it is useful to talk in this way of “levels”.27
But just the two kinds of reason I have outlined suffice to highlight the utility of talking
about “levels” in areas of science where plural, integrated compositional
models/explanations are common.
Part 7 – The New Philosophical Skepticism about Levels: Why Scale Fails as a
Replacement for Level in Scientific Practice
Though we now have a strong case that the concept of a “compositional level
plays a useful role in actual practice, some philosophers of sciences have argued that this
practice should be revised. A number of writers have recently advocated replacing the
concept of “level” with the notion of “scale”, including Potochnik ((2010), (2012)) and
Eronen ((2013), (2015)), amongst others. So we need to briefly assess this proposal.
When we seek a replacement for some successful scientific notion X, then the
main work one needs to do is to show how the broached replacement serves the purpose
of notion X just as well or better. One can therefore feel some sympathy for the recent
proponents of scale as a replacement for level because we have not had a clear idea of
either the nature of a level or the purpose levels serve for working scientists. And so it
has not been easy to make the required case that scale serves the purpose of level just as
well or better.28
Fortunately, our earlier work about the nature and purpose of levels now means
we can more carefully assess whether scale can indeed replace level. We have now seen
that “level” plausibly refers to the compositional arrays to which integrated
compositional explanations/models are committed. And we saw that talking of “levels”
allows scientists to efficiently communicate about arrays, and to track what entities are in
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
27 For example, Brooks and Eronen (2018) provide a plausible account of how a broader
notion of “level” frames problems and aids problem solving. This purpose is also
plausibly served by the narrower “levels” described by the Integrative Account.
28 This may explain why critics often focus on philosophical concepts of level (Brooks
(2017), rather than the notions of level found in scientific practice
Carl%Gillett,%Nov%‘19%
31"
arrays, amongst other useful roles. Focusing on these purposes that the notion of “level”
serves, we can thus better assess how well “scale” does as a replacement.
Unfortunately, our work highlights a couple of foundational reasons why “scale
fails as a successful replacement for “level” given the purposes of scientists. First, in 4.2,
we saw that proponents of the scale idea, like Eronen, highlight cases where we find
individuals of the same kind, and hence same scale, in different compositional arrays of
the same integrated compositional models/explanations. Consequently, the scale of some
entity does not always track which compositional array that entity is in and we outlined
the deeper reasons for this mismatch above in 4.2. So “scale” cannot successfully replace
the notion of “level” for scientists in tracking the commitments of their plural, integrated
compositional models/explanations.
Second, and more significantly, an entity being at some scale does not even
suffice for this entity to be a component of some terminal whole or to be in any
compositional array at all. As we have seen, a virus can be inside a cell, and be of the
same scale as its sub-cellular parts, but the virus is not a part of the cell at all, nor hence
in any compositional array. Why is that so? Plausibly because an individual is a part of
another, in the sense used in compositional models/explanations, when the individual’s
properties or activities realize or implement properties or activities of the relevant
whole.29 Simply being of a certain scale thus does not leave an individual in any
compositional array. The notion of “scale” is therefore too blunt a tool to track for
scientists whether an entity is even in a compositional array at all, let alone which of
them it is in.
Overall, we thus find that the concept of “scalefails for scientists in key roles
that the concept of “level” plays for them. And similar points can be used to show that
scale” also fails to allow scientists to communicate about compositional arrays as well,
again unlike “level”. But “scale” can only replace “level” in actual practice if “scale
serves the purposes of “level” as well or better -- so “scale” plausibly cannot replace
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
29 I noted the circumstantial descriptive evidence supporting this point in Part 2 above,
but for a more detailed defense see Gillett (2013), (2016), chapter 2, and (Unpublished),
chapter 7.
Carl%Gillett,%Nov%‘19%
32"
“level” in actual practice. This does not mean it is never useful to use the concept of
scale” as well as, or even instead of, the notion of “level”, since the purposes of
researchers are many. But it scuppers the claim that we can replace level with scale in
scientific practice, hence undercutting the Eronen-Potochnik proposal.
Conclusion
Compositional or organizational levels in the sciences can only be properly
appreciated once we finally begin to understand compositional models and explanations,
since the Integrative Account highlights how scientists use the term “level” to talk about
the complex ontological commitments of their plural, integrated compositional
models/explanations. I have sketched why scientists use the term “level” in this way for a
variety of practical reasons that cannot be duplicated by the concept of “scale” – hence
rebutting the recent Eronen-Potochnik proposal. Furthermore, the Integrative Account
frames precise ascriptions of same, higher and lower level in ways that capture actual
scientific practice, thus also rebutting the Kitcher-Schaffner objection to levels.
As we grapple with the broader scientific practices associated integration between
different areas of science, and their products, we now need to explore how the Integrative
notion of level, used with compositional models, is related to other notions of “level”,
including those used in these other areas of the sciences.30 For instance, evolutionary
biology use explanations, and practices, and associated notions like “levels of selection”,
that are often focused on historically individuated entities – and hence not the
ahistorically individuated entities of compositional models/explanations. However, an
interesting option to explore is whether the Integrative notion of a compositional level in
some way informs the notion(s) of “level” at play in evolutionary biology.31 One can only
be excited at such new research questions that are revealed once we take compositional
models/explanations seriously.
""""""""""""""""""""""""""""""""""""""""""""""""""""""""
30 Helpful starting points are Brooks an Eronen (2018) and Brooks (This Volume).
31 For discussion of closely related questions see DiFrisco (This Volume).
Carl%Gillett,%Nov%‘19%
33"
References
Aizawa, K. and Gillett, C. (eds.) 2016a: Scientific Composition and Metaphysical
Grounding. New York: Palgrave MacMillan.
Aizawa, K. and Gillett, C. 2016b: “Vertical Relations in Science, Philosophy and the
World”. In Aizawa and Gillett (2016a).
Aizawa, K. and Gillett, C. 2019: “Defending Pluralism about Compositional
Explanations”. Studies in the History and Philosophy of Science, Part C.
Amundsen, R. and Lauder, R. 1994: “Function without Purpose”. Biology and
Philosophy, v.9, pp.443-69.
Barberis, S. 2017: Multiple Realization, Levels, and Mechanisms”. Teorema, v.xxvi,
pp. 53-68.
Brooks, D. 2017: “In Defense of Levels: Layer Cakes and Guilt by Association”.
Biological Theory, v.12.
Brooks, D. This Volume: “’Levels of Organization’ as Tool and Doctrine in Biology”.
Brooks, D. and Eronen, M. 2018: “The significance of levels of organization for
scientific research: A heuristic approach”. Studies in History and Philosophy of
Biological and Biomedical Sciences, 68-69, pp.34-41.
Baumgartner, M. and Gebharter, A. 2016: “Constitutive Relevance, Mutual
Manipulability and Fat handedness”. British Journal for Philosophy of Science,
v.67, pp.731-56.
Betts, J.G. et al. (eds.) 2013: Anatomy and Physiology. BC Campus. Retrieved from:
https://openstax.org/books/anatomy-and-physiology/pages/1-introduction.
Craver, C. 2007: Explaining the Brain. Oxford: Oxford University Press, USA.
Craver, C. F., & Bechtel, W. 2007: “Top-down causation Without Top-down
Causes”. Biology and Philosophy, 20, pp. 715-734.
DiFrisco, J. This Volume: “Composition and Process in Levels of Developmental
Evolution”.
Eronen, M. I. 2013. “No Levels, No Problems: Downward Causation in
Neuroscience”. Philosophy of Science, 80(5), pp. 1042-1052.
Eronen, M. I. 2015: “Levels of Organization: A deflationary account”. Biology and
Carl%Gillett,%Nov%‘19%
34"
Philosophy, 30(1), pp. 39-58.
Giere, R. 2006: Scientific Perspectivism. Chicago: Univ. Chicago Press.
Gillett, C. 2002: “The Dimensions of Realization: A Critique of the Standard
View”. Analysis, v.62, pp.316-23.
Gillett, C. 2003: “The Metaphysics of Realization, Multiple Realizability and
the Special Sciences”. Journal of Philosophy, pp.591-603.
Gillett, C. 2007: “Hyper-Extending the Mind? Setting Boundaries in the Special
Sciences”. Philosophical Topics, v.351, pp.161-188.
Gillett, C. 2010: “Moving Beyond the Subset Model of Realization: The Problem of
Qualitative Distinctness in the Metaphysics of Science”. Synthese, vol.177,
pp.165-192.
Gillett, C. 2013: “Constitution, and Multiple Constitution, in the Sciences: Using the
Neuron to Construct a Starting Framework”. Minds and Machines.
Gillett, C. 2016a: Reduction and Emergence in Science and Philosophy. New York:
Cambridge University Press.
Gillett, C. 2016b: “The Metaphysics of Nature, Science, and the Rules of
Engagement”. In Aizawa and Gillett (2016a), pp.205-248.
Gillett, C. 2020: “Why Constitutive Mechanistic Explanation Cannot be Causal”.
American Philosophical Quarterly.
Gillett, C. Unpublished: Compositional Explanation and the Parts of Science.
Glennan, S. 2017: The New Mechanical Philosophy. New York: Oxford University
Press.
Godfrey-Smith,"P."2010:"“Causal"Pluralism”."In"Beebee,"Hitchcock"and"Menzies"(eds.)"
Oxford%Handbook%of%Causation,"pp."326-37."New"York:"Oxford"Univ."Press."
Healey, R. 2013: “Physical Composition”. Studies in History and Philosophy of Science
Part B: Studies in History and Philosophy of Modern Physics, v.44, pp.48-62.
Kaiser, M. 2018: “Individuating Part-Whole Relations in the Biological World”. In
Bueno, Chen and Fagan (eds.), Individuation Across Experimental and
Theoretical Sciences. New York: Oxford University Press.
Love, A. and Huttemann, A. 2011: “Comparing Part-Whole Reductive Explanations in
Biology and Physics”, pp.183-202, in Dieks et al (eds.), Explanation, Prediction,
Carl%Gillett,%Nov%‘19%
35"
and Confirmation. New York: Springer.
Machamer, P. 2004: “Causation and Activities”. International Studies in the Philosophy of
Science, v.18, pp.27 39.
Machamer, P., Darden, L. and Craver, C. 2000: “Thinking about Mechanisms”.
Philosophy of Science, vol.67, pp-1-25.
Mitchell, S. 2002: “Integrative Pluralism”. Biology and Philosophy, v.17, pp.55-70.
Mitchell, S. 2003: Biological Complexity and Integrative Pluralism. New York:
Cambridge University Press.
Novikoff, A. 1945: “The Concept of Integrative Levels and Biology”. Science, v.101,
pp. 209-215.
Potochnik, A. 2010: “Levels of Explanation Reconceived”. Philosophy of Science,
77(1), pp. 59-72.
Potochnik, A. 2017: Idealization and the Aims of Science. Chicago: University of
Chicago Press.
Potochnik, A., and McGill, B. (2012). “The Limitations of Hierarchical Organization”.
Philosophy of Science, 79(1), pp. 120-140.
Romero, F. 2015: “Why there is not inter-level causation in mechanisms”. Synthese
v.192, pp.3731-55.
Schaffner,"K."1993:%Discovery%and%Explanation%in%Biology%and%Medicine.""Chicago:""
University"of"Chicago"Press."
Silverthorn, D. 2013: Human Physiology: An Integrated Approach. San Francisco:
Pearson.
Wimsatt, W. 1994: “The Ontology of Complex Systems: Levels of Organization,
Perspectives and Causal Thickets”. Canadian Journal of Philosophy, supp. vol.
20, pp.207-74.
!
Carl%Gillett,%Nov%‘19%
36"
Figure 1. The basis of the famous sliding filaments model of skeletal muscle contraction.
(Wikimedia commons image created by Raul654 distributed under CC-BY 3.0 license:
https://en.wikipedia.org/wiki/Muscle_contraction#/media/File:Skeletal_muscle.jpg)
Carl%Gillett,%Nov%‘19%
37"
Figure 2. A textbook diagram of the sliding filament model of muscle contraction and a
Dynamic compositional model. (From (Betts (2013), Ch.10, sec 10.3, Fig.1)
Carl%Gillett,%Nov%‘19%
38"
Figure 3. A textbook diagram of the composition of a skeletal muscle at tissue and
cellular levels in muscles cells/fibers, and hence an Analytic model of it. (From (Betts
(2013), Ch.10, sec 10.2, Fig.1)
Carl%Gillett,%Nov%‘19%
39"
Figure 4. At textbook diagram of a muscle cell/fiber at the organelle level, and hence an
Analytic model of it. (From (Betts (2013), Ch.10, sec 10.2, Fig.2)
!
Carl%Gillett,%Nov%‘19%
40"
!
!
Figure!5."Textbook"diagram"of"the"composition"of"a"myofibril"at"the"molecular"level,"
and"hence"an"Analytic"model"of"it."(From"(Betts"(2013),"Ch.10,"sec"10.2,"Fig.3)
!
Carl%Gillett,%Nov%‘19%
41"
Figure 6. Diagram of the combined ontological commitments of scientists when
incorporating their compositional models and models positing thin causal relations
between parts and wholes. Circles are individuals, thick horizontal lines are activities, the
thin diagonal line are thin causal relations and vertical hollow arrows are compositional
relations.
ResearchGate has not been able to resolve any citations for this publication.
Chapter
Full-text available
Scientific philosophers examine the nature and significance of levels of organization, a core structural principle in the biological sciences. This volume examines the idea of levels of organization as a distinct object of investigation, considering its merits as a core organizational principle for the scientific image of the natural world. It approaches levels of organization—roughly, the idea that the natural world is segregated into part–whole relationships of increasing spatiotemporal scale and complexity—in terms of its roles in scientific reasoning as a dynamic, open-ended idea capable of performing multiple overlapping functions in distinct empirical settings. The contributors—scientific philosophers with longstanding ties to the biological sciences—discuss topics including the philosophical and scientific contexts for an inquiry into levels; whether the concept can actually deliver on its organizational promises; the role of levels in the development and evolution of complex systems; conditional independence and downward causation; and the extension of the concept into the sociocultural realm. Taken together, the contributions embrace the diverse usages of the term as aspects of the big picture of levels of organization. Contributors Jan Baedke, Robert W. Batterman, Daniel S. Brooks, James DiFrisco, Markus I. Eronen, Carl Gillett, Sara Green, James Griesemer, Alan C. Love, Angela Potochnik, Thomas Reydon, Ilya Tëmkin, Jon Umerez, William C. Wimsatt, James Woodward
Chapter
Full-text available
Scientific philosophers examine the nature and significance of levels of organization, a core structural principle in the biological sciences. This volume examines the idea of levels of organization as a distinct object of investigation, considering its merits as a core organizational principle for the scientific image of the natural world. It approaches levels of organization—roughly, the idea that the natural world is segregated into part–whole relationships of increasing spatiotemporal scale and complexity—in terms of its roles in scientific reasoning as a dynamic, open-ended idea capable of performing multiple overlapping functions in distinct empirical settings. The contributors—scientific philosophers with longstanding ties to the biological sciences—discuss topics including the philosophical and scientific contexts for an inquiry into levels; whether the concept can actually deliver on its organizational promises; the role of levels in the development and evolution of complex systems; conditional independence and downward causation; and the extension of the concept into the sociocultural realm. Taken together, the contributions embrace the diverse usages of the term as aspects of the big picture of levels of organization. Contributors Jan Baedke, Robert W. Batterman, Daniel S. Brooks, James DiFrisco, Markus I. Eronen, Carl Gillett, Sara Green, James Griesemer, Alan C. Love, Angela Potochnik, Thomas Reydon, Ilya Tëmkin, Jon Umerez, William C. Wimsatt, James Woodward
Article
Full-text available
The concept of 'levels of organization' has come under fire recently as being useless for scientific and philosophical purposes. In this paper, we show that 'levels' is actually a remarkably resilient and constructive conceptual tool that can be, and in fact is, used for a variety of purposes. To this effect, we articulate an account of the importance of the levels concept seen in light of its status as a major organizing concept of biology. We argue that the usefulness of ‘levels’ is best seen in the heuristic contributions the concept makes to treating and structuring scientific problems. We illustrate this with two examples from biological research.
Article
Full-text available
Despite the ubiquity of “levels of organization” in the scientific literature, a nascent “levels skepticism” now claims that the concept of levels is an inherently flawed, misleading, or otherwise inadequate notion for understanding how life scientists produce knowledge about the natural world. However, levels skeptics rely on the maligned “layer-cake” account of levels stemming from Oppenheim and Putnam’s defense of the unity of science for their critical commentary. Recourse to layer-cake levels is understandable, as it is arguably the default conception of levels in philosophy. However, relying on this conception of levels undermines the initial plausibility of general dismissals of the concept, because the problems skeptics identify within the “basic idea” of the concept “levels” are merely ones already widely acknowledged in this conception. I illustrate this “guilt by association” by looking at the embedded role of “levels” in articulating intertheoretical reductionism during the latter part of the 20th century. I conclude by suggesting a methodological framework focusing on local usage patterns as a promising means for future analysis of the levels concept.
Book
Grand debates over reduction and emergence are playing out across the sciences, but these debates have reached a stalemate, with both sides declaring victory on empirical grounds. In this book, Carl Gillett provides new theoretical frameworks with which to understand these debates, illuminating both the novel positions of scientific reductionists and emergentists and the recent empirical advances that drive these new views. Gillett also highlights the flaws in existing philosophical frameworks and reorients the discussion to reflect the new scientific advances and issues, including the nature of 'parts' and 'wholes', the character of aggregation, and thus the continuity of nature itself. Most importantly, Gillett shows how disputes about concrete scientific cases are empirically resolvable and hence how we can break the scientific stalemate. Including a detailed glossary of key terms, this volume will be valuable for researchers and advanced students of the philosophy of science and metaphysics, and scientific researchers working in the area.
Article
In his “New Consensus” on explanation, Wesley Salmon (1989) famously argued that there are two kinds of scientific explanation: global, derivational, and unifying explanations, and then local, ontic explanations backed by causal relations. Following Salmon’s New Consensus, the dominant view in philosophy of science is what I term “neo-Causalism” which assumes that all ontic explanations of singular fact/event are causal explanations backed by causal relations, and that scientists only search for causal patterns or relations and only offer causal explanations of singular facts/events. I argue that there are foundational, and fatal, flaws in the neo-Causal picture. The relations backing constitutive mechanistic explanations of activities of wholes using activities of parts, as well as other species of compositional explanation, cannot be causal relations. Treating them as causal or causation-like is therefore plausibly a category mistake. Compositional explanations in the sciences represent instead a sui generis kind of ontic explanation of singular fact/event backed by sui generis compositional relations. We thus need a pluralistic revision of Salmon’s New Consensus on explanation to reflect these findings.
Article
In the New Mechanist literature, most attention has focused on the compositional explanation of processes/activities of wholes by processes/activities of their parts. These are sometimes called “constitutive mechanistic explanations.” In this paper, we defend moving beyond this focus to a Pluralism about compositional explanation by highlighting two additional species of such explanations. We illuminate both Analytic compositional explanations that explain a whole using a compositional relation to its parts, and also Standing compositional explanations that explain a property of a whole using a compositional relation to the properties of its parts. We also highlight how adopting a Pluralism about compositional explanations justifies a more Pluralist view of the ontological posits of such explanations and opens up a range of new research questions.
Chapter
Compositional explanations in the sciences have been amazingly successful. Given their success, I outline a strategy focused on such compositional explanation that I term “Engagement” that I argue offers the best prospects for illuminating the compositional concepts of such explanations and also the metaphysics of nature itself. In contrast, I show the prominent philosophical accounts of verticality all fail to follow Engagement and hence provide inadequate accounts of verticality in compositional explanation and nature. I also detail how a minority approach in philosophy, that does follow Engagement, offers the most promising account of verticality in science and nature. My conclusion is that in the future, philosophers interested in the metaphysics of nature, and scientific composition, need to follow the rules of Engagement.
Article
What distinguishes good explanations in neuroscience from bad? This book constructs and defends standards for evaluating neuroscientific explanations that are grounded in a systematic view of what neuroscientific explanations are: descriptions of multilevel mechanisms. In developing this approach, it draws on a wide range of examples in the history of neuroscience (e.g., Hodgkin and Huxley's model of the action potential and LTP as a putative explanation for different kinds of memory), as well as recent philosophical work on the nature of scientific explanation.