Di Frisco Et Al. 2020

Biology & Philosophy (2020) 35:44
https://doi.org/10.1007/s10539-020-09762-2
Character identity mechanisms: a conceptual model

for comparative‑mechanistic biology
James DiFrisco1 · Alan C. Love2 · Günter P. Wagner3
Received: 4 May 2020 / Accepted: 17 July 2020 / Published online: 30 July 2020
© Springer Nature B.V. 2020
Abstract
There have been repeated attempts in the history of comparative biology to provide
a mechanistic account of morphological homology. However, it is well-established
that homologues can develop from diverse sets of developmental causes, appear-
ing not to share any core causal architecture that underwrites character identity. We
address this challenge with a new conceptual model of Character Identity Mecha-
nisms (ChIMs). ChIMs are cohesive mechanisms with a recognizable causal pro-
file that allows them to be traced through evolution as homologues despite having
a diverse etiological organization. Our model hypothesizes that anatomical units at
different levels of organization—cell types, tissues, and organs—have level-specific
ChIMs with different conserved parts, activities, and organization. Relying on a
methodology of conceptual engineering, we show how the ChIM concept advances
our understanding of the developmental basis of morphological characters, while
forging an important link between comparative and mechanistic biology.
Keywords Homology · Developmental mechanisms · Comparative biology · Cell

types · Characters · Levels of organization
* James DiFrisco
[email protected]
Alan C. Love
[email protected]
Günter P. Wagner
[email protected]
1
Institute of Philosophy, KU Leuven, Leuven, Belgium
2
Department of Philosophy, Minnesota Center for Philosophy of Science, University
of Minnesota, Minneapolis, MN, USA
3
Department of Ecology & Evolutionary Biology; Systems Biology Institute, Yale University,
New Haven, USA
13
Vol.:(0123456789)
44 Page 2 of 32 J. DiFrisco et al.
Wanted: a mechanistic account of morphological homology
Throughout the history of comparative biology, there have been repeated attempts
to provide a mechanistic account of morphological homology. However, these
attempts have foundered on a robust evolutionary pattern: homologous characters
develop from different and often diverse sets of causes, appearing not to share any
core causal architecture that underwrites trait identity (de Beer 1971; Spemann
1915). This conundrum is related to a longstanding question in evolutionary
biology: what is a trait? (Wagner 2001) As Gould and Lewontin (1978) argued,
adaptationist reasoning often involves an initial step of decomposing an organism
arbitrarily into traits thought to be individually optimized and explaining their
modification in a population in terms of natural selection. Yet it is often crucial to
understand the developmental mechanisms that control a trait in order to explain
how it can be modified evolutionarily. In many cases, what appears to be a trait
is only a by-product or intersection of genuinely individualized traits. The human
chin, for example, is not a distinct, individualized character, but rather a side
effect of the relative sizes of the alveolar and mandibular growth fields (Gould
1977; Coquerelle et al. 2013).
A mechanistic account of morphological homology would contribute substan-
tially to a developmental understanding of trait identity. This type of account
would address how traits can have an individuality that makes them traceable in
evolution, despite varying significantly across taxonomic groups and being inter-
mittently present across generations. To advance this comparative-mechanistic
research program, we introduce the idea of “Character Identity Mechanisms”
(ChIMs). ChIMs are cohesive units with a recognizable mechanistic architecture
that is traceable through evolution even though this architecture can be multiply
realized and exhibit diverse etiological organization.
The challenges from multiple realization and diverse etiologies are severe.
Empirical studies in evolutionary developmental biology have strengthened the
case that homologous traits are induced by different signaling pathways in differ-
ent species or arise from different germ layers and cell populations. Many appeals
to conserved developmental mechanisms shared among homologues have been
undermined by extensive evidence for developmental system drift (DSD)—the
evolutionary rearrangement of components and processes underlying homolo-
gous characters (True and Haag 2001; Haag and True 2018). Some of the mech-
anisms underlying the development of morphological characters (body parts,
organs, and cell types) are unexpectedly variable among species, even in cases
where the homology of the resultant characters is not in doubt (Hall 1995, 2003;
Haag 2014). This phenomenon has encouraged theorizing about how organiza-
tional properties of organisms might account for morphological stability (despite
variation) and therefore identity of homologous traits (Müller 2003). At the other
extreme, some developmental mechanisms for metazoan traits such as eyes and
limbs—characters whose homology was and is very much in doubt—appear more
conserved than initially thought, exhibiting substantive similarity across wide
phylogenetic distances (so-called “deep homology”; Shubin et al. 1997, 2009).
13
Character identity mechanisms Page 3 of 32 44
In between these extremes are examples of conserved developmental mechanisms

that correspond to clearly homologous characters (Davidson and Erwin 2006;
Hobert 2008; Wagner 2007). The idea of ChIMs is situated in this intermediate
position on the spectrum between conservation and variability.
There are three groups of empirical facts that point towards the ChIM hypoth-
esis of a separate mode of developmental regulation that determines character
identity quasi-independently of character phenotype (see Wagner 2014 for full
details). The first relates to the paradoxical pattern of interspecific variation in
developmental mechanisms just described: some parts of a developmental control
network for a body part are highly variable even for clearly homologous charac-
ters, while other mechanisms are highly conserved even without a clear relation
to homologous body parts. Closer inspection of the examples suggests the exist-
ence of a core causal architecture that can be activated by variable inputs and
control the execution of character-specific developmental processes. This archi-
tecture is strongly associated with morphological homology and appears to be a
distinct aspect of the developmental process. Second, the most direct experimen-
tal evidence for the existence of ChIMs comes from the recognition that the role
of Ultrabithorax is to control character identity rather than phenotype (Deutsch
2005; Wagner 2007, 2014). The critical observation was that a knockdown
of Ultrabithorax in the beetle Tribolium led to the development of a forewing
(elytra) at the location of the hindwing. Third, detailed mechanistic accounts
from cell biology have shown that the cell phenotype is controlled by a flat hier-
archy of control genes that “micro-manage” the expression of realizer genes (e.g.,
Hobert 2008), and that this level of control is more conserved evolutionarily than
cell morphology (e.g., Brunet et al. 2016).
These three groups of empirical facts provide prima facie evidence for the exist-
ence of ChIMs: conserved mechanisms that have a distinctive biological role in trait
individuation. Their conservation has been explained by mutual interdependen-
cies of gene expression in positive feedback loops (Wagner 2007, 2014). However,
a narrow focus on gene regulatory networks (GRNs) does not fully address phe-
nomena such as DSD or elucidate other systemic organizational properties, both of
which involve heterogeneous features at multiple levels of organization. Our abstract
account of ChIMs establishes a more general theoretical basis for character trace-
ability that can accommodate a variety of molecular and cell biological mechanisms
contributing to character identity—differently constituted mechanisms with diverse
etiological organization—while facilitating empirical investigation (i.e., the accom-
plishment of tracing homology).
Philosophical analyses of homology have viewed these ideas primarily as vari-
ants of a theoretical perspective on developmental homology, which contrasts with
genealogical views of homology more predominant in phylogenetic systematics and
other areas of evolutionary biology (Bock 1973; Patterson 1988). The task has been
to understand how these two broad perspectives relate to one another, especially
whether a unified account is possible (e.g., Brigandt 2007; Currie 2014; Novick
2018; Ramsey and Petersen 2012; Suzuki and Tanaka 2017). To date, these discus-
sions have neglected the broader question of what counts as a morphological trait.
This is surprising because David Hull recognized more than fifty years ago that the
13
association between questions about traits and questions about homology was con-
ducive to philosophers making a contribution to biology.
A second instance in which philosophers could have been of some service to
biologists is afforded by the question—what is a character?… Closely associ-
ated with this problem is the question of homology—when are two instances
of a character to be considered instances of the same character and in what
sense [the] same?” (Hull 1969, 262).
In addition to this neglect of the character question by philosophers analyzing
homology, most analyses have tended to ignore the pragmatic importance of tracing
these developmental mechanisms empirically in comparative, experimental practice.
How does a homology concept enable practices of establishing continuity between
characters? Philosophical approaches that concentrate only on unifying different
perspectives on homology overlook the practical application of the concept.
We begin our analysis with a review of extant conceptual resources available for
a mechanistic model of homology to identify the outstanding needs for an adequate
conceptual framework, with special attention to traceability. Next, we distinguish
ChIMs from developmental mechanisms more generally and show how they can
provide a basis for the construction and individuation of traits, which illuminates
how homology is possible. With this framework in place, we describe how ChIMs
are realized at different levels of organization (cell types, tissues, and organs) and
then examine how different aspects of mechanisms—outcomes, parts, activities,
and organization—contribute to the traceability of ChIMs. In conclusion, we show
how our method of conceptual engineering yields novel biological hypotheses and
advances our understanding of the developmental basis of morphological characters.
Developmental approaches to homology: extant conceptual

resources and outstanding needs
Beginning in the 1980s, the introduction of genetic approaches to developmental

biology brought a rapid expansion in our understanding of the mechanistic basis of
trait development. This occurred in large part through the discovery of conserved
regulatory genes that played similar roles across wide phylogenetic distances in
the ontogeny of model organisms (e.g., eyeless/Pax6 in fruit fly and mouse; Hal-
der et al. 1995; Quiring et al. 1994). Two kinds of patterns coalesced in compara-
tive developmental biology as these genes were studied in non-model organisms.
On the one hand, increasingly fine-grained detail confirmed the conclusion of earlier
research that some of the mechanisms underlying the development of morphologi-
cal characters are highly variable among species, even in cases where the homology
of the resultant characters is not in doubt. This pattern was particularly striking in
species where these types of evolutionary changes were unexpected, such as vulva
development in nematode worms (Sommer 2012). On the other hand, there was the
discovery of highly conserved networks of gene regulation where homology of the
morphological characters was debated or highly suspect, such as the shared role of
distalless in appendage outgrowth and proximal specification by Meis1/2 across
13
vertebrates and insects (Mercader et al. 1999; Shubin et al. 1997) or shared genetic
enhancers involved in the development of fin rays and tetrapod digits (Gehrke et al.
2015; Nakamura et al. 2016). This second pattern led to the formulation of the con-
cept of “deep homology” (Shubin et al. 1997, 2009). Similar patterns of deeply con-
served regulatory network architecture were discovered in the specification of devel-
opmental characters, such as the endoderm germ layer or the heart field, and labeled
“kernels” by Davidson and Erwin (2006).
These robust developmental evolutionary patterns are in tension. The first sug-
gests incredible lability in the developmental basis of traits, even over short evolu-
tionary times, whereas the second pattern suggests the reverse: incredible stability
in the developmental basis of traits over very long evolutionary times. A synthetic
review of the evidence for these patterns offered a potential resolution of the tension
(Wagner 2007). The majority of variable mechanisms contributing to homologous
characters are found among signaling systems that initiate the development of body
parts. In contrast, many of the stable, phylogenetically deep mechanisms respond to
these inductive signals and control the expression of downstream genes that realize
the final phenotype of the character, such as the network of transcription factors and
cofactors that induce eye formation (Donner and Maas 2004). The variable mecha-
nisms consist mostly of signaling molecules, whereas the stable mechanisms consist
primarily of transcription factor genes and allied proteins.
This line of reasoning led to the Character Identity Network (ChIN) model (Wag-
ner 2007, 2014; Fig. 1). Early in development there are functionally redundant and
variable signaling inputs that convey information about cell fate. This is followed by
a middle layer of highly conserved GRNs that are activated by these signals. Finally,
downstream genes are turned on that realize the phenotype of the character. The lat-
ter layer also exhibits variability, which makes it possible to meet the shifting func-
tional demands posed by natural selection. In parallel, others proposed a strikingly
similar model—terminal selector regulons—for cell fate determination in the nerv-
ous system based on GRNs (e.g., Hobert 2008, 2011; Fig. 2). Subsequently, cell
Fig. 1 Character Identity

Networks superimposed on
an hourglass shape, in which
greater width represents greater
evolutionary lability (Wagner
2014)
13
Fig. 2 Terminal selector regu-

lons (from Hobert 2011)
type differentiation studies demonstrated a critical role for protein complexes asso-
ciated with transcription factors beyond the GRN in specifying cell-type identity.
These core-regulatory (transcription factor) complexes (CoRCs) were incorporated
into a revised model of cell type identity and evolution (Arendt et al. 2016).
Although the ChIN/CoRC models offer an explanation for the paradoxical pattern
of variable and conserved aspects of character development, empirical challenges
remain. Highly conserved cell–cell signaling networks, such as in the initial devel-
opment of limbs or the segment polarity network of arthropods, do not fit the predic-
tion of an early layer of variable cell fate signals. This challenge arises from the fact
that ChINs are envisioned as a network of transcription factor genes even though
multicellular systems typically include cell–cell signaling. Additionally, existing
models have ignored the role of non-coding RNA in gene regulation, which is likely
relevant for cell-type and character identity (Makeyev and Maniatis 2008). These
challenges and open questions suggest a more general model is needed, but there
also is the pragmatic role for a model to guide comparative investigation. Whatever
features are being evaluated to determine whether two traits correspond across indi-
viduals or taxa, they must be traceable in practice. The tracing practices of research-
ers reflect commitments both about what is being traced (the individuation of a unit)
and how to trace (the relevant features associated with unit identity) (Griesemer
2007, 2018).1 A simple metaphysical requirement of “continuity” is uninformative
because it cannot address either of these commitments.
Ideally, a model of character identity can supply criteria of traceability that
capture both what is being traced and how to establish sameness in comparative
1
The traceability of organisms through a life history in order to count them or lineages through a phy-
logeny as historical individuals have been analyzed extensively (e.g., Ghiselin 1987; Pepper and Herron
2008). However, the traceability of parts, characters, or mechanisms has been largely neglected (but see
Love 2018a, b). Traceability and tracking practices are relevant for diverse sciences, from tracing alleles
in a pedigree (transmission genetics) to tracking a comet through space and time (astronomy).
13
reasoning about homologues. This is especially salient because homologues have

to be traced through variations in their appearance and behavior, as well as across
discontinuous periods of observation (e.g., transience during ontogeny and across
generations, such as atavisms where characters reappear in a lineage). The trace-
ability of homologues facilitates the construction of trait phylogenies even when the
characters are highly dissimilar (e.g., the transformation of the jaw joint in reptiles
into ear ossicles in mammals).
Comparative reasoning about morphological structures prior to an evolution-
ary framing of the question of homology relied implicitly on tracing practices even
though there was not clarity about the process that could generate ordered traces. In
this period, empirical evidence for homology was limited to various criteria of simi-
larity, such as topological position or shared developmental precursors, which were
either agnostic or flexible as to the process generating those similarities (Owen 2007
[1849]). This meant there was not an independent justification for which similarities
were most informative or which traits should be candidates to stand in relations of
homology. Some traits, such as size or shape, are not homologues and traceable only
as properties of individualized traits. Even after an evolutionary perspective came to
predominate, these criteria remained central (Remane 1952; Rieppel 1988). How-
ever, a partial justification for informative similarities emerged through cladistic
methodology, which facilitated testing candidate homologies among characters by
evaluating the consistency of different phylogenetic relationships based on implied
transformations. Yet this procedure suffered from ambiguities related to whether
phylogenetic relationships were sufficiently resolved to accomplish the testing (see
discussion in Richter 2005).
In the midst of these advances, little consideration has been given to the trace-
ability of the mechanistic constituents of homologues. DSD suggests traceability of
mechanisms is problematic because constituents will not necessarily co-vary with
a homologue, and over-eager claims of character homology based on shared regu-
latory gene expression cast doubt on the value of tracing constituents to establish
homologous traits (Abouheif et al. 1997). However, the explanatory advances found
in ChIN/CoRC models encourage a renewed attempt at isolating traceable constitu-
ents of mechanisms whose similarities are independently justified. A more general
model that can explain the pattern of variable and conserved aspects of character
development could also provide an improved perspective on the developmental and
phylogenetic traceability of character mechanisms.
The causal profile of character identity mechanisms
How can characters be traceable in evolution given that they are only intermittently
present across generations and can vary significantly across taxonomic groups? The
idea of “Character Identity Mechanisms” (ChIMs) is the hypothesis that (1) there
is a general, recognizable mechanistic architecture in development that explains the
traceability of characters, and (2) these mechanisms are themselves cohesive units
that can be traced as homologues in evolution. In this section, we offer a basic char-
acterization of ChIMs in development and evolution. The aim is to differentiate
13
ChIMs from the broad and heterogeneous category of “developmental mechanisms”

(see Love 2018a, 2020), while also capturing what is common to ChIMs across
different kinds of characters at different levels of organization (cell types, tissues,
and morphological structures). What follows is not a set of necessary conditions or
defining features that must hold for every ChIM and character, but rather a “causal
profile” or syndrome of properties that tend to be co-instantiated due to their inter-
dependencies (cf. Pepper and Herron 2008; Wagner 2014).
ChIMs are modular due to factors like positive feedback between components,
which contribute to unit integrity through internal cohesion, and negative interac-
tion or inhibition exercised on alternative physiological states, cell fates, or develop-
mental outcomes, which maintain the distinctness of the module. This is a generali-
zation of what makes ChINs modular: cross-regulatory connections between genes
that sustain each other’s expression and block the expression of alternative regula-
tory genes. Not only are ChIMs modules themselves but they also exhibit complex
organization, whereby heterogeneous components of different types are arranged to
yield activities that are not a simple consequence of primary elements (e.g., “master
control genes”). The basis of modularity in ChIMs (internal integration and main-
tenance of distinctive states) combined with complex organization means that the
activities of ChIMs are often necessary for particular functional outcomes. When a
characteristic activity is blocked (e.g., via knockout perturbation), the morphological
character or cell type associated with the ChIM either does not manifest or develops
into something else. The central role of a ChIM in triggering the morphogenesis or
differentiation of a character or cell type tends to be causally non-redundant: other
causes do not influence the effect in the same way.2 Just as a transcription factor dis-
tinctively binds to specific DNA sequences in coordination with sets of cofactors, so
also a ChIM yields a morphological outcome in coordination with contextual signals
from other mechanisms operating during ontogeny. Unlike the ChIM, these con-
textual signals are more functionally redundant, operating in a graded fashion, and
multiple factors can play the same causal role. Similarly, downstream effector and
structural genes often have cumulative effects on phenotypic aspects of a character
like size, shape, or color (Boyle et al. 2017). Eyes can be initiated in diverse bodily
locations and vary tremendously in their character states (from light detecting spots
to a camera lens), but their identity as eyes is narrowly specified.
These four characteristic features of ChIMs (modularity, complex organization,
necessity, and non-redundancy) yield a number of consequences. First, ChIMs are
less replaceable in evolution than their upstream signaling inputs and downstream
effector mechanisms. Second, as a result, ChIMs have a high degree of burden
(Riedl 1978) or are generatively entrenched (Wimsatt 1986, 2015), which means
the developmental outcomes of ChIMs are subject to strong stabilizing selection
2
Causal non-redundancy is different from causal specificity (Woodward 2010), which concerns the
bijectivity of mapping between states of a pair of variables (X→Y) rather than the presence or absence of
other causes (Z→Y, W→Y,…) with similar effects. The causal relations between ChIMs and phenotypic
characters are often non-specific: given that activities of parts are often necessary for the operation of the
ChIM as a cohesive unit, perturbing them will tend to “turn off” the effect completely.
13
and that ChIMs therefore are more likely to be evolutionarily conserved than other
developmental mechanisms. In turn, this implies that the historical association
between ChIMs and homologous characters is more likely to be conserved than
historical associations between other developmental mechanisms and traits. Alto-
gether, these features in combination with their consequences support the idea that
a ChIM will be traceable within and across individuals, as well as across species,
through perturbations or substitutions of its components or other aspects. ChIMs are
relatively stable nodes between upstream causal features that are subject to DSD and
divergent downstream factors that are responsive to the changing demands of selec-
tion on resulting phenotypes. Phenomenologically, ChIMs are the central knot in
this “bow-tie” pattern of ontogeny—the mechanistic analogue of the conserved phy-
lotypic stage of the metazoan body plan depicted in the hourglass model (Duboule
1994; Raff 1996).
Although the ChIN model (Wagner 2007, 2014) exemplifies this core characteri-
zation of ChIMs, mediating between variable upstream inputs of positional informa-
tion or signaling molecules and variable downstream outputs of realizer or effector
genes via a complex organization of conserved regulatory genes, it is just one pos-
sible realization of a more general mechanistic architecture. The conceptual innova-
tion of a generalized ChIM model is twofold. First, the unit that accounts for the
traceability of characters—and that is itself traceable—is a mechanism rather than
a network or a specific set of components.3 As such, given the described features
and consequences of ChIMs, they often can be traced despite changes in their com-
ponent parts because the number and type of traceable features has increased. The
second conceptual innovation of the ChIM model is that it captures how the types
of mechanistic components and activities can differ for characters at different levels
of organization, whether they are regulatory genes for cell types or signaling centers
for morphological structures, while still exhibiting the same causal profile. These
two innovations considerably expand the scope of application for the model, provide
more detailed resources for using the model in biological practice, and indicate that
the model is not limited to a small set of lower-level causes and therefore not tacitly
committed to a form of molecular reductionism.
ChIMs for anatomical units at different levels
Multicellular life is structured into anatomical units at different levels of organiza-

tion, such as cell types, tissues, and organs.4 We posit that these anatomical units
have level-specific forms of ChIMs. For example, the molecular mechanisms of cell
type identity are different from, though related to, those for tissue types. The causal
3
We refer to a consensus viewpoint on “mechanisms” (Craver and Tabery 2016), though this explic-
itly includes a dynamical perspective (DiFrisco and Jaeger 2019; McManus 2012) and therefore does
not correspond to what has been termed “machine mechanisms” (Nicholson 2012). Further details are
worked out below (see “Tracing ChIMs in evolution: outcomes, parts, activities, and organization”).
4
Some units, like zooids, only apply to animals and plants that consist of the integration of multiple
organisms (e.g., polyps in the case of siphonophores).
13
profile delineated above allows us to understand ChIMs in terms of level-specific

modes of integration. To demonstrate this, we briefly discuss candidate ChIMs for
cell types, tissues, and organs.
Cell types
The cells of a multicellular organism fall into relatively well-defined groups of simi-
lar cells, often dedicated to the same kinds of functional tasks, which are the basis
for the cell type concept. Cell types exhibit differential gene expression, and their
identity is based on the ability to regulate different sets of effector genes. New chal-
lenges arose in comparative biology when it became clear that the same cell types
can change function through evolution and thus assume different phenotypes (Arendt
2008). This problem is identical to what had been long recognized for morphologi-
cal characters: homologous body parts can assume different shapes and functions in
evolution, but are still traceable as the same historical individuals (Remane 1952). A
potential solution to this conundrum is the three layer model of character develop-
ment (Fig. 1), with cell type identity being implemented through the activation of a
core GRN (terminal selector modules, kernels or ChINs), while the phenotype of the
character is determined by a downstream layer of genes (Wagner 2007, 2014). As
the phenotype is downstream of the identity mechanism, the functionally relevant
phenotype can be decoupled and thus evolve independently from cell type identity.
Hence the conceptual distinction between cell type identity (“character”) and cell
type shape and function (“character state”) finds its counterpart in two levels of gene
regulation: the core regulatory network and the effector genes.
The need for a more abstract model that mechanistically explains cell type iden-
tity beyond the ChIN model arose from the realization that different kinds of molec-
ular mediators are critical for cell type identity. Cell type-specific transcription factor
protein complexes (CoRCs) may be more relevant than the transcriptional network
regulating these genes (Wagner 2014; Arendt et al. 2016). In addition, there is the
potential role of non-coding RNA as regulatory agents within the cell (Makeyev and
Maniatis 2008). Another dimension of the implementation and maintenance of cell
type identity is autocrine signaling, where a cell secretes signaling molecules that
act on receptors of the same cell (e.g., Locker et al. 2004; Hemmingsen et al. 2013).
Autocrine signaling allows transient external signals to be replaced by signals origi-
nating from the cells themselves, and is an example of an activity that maintains the
modularity of a character by positive feedback (see above, “The causal profile of
character identity mechanisms”).
All these models are consistent with the basic idea that there is a level of gene
regulation that determines the developmental individuality of cells (i.e., enables dif-
ferential gene expression) without determining their specific phenotype. The dis-
sociation between identity mechanisms and phenotype facilitates the adaptation of
a cell type to different functional demands. The ChIM concept further emphasizes
that one cannot predict the specific molecular realization of such an identity-confer-
ring mechanism. This lesson was learned during the introduction of genetic meth-
ods into developmental biology, where it became clear that the same developmental
13
task (e.g., pattern formation) can have many different molecular and cell-biological
realizations (see “Developmental approaches to homology”). A simple and general
theory of development is impossible at the level of molecular mechanisms.
Another contribution of the ChIM concept is resolution of the discrepancy
between developmental cell lineages and phylogenetic relationships among cell
types (Arendt et al. 2016). The developmental origin of a cell can be traced in the
embryo with a variety of methods, and the developmental origin of a cell type can
be conserved and thus carry information about its evolutionary origin. However, the
same cell type can arise from different parts of the embryo or adult organism (see
Arendt 2008; Wagner 2014), and patterns of evolutionary origination often do not
match patterns of developmental origination (see Arendt et al. 2016). These facts
are particularly confusing if cell type identity is linked to ontogenetic origin rather
than gene regulatory state. If cell type identities are linked to the activation of spe-
cific cell type identity mechanisms—ChIMs—then their conservation and trace-
ability despite diverse cell lineage origins becomes intelligible, and the discrepancy
between ontogenetic and phylogenetic continuity becomes a non-issue.
Tissues
Tissues have long been recognized as a distinct unit of organization, being the defin-
ing subject of histology, and can be traced among distantly related organisms—for
example, cartilage in invertebrates is likely homologous to vertebrate cartilage (Hall
2015). Here we focus on tissue types, such as cartilage and bone, rather than kinds
of tissue organization like epithelium or mesenchyme. Epithelial organization of
cells belongs to many different kinds of tissues (e.g., lung, liver and kidney), and
almost all vertebrate tissues have a mesenchymal component. Like cell types, tissue
types tend to be associated with specific functions or classes of functions, such as
body support for skeletal tissue or information transmission for nerve tissue.
Compositionally, a tissue consists of a number of different cell types and an extra-
cellular matrix (ECM) (Ross and Pawlina 2011). The precise complement of cell
types and the amount and nature of the ECM are regulated to meet the functional
demands on the tissue. Skeletal muscle tissue consists mostly of multinucleated cells
specialized in contractility (myotubes). Bone is dominated by a mineralized ECM.
Tissue ChIMs control their local environment through the homeostatic maintenance
of a certain composition of cell types and ECM, and active exclusion of other cell
types, which gives tissues a degree of structural and functional modularity. In gen-
eral, tissues are composed of more than one cell type—usually four or five—but
there are exceptions.
Simple tissues
Cartilage consists of only one cell type, the chondrocyte, and a special form of ECM
that includes a typical fibrillar collagen and complex glucosaminoglycans (Fig. 3).
Vertebrate and invertebrate cartilage share many molecular and gene regulatory fea-
tures, and are likely homologous (Hall 2015; Tarazona et al. 2016).
13
Fig. 3 Cartilage, a simple tissue,

consists of one cell type—the
chondrocyte—and secreted
ECM. Cartilage controls its
local tissue environment by
the mutual maintenance of
chondrocytes and ECM, and the
exclusion of blood vessels
One perspective on cartilage is that the mechanisms of tissue identity coincide

with those of cell type identity (i.e., the identity of the chondrocyte). Why then think
about cartilage as a tissue type rather than just an aggregation of a cell type? One
reason is that cartilage takes control of or creates its own local environment by the
secretion of copious amounts of ECM. Another relevant facet of local environmen-
tal regulation is that chondrocytes live at very low oxygen tensions, which excludes
blood vessels and thus other cell types. This is an example of how tissue-level
ChIMs increase the modularity of the associated character (i.e., cartilage) by inhibit-
ing alternative developmental states.
The vertebrate body harbors different kinds of cartilage. Prototypical (hyaline)
cartilage is found at the end of long bones and forms the articulation surfaces. It is
dominated by large amounts of hyaluronic acid and chondroitin sulfate. Other forms
of cartilage are characterized by different ECM composition, like elastin in the elas-
tic cartilage of the outer ear pinnae. By implication, one might expect that different
kinds of cartilage are formed by different chondrocyte cell types, but we are not
aware that this has been demonstrated experimentally.
Composite tissues
Most tissue types are composed of more than one cell type. Ruslan Medzhitov and
colleagues have identified a minimal set of compositional units for tissue organiza-
tion that consists of four to five functional classes of cell types: parenchymatous
cells, fibroblast or stromal cells, endothelial cells, tissue-resident and/or tissue-spe-
cific macrophages, and sometimes an ancillary cell type (Meizlish et al. in press;
Fig. 4).
Parenchymatous cells, such as muscle or liver cells, are dedicated to the specific
functional role of the tissue. Sometimes they are associated with helper cells (i.e.,
ancillary cell types) that are themselves specialized to facilitate the function of the
parenchymatous cell (akin to an administrative assistant who relieves a manager
from routine tasks). For example, Schwann cells wrap around the axon of nerve cells
in the peripheral nervous system and aid in both action-potential transmission and
axonal maintenance.
Many tissues also have tissue-resident macrophages and, in some cases, highly
specialized (tissue-specific) macrophages. Macrophages are a type of white blood
cell that operates in the innate immune system (Kotas and Medzhitov 2015). Within
tissues, macrophages not only guard the tissue against infections but also monitor
13
Fig. 4 Outline of the general organization of a composite tissue, with cell–cell interactions maintaining
homeostasis of cell type composition. Re-drawn after a figure in Meizlish et al. in press with permission
tissue integrity and elicit an inflammatory response during an infection or other

threat by recruiting granulocytes and lymphocytes when integrity has been com-
promised (Medzhitov 2008; Okabe and Medzhitov 2016). In addition to generic
tissue-resident macrophages, some tissues contain cells phylogenetically related to
macrophages that have acquired specialized functions and identities. For instance,
cellular bone contains osteoclasts, a type of macrophage, which play an important
role in bone remodeling by removing bone substance, which is eventually replaced
by the ECM secreted by osteoblasts. Other examples include the Kupffer cells of
the liver, alveolar macrophages in the lung, and the microglia of the central nervous
system.
Another common cell type constituent of tissues is the fibroblast, which is the
main source of collagen that provides structural support for a tissue. Fibroblast cells
also play a role in monitoring tissue integrity and can elicit inflammatory reactions
(Smith et al. 1997). Some tissues have tissue-specific fibroblast types, such as liver
stellate cells that store fat and vitamin A, and replace endothelial cells in the hepatic
blood sinus. They are activated like normal fibroblasts after tissue damage and can
lead to liver fibrosis. Finally, all vascularized tissues have endothelial cell types that
form the inner lining of blood capillaries and other blood vessels.
Composite tissues are organized and maintained by chemical and mechani-
cal communication among cell types to ensure the appropriate proportions of cells
of different types. For example, low oxygen in a tissue often leads to secretion of
vascular endothelial growth factor (VEGF), which stimulates the proliferation of
endothelial cells, increasing vascularization and thereby the availability of oxy-
gen (see Fig. 4). Similar feedback has been demonstrated between fibroblasts and
macrophages (Zhou et al. 2018). These feedback relationships are examples of how
13
Fig. 5 Complex composite tissues have additional scales of organization between cells and the tissue.
Left: kidney nephrons, the unit of structure and function in the kidney. Right: hepatic lobules, the unit of
structure and function in the liver
ChIMs actively maintain the modularity, integrity, and cohesiveness of their asso-
ciated anatomical units. Based on the these considerations, a candidate ChIM for
a composite tissue consists of the cell types that mutually maintain each other in
appropriate proportions via direct cell–cell signaling (Cerchiari et al. 2015; Zhou
et al. 2018), in combination with the ECM that aids in tissue function and exerts
control over the tissue-resident cell types in the space occupied by the tissue.
Composite tissues have different degrees of internal organization. Minimally,
we can recognize aggregative composite tissues as aggregations of cell types and
ECM with little or no internal spatial organization. Examples include bone tissue,
especially immature woven bone before remodelling, or smooth muscle tissue where
spatial organization is limited to smooth muscle cell orientation and distribution of
nerves and blood supply. Maximally, we can recognize complex composite tissues
with highly intricate internal structure, such as kidney tissue. Complex composite
tissues display additional scales of organization between the cellular level (inclusive
of the ECM) and the tissue level. This is observable when there are units of function
larger than the parenchymatous cell or ECM. For instance, the primary functional
unit of kidney tissue is the nephron, consisting of the glomerulus, the nephric cap-
sule, and several kinds of nephric tubules (Fig. 5). Kidney tissues of different sizes
have different numbers of nephrons. The functional unit for liver tissue is the liver
lobule, which consists of (1) hepatic cells that surround a blood sinus, (2) blood ves-
sels that bring the blood to and from the liver sinus, and (3) spaces that collect bile
secretion from liver cells and lead it to the bile ducts (Fig. 5).
The intricate internal organization of complex composite tissues makes the dis-
tinction between organs and tissues seem blurry. The difference between complex
tissues and organs is that complex composite tissues create spatial structures largely
due to local self-organization, such as mammary gland tissue driven by the affin-
ity of myo-epithelial cells and the basal membrane (Cerchiari et al. 2015). Though
13
spatially complex, the drivers of morphogenesis and tissue identity are still cell
types, their interrelationships, and relations with the ECM. Organs, the next higher
level of organization, require signaling centers for their spatial organization and are
usually composed of multiple different tissues. Thus, they are more flexible regard-
ing the range of final phenotypes that can be manifested compared to tissues struc-
tured largely via self-organization.
Although there is an emerging understanding of the development of aggregative
tissues (e.g., Cerchiari et al. 2015; Zhou et al. 2018), we are not aware of models
that explain the origin and maintenance of the spatially organized functional units
of complex composite tissues. The regenerative capabilities of these tissues provide
evidence of the existence of active maintenance mechanisms (i.e., ChIMs) that con-
trol their shape and function.
Organs
The identity of an organ manifests itself in control over the spatial organization of
tissues within its domain. This ability is achieved by the deployment of mutually
interdependent signaling centers, which involves both co-induction and co-mainte-
nance of their activity over certain periods of time. The signaling centers jointly
determine the identity as well as the location of downstream elements of the organ,
which differentiate within its spatial domain. To illustrate these points, we identify
the putative ChIMs for the development of limb and fin buds in vertebrates and early
differentiation of the chordate brain (Fig. 6).
Paired appendages of vertebrates—fins and limbs—develop from a local pro-
liferation of mesenchyme originating from the lateral plate mesoderm (Hinchliffe
and Johnson 1980). As the mesenchyme and its overlying ectoderm form a pocket
growing out from the embryonic body wall, four signaling centers emerge, of which
the best known are the Apical Ectodermal Ridge (AER) and the Zone of Polarizing
Fig. 6 Organ ChIMs control the tissue composition in their spatial domain. This abstract model can be
filled in with the examples of vertebrate limb/fin bud development or chordate brain differentiation (see
text). Tissue types are indicated as colored rectangles, and the signaling centers are circles with graded
signals emanating from them. Small bi-directional arrows indicate local paracrine interactions among
tissue types. Note that tissue types can vary over evolution in the same organ, pointing to the quasi-inde-
pendence of the ChIM from the phenotype of the organ
13
Activity (ZPA) (Zeller et al. 2009; Delgado and Torres 2017). The AER is a popula-
tion of ectodermal cells that forms at the boundary of the dorsal and ventral com-
partments of the body wall and is located at the tip of the limb bud. The AER signals
to the underlying mesenchyme to keep proliferating via FGF secretion. The ZPA is
a cluster of mesenchymal cells at the posterior edge of the limb bud, which signals
through the Sonic Hedgehog pathway and interacts with the posterior part of the
AER. The ZPA both maintains the signaling activity of the AER and is maintained
by the signals from the AER (Zeller et al. 2009). Together, they establish a 2D coor-
dinate system along the anterior–posterior and proximal–distal axes of the limb bud
within which the different parts of the limb differentiate. The third dimension of
the limb bud—dorso-ventral polarity—is determined by Wnt7A signaling from the
dorsal ectoderm and BMP signaling from the ventral ectoderm to the underlying
mesenchyme. It is responsible for differentiating the palmar side of the hand and
the back of the hand. The signaling system of AER, ZPA, and dorsal/ventral ecto-
derm is highly conserved and responsible for the spatial patterning of fins as well as
limbs (Mercader 2007). We propose that the paired gnathostome appendage ChIM
is constituted by these four signaling centers and their interdependent activities and
organization.
All vertebrate brains are composed of five fundamental building blocks, arranged
along the anterior–posterior extent of the neural tube (Nieuwenhuys 1994). These
are the forebrain, consisting of the paired telencephalon and the diencephalon, the
midbrain (mesencephalon), the metencephalon with the cerebellum as the most
prominent part in jawed vertebrates, as well as the myelencephalon or brainstem.
Before these subdivisions develop, the future brain regions are mapped out through
the activity of three signaling centers: the Anterior Neural Ridge (ANR), Zona Limi-
tans Intrathalamica (ZLI), and Midbrain-Hindbrain Boundary (MHB) (Butler 2000;
Rodrigues et al. 2009; Holland 2015). These centers are widely conserved and pre-
sent in basal chordates (e.g., amphioxus); thus, they were present in the most recent
common ancestor of all chordates, though there have been losses in some lineages
with a less organized central nervous system (Holland 2015). In addition, there is
evidence that these centers also exist in the skin of hemichordates, suggesting that
the evolutionary origin of these centers is related to the tripartite body organization
of early deuterostome animals (Pani et al. 2012). Hence, a vertebrate brain ChIM
would likely include these three signaling centers.
Again, the abstract nature of the ChIM concept allows us to hypothesize level-
specific parts of the ChIM, appropriate for the focal anatomical unit: transcription
factors for cell types, cell types for tissue types, and signaling centers for organs.
In contrast, insisting on a GRN as the basis of character identity in general would
not respect the different levels of organization that these anatomical units repre-
sent. Signaling centers in an organ primordium like the limb bud have an underly-
ing genetic basis, which means it is reasonable to expect conservation of elements
across multiple levels. However, a systematic correspondence of parts across all lev-
els is not a prediction of our model. Instead, the model predicts that ChIMs will be
conserved at the level of organization appropriate to the kind of anatomical unit they
control (cell type, tissue, organ). These candidate ChIMs are empirical hypotheses
that are open to revision based on new experimental findings. Alternative molecular
13
Table 1 Character identity mechanisms (ChIMs) for anatomical units at different levels of organiza-
tion (see text for details). Just as there are different kinds of traceable characters at different levels, the
ChIMs for those characters comprise level-specific parts, activities, and control outcomes. (Filled cells
are intended to be illustrative not exhaustive)
Anatomical Control outcome achieved by ChIM Parts of the ChIM Activities of the ChIM
Unit
Cell type Reaction norm in response to signals Genes (ChINs, cell Cross-regulatory activa-
and other environmental stimuli type identity net- tion or repression
Enabling differential gene expression work) Autocrine signaling
TF (CoRC)
Non-coding RNA
Tissue type Local environment of cell types and Cell types Cell–cell signaling
ECM within tissue ECM ECM production
Signaling molecules
Organ Spatial arrangement of tissues in a Signaling centers Mutually interdepend-
specific domain of the embryo ent signaling activities
among signaling
centers
realizations of the ChIM for cell types, tissues and organs are compatible with the
ChIM concept developed herein (Table 1).
Tracing ChIMs in evolution: outcomes, parts, activities,

and organization
Having laid out the causal profile of ChIMs and provided empirical illustrations at
different levels of biological organization, it is now necessary to investigate how
ChIMs can be traced as homologues in comparative biology. ChIMs provide multi-
ple “handles for homology” that correspond to the items broadly understood to com-
prise the concept of a mechanism: outcome, parts, activities, and organization.
Outcome
Mechanisms are mechanisms “for” something. The outcome-relative nature of

mechanisms is sometimes expressed by saying that mechanisms are “functionally
individuated” or “individuated by their phenomena” (Glennan 1996; Craver and
Tabery 2016). ChIMs are mechanisms for character identity: the outcome they are
supposed to explain is the individuality and distinctness of phenotypic characters
in development and evolution. As a consequence, criteria for tracing ChIMs can
include reference to the outcome of the mechanism’s operation. For cell types, the
outcome is a reaction norm that enables differential gene expression in response
to signals and other environmental inputs. For tissue types, the outcome is a stable
tissue organization with a tissue-specific distribution of cell types and ECM. For
organs, it is a coordinated arrangement of tissues within a region of the embryo.
13
Each of these outcomes can be traced ontogenetically and phylogenetically, and

therefore provide evidence in favor of the presence of a ChIM.
In characterizing the outcomes of ChIMs, it is important to emphasize a subtle
difference with other developmental mechanisms. The ChIM model is based on the
hypothesis that the mechanistic architecture of character identity is distinct from,
and upstream of, the mechanistic architecture of phenotypic differences generally.
For example, in insects, Ubx controls wing identity but not wing phenotype (see
Deutsch 2005; Wagner 2007, 2014). Perturbing the gene does not alter the wing phe-
notype—its size, shape, or histology—but it alters the wing identity (i.e., wing, hal-
tere, elytron) that is present in a given position (i.e., forewing, hindwing). Another
example of the distinction between character identity and character state comes from
cell types. Vertebral skeletal myocytes have a striated phenotype, whereas smooth
myocytes are non-striated and fibroblast-like. In Drosophila, however, the “smooth”
myocytes found in the gut have a striated appearance. Although these cells are phe-
notypically more similar to vertebral skeletal myocytes than smooth myocytes in
this respect, their identity as cell types, in terms of transcription factor control over
gene expression, is still that of a smooth myocyte (Brunet et al. 2016; Arendt et al.
2016). In this case, the execution of the striated muscle module came under the con-
trol of the smooth muscle ChIM. What ChIMs enable is the execution of a develop-
mental process that can differ phenotypically from what is happening in other parts
of the body or in other species. Hence, viewed from the perspective of phenotypic
differences, the biological role of a ChIM is to provide the conditions for a differ-
ence rather than a particular phenotypic difference. In some cases, however, ChIMs
do not have a completely distinct architecture, but also contribute to phenotypic dif-
ferences. The cell types and ECM that form ChIMs for tissues also are parts of the
tissue phenotype, as are signaling centers in an organ.
The outcome-relative nature of mechanisms raises an immediate obstacle for
the application of the homology concept. Homologues are, by definition, the same
“under every variety of form and function” (Owen 1843). Homologues, such as the
tetrapod limb, form lineages of characters that are traceable irrespective of evolu-
tionary changes in the function of the character (swimming, walking, flying, or dig-
ging). By contrast, if “mechanism” is conceptually tied to a functional role, then the
category of mechanisms would seem to be an inappropriate candidate for standing
in relations of homology (see Love 2018a).
One response to this problem is to note that there are multiple senses of “func-
tion” (see Wouters 2003; Griesemer 2006; Love 2007; Brigandt 2017) and not all
are in tension with homology. Following Wouters (2003), a “function” can be an
activity (what something does), a causal role (contribution to a system property,
activity, or capacity), a fitness contribution (the adaptive value of something), or a
selected effect (i.e., an effect targeted by natural selection) (see Love 2007, 695). The
latter two—fitness contributions and selected effects of a character—are excluded
as conceptual determinants of homology.5 However, activities can be homologues
5
Natural selection “sees” the character state but not the character. Stabilizing selection on a character
state indirectly stabilizes—though does not determine—character identity.
13
that are traceable either within ChIMs or in their own right (see “Activities” below).
Causal roles, the outcomes of a mechanism in a more encompassing system, would
be problematic if they were the only determinants of ChIM homology—then any
two mechanisms that cause homologous characters would be homologues, no mat-
ter how structurally divergent and regardless of their evolutionary history. Likewise,
mechanisms co-opted into new causal roles would not be homologous. However,
because ChIMs are traceable via other aspects of mechanisms (parts, activities, and
organization), these difficulties are avoidable.
In the context of the ChIM model, outcome-functions do not individuate ChIMs,
but rather serve to establish the phenomenon of character identity for which a mech-
anistic explanation is given. The ChIM model of character identity is the hypoth-
esis that a typical structure—developmental mechanisms that conform to the causal
profile—will perform a function or outcome. In this way, it is not susceptible to the
circularity or triviality that is sometimes present when functionally-defined con-
cepts are used for individuation or explanation (DiFrisco 2017). A prediction of the
ChIM model is that historical associations between ChIMs and characters will be
conserved. This is due to burden or generative entrenchment, as well as the apparent
difficulty of replacing one ChIM with a structurally distinct but functionally equiva-
lent one. Sameness of outcome therefore provides strong evidence for homology of
ChIMs and should be a reliable heuristic for tracing them through changes in parts,
activities, and organization.
Parts
Among the various elements of a ChIM, the parts are the most straightforwardly
evaluated determinants of ChIM homology—especially component genes and their
products. Homology between individual genes and proteins has consensus criteria
(Patterson 1988), and existing methods of molecular phylogenetics offer tools for
inferring phylogenies between gene groups, as well as between characters and spe-
cies (Yang and Rannala 2012). In the ChIN/CoRC models, the crucial elements are
gene components of ChINs, kernels, or terminal selector modules, as well as the
transcription factors of CoRCs, cis-elements, and protein interaction domains that
form the regulatory network core. From the point of view of mechanistic expla-
nation, much of the difference-making power of a developmental mechanism for
a given phenomenon derives from differences between the parts (e.g., base pair
sequence).
Although homology between the parts of a ChIM contributes substantially to
its traceability, ChIM homology cannot be based solely on homology of the parts.
ChIMs are especially conserved in evolution because of their mechanistic activity
rather than structural identity of components. The modularity of a ChIM is not the
default state of a set of components, but must be actively maintained by positive
feedback between internal components and negative feedback that suppresses alter-
native network behaviors and blocks interference from external elements. As long as
the appropriate activities and organization of the mechanism are maintained, com-
ponents can be replaced, especially over long evolutionary time scales.
13
An important example of component replacement is gene duplication and sub-

functionalization (Ohno 1970; Force et al. 1999; Crow and Wagner 2006; Wapinski
et al. 2007; Van de Peer et al. 2009), which initially increases the causal redundancy
of the duplicated gene and thereby allows for subsequent structural divergence.6 In
general, the key determinant of evolutionary lability for ChIM components is their
degree of non-redundancy. Although there is little direct data for this variable in
a wide sample of developmental systems, available evidence suggests different
degrees of lability for mechanistic components depending on the developmental
stage in which the mechanism is active. Investigations of the maximally conserved
“phylotypic stage” in different taxa have accumulated transcriptomic and genomic
data in support of the differential lability of regulators in different stages of devel-
opment (Kalinka et al. 2010; Piasecka et al. 2013). However, less is known about
lability and redundancy of ChIM components at higher levels of organization, such
as cell types and signaling centers. In general, where redundancy is expected to be
greater, more weight should be placed on tracing ChIMs in terms of their activities
and organization rather than components.
There are three additional reasons why ChIM homology cannot be based solely
on the homology of components. First, in comparing two regulatory networks, some
genes may be homologous while others are not. Without some consideration of the
activity or behavior of the gene network, there is no non-arbitrary way to individuate
networks into modules. Second, genes are routinely co-opted into novel roles (True
and Carroll 2002). A network or character that is newly regulated by a co-opted gene
shouldn’t automatically count as homologous with whatever the same gene previ-
ously regulated. Third, even if all of the component genes in a network are homolo-
gous, the same network can have a distinct regulatory organization, and—depending
on its cellular context and dynamical properties—can engage in different dynamical
behaviors with different phenotypic effects (Jiménez et al. 2017; DiFrisco and Jaeger
2019). The activities and organization that canalize a set of components into a spe-
cific behavior are indispensable to the developmental and evolutionary identity of a
ChIM.
Activities
ChIMs are not sets of parts correlated with sets of phenotypic outcomes, like sta-
tistical genotype–phenotype maps. Instead, they are dynamic entities and their
causal activity is what binds together specific mechanisms and characters in devel-
opment and evolution. Activities are not emphasized in many explanatory contexts
in biology. This is justifiable when there is a strict correspondence between the
activity, structural feature that performs it, and outcome, as in simple genetic dis-
eases or Mendelian traits. However, when the activity of a mechanism can vary in a
6
This case is more complicated, as duplicated genes are still traceable as paralogs and will likely
retain the same cis-regulatory and protein–protein interaction domains. But a sufficiently high degree
of sequence divergence can change these features and make the true gene phylogeny impossible to trace.
13
quasi-independent fashion from its structural parts, it is often necessary to treat the
activity as an independent causal variable (DiFrisco and Jaeger 2020). The same
result applies to treating activities as homologues within an encompassing ChIM.
When component genes in a ChIM change their structural identity, ChIM homology
may still be traced with respect to the activity of the mechanism. But how exactly
are activities traceable as homologues?
The inclusion of activities as determinants of ChIM homology connects to the
unresolved conceptual problem of “process homology.” Although we cannot engage
this problem here, it raises a central difficulty relevant to our discussion. Activities
appear to lack the individuality needed to make them traceable units in evolution.
This is why extant treatments of activity or process homology tend to run into a
dilemma: activities are individuated either in terms of (1) participant structure or
parts (Gilbert and Bolker 2001), or (2) functional outcome or developmental role of
the activity in a wider system (Manak and Scott 1994; see Love 2007). Though both
parts and outcomes can be relevant for assessing activity homology, activity homol-
ogy should not be defined in terms of either one alone. In the case of parts, activities
would not make ChIMs traceable through changes in the component parts. In the
case of outcomes, we would be unable to recognize and describe situations in which
the same activity gets co-opted into novel developmental roles. In both cases, activi-
ties would be collapsed into another element of mechanisms—parts or outcomes.
To address this difficulty, we need a distinction between simple and complex
activities. Simple activities are behaviors of individual components, such as a tran-
scription factor binding to a DNA segment. Complex activities, by contrast, result
from interactions between multiple components of the same mechanism, such as the
dynamical behavior of a GRN or coordinated cell and tissue changes in morpho-
genesis. Complex activities are based on the integration of multiple simple activi-
ties and have the property of organization, which may introduce complex causal
patterns such as nonlinearity. Due to their constitutive relationship, the designation
“simple” or “complex” is relative to a level and a grain of description. The activi-
ties of integral ChIMs are complex, but they result from the integration of simple
activities such as transcription factor binding and signal diffusion. In turn, simple
activities such as one gene repressing another can be decomposed into complex bio-
chemical activities. Simple activities are unlikely to have a meaningful degree of
quasi-independence in development and evolution from the parts that perform them;
complex activities, on the other hand, are more likely to exhibit quasi-independence
that is both ontogenetically and phylogenetically stable, such as the gene expression
regimes associated with cell type identity.
This distinction points toward a solution to the dilemma for activity homology
(Table 2). A simple activity can be traced via the identity of the part that performs it.
Homology of simple activities is therefore established via existing homology crite-
ria for the parts (e.g., sequence similarity for genes). Complex activities, by contrast,
can be traced via the identity of each or several of the aspects of the mechanism: the
parts, the activity itself, the organization, or the outcome (Table 2). When two com-
plex activities diverge in their parts (e.g., due to gene duplication or substitution),
the complex activity can be traced via its own intrinsic features, or via the organi-
zation or outcome of the ChIM. This makes it possible to trace complex activities
13
Table 2 Different types of ChIM activities are traceable via different aspects of mechanisms that pos-
sess distinctive criteria of homology. Because the activities of whole ChIMs are complex, the aspect of
organization becomes especially important for tracing ChIM homology (see text for details)
Activity type Traceable via Homology criteria
Simple Part Sequence similarity

Signal identity
Signaling center identity
Complex Outcome Morphological homology criteria, transitional forms
Parts Sequence similarity, special quality
Activity Dynamical regularity, morphological temporal pattern
Organization Complexity, topological position
through evolutionary changes in parts, unlike in Gilbert and Bolker’s (2001) model
of process homology. When two complex activities diverge in their outcome (e.g.,
due to being co-opted into a new role), they can be traced via the parts, organiza-
tion, or the activity itself. Since parts and organization are treated elsewhere, here
we examine how intrinsic features of a complex activity can be compared and traced
across divergent instances.
The intrinsic characteristics of activities can be represented by (1) variables and
parameters that are used to measure or register change, and (2) regularities or pat-
terns of change that groups of variables and parameters exhibit. As an example of
the latter, dynamical behaviors of networks of genes or signaling molecules can be
used to compare activities. In this case, formal tools are available for classifying
dynamical behaviors of networks, such as network motifs (e.g., feed-forward loops)
or dynamical regimes (e.g., oscillation or multistability). A good illustration is the
oscillatory dynamics of cyclical genes during vertebrate somitogenesis. Although
the individual cycling genes involved are highly divergent between mouse, chicken,
and zebrafish, the oscillatory behavior that guides the spatial patterning of somites is
conserved, as are certain signaling pathways (see Krol et al. 2011). A different kind
of regularity of activity that can be traced evolutionarily is morphogenetic changes,
where homology of activity can be established by applying existing morphological
criteria of homology not to structures or end-states (e.g., neural tube) but to entire
sequences of states leading to those end-states (e.g., neurulation).
Both kinds of activity—dynamical behaviors and morphological changes—are
currently theoretically underdeveloped as traceable homologues with distinctive
criteria. Fortunately, ChIM homology does not require putting much weight on the
inherent traceability of activities: there are no obvious examples of a ChIM activity
that varies quasi-independently of parts and organization and outcomes. Somitogen-
esis and neurulation have characteristic signaling pathways as well as outcomes of
somite and neural morphology.
ChIM activities vary across levels of organization. Cross-regulatory interactions
of activation of ChIM components and repression of alternative regulatory processes
have a modularizing effect at the level of transcriptional regulation. The active crea-
tion of boundaries between cell populations commonly involves mutual inhibition
13
motifs, as in the gap gene system in insect segmentation, which combine with posi-
tive feedbacks that “lock in” a cell population to a distinctive fate and yield spatial
patterning at the tissue level. Within tissues, a similar modularizing effect derives
from activities of cell–cell signaling that maintain specific proportions of cell types
and control the secretion of tissue-specific ECM. At the level of organs, control of
the spatial arrangement of tissues in an embryo is based on the activities of interde-
pendent signaling centers, which induce and maintain each other while also coop-
erating to determine the identity of downstream differentiating elements within the
organ.
Organization
Organization includes the qualitative connections between parts and activities (e.g.,
activation, repression), their internal spatiotemporal arrangement, and the external
context of a mechanism. The conceptual boundary between organization and com-
plex activities is not always straightforward—the interactions between genes in a
gene network can be viewed as the network organization or as a complex activity of
the network. However, it is useful to treat organization as a distinct aspect of ChIMs
to capture the fact that the same individual parts and activities in a different rela-
tional configuration can generate different phenotypic outcomes.
ChIMs are organized such that their parts interact in a non-aggregative fashion
(Wimsatt 1997). This requires heterogeneous part types and “modularizing” non-
linear feedback activities between them. Nonlinear feedback activities can include
a variety of causal dynamics, including self-organization among interacting parts.
These features distinguish the ChIM hypothesis from the “master control gene” par-
adigm, which attributes causation of complex phenotypes to individual genes. They
also distinguish the ChIM hypothesis from “genetic program” thinking, which con-
strues regulatory mechanisms as properties of the genome rather than the genome
within cell-biological and tissue forms of organization. Following the causal profile
of ChIMs, the property of organization helps to explain why ChIMs persist as dis-
tinct modules at higher levels of organization, and why they can be traced through
changes in the parts and activities even though those are likely to be conserved.
The organization of a gene network, such as a ChIN, is its regulatory topol-
ogy, considered separately from the identities of the constituent genes. Similarly,
in a ChIM for tissue identity, organization refers to the qualitative signaling con-
nections between cell types (e.g., activation, repression), the spatial configuration of
cell types, and the temporal sequence of signaling and secretory activities. Temporal
sequence is particularly important when the ChIM depends on changes in the organ
or tissue environment having “downward” effects on signaling dynamics and gene
expression, which may often be the case in the development of complex morpho-
logical structures (Salazar-Ciudad 2010).
Unlike parts and activities, organization is a property of ChIMs and not a trace-
able unit in its own right. However, organization is crucial to the tracing of ChIMs
due to its direct link to the traditional homology criteria of complexity and topol-
ogy. The use of complexity as a criterion of homology (Riedl 1978) is based on
13
the recognition that similarities in the “special quality” (structural details) (Remane
1952) of two characters can result from convergent evolution rather than inherit-
ance from a common ancestor. Similarities of structural detail therefore carry more
phylogenetic information to the extent that they are complex. Convergent complex
structures are unlikely to have evolved by means of the same structural modifica-
tions. These principles also apply when the structure is a developmental mechanism.
Two mechanisms might agree in their “special quality”—identities of parts and
activities—as a result of convergent evolution. The more complex the organization
of the mechanism, however, the less likely those similarities are due to convergence
rather than homology. These assessments depend on background assumptions about
the way evolution proceeds and mechanistic details of how the character in question
develops. At present, these details are better known for cell types than for tissues
and organs.
“Topology” refers to the position of a character relative to other characters,
including both its internal connections between parts and external connections to
the rest of the body plan. Early anatomical work by figures such as Geoffroy (1818)
and Owen (1866) found that classification schemes based on topological position
of characters were more orderly and inferentially reliable than classifications based
on form and function (Panchen 1994). It was only with the rise of developmental
genetics that the conserved positions of major body parts could start to be explained
mechanistically by factors like Hox genes. Similar to Hox genes, ChIMs are active in
specific regions of the developing embryo where they control the fate of certain cell
populations, sometimes by the active creation of histological boundaries. The rela-
tive position and developmental timing of a ChIM are therefore sound criteria to use
for tracing ChIMs across phylogenetic groups—with some qualifications.
These criteria have to be appropriately modified for serial homologues, which
often develop sequentially (e.g., somites, short-germ insect segments), and develop
in the “same” positions only in the sense that the positions are symmetrical relative
to the major body axes. Heterotopy and heterochrony alter the position and timing
of a ChIM, respectively, but the other criteria (related to internal organization, parts,
and activities) can permit tracing the ChIM through such changes. Identification of
a change as heterotopy or heterochrony is, in fact, dependent on successful tracing.
Summary and synthesis
The results of our analysis of how ChIMs can be traced in comparative biology are
summarized in Table 3. Although these results also can be applied to homology of
developmental mechanisms generally, rather than just homology of ChIMs, the for-
mer broad category has further issues related to criteria for “sameness of mecha-
nisms” (Love 2018a). Unlike mechanisms in general, ChIMs have more determinate
criteria of traceability and a theoretically-grounded propensity to persist in evolution
due to their distinctive causal profile in development.
One key advance of the ChIM model is that taking mechanisms as the central
units of developmental homology, as opposed to genes or GRNs, allows these
units to be traced as homologues despite changes or non-homologies in particular
13
Table 3 Summary of the Aspect of ChIM Homology criteria

contribution of different
aspects of mechanisms to the Outcome Morphological char-
traceability of ChIMs. When acter identity (cell
tracing changes in a ChIM in type, tissue, organ)
evolution, non-homology of
one aspect (e.g., parts) can be Parts Sequence similarity
compensated by homology in Signal identity
another aspect (e.g., activity, Cell type
organization) without disrupting Activities Regularity of dynami-
the individuality of the ChIM cal regime or mor-
phological temporal
pattern
Organization Complexity
Topological position
elements of the mechanism. Even with this notion of compensatory tracing, how-
ever, the ChIM model does not specify exactly how to weigh the different aspects of
conservation in a mechanism in order to evaluate whether particular cases are trans-
formational homologues. The extent of changes that are allowed and expected must
be informed by what is known about the evolutionary biology of the character and
mechanism under consideration. These material factors can inform applications of
the model in the context of local realizations of the causal profile of ChIMs.
For example, some mechanisms may exhibit more redundancy of components
than others, and in such cases more weight should be placed on tracing the ChIM
via activities, organization, and outcomes. Likewise, some characters may be more
highly burdened than others, and this may stabilize certain aspects of the mecha-
nism more than others. Both of these local differences are found in the case of ver-
tebrate somitogenesis (Krol et al. 2011). The genes that generate oscillations in
the presomitic mesoderm are highly redundant and have been replaced in different
vertebrate groups. The burden in this system, which arises in part from self-organ-
izational dynamics (Tsiairis and Aulehla 2016), seems inherent to the oscillatory
activity and spatiotemporal organization of the somitogenesis mechanism more than
its specific components. Somites are also an example of transient embryonic charac-
ters that are present before the definitive organs that characterize vertebrates. These
classes of characters may have different mechanistic architectures and propensities
for conservation (Wagner 2014).
Some traceable body parts, such as elements of the vertebrate vascular system,
may not be endowed with an identifiable ChIM (see Wagner 2014, 76ff.). While the
central parts of the vascular system like the heart and large arteries are controlled
by specific genes (such as Nkx2.5/tinman), as is the distinction between arteries and
veins (Red-Horse and Siekmann 2019), more peripheral but still conserved parts
may not be determined genetically (Poduri et al. 2017; Sharma et al. 2017). Instead,
many elements of the vascular system develop from a network of blood vessels that
are shaped by epigenetic factors, such as shear stress caused by blood flow and pres-
sure differences. In these cases, an anatomical unit that bears a name and can be
compared across divergent lineages does not have a specific ChIM. Such examples
13
can be captured by the “organizational” homology concept (Müller 2003), which

holds that homologous body parts are traceable in virtue of their being embedded
in a conserved and burdened body plan. However, body parts with organizational
identities are limited in their potential for independent evolutionary change, and thus
have limited developmental and evolutionary individuality in spite of their traceabil-
ity. Working out exactly how the causal profile of ChIMs is realized (or not realized)
in different kinds of characters—granting the characters different forms of traceabil-
ity—is an open problem for future work in comparative-mechanistic biology.
Conclusion: re‑engineering homology for finite beings
We have referred to the ChIM concept as a model and a hypothesis rather than an
account or conception of homology. This is intentional and reflects our investigative
methodology. The theme of character identity has all of the ingredients of a problem
that calls for conceptual engineering (sensu Wimsatt 2007; see Brigandt 2011; Wil-
son 2018) rather than conceptual analysis. The systems under scrutiny are evolved,
natural systems rather than constructed, formal systems. Variations, exceptions, and
counterexamples are to be expected and do not operate as the primary driver of con-
ceptual refinement. These systems also are historical, and therefore partly inacces-
sible to finite beings. They are not amenable to the epistemology of “in principle”
considerations or the image of cognitive agency that is not constrained by historical
situatedness and limited information. Finally, the systems in which character identity
is traced exhibit high levels of descriptive and interactional complexity (Wimsatt
1972). Understanding complex systems requires strategies and heuristics to reduce
complexity, such as idealized models. These strategies offer partial and incomplete
perspectives, which may require supplementation from other models and perspec-
tives. There is no expectation of an all-encompassing or definitive “account” of the
target phenomenon (i.e., homology or biological characters).
These highlighted features influence the appropriate methodology for concep-
tual inquiry as well as its conditions of adequacy. Conceptual engineering aims to
provide a characterization of the important features of a concept and their inter-
dependencies, targeting high-level theoretical insights or principles that can be used
to guide research at multiple resolutions of inquiry. The adequacy of conceptual
engineering is evaluated largely by the productivity of research programs it inspires
(Wagner 2014, 245) rather than by descriptive adequacy in handling all cases, real
or imagined. Definition and counterexample-based reasoning can play a role in the
engineering of scientific concepts, but it is neither the sole nor definitive arbiter of
theoretical adequacy (Brigandt and Love 2012).
The ChIM model developed in this paper exemplifies this methodology of con-
ceptual engineering. The model is intended to describe, explain, and predict aspects
of developmental homology, but not to re-define homology. It describes a general
mechanistic architecture of ChIMs by characterizing their most important features
and interdependencies (a “causal profile”), which is realized differently in differ-
ent anatomical units. It explains why particular elements in the mechanistic archi-
tecture of development are more conserved than others and, on the basis of this
13
interdependent architecture, accounts for how morphological characters can possess

an individuality that allows them to form traceable character phylogenies. These fea-
tures allow the model to predict patterns of conservation not yet observed, as well
as to generate hypotheses about the effects of experimental manipulations. Yet the
ChIM model does not propose a new definition or conception of homology that
competes with other homology concepts. The latter are still necessary for establish-
ing phylogenies of genes and characters that are explananda of the ChIM model. At
the same time, the dependence is not asymmetric because the ChIM model sheds
light on the preconditions for characters to form stable phylogenies, and also can
contribute evidence for phylogenetic hypotheses.
As a model, the ChIM concept makes use of several idealizations. One is the pos-
tulate that character identity has a distinct or distinguishable mechanistic architec-
ture from that for phenotypic differences. In some cases, the two are intermingled.
A related idealization is the assumption that ChIMs are highly modular and isolable
from other mechanisms. In reality, many mechanisms and networks may turn out to
be more context-dependent and interactionally complex, drawing on diverse devel-
opmental resources in different cases. These idealizations are themselves hypotheses
that can be investigated, potentially leading to revision or supplementation of the
model, such as a more precise characterization of its empirical domain of applicabil-
ity. Are there certain kinds of characters that conform to the ChIM model better than
others? An example of this sort of characterization is Wagner’s (2014) restriction
of the ChIN model to individualized morphological characters that only appear at a
certain stage of development (roughly, organogenesis), rather than transient embry-
onic characters or physiological and behavioral characters. Similar considerations
also could inform the differential weighting of different aspects of mechanisms for
tracing ChIMs. These open issues require more progress on the question “what is
a trait?”, which extends beyond the themes of developmental individualization and
homology.
Our analysis of the ChIM model builds on the general motivations for develop-
mental approaches to homology—that there are likely to be conserved developmen-
tal constraints, mechanisms, or information that underwrite the evolutionary conser-
vation of morphological characters (Spemann 1915; De Beer 1971; Riedl 1978; Van
Valen 1982; Roth 1988; Wagner 1989). Our hypothesis is that this developmental
substrate cannot be identified with a specific type of part (e.g., sets of transcription
factors), but has a dynamic and emergent (non-aggregative) nature that can be cap-
tured using the notion of mechanism. The factors that account for why a mechanistic
architecture is likely to stay conserved—the causal profile of ChIMs—permit this
architecture to have diverse etiological organization with different realizations at dif-
ferent levels of anatomical organization. This generates the expectation that different
aspects of a ChIM may be more or less conserved in different cases, depending on
their connection to that causal profile (e.g., non-redundancy and burden of parts ver-
sus activities or complexity of organization). Despite this potential to vary, ChIMs
are nonetheless traceable in evolution using specific criteria of homology. The prac-
tice of tracing ChIMs is more complicated than tracing individual genes, but this is a
reflection of developmental causation of morphological characters being realized at
a higher level of organization than individual genes (Wagner 2014).
13
The ChIM model also exemplifies how philosophical work on mechanism can be
useful when constrained to more specific investigative contexts than the general con-
trast of mechanisms versus non-mechanisms and the nature of explanation. ChIMs
are a subset of developmental mechanisms that have a well-defined biological role
and recognizable cluster of features. This is what makes ChIMs determinate enough
to play a descriptive role as traceable units of homology, in addition to the explana-
tory role commonly attributed to mechanisms. Crucially, having these dual descrip-
tive and explanatory roles is what enables ChIMs to serve as a bridge between
comparative and mechanistic biology. A mechanism that explains how a character
becomes individualized in development provides projectible insights about an evo-
lutionarily related character only if that mechanism is itself traceable in evolution.
This line of reasoning suggests that the ChIM model holds promise for estab-
lishing robust correspondence principles between developmental and evolutionary
patterns. Such principles have long been sought in biological theory, but typically
were frustrated by the fact that developmental and evolutionary changes frequently
fail to covary in a simple fashion. One response to this situation is to adopt a skepti-
cal empiricism that abandons any theorized correspondence between developmen-
tal and evolutionary characters across taxa. A more constructive path forward is to
investigate how variability in the correspondence between development and evolu-
tion might itself exhibit patterns. The ChIM model describes and predicts a general
pattern of this kind on the basis of underlying causal dependencies in development,
dependencies that maintain structures through the history of life—homologues—
while enabling the variation that fuels evolutionary processes of change.
Acknowledgements ACL and GPW gratefully acknowledge the financial support of the John Templeton
Foundation (Grant Number 61329). The opinions expressed in this paper are those of the authors and
not those of the JTF. JD thanks the Research Foundation—Flanders (FWO) for financial support (Grant
Number 41277) and for funding a research stay at Yale University in Spring 2020, where most of this
paper was written.
Funding ACL and GPW are supported by a John Templeton Foundation Grant, Number 61329. JD is
supported by the Research Foundation—Flanders (FWO), Grant Number 41277.
Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
References
Abouheif E, Akam M, Dickinson WJ, Holland PWH, Meyer A, Patel NH, Raff RA, Roth VL, Wray GA
(1997) Homology and developmental genes. Trends Genet 13:432–433
Arendt D (2008) The evolution of cell types in animals: emerging principles from molecular studies. Nat
Rev Genet 9:868–882
Arendt D, Musser JM, Baker CVH, Bergman A, Cepko C, Erwin DH, Pavlicev M, Schlosser G, Wid-
der S, Laubichler MD, Wagner GP (2016) The origin and evolution of cell types. Nat Rev Genet
17:744–757
Bock WJ (1973) Philosophical foundations of classical evolutionary classification. Syst Biol 22:375–392
Boyle EA, Li YI, Pritchard JK (2017) An expanded view of complex traits: from polygenic to omnigenic.
Cell 169:1177–1186
13
Brigandt I (2007) Typology now: homology and developmental constraints explain evolvability. Biol
Philos 22:709–725
Brigandt I (2011) Natural kinds and concepts: A pragmatist and methodologically naturalistic account.
In: Knowles J, Rydenfelt H (eds) Pragmatism, science and naturalism. Peter Lang, Frankfurt am
Main, pp 171–196
Brigandt I (2017) Bodily parts in the structure-function dialectic. In: Lidgard S, Nyhart LK (eds) Bio-
logical individuality: integrating scientific, philosophical, and historical perspectives. University of
Chicago Press, Chicago, pp 249–274
Brigandt I, Love AC (2012) Conceptualizing evolutionary novelty: moving beyond definitional debates.
Journal of Experimental Zoology B: Molecular and Developmental Evolution 318:417–427
Brunet T, Fischer AH, Steinmetz PR, Lauri A, Bertucci P, Arendt D (2016) The evolutionary origin of
bilaterian smooth and striated myocytes. eLife 5:e19607
Butler AB (2000) Chordate evolution and the origin of craniates: an old brain in a new head. Anat Rec
261:111–125
Cerchiari AE, Garbe JC, Jee NY, Todhunter ME, Broaders KE, Peehl DM, Desai TA, LaBarge MA,
Thomson M, Gartner ZJ (2015) A strategy for tissue self-organization that is robust to cellular het-
erogeneity and plasticity. Proc Natl Acad Sci USA 112:2287–2292
Coquerelle M, Prados-Frutos JC, Rojo R, Mitteroecker P, Bastir M (2013) Short faces, big tongues:
developmental origin of the human chin. PLoS ONE 8(11):e81287
Craver CF, Tabery J (2016) Mechanisms in science. In: Zalta EN (ed) The Stanford encyclopedia of phi-
losophy. http://plato.stanford.edu/archives/spr2016/entries/science-mechanisms/
Crow KD, Wagner GP (2006) What is the role of genome duplication in the evolution of complexity and
diversity? Mol Biol Evol 23:887–892
Currie AM (2014) Venomous dinosaurs and rear-fanged snakes: homology and homoplasy characterized.
Erkenntnis 79:701–727
Davidson EH, Erwin DH (2006) Gene regulatory networks and the evolution of animal body plans. Sci-
ence 311:796–800
De Beer GR (1971) Homology, an unsolved problem. Oxford University Press, London
Delgado I, Torres M (2017) Coordination of limb development by crosstalk among axial patterning path-
ways. Developmental Biology 429:382–386
Deutsch J (2005) Hox and wings. BioEssays 27:673–675
DiFrisco J (2017) Functional explanation and the problem of functional equivalence. Stud Hist Philos
Biol Biomed Sci 65:1–8
DiFrisco J, Jaeger J (2019) Beyond networks: mechanism and process in evo-devo. Biol Philos 34:54
DiFrisco J, Jaeger J (2020) Genetic causation in complex regulatory systems: an integrative dynamic per-
spective. BioEssays 42:1900226
Donner AL, Maas RL (2004) Conservation and non-conservation of genetic pathways in eye specifica-
tion. Int J Dev Biol 48:743–753
Duboule D (1994) Temporal colinearity and the phylotypic progression: a basis for the stability of
a vertebrate Bauplan and the evolution of morphologies through heterochrony. Development
(supplement):135–142
Force A, Lynch M, Pickett FB, Amores A, Yan YL, Postlethwait J (1999) Preservation of duplicate genes
by complementary, degenerative mutations. Genetics 151:1531–1545
Gehrke AR, Schneider I, de la Calle-Mustienes E, Tena JJ, Gomez-Marin C, Chandran M, Nakamura T,
Braasch I, Postlethwait JH, Gómez-Skarmeta JL, Shubin NH (2015) Deep conservation of wrist
and digit enhancers in fish. Proceedings of the National Academy of Sciences USA 112:803–808
Geoffroy Saint-Hilaire E (1818) Philosophie Anatomique, vol 1, Des Organes Respiratoires sous le Rap-
port de la Détermination et de l’Identité de leurs Pièces Osseuses. J. B. Ballière, Paris
Ghiselin MT (1987) Species concepts, individuality, and objectivity. Biol Philos 2:127–143
Gilbert SF, Bolker JA (2001) Homologies of process and modular elements of embryonic construction.
In: Wagner GP (ed) The character concept in evolutionary biology. Academic Press, San Diego, pp
437–456
Glennan S (1996) Mechanisms and the nature of causation. Erkenntnis 44:50–72
Gould SJ (1977) Ontogeny and phylogeny. Harvard University Press, Cambridge
Gould SJ, Lewontin R (1978) The spandrels of San Marco and the Panglossian paradigm: a critique of
the adaptationist programme. Proc R Soc Lond B 205:581–598
13
Griesemer J (2006) Genetics from an evolutionary process perspective. In: Neumann-Held EM,
Rehmann-Sutter C (eds) Genes in development: re-reading the molecular paradigm. Duke Univer-
sity Press, Durham, pp 199–237
Griesemer J (2007) Tracking organic processes: Representations and research styles in classical embryol-
ogy and genetics. In: Laubichler MD, Maienschein J (eds) From embryology to evo-devo: a history
of developmental evolution. MIT Press, Cambridge, pp 375–433
Griesemer J (2018) Individuation of developmental systems. In: Bueno O, Chen R-L, Fagan MB (eds)
Individuation, process and scientific practices. Oxford University Press, Oxford, pp 137–164
Haag ES (2014) The same but different: worms reveal the pervasiveness of developmental system drift.
PLoS Genet 10:e1004150
Haag ES, True JR (2018) Developmental system drift. In: Nuño de la Rosa L, Müller GB (eds) Evolu-
tionary developmental biology: a reference guide. Springer, Cham
Halder G, Callaerts P, Gehring WJ (1995) Induction of ectopic eyes by targeted expression of the eyeless
gene in Drosophila. Science 267:1788–1792
Hall BK (1995) Homology and embryonic development. In: Hecht MK, Macintyre RJ, Clegg MT (eds)
Evolutionary biology, vol 28. Plenum Press, New York, pp 1–37
Hall BK (2003) Descent with modification: the unity underlying homology and homoplasy as seen
through an analysis of development and evolution. Biol Rev Camb Philos Soc 78:409–433
Hall BK (2015) Bone and cartilage. Academic Press, Tokyo
Hemmingsen M, Vedel S, Skafte-Pedersen P, Sabourin D, Collas P, Bruus H, Dufva M (2013) The role
of paracrine and autocrine signaling in the early phase of adipogenic differentiation of adipose-
derived stem cells. PLoS ONE 8(5):e63638
Hinchliffe JR, Johnson DR (1980) The development of the vertebrate limb. Oxford University Press, New
York
Hobert O (2008) Regulatory logic of neuronal diversity: terminal selector genes and selector motifs. Pro-
ceedings of the National Academy of Sciences USA 105:20067–20071
Hobert O (2011) Regulation of terminal differentiation programs in the nervous system. Annu Rev Cell
Dev Biol 27:681–696
Holland LZ (2015) The origin and evolution of chordate nervous systems. Philosophical Transactions of
the Royal Society B: Biological Sciences 370:20150048
Hull D (1969) What the philosophy of biology is not. J Hist Biol 2:241–268
Jiménez A, Cotterell J, Munteanu A, Sharpe J (2017) A spectrum of modularity in multi-functional gene
circuits. Molecular Systems Biology 13:925
Kalinka A, Varga KM, Gerrard DT, Preibisch S, Corcoran DL, Jarrells J, Ohler U, Bergman CM, Toman-
cak P (2010) Gene expression divergence recapitulates the developmental hourglass model. Nature
418:811–814
Kotas ME, Medzhitov R (2015) Homeostasis, inflammation, and disease susceptibility. Cell 160:816–827
Krol AJ, Roellig D, Dequéant M-L, Tassy O, Glynn E, Hattem G, Mushegian A, Oates AC, Pourquié O
(2011) Evolutionary plasticity of segmentation clock networks. Development 138:2783–2792
Locker M, Kellermann O, Boucquey M, Khun H, Huerre M, Poliard A (2004) Paracrine and autocrine
signals promoting full chondrogenic differentiation of a mesoblastic cell line. J Bone Miner Res
19:100–110
Love AC (2007) Functional homology and homology of function: biological concepts and philosophical
consequences. Biol Philos 22:691–708
Love AC (2018a) Developmental mechanisms. In: Glennan S, Illari P (eds) The Routledge handbook of
the philosophy of mechanisms and mechanical philosophy. Routledge, New York, pp 332–347
Love AC (2018b) Individuation, individuality, and experimental practice in developmental biology. In:
Bueno O, Chen R-L, Fagan MB (eds) Individuation, process and scientific practices. Oxford Uni-
versity Press, Oxford, pp 165–191
Love AC (2020) Developmental biology. In: Zalta EN (ed) The Stanford encyclopedia of philosophy:
https://plato.stanford.edu/archives/spr2020/entries/biology-developmental/
Makeyev EV, Maniatis T (2008) Multilevel regulation of gene expression by microRNAs. Science
319:1789–1790
Manak JR, Scott MP (1994) A class act: conservation of homeodomain protein functions. Development
(supplement):61–71
McManus F (2012) Development and mechanistic explanation. Stud Hist Philos Biol Biomed Sci
43:532–541
Medzhitov R (2008) Origin and physiological roles of inflammation. Nature 454:428–435
13
Mercader N (2007) Early steps of paired fin development in zebrafish compared with tetrapod limb devel-
opment. Dev Growth Differ 49:421–437
Mercader N, Leonardo E, Azpiazu N, Serrano A, Morata G, Martínez C, Torres M (1999) Conserved
regulation of proximodistal limb axis development by Meis1/Hth. Nature 402:425–429
Meizlish M, Zhou X, Franklin RA, Medzhitov R (in press) Tissue homeostasis and inflammation. Annual
Rev Immunology
Müller GB (2003) Homology: The evolution of morphological organization. In: Müller GB, Newman SA
(eds) Origination of organismal form: beyond the gene in developmental and evolutionary biology.
MIT Press, Cambridge, pp 51–69
Nakamura T, Gehrke AR, Lemberg J, Szymaszek J, Shubin NH (2016) Digits and fin rays share common
developmental histories. Nature 537:225–228
Nicholson DJ (2012) The concept of mechanism in biology. Stud Hist Philos Biol Biomed Sci 43:152–163
Nieuwenhuys R (1994) Comparative neuroanatomy: place, principles, practice and programme. Eur J
Morphol 32:142–155
Novick A (2018) The fine structure of ‘homology’. Biol Philos 33:6
Ohno S (1970) Evolution by gene duplication. Springer-Verlag, New York
Okabe Y, Medzhitov R (2016) Tissue biology perspective on macrophages. Nat Immunol 17:9–17
Owen R (1843) Lectures on the comparative anatomy and physiology of the invertebrate animals. Long-
man, Brown, Green, and Longmans, London
Owen R (1866) On the anatomy of vertebrates, Volumes I-II. Longmans, Green, London
Owen R (2007 [1849]) On the nature of limbs: a discourse. University of Chicago Press, Chicago
Panchen AL (1994) Richard Owen and the concept of homology. In: Hall BK (ed) Homology: the hierar-
chical basis of comparative biology. Academic Press, London, pp 21–62
Pani AM, Mullarkey EE, Aronowicz J, Assimacopoulos S, Grove EA, Lowe CJ (2012) Ancient deuteros-
tome origins of vertebrate brain signalling centres. Nature 483:289–294
Patterson C (1988) Homology in classical and molecular biology. Mol Biol Evol 5:603–625
Pepper JW, Herron MD (2008) Does biology need an organism concept? Biol Rev 83:621–627
Piasecka B, Lichocki P, Moretti S, Bergmann S, Robinson-Rechavi M (2013) The hourglass and the early
conservation models: co-existing patterns of developmental constraints in vertebrates. PLoS Genet
9:e1003476
Poduri A, Chang AH, Raftrey B, Rhee S, Van M, Red-Horse K (2017) Endothelial cells respond to the
direction of mechanical stimuli through SMAD signaling to regulate coronary artery size. Devel-
opment 144:3241–3252
Quiring R, Walldorf U, Kloter U, Gehring WJ (1994) Homology of the eyeless gene of Drosophila to the
Small eye gene in mice and Aniridia in humans. Science 265:785–789
Raff RA (1996) The shape of life: genes, development and the evolution of animal form. University of
Chicago Press, Chicago
Ramsey G, Petersen AS (2012) Sameness in biology. Philos Sci 79:255–275
Red-Horse K, Siekmann AF (2019) Veins and arteries build hierarchical branching patterns differently:
bottom-up versus top-down. BioEssays 41:e1800198
Remane A (1952) Die Grundlagen des natürlichen Systems, der vergleichenden Anatomie und der Phylo-
genetik. Akademische Verlagsgesellschaft Geest & Portig K.G, Leipzig
Richter S (2005) Homologies in phylogenetic analyses—concept and tests. Theory Biosci 124:105–120
Riedl R (1978) Order in living organisms: a systems analysis of evolution. Wiley, New York
Rieppel OC (1988) Fundamentals of comparative biology. Birkhäuser Verlag, Berlin
Rodríguez-Seguel EP, Alarcón P, Gómez-Skarmeta. JL (2009) The Xenopus Irx genes are essential for
neural patterning and define the border between prethalamus and thalamus through mutual antago-
nism with the anterior repressors Fezf and Arx. Developmental Biology 329:258–268
Ross MH, Pawlina W (2011) Histology: a text and atlas. Wolters Kluwer, Philadelphia
Roth VL (1988) The biological basis of homology. In: Humphries CJ (ed) Ontogeny and systematics.
Columbia University Press, New York, pp 1–26
Salazar-Ciudad I (2010) Morphological evolution and embryonic developmental diversity in metazoa.
Development 137:531–539
Sharma B, Chang A, Red-Horse K (2017) Coronary artery development: progenitor cells and differentia-
tion pathways. Annu Rev Physiol 79:1–19
Shubin N, Tabin C, Carroll SB (1997) Fossils, genes and the evolution of animal limbs. Nature
388:639–648
13
Shubin N, Tabin C, Carroll SB (2009) Deep homology and the origins of evolutionary novelty. Nature
457:818–823
Smith RS, Smith TJ, Blieden TM, Phipps RP (1997) Fibroblasts as sentinel cells. Synthesis of
chemokines and regulation of inflammation. Am J Pathol 151:317–322
Sommer RJ (2012) Evolution of regulatory networks: Nematode vulva induction as an example of devel-
opmental system drift. In: Soyer OS (ed) Evolutionary systems biology. Springer, New York, pp
79–91
Spemann H (1915) Zur Geschichte und Kritik des Begriffs der Homologie. In: Chun C, Johannsen W
(eds) Allgemeine biologie (Vol 3). Teubner, Leipzig, pp 63–86
Suzuki DG, Tanaka S (2017) A phenomenological and dynamic view of homology: homologs as persis-
tently reproducible modules. Biological Theory 12:169–180
Tarazona OA, Slota LA, Lopez DH, Zhang G, Cohn MJ (2016) The genetic program for cartilage devel-
opment has deep homology within Bilateria. Nature 533:86–89
True JR, Carroll SB (2002) Gene co-option in physiological and morphological evolution. Annu Rev Cell
Dev Biol 18:53–80
True JR, Haag ES (2001) Developmental system drift and flexibility in evolutionary trajectories. Evolu-
tion & Development 3:109–119
Tsiairis CD, Aulehla A (2016) Self-organization of embryonic genetic oscillators into spatiotemporal
wave patterns. Cell 164:656–667
Van de Peer Y, Maere S, Meyer A (2009) The evolutionary significance of ancient genome duplications.
Nat Rev Genet 10:725–732
Van Valen L (1982) Homology and causes. J Morphol 173:305–312
Wagner GP (1989) The origin of morphological characters and the biological basis of homology. Evolu-
tion 43:1157–1171
Wagner GP (ed) (2001) The character concept in evolutionary biology. Academic Press, San Diego
Wagner GP (2007) The developmental genetics of homology. Nat Rev Genet 8:473–479
Wagner GP (2014) Homology, genes, and evolutionary innovation. Princeton University Press, Princeton
Wapinski I, Pfeffer A, Friedman N, Regev A (2007) Natural history and evolutionary principles of gene
duplication in fungi. Nature 449:54–61
Wilson M (2018) Physics avoidance: essays in conceptual strategy. Oxford University Press, Oxford
Wimsatt WC (1972) Complexity and organization. In: Schaffner KF, Cohen RS (eds) PSA 1972. D. Rei-
del, Dordrecht, pp 67–86
Wimsatt WC (1986) Developmental constraints, generative entrenchment and the innate-acquired distinc-
tion. In: Bechtel W (ed) Integrating scientific disciplines. Martinus Nijhoff, Dordrecht, pp 185–208
Wimsatt WC (1997) Aggregativity: reductive heuristics for finding emergence. Philos Sci 64:S372–S384
Wimsatt WC (2007) Re-engineering philosophy for limited beings. Harvard University Press, Cambridge
Wimsatt WC (2015) Entrenchment as a theoretical tool in evolutionary developmental biology. In: Love
AC (ed) Conceptual change in biology: scientific and philosophical perspectives on evolution and
development. Springer, Dordrecht, pp 365–402
Woodward J (2010) Causation in biology: stability, specificity, and the choice of levels of explanation.
Biol Philos 25:287–318
Wouters A (2003) Four notions of biological function. Stud Hist Philos Biol Biomed Sci 34:633–668
Yang Z, Rannala B (2012) Molecular phylogenetics: principles and practice. Nat Rev Genet 13:303–314
Zeller R, Lopez-Rios J, Zuniga A (2009) Vertebrate limb bud development: moving towards integrative
analysis of organogenesis. Nat Rev Genet 10:845–858
Zhou X, Franklin RA, Adler M, Jacox JB, Bailis W, Shyer JA, Flavell RA, Mayo A, Alon U, Medzhitov R
(2018) Circuit design features of a stable two-cell system. Cell 172:744–757
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.
13

Di Frisco Et Al. 2020

Uploaded by

Copyright:

Available Formats

Di Frisco Et Al. 2020

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Di Frisco Et Al. 2020

Uploaded by

Copyright:

Available Formats

Biology & Philosophy (2020) 35:44

Character identity mechanisms: a conceptual model

James DiFrisco1 · Alan C. Love2 · Günter P. Wagner3

Keywords Homology · Developmental mechanisms · Comparative biology · Cell

Wanted: a mechanistic account of morphological homology

In between these extremes are examples of conserved developmental mechanisms

Developmental approaches to homology: extant conceptual

Beginning in the 1980s, the introduction of genetic approaches to developmental

Fig. 1 Character Identity

Fig. 2 Terminal selector regu-

reasoning about homologues. This is especially salient because homologues have

The causal profile of character identity mechanisms

ChIMs from the broad and heterogeneous category of “developmental mechanisms”

ChIMs for anatomical units at different levels

Multicellular life is structured into anatomical units at different levels of organiza-

profile delineated above allows us to understand ChIMs in terms of level-specific

Fig. 3 Cartilage, a simple tissue,

One perspective on cartilage is that the mechanisms of tissue identity coincide

tissue integrity and elicit an inflammatory response during an infection or other

Tracing ChIMs in evolution: outcomes, parts, activities,

Mechanisms are mechanisms “for” something. The outcome-relative nature of

Each of these outcomes can be traced ontogenetically and phylogenetically, and

An important example of component replacement is gene duplication and sub-

Simple Part Sequence similarity

Summary and synthesis

Table 3 Summary of the Aspect of ChIM Homology criteria

can be captured by the “organizational” homology concept (Müller 2003), which

Conclusion: re‑engineering homology for finite beings

interdependent architecture, accounts for how morphological characters can possess

Compliance with ethical standards

You might also like