15206505, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/evan.21931 by Universidad Nacional Autonoma De Mexico, Wiley Online Library on [25/01/2023].

See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Received: 15 May 2021 Revised: 4 September 2021 Accepted: 19 November 2021
DOI: 10.1002/evan.21931

CLASSIC CONTRIBUTIONS

Leveling with Tinbergen: Four levels simplified to causes and

consequences

Thore J. Bergman1,2 | Jacinta C. Beehner1,3

1
Department of Psychology, University of
Michigan, Ann Arbor, Michigan, USA Abstract
2
Department of Ecology and Evolutionary In 1963, Niko Tinbergen published his foundational manuscript identifying the four
Biology, University of Michigan, Ann Arbor,
questions we ask in animal behavior—how does the behavior emerge across the lifespan
Michigan, USA
3
Department of Anthropology, University of (development); how does it work (mechanism); how and why did it evolve (evolution);
Michigan, Ann Arbor, Michigan, USA and why is it adaptive (function). Tinbergen clarified that these ‘levels of analysis’ are

Correspondence complementary, not competing, thereby avoiding many fruitless scientific debates.
Jacinta C. Beehner, Department of However, the relationships among the four levels was never established. Here, we
Psychology, University of Michigan, Ann
Arbor, MI 48109, USA. propose ‘leveling’ Tinbergen's questions to a single temporal timescale divided into
Email: jbeehner@umich.edu causes (encompassing mechanism, development, and evolution) and consequences
Funding information (encompassing function). Scientific advances now seamlessly link evolution, develop-
University of Michigan ment, and mechanism into a continuum of ‘causes’. The cause–consequence distinc-
tion separates the processes that precede (and lead to) a behavior, from the
processes that come after (and result from) a behavior. Even for past behaviors, the
functional outcomes are (historical) consequences of the causes that preceded them.

KEYWORDS
behavior, how questions, levels of analysis, proximate, ultimate, why questions

ON AIMS AND METHODS IN ETHOLOGY the immediate causes for a behavior; these are the physiological pro-
cesses (e.g., neural, muscular, and hormonal) that allow an animal to

BY Nikolaas Tinbergen express a behavior. These physiological processes are often triggered
by external stimuli and contexts, such as social and ecological cues.
(1963) Z Tierpsychol. 20: 410–433. The development level examines the ontogenetic changes that cause a
behavior; these are often the same (physiological) processes as the
mechanism level but at an earlier stage in the individual's lifetime, lay-
ing down the architecture (neural or otherwise) for supporting later
1 | I N T RO DU CT I O N behaviors. The evolutionary history level examines the trajectory of a
behavior across many generations and how that behavior has changed
Nearly 60 years ago, Nikolaas (‘Niko’) Tinbergen identified four differ- or has been maintained throughout a phylogenetic lineage. The func-
ent ways to correctly answer behavioral questions in his classic manu- tion level examines the fitness consequences of a behavior; these con-
script, On Aims and Methods of Ethology. Most students of animal sequences are primarily ‘seen’ by evolution when they contribute to
behavior today have internalized the foundational framework he differences in survival or reproductive success. Tinbergen taught us
established—variously called Tinbergen's four questions or that a full understanding of each instance of behavior requires more
Tinbergen's four levels of analysis (Figure 1). These levels give equal than one type of answer. Not only are the molecular geneticists and
honor to the pursuit of four different but complementary approaches the field biologists each doing valid science, their work synergistically
to ethology—each with its own methods and specialized knowledge— can be brought to bear on questions about animal behavior—allowing
making larger questions more tractable. The mechanism level examines us to answer old questions in new ways.

12 © 2021 Wiley Periodicals LLC wileyonlinelibrary.com/journal/evan Evolutionary Anthropology. 2022;31:12–19.

 15206505, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/evan.21931 by Universidad Nacional Autonoma De Mexico, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BERGMAN AND BEEHNER 13

F I G U R E 1 Tinbergen's four levels of

analysis, modified from Nesse (2013)2

At the time of Tinbergen's publication (1963), ethology was strug- often grouped as historical sequences (the short-term sequence of
gling to both define the scope of the field and to seek theoretical con- development and the long-term sequence of evolution) versus a slice-
nections between disparate research approaches. Although Tinbergen in-time (the underlying mechanism or function at the time of the
was focused on explaining behavior, the four questions help biologists behavior).2 Recently, Sapolsky narrated a scenario where we can iden-
explain any phenotypic trait. Modifying Julian Huxley's three major tify different ‘causes’ of a specific behavior by zooming in (to identify
1
problems for biology (causation, survival value, and evolution) and specific neurons firing) or zooming out (to identify early life develop-
sprinkling in a bit of Ernst Mayr's differentiation of how versus why mental effects on the individual) allowing us to view the causes of
questions,3 Tinbergen set a broad agenda for ethology, defining it as a behavior at different timescales. In this way, we are able to blur the
science that spans timescales from milliseconds to millennia and phys- line between what counts as developmental and what counts as
ical scales from molecules to biomes. mechanistic into one continuum.8 Similarly, the utility of the proxi-
Tinbergen's framework took root and has been the organizing mate/ultimate dichotomy has repeatedly been questioned.9,10
structure for studies of animal behavior ever since. The rise of integra- More problematic, researchers continue to confuse explanations
tive biology is a testament to the power of this type of thinking. By at different levels; they contribute an explanation at one level for a
clarifying that hypotheses at different levels are complementary (not question posed at another (we detail an example of this at the end).
competing), this classic manuscript has averted fruitless debates. For This occurs most commonly between ‘how’ and ‘why’ questions, with
example, we cannot ask do chimpanzees eat fruit because they find the people giving a how-answer (mechanism) to a why-question (func-
sweet taste rewarding (mechanism) or because it provides energy for sur- tion). Consider the following question: Why did the chicken cross the
vival (function)? These are not mutually exclusive explanations. An road? Because her legs carried her? Or, because she had to get away
explanation at one level cannot exclude a different explanation at from the farmer? Why questions can be answered correctly with both
another level. proximate and ultimate explanations, which means the how/why dis-
tinction is not overly helpful. This very ambiguity adds humor to the
why-do-chickens-cross-roads jokes. Sapolsky uses this question to
2 | P R O B L E M S A N D SO L U TI O N S open his popular book Behave,8 and he answers this with a narrative
of explanations that span from the evolutionary to the mechanistic—a
2.1 | First problem—there remains conceptual narrative that inspired us to reconsider the four levels in the first
ambiguity between ‘how’ and ‘why’ questions place.

Tinbergen's four levels have aged remarkably well. They remain very
influential, appearing in (and often organizing) most animal behavior 2.2 | Second problem—scientific discovery has
4,5
textbooks. However, there continues to be ongoing debate and broken down the boundaries between questions
confusion about the relationship among the four levels. Tinbergen did
not attempt to organize them in his original publication, and this has Innovative technology and scientific advances have eroded the tem-
left the topic open for debate. Most commonly, the levels are grouped poral boundaries between the evolution, development, and mecha-
as proximate (mechanism, ontogeny) and ultimate (function, evolution) nism levels. At the time of Tinbergen's publication, separating
explanations for behavior,6,7 which often are equated to ‘how’ (proxi- evolution from development was justified, but our current under-
mate) and ‘why’ (ultimate) questions. Additionally, the four levels are standing is far more sophisticated. Discoveries in evolutionary-
 15206505, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/evan.21931 by Universidad Nacional Autonoma De Mexico, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
14 BERGMAN AND BEEHNER

development and gene expression (e.g., epigenetics) have made it the effects of what happens influence survival (including
clear that within- and between-generational processes overlap. Simi- reproduction)?" The first question can be roughly divided
larly, at shorter time scales, our increasing temporal resolution for into three separate questions, differing in the time scale
measuring physiology and the brain has made it increasingly difficult involved.16
to separate developmental processes from more-immediate mecha-
nisms that cause a behavior to occur. We know that gene These “three separate questions, differing in the time scale
11 12 13
expression, developmental plasticity, and social experience can involved” are three of the four levels combined (evolution, develop-
produce permanent and irreversible brain organization that ‘cause’ ment, mechanism) in the yellow-orange arrow of Figure 2. This was
behaviors once animals are adults.14 We also know that other forms published in the author notes with a volume of Tinbergen's articles
of brain plasticity and hormonal regulation continue well into adult- and seems to have largely been lost (although see Shettleworth17). It
15
hood in temporary and reversible ways making it again difficult to is unfortunate that this conceptualization never caught on, while the
separate development from mechanism. much less clear proximate-ultimate grouping did.18 As we argue
below, ‘proximate’ explanations for behaviors may actually not be so
proximate after all, especially when we consider that gene regulation
2.3 | The solution—a two-level framework in one generation can be implicated in the behavior of their grand-off-
comprising cause and consequence spring.19 Rather than four separate time points (i.e., evolutionary past,
developmental past, immediately-preceding-the-behavior past, and
In the absence of a conceptual structure from Tinbergen, and to the following-the-behavior future), we can now think about a tempo-
refine, integrate, and extend conceptual arrangements suggested by ral continuum allowing researchers to zoom in or out for any single
others, we propose a simplified framework. Both the conceptual ambi- instance of a behavior to study the causes and consequences at dif-
guity and the breakdown of temporal boundaries can be solved by ferent time scales. Note that Figure 2 depicts a behavior once it has
moving to a two-level framework surrounding any single instance of a already occurred with hypothetical causes and consequences that
behavior (Figure 2): causes (encompassing mechanism, development, have emerged; the reader should keep in mind that the causes that
and evolution) and consequences (function). The cause-consequence contribute to (and the consequences that emerge from) any particular
distinction neatly separates the processes that precede (and, there- behavior are probabilistic (not deterministic) in nature.
fore, can lead to) a specific behavior, from the processes that come
after (and could possibly result from) the behavior. The moment the
specific behavior occurs separates causes from consequences. We 3 | CAUSES AND CONSEQUENCES
have leveled the four questions to a temporal timescale (before and
after the behavioral event). Indeed, Tinbergen himself proposed 3.1 | Causes
almost this exact scheme:
The two-level framework breaks down the barriers, not just between
I have always found it helpful to think of biology as con- the development and mechanism levels but also between an individ-
cerned…with two problems; that of causation and that of ual's lifetime and evolutionary history (Figure 2). Evolutionary history
function in the sense of survival value. By this I mean is the broadest timescale for how things came to be. Although it is
that…we ask "what makes this happen?" and "how do cross-generational, it maintains a connection from one generation to

F I G U R E 2 Tinbergen's four levels, reduced to a single temporal continuum separated into the processes that precede (and, therefore, can
lead to) a single instance of a behavior, and those that come after (and, could possibly, result from) this particular behavior. Examples of causal
processes and their approximate timescales are in gray below the arrows
 15206505, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/evan.21931 by Universidad Nacional Autonoma De Mexico, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BERGMAN AND BEEHNER 15

the next via direct genetic transmission. The entire causal chain con- in the field of animal behavior and is captured in Tinbergen's ‘func-
nects generations via genetic inheritance, connects neighboring gen- tion’ level of analysis (Figure 1).
20
erations via epigenetics, connects early life experiences to reactivity But, very often, we are interested in how natural selection in the
as an adult via developmental plasticity,21 connects something that past shaped the trait that we are observing now. This is not the pro-
happened that morning to a hormone state later that afternoon via cess of documenting the sequence of evolutionary steps that led to
regulatory changes,22 connects a hormone state that afternoon to a the trait, which is a simple, historical process that belongs squarely on
23
sensory neuron being more likely to fire, and so on—until we reach the cause side of the analysis. By contrast, if we are to understand
the shortest timescale (on the order of milliseconds) connecting a how natural selection operated in the past, we need to understand
motor neuron firing as a chicken dodges a farmer (Figure 2). This con- how the trait previously affected survival and reproduction. This
nection across timescales makes categorical thinking obsolete. Our belongs squarely on the consequence side of the analysis. When a trait
explanation for what caused the chicken to cross the road will simply evolves by natural selection, this means that, in previous generations,
depend on how far back in time we, as scientists, are willing to look. A the trait had a net fitness benefit, a consequence of the trait at that
cause at one less-proximate point in time is itself directly linked to a point in time. One of Tinbergen's greatest mistakes was not making
cause at a more-proximate point in time. the distinction clear between these two processes within his evolution
level. Tinbergen (1963, p. 428)1 described evolution as encompassing
both the evolutionary history (how animals got their forms) and the
3.2 | Consequences selection that shaped the trait (why evolution proceeded the way it
did). In short, Tinbergen wanted his evolution level to ‘do double-
The other side of the behavior—the aftermath—is a bit more difficult duty’, simultaneously calling upon this level to answer both cause and
to grasp conceptually (Figure 2, blue arrow). When we think about consequence questions about the past. Mayr3 attempts to reconcile
consequences, we are concerned with the effect of the trait, usually this with his term, ‘ultimate causation’, to refer to (what we call) the
in terms of how it relates to survival and reproduction. Logically, con- historical consequences of past selection. Ultimate causation is a prob-
sequences come after the behavior. And, they do. With current lematic concept that conflates past consequences with past causes.24
behavior, the current consequences can be measured and analyzed in To understand why this is problematic, an analogy might be use-
a relatively straightforward manner. This is the primary level of inquiry ful. Imagine that selection is a series of sieves with different-sized

F I G U R E 3 Tinbergen did not distinguish between cause and consequence processes with his evolution level. (a) A change in phenotype due
to the causal mechanisms of mutation, recombination, differential gene expression, and inheritance. (b) A change in phenotypic frequency due to
the consequences of selection
 15206505, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/evan.21931 by Universidad Nacional Autonoma De Mexico, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
16 BERGMAN AND BEEHNER

openings in the mesh. Each ‘selection’ sieve only allows stones to (e.g., a history of directional selection followed by stabilizing selec-
move on to the next sieve if they can pass through the openings in tion). Historical and current consequences are two separate, indepen-
the previous one. Starting with a jumble of different-sized stones, the dent questions. The comparative method is a powerful tool, but it
process ends up with a relatively homogenous pile of similar-sized does not identify the ‘ultimate causation’ of a behavior, it identifies
stones in each pile. If you reach into one of those piles and grab a the historical consequences.
stone, you could ask: Why is this stone this size? (Figure 3). Just as with Finally, it is critical to keep in mind that the fitness consequences
all ‘why’ questions about animal behavior, this can be answered in of a behavior are not always beneficial. Traits can be neutral—or even
two logically distinct ways; but one is a ‘how’ answer and one is a detrimental—to fitness. Another benefit of using the term ‘conse-
‘why’ answer. The ‘how’ answer satisfies the question: How did this quences’ (rather than the weighted terms ‘survival value’, ‘function’,
stone come to have this particular size? (Figure 3a). For this answer, we or ‘current utility’) is that the valence of this term leaves open the
need to know the history of the stone and the breakage and erosion possibility for positive, neutral, or negative effects.
events that caused it to arrive at the size it has. This is a causal ques-
tion about each individual stone that asks what events directly led to
the stone having its current form. This is equivalent to Tinbergen's 3.3 | Consequence ripples
how-animals-got-their-form side of the evolution level. In this case,
the filtering process is irrelevant. Alternatively, the ‘why’ answer sat- Generally, the environment (context, social, ecological, and physical) is
isfies the question: Why did I grab a stone of this size? (Figure 3b). For what determines the fitness consequences of a behavior. But, behav-
this answer, we need to know the process of stone-sorting that hap- iors can also alter the environment in ways that change future fitness
pened in the past. This is a population-level sampling question that consequences. This bi-directional relationship was termed ‘reciprocal
depends entirely on the size of the holes in the sieve and the sorting causation’ and can be seen in processes such as niche construction,
process. The output of the sorting process is a consequence of each coevolution, habitat selection, and cultural evolution.9 Certainly, the
stone's size. This is equivalent to Tinbergen's why-evolution- effect of behavior on the environment is an important part of the
proceeded-the-way-it-did side of the evolution level. Keep in mind, story. However, the term ‘reciprocal causation’ has the same prob-
this process tells us nothing about how each stone came to have its lems as the term ‘ultimate causation’—mainly, it is not causal at all.24
size. Natural selection (or drift, or any evolutionary filter) does not Phenotypes can (and do) cause changes in the environment, which
cause individual traits any more than the size of the holes in a sieve then alter the selective pressures on themselves (and other pheno-
cause the sizes of the stones that pass through them. It is possible for types); but this is very different from selection causing a phenotype.
“filters” to change the things that pass through them (like potato Rather, consequences that change the environment for future behav-
ricers) where the shape of the trait is caused by the filter, but we iors may be thought of as consequence ripples that extend forward in
know that is not how natural selection operates. In biology, variation time. Consequences are often ephemeral and are essentially reset
is caused by only two generative processes - mutation and recombina- with each instance of a behavior. For example, having escaped a lion
tion—which can variably be expressed through epigenetic processes yesterday has little bearing on your chances of escaping a different
that we simplify here as gene expression. Importantly, the consequence lion today. However, consequences can persist and alter future conse-
of the sorting (the traits that remain after population-level sorting) are quences. For example, the nest you build today continues to provide
not the cause for the trait. The sieve as an analogy is useful in that it nest-related benefits in the future, even for other individuals and
helps us understand the process of selection, but it is imperfect in that future generations. In sum, current behaviors change the future con-
it suggests a teleological process with a particular goal (size of stone) sequences of other behaviors (they do not cause those behaviors).
as the outcome. Natural selection is a filter but not a goal-directed
one; it emerges from an ever-changing environment.
Evolution by natural selection is an iterative process. The varia- 3.4 | Cause-driven and consequence-driven traits
tion that is present in one generation results, in part, from natural
selection in the previous generation. However, as soon as we start To further understand the difference between cause and conse-
describing processes in this way, we have moved from individual-level quence in the evolution of a trait, it helps to understand the two
thinking to population-level thinking.10 We have started to answer extremes—in what we will call cause-driven traits versus consequence-
the why-did-I-grab-a-stone-of-this-size question, which involves looking driven traits. Consider an evolutionarily stable trait. This trait does not
at the prior consequences of the behavior (i.e., the historical conse- change across large fluctuations in the environment, and it does not
quences, Figure 3b). Although current consequences of behaviors are respond to selection. In such cases, causal processes either (1) con-
the stuff of most animal behavior manuscripts, historical conse- strain variation in a trait, so there is no raw material for selection to
quences can also be studied but only indirectly using comparative act on,26 or (2) yoke variation in a trait to negative consequences of
methods to identify the selective pressures that shaped the trait in other traits, so any potential benefits of variations in the trait are
25
the past. Historical consequences of a behavior could be the same invisible to selection.27 In short, this trait has low evolvability.26 For
as current consequences (e.g., a history of directional selection that example, all primates have four limbs. There is no variation in this trait
continues in the present), but they may also be entirely different (four limbs), and it appears to be highly constrained across primates.
 15206505, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/evan.21931 by Universidad Nacional Autonoma De Mexico, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BERGMAN AND BEEHNER 17

Such traits are cause-driven because the trait we see today is driven in human stature in terms of estrogens and sexual selection are not
by the process of producing the trait (Figure 3a) rather than the pro- mutually exclusive but are answers to different questions.33 In short,
cess of sorting it (selection). Note that cause-driven traits still have support for a mechanism explanation cannot reject a functional one.
consequences (just as consequence-driven traits still have causes, see The cause-consequence framework can help clarify this debate in
below). To measure these consequences; however, we would need to two ways. First, the framework highlights the temporal relationship
be able to isolate and compare variants of the trait (which is often between a cause and a consequence making it clear why one can
impossible for the same reasons that natural selection cannot act on never be substituted for the other. The pattern of estrogens secretion
the trait). Five limbs might be better than four, but we have no way of precedes, and is therefore a potential cause of, adult stature (i.e., it
studying this. stops further growth). By contrast, sexual selection follows, and is
By contrast, a trait with high-evolvability—heavily shaped by nat- therefore a potential consequence of, the preexisting stature. Once
ural selection with a high degree of variation that maps onto the the adult stature is achieved and the phenotype is active in the envi-
environment—is a consequence-driven trait. The process that shaped ronment, the process of how that phenotype came about (the cause)
the trait (the cause) takes a backseat to understanding the exact form is largely invisible to the selection process (that yields the conse-
of the trait we see today. Instead, because there are so many variants quences). Any selection acting on the length of a giraffe's neck does
to sort (e.g., Figure 3b), it is the sorting process (selection) that is not ‘care’ if the neck is long because it has extra vertebrae or because
largely responsible for the traits that persist. For example (returning to it has longer vertebrae. Any selection acting on sex differences in body
the limbs of primates), while the number of limbs is invariable, the pro- size does not ‘care’ if men are taller than women because of differen-
portions of limb lengths across primates varies considerably. These tial estrogen secretion or (hypothetically) growth hormone secretion.
limb proportions covary with the locomotor style and habitat use of Even if we are able to reject the growth hormone hypothesis for why
different species in ways that are adaptive.28 The consequences from men are taller than women, this does not make the sexual selection
having different limb proportions in the past contributed to which hypothesis any more (or less) likely. The hormonal causes are entirely
limb proportions we see today. For consequence-driven traits with a orthogonal to testing the consequences of differential growth.
high degree of variability, this is where we need to address our adap- Dunsworth does make the important point that phenotypes do not
tive hypotheses. The extent to which phenotypes are shaped by always have an adaptive explanation.30 Certainly, traits are not always
selection (consequence-driven) remains an open question, and this is adaptive. They could emerge simply as a byproduct of another trait34
an active area of debate in evolutionary thinking.29 or by chance.35 But, such traits still produce consequences—adaptive,
neutral, or detrimental to fitness. This then raises the question, if traits
are not driven or maintained by natural selection, how do they persist
4 | H U M A N SE X D I F F E R E NC E S E X A M P L E over evolutionary time?
In cases where selection is unable to act on a trait, we consider
We end with an example of how this framework can clarify confusion these cause-driven traits (e.g., the four limbs present in all primates). A
about answers from different levels of analysis. A recent analysis, cause-driven phenotype is likely what Dunsworth30 is arguing for the
from this journal, addressed the question, why are there sex differences estrogens explanation for sex differences in human stature. This
in human stature?30 The author, Dunsworth, focused analyses on both would mean that human stature is largely a product of constraints
stature and pelvic shape, but for simplicity, we focus on the stature (e.g., relating to reproduction and estrogen secretion) rather than
question because the logic is the same. Dunsworth states that sexual selective consequences (e.g., relating to sexual selection). Like the four
selection explanations for sex differences in human stature (e.g., that limbs in primates, a cause-driven hypothesis predicts that stature
male competition favors the evolution of larger men, for example, dimorphism will show little variation from humans to apes to mon-
31
Puts ) have been over-emphasized in the story of human evolution. keys. Comparative data do not, however, support this prediction. Size
Instead, they propose that sex differences in human stature are due dimorphism is immensely variable both within humans and across pri-
to differential estrogen secretion (because estrogens fuse the epiphy- mates. Across primates, females are larger than males in some species
ses of long bones): and males more than three times the size of females in others, and
these differences closely map onto different social and mating sys-
For humans and likely other hominids, male skeletons tems.33,36 Contrary to a cause-driven hypothesis, these data suggest
continue to grow after females' stop because their bodies that differences in body size (across primates, and even across verte-
take longer to produce enough estradiol to surpass the brates) are enormously plastic and what we would consider to be con-
amount that stimulates continued growth and to achieve sequence-driven, with very high evolvability.
a level that closes long bone epiphyses,30 p. 111). Although support for the estrogens hypothesis explaining differ-
ences in human stature cannot be used to reject the sexual selection
They additionally state that the estrogens explanation means that hypothesis, the high evolvability in primate body size dimorphism
“the sexual selection perspective on male height seems unneces- actually supports Dunsworth's primary claim that sexual selection plays
sary”30 (p.110). Two published responses have already disputed the a reduced role in the recent history of humans. Indeed, other authors
logic of this approach,32,33 saying that explanations for sex differences have successfully argued using comparative datasets that sexual
 15206505, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/evan.21931 by Universidad Nacional Autonoma De Mexico, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
18 BERGMAN AND BEEHNER

selection, if anything, is very much relaxed in humans compared to [7] Alcock J. Sinauer Associates, Inc; 1975. Animal Behavior: An Evolu-
other closely related primates.36 tionary Approach. .
[8] Sapolsky RM. Behave: The Biology of Humans at Our Best and
Worst. Penguin; 2017.
[9] Laland KN, Sterelny K, Odling-Smee J, Hoppitt W, Uller T. Cause and
5 | C O N CL U S I O N effect in biology revisited: is Mayr's proximate-ultimate dichotomy
still useful? Science. 2011;334:1512-1516.
[10] Ariew A. Ernst Mayr's “ultimate/proximate” distinction reconsidered
Sapolsky warns us against categorical thinking, what he calls ‘thinking
and reconstructed. Biol Philos. 2003;18:553-565.
in bins’. We wholeheartedly agree—not just because thinking across [11] Arnold AP, Chen X. What does the “four core genotypes” mouse
bin boundaries is necessary for integrative science; but also because model tell us about sex differences in the brain and other tissues?
now that we understand so much more about the processes that con- Front Neuroendocrinol. 2009;30:1-9.
[12] Nettle D, Bateson M. Adaptive developmental plasticity: what is it,
tribute to behavioral outcomes, the bins themselves are confusing.
how can we recognize it and when can it evolve? Proc R Soc Lond B
Moving away from conceptual bins to a temporal continuum is more Biol Sci. 2015;282:20151005.
compatible with our current understanding of integrative biology. A [13] Champagne FA. Interplay between social experiences and the
temporal continuum makes a fundamental distinction between pro- genome: epigenetic consequences for behavior. Adv Genet. 2012;77:
33-57.
cesses that precede (and could cause) a phenotype and processes that
[14] Phoenix CH, Goy RW, Gerall AA, Young WC. Organizing action of
come after (and could be consequences of) a phenotype. Given the
prenatally administered testosterone propionate on the tissues medi-
iterative nature of natural selection, this distinction is particularly ating mating behavior in the female Guinea pig. Endocrinology. 1959;
important. Natural selection links cause and consequence because 65:369-382.
current fitness consequences determine which ‘causes’ persist into [15] Goldman SA, Nottebohm F. Neuronal production, migration, and dif-
ferentiation in a vocal control nucleus of the adult female canary
the future. Despite this link, population-level sorting processes (con-
brain. Proc Natl Acad Sci USA. 1983;80:2390-2394.
sequences) remain logically distinct from individual-level determinants [16] Tinbergen N. The Animal in its World: Explorations of an Ethologist,
(causes). Tinbergen himself said that there are only two problems in 1932–1972. Harvard University Press; 1972.
biology—that of causation and that of function.16 Therefore, we rec- [17] Shettleworth SJ. Function, causation, evolution, and development of
behavior: a review of the animal in its world, by N. Tinbergen. J Exp
ognize that a more appropriate title for our manuscript might have
Anal Behav. 1974;22:581-590.
been “Leveling along with Tinbergen…”, since we are simply advocat- [18] Sherman PW. The levels of analysis. Anim Behav. 1988;36:616-619.
ing what he first championed almost 50 years ago. [19] D'Urso A, Brickner JH. Mechanisms of epigenetic memory. Trends
Genet. 2014;30:230-236.
[20] Zambrano E, Martinez-Samayoa PM, Bautista CJ, et al. Sex differ-
ACKNOWLEDGMENTS
ences in transgenerational alterations of growth and metabolism in
We thank our graduate school professors (Garland Allen, Jim progeny (F2) of female offspring (F1) of rats fed a low protein diet
Cheverud, Cliff Jolly, Allan Larson, Jonathan Losos, Jane Phillips-Con- during pregnancy and lactation. J Physiol. 2005;566:225-236.
roy, Alan Templeton) for giving us a strong foundation in evolutionary [21] Ader R, Grota LJ. Effects of early experience on adrenocortical reac-
tivity. Physiol Behav. 1969;4:303-305.
biology; and we thank our postdoctoral mentors (Jeanne Altmann,
[22] Oyegbile TO, Marler CA. Winning fights elevates testosterone levels
Dorothy Cheney, and Robert Seyfarth) for giving us a strong founda- in California mice and enhances future ability to win fights. Horm
tion in animal behavior. Behav. 2005;48:259-267.
[23] Joëls M, de Kloet ER. Control of neuronal excitability by corticoste-
roid hormones. Trends Neurosci. 1992;15:25-30.
DATA AVAI LAB ILITY S TATEMENT
[24] Dewsbury DA. The proximate and the ultimate: past, present, and
Data sharing is not applicable to this article as no new data were cre-
future. Behav Process. 1999;46:189-199.
ated or analyzed in this study. [25] Martins EP. Adaptation and the comparative method. Trends Ecol
Evol. 2000;15:296-299.
ORCID [26] Kirschner M, Gerhart J. Evolvability. Proc Natl Acad Sci USA. 1998;
95:8420-8427.
Jacinta C. Beehner https://orcid.org/0000-0001-6566-6872
[27] Futuyma DJ. Evolutionary constraint and ecological consequences.
Evolution. 2010;64:1865-1884.
RE FE R ENC E S [28] Jungers WL. Body size and scaling of limb proportions in primates. In:
[1] Tinbergen N. On aims and methods of ethology. Z Tierpsychol. 1963; Jungers WL, ed. Size and Scaling in Primate Biology; Boston, MA:
20:410-433. Springer; 1985:345-381.
[2] Nesse RM. Tinbergen's four questions, organized: a response to [29] Shanahan T. Selfish Genes and Lucky Breaks: Richard Dawkins' and
Bateson and Laland. Trends Ecol Evol. 2013;28:681-682. Stephen jay Gould's Divergent Darwinian Agendas. The Darwinian
[3] Mayr E. Cause and effect in biology. Science. 1961;134:1501-1506. Tradition in Context. Springer International Publishing; 2017:11-36.
[4] Rubenstein DR, Alcock J. Animal Behavior: An Evolutionary [30] Dunsworth HM. Expanding the evolutionary explanations for sex dif-
Approach. Sinauer Inc.; 2018. ferences in the human skeleton. Evol Anthropol. 2020;29:108-116.
[5] Dugatkin LA. Principles of Animal Behavior. 4th ed. University of Chi- [31] Puts D. Human sexual selection. Curr Opin Psychol. 2016;7:28-32.
cago Press; 2020. [32] Galipaud M, Kokko H. Adaptation and plasticity in life-history theory:
[6] Klopfer PH, Hailman JP, eds. Function and Evolution of Behavior. how to derive predictions. Evol Hum Behav. 2020;41:493-501.
Addison-Wesley; 1972.
 15206505, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/evan.21931 by Universidad Nacional Autonoma De Mexico, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BERGMAN AND BEEHNER 19

[33] Font E, Carazo P. False dichotomies and human sexual size dimor- Taboga Research Project in Costa Rica (studying white-faced
phism: a comment of Dunsworth (2020). Evol Hum Behav. 2021;42: capuchins).
176-178.
[34] Paaby AB, Rockman MV. The many faces of pleiotropy. Trends Jacinta C. Beehner is a Professor in the Department of Psychology
Genet. 2013;29:66-73. and in the Department of Anthropology at the University of Michigan.
[35] Gould SJ, Lewontin RC. The spandrels of san Marco and the Pan-
She is broadly interested in hormones and behavior, specifically as
glossian paradigm: a critique of the adaptationist programme. Proc R
Soc Lond B Biol Sci. 1979;205:581-598. they relate to reproductive success. She founded and currently directs
[36] Lindenfors P, Tullberg BS. Phylogenetic analyses of primate size evo- two field sites focused on wild primates: the Simien Mountains Gelada
lution: the consequences of sexual selection. Biol J Linn Soc Lond. Research Project in Ethiopia (studying geladas) and the Capuchins at
1998;64:413-447.
Taboga Research Project in Costa Rica (studying white-faced capu-
chins). She also directs two hormone laboratories—one at the Univer-
AUTHOR BIOGRAPHI ES sity of Michigan (Beehner Endocrine Laboratory) and another at the
Capuchins at Taboga field site (TREX Endocrine Laboratory).
Thore J. Bergman is a Professor in the Department of Psychology and
in the Department of Ecology and Evolutionary Biology at the Univer-
sity of Michigan. He is broadly interested the evolution of cognition How to cite this article: Bergman TJ, Beehner JC. Leveling
and communication, specifically as they relate to social knowledge with Tinbergen: Four levels simplified to causes and
and evolutionary fitness. He founded and currently directs two field consequences. Evolutionary Anthropology. 2022;31:12–19.
sites focused on wild primates: the Simien Mountains Gelada https://doi.org/10.1002/evan.21931
Research Project in Ethiopia (studying geladas) and the Capuchins at