PreprintPDF Available

Abstract

This article presents a unifying theory of the embodied, situated human brain called the Hierarchically Mechanistic Mind (HMM). The HMM describes the brain as a complex adaptive system that actively minimises the decay of our sensory and physical states by producing adaptive action-perception cycles via dynamic interactions between hierarchically organised neurocognitive mechanisms. This theory synthesises the free-energy principle (FEP) in neuroscience with an evolutionary systems theory of psychology that explains our brains, minds, and behaviour by appealing to Tinbergen's four questions: adaptation, phylogeny, ontogeny, and mechanism. After leveraging the FEP to formally define the HMM across different spatiotemporal scales, we conclude by exploring its implications for theorising and research in the sciences of the mind and behaviour.
1
Accepted manuscript (post-peer-review, pre-copyedit)
Physics of Life Reviews. Special Issue: Physics of Mind
Please do not cite this version.
Title:
The Hierarchically Mechanistic Mind: A Free-Energy Formulation of the
Human Psyche
Authors:
* Paul Benjamin Badcock 1,2,3
Karl John Friston 4
Maxwell James Désormeau Ramstead 4,5,6
Affiliations:
1 Centre for Youth Mental Health, The University of Melbourne, Melbourne,
Australia, 3052.
2 Melbourne School of Psychological Sciences, The University of Melbourne,
Melbourne, Australia, 3010.
3 Orygen, the National Centre of Excellence in Youth Mental Health, Melbourne,
Australia, 3052.
4 Wellcome Trust Centre for Neuroimaging, University College London, London, UK,
WC1N3BG.
5 Department of Philosophy, McGill University, Montreal, Quebec, Canada, H3A
2T7.
6 Division of Social and Transcultural Psychiatry, Department of Psychiatry, McGill
University, Montreal, Quebec, Canada, H3A 1A1.
* Corresponding author
Correspondence to: pbadcock@unimelb.edu.au
2
Abstract:
This article presents a unifying theory of the embodied, situated human brain called
the Hierarchically Mechanistic Mind (HMM). The HMM describes the brain as a
complex adaptive system that actively minimises the decay of our sensory and
physical states by producing adaptive action-perception cycles via dynamic
interactions between hierarchically organised neurocognitive mechanisms. This
theory synthesises the free-energy principle (FEP) in neuroscience with an
evolutionary systems theory of psychology that explains our brains, minds, and
behaviour by appealing to Tinbergen’s four questions: adaptation, phylogeny,
ontogeny, and mechanism. After leveraging the FEP to formally define the HMM
across different spatiotemporal scales, we conclude by exploring its implications for
theorising and research in the sciences of the mind and behaviour.
Keywords:
Behaviour; Brain; Cognition; Complex Adaptive Systems; Evolutionary Systems
Theory; Hierarchically Mechanistic Mind; Free-Energy Principle; Neuroscience;
Psychology.
Highlights:
We present an interdisciplinary theory of the embodied, situated human brain
called the HMM.
We describe the HMM as a model of neural architecture.
We explore how the HMM synthesises the free-energy principle in
neuroscience with an evolutionary systems theory of psychology.
We translate our model into a new heuristic for theorising and research in
neuroscience and psychology.
Colour Figure Guidelines: Colour should not be used for all figures in print.
Copyright: Our figures have been adapted from our previous work.
3
Life is poised on the edge of chaos.
Stuart Kauffman
Introduction:
The aim of our review is to unify dominant schools of thought spanning
neuroscience and psychology by presenting a new theory of the human brain called
the hierarchically mechanistic mind (HMM). Originally proposed to reconcile
paradigmatic divisions within psychology [1], the HMM offers an integrative
perspective of the brain, cognition and behaviour that has since been leveraged to
explain our species-typical capacity for depression [2], and to exemplify a new,
transdisciplinary approach to the study of living systems called variational
neuroethology [3, 4]. The HMM defines the embodied, situated brain as a complex
adaptive system that actively minimises the entropy (i.e., the spread or decay) of
human sensory and physical states by generating action-perception cycles that emerge
from dynamic interactions between hierarchically organised neurocognitive
mechanisms.
The HMM leverages evolutionary systems theory 1 (EST) to bridge two
complementary perspectives on the brain. First, it subsumes the free-energy principle
(FEP) in neuroscience and biophysics to provide a biologically plausible,
mathematical formulation of the evolution, development, form, and function of the
1 EST is a transdisciplinary paradigm that harkens from [5] Schrödinger E. What is life? Cambridge:
Cambridge University Press; 1944. EST explains all adaptive dynamic systems in terms of circular
interactions between self-organisation and general selection (e.g., natural selection) within and across
(hierarchically nested) spatiotemporal scales [1] Badcock PB. Evolutionary systems theory: a unifying
meta-theory of psychological science. Review of General Psychology. 2012;16:10-23, [3] Ramstead
MJD, Badcock PB, Friston KJ. Answering Schrödinger's question: A free-energy formulation. Physics
of Life Reviews. 2018;24:1-16. This universal, dynamical process creates complex adaptive systems,
which adapt to the environment through an autonomous process of selection that recruits the outcomes
of locally interacting components within that system to select a subset of those components for
replication or enhancement [6] Levin S. Complex adaptive systems: exploring the known, the unknown
and the unknowable. Bulletin of the American Mathematical Society. 2003;40:3-19. Of particular
relevance here, a widely cited example is the brain [7] Haken H. Principles of brain functioning: a
synergetic approach to brain activity, behaviour and cognition. Berlin: Springer-Verlag; 1996, [8]
Kelso JS. Dynamic patterns: the self-organization of brain and behavior. Cambridge, MA: MIT Press;
1995. For in-depth discussions of the dynamical causal relationship between selection and self-
organisation, see [9] Depew DJ, Weber BH. Darwinism evolving: systems dynamics and the genealogy
of natural selection. Cambridge, MA: MIT Press; 1995, [10] Eigen M, Schuster P. The hypercycle: a
principle of natural self-organisation. Berlin: Springer-Verlag; 1979, [11] Holland JH. Hidden order:
how adaptation builds complexity. New York: Basic Books; 1995, [12] Kauffman SA. The origins of
order: self-organization and selection in evolution. Oxford: Oxford University Press; 1993. For a
succinct treatment of the underlying mathematics, also see [13] Ao P. Emerging of stochastic
dynamical equalities and steady state thermodynamics from Darwinian dynamics. Communications in
theoretical physics. 2008;49.
4
brain [14-16]. Second, it follows an EST of psychology by recognising that neural
structure and function arise from a hierarchy of causal mechanisms that shape the
brain-body-environment system over different timescales [1, 2]. Under this view,
human neural dynamics can only be understood by considering the broader context of
our evolution, enculturation, development, embodiment, and behaviour. After
describing the architectural claim that underpins the HMM, we consider these two
perspectives in turn, before bringing them together with a formal definition of
multiscale neural dynamics. We conclude by considering the implications of our
model for theorising and research in neuroscience and psychology.
In a nutshell, the HMM rests on two cardinal elements: an EST of human
cognition and behaviour that draws on four intersecting levels of explanation in
psychology; and a mathematical formulation of multiscale neural dynamics based on
the FEP. Our central claim is that the FEP and EST of psychology reflect two sides of
the same coin the former furnishes a non-substantive formal theory of neural
structure, function, and dynamics; the latter affords a substantive evolutionary theory
that can explain the particular manifestations of the FEP observed in Homo sapiens.
By leveraging theories, frameworks, and methods originally drawn from physics and
biology (which also yield EST and the FEP; [3, 4]), the HMM synthesises psychology
and neuroscience with a systematic framework to formulate multilevel models of the
extraordinary nexus between the brain, our minds and behaviour.
1. The HMM as a Model of Neural Structure
The HMM rests on the architectural claim that the human brain is a
hierarchically organised system of neurocognitive mechanisms that interact in a
dynamic, reciprocal fashion. The lowest or most peripheral levels of this hierarchy
comprise relatively segregated, highly specialised neural mechanisms responsible for
sensorimotor processing (‘domain-specific’ systems), while its higher, deeper or more
central layers consist of developmentally plastic, highly integrated (‘domain-general’)
mechanisms. The latter are widely distributed subsystems that respond flexibly to
input received from multiple lower levels, feed information downstream for further
processing, and underlie the executive cognitive functions unique to humans [1].
There are two important distinctions here. First, although there are many
interpretations of the neural hierarchy, here we refer to a fractal or nested modular
hierarchy, which entails the repeated encapsulation of smaller (neuronal) elements in
5
larger ones across different spatial, temporal, topological, and functional neural scales
(i.e., ‘modules within modules’; [17-19]). The second is that neurocognitive
mechanism is defined as a neural subsystem that operates at any spatiotemporal scale,
ranging from a particular neuronal population through to macroscopic brain regions.
Such mechanisms involve a dynamic, bidirectional relationship between specialised
functional processing mediated by dense, short-range connections intrinsic to that
scale (i.e., its local integration); and its global (functional) integration with other
neural subsystems via relatively sparse, long-range (e.g., extrinsic cortico-cortical)
connections [20])2. Accordingly, the HMM implies a complementary relationship
between functional segregation and integration: all neurocognitive mechanisms
involve a sub-population of cells that have a common, specialised function, but they
are also functionally integrated because of their distal connections with other
subsystems [20, 30]. At the same time, it also recognises that some neural subsystems
will be more integrated than others.
The type of neural architecture described here echoes a growing consensus
that human cognition and behaviour emerge from the integrated dynamics of
hierarchical networks of (functionally segregated and differentially integrated) neural
processing mechanisms [20, 25, 31-40]. There is nothing controversial about this
claim. The idea that the brain exhibits a hierarchical structure that progresses from
relatively ‘domain-specific’ systems through to highly integrated, ‘domain-general’
regions is far from new, having long been recognised by influential perspectives such
as global neuronal workspace theory [41, 42] and the dual process theory of reasoning
[43, 44]. More recently, sophisticated structural and functional imaging studies in
network neuroscience have furnished extensive evidence that the brain exhibits a
nested, fractal-like structure; extending from cellular microcircuits in cortical columns
2 In network neuroscience, this kind of subsystem is called a module [21] Sporns O, Betzel RF.
Modular brain networks. Annual review of psychology. 2016;67:613-40. We have avoided the term
here because of its potential confusion with massive modularity, domain-specificity, and informational
encapsulation [22] Barrett HC, Kurzban R. Modularity in cognition: framing the debate. Psychological
review. 2006;113:628. [23] Barrett HC. A hierarchical model of the evolution of human brain
specializations. Proceedings of the national Academy of Sciences. 2012;109:10733-40. [24] Fodor JA.
The modularity of mind: An essay on faculty psychology. Cambridge, MA: MIT press; 1983. For
incisive critiques of the use of ‘modularity’ in psychology, see [25] Anderson ML, Finlay BL.
Allocating structure to function: the strong links between neuroplasticity and natural selection.
Frontiers in human neuroscience. 2014;7, [26] Chiappe D, Gardner R. The modularity debate in
evolutionary psychology. Theory & Psychology. 2012;22:669-82, [27] Colombo M. Moving forward
(and beyond) the modularity debate: A network perspective. Philosophy of Science. 2013;80:356-77,
[28] Frankenhuis WE, Ploeger A. Evolutionary psychology versus Fodor: Arguments for and against
the massive modularity hypothesis. Philosophical Psychology. 2007;20:687-710, [29] Zerilli J. Against
the “system” module. Ibid. 2017;30:235-50.
6
at the lowest level, to cortical areas at intermediate levels, through to distributed
clusters of highly interconnected brain regions at the global level [45-47]; see Figure
1a.3 Notably, a hierarchical neural structure is also central to the theory of predictive
processing, an increasingly popular scheme that describes the brain as a Bayesian
‘inference machine’ that minimises discrepancies between incoming sensory inputs
and top-town predictions (see Figure 1b). Since the literature on the brain’s
hierarchical organisation has already been reviewed elsewhere (e.g., [18, 21, 31, 33,
49, 50]), we will not dwell on it here. Instead, we will now concentrate on the more
contentious issue of why the brain is structured in this way.
2. The Variational Free-Energy Formulation
The FEP is a simple mathematical postulate that draws from statistical
thermodynamics and machine leaning to explain how living systems maintain their
physical integrity by revisiting a small number of characteristic, phenotypic states
[15]. It rests on the elegant premise that biotic agents actively reduce the entropy (i.e.,
the decay or dispersion) of their sensory and phenotypic states by minimising their
variational free-energy. Technically, variational free-energy is an information
theoretic quantity that bounds or limits (by being greater than) the entropy of a brain’s
sensations or sensory samples from the environment. In this context, entropy is a
measure of information that refers to the long-term average of surprise: a statistical
measure of the probability (technically, the negative log probability) of sensory
samples sampled by an agent.4
3 Note that the levels of organisation listed here are a gross approximation. A hierarchically nested
organisation is likely to produce different topologies at different neural scales, and the precise,
hierarchical organisation of neural elements at different levels remains open to question [48] Hilgetag
CC, Goulas A. Is the brain really a small-world network? Brain Structure and Function.
2016;221:2361-6., [45] Kaiser M. A tutorial in connectome analysis: topological and spatial features of
brain networks. Neuroimage. 2011;57:892-907.
4 Please note an important caveat about the relationship (and major differences) between the variational
formulation described here and statistical thermodynamics: variational free-energy should not be
conflated with thermodynamic free-energy. The free-energy formulation is a mathematical description
of the dynamics of systems at nonequilibrium steady-state, and should not be confused with the second
law of thermodynamics. The FEP deals with information theoretic measures (e.g., variational free-
energy, mutual information, relative entropy, self-information, surprisal, information gain, Bayesian
surprise, Shannon entropy, etc.). The connection between the variational free-energy for
nonequilibrium steady-state systems and thermodynamic free-energy in statistical thermodynamics
remains an outstanding question. For more details, see [4] Ramstead MJ, Badcock PB, Friston KJ.
Variational neuroethology: Answering further questions: Reply to comments on “Answering
Schrödinger's question: A free-energy formulation”. Physics of life reviews. 2018;24:59-66. For an
account of this connection by appeal to Landauer’s principle and the Jarzynski equality, consider [51]
7
The FEP flows from the idea that living systems can be distinguished from
other self-organising systems because they actively avoid deleterious phase-
transitions by bounding the entropy of their sensory and physical states under the
FEP, to be alive simply means to revisit a bounded set of states with a high
probability [5, 52]. Here, a deleterious phase transition is cast as a surprising one (i.e.,
a low probability event, given that the creature in question continues to survive).
Because the repertoire of states an organism occupies is limited, the probability
density over these states must have low entropy (i.e., they are found in characteristic,
unsurprising states). Heuristically, we can think of the expectations of an organism as
having an evolutionary or adaptive value, in that organisms expect to remain within
their most probable (characteristic or phenotypic) states: those which make it the kind
of creature that it is. In this specific sense, surprising or unexpected states (i.e., those
incongruous with the expectations of the organism; e.g., a fish out of water; [15]) are
deleterious, and must therefore be avoided. Hence, an organism’s ultimate,
evolutionary imperative of maintaining its repertoire of functional states within
physiological bounds (i.e., survival, homeostasis, and allostasis) translates into a
proximal avoidance of surprising states [15]. Following EST, this propensity to avoid
surprise is the product of selection: self-organising systems that can avoid surprising
phase-transitions have been favoured by natural selection over those that could not
[16].
So how does the FEP pertain to hierarchical neural dynamics? The FEP aligns
with the theory of predictive processing by casting the brain as a hierarchically-
organised ‘inference machine’ that optimises the evidence for an organism’s model of
the world by minimising variational free-energy (see Figure 2). When relating the
FEP to prediction and inference, the key move is to note that surprise is the negative
logarithm of Bayesian model evidence. It therefore follows that any creature that
minimises surprise is simply optimising Bayesian model evidence.5 On this view,
brain dynamics (i.e., the general ‘behaviour’ or ‘ensemble dynamics’ of neural
Sengupta B, Stemmler MB, Friston KJ. Information and efficiency in the nervous systema synthesis.
PLoS computational biology. 2013;9:e1003157.
5 Technically, variational free energy provides an upper bound on surprise or self information, via the
addition of a nonnegative Kullback-Leibler divergence to surprise that induces an approximate
posterior density over the causes of sensory states. This means that the negative free-energy lower
bounds log evidence. In this sense, it is sometimes referred to as an evidence lower bound (ELBO) in
machine learning and statistics. It also means that minimising surprise effectively means minimising
the variational free-energy via an approximate posterior encoded by the brain’s internal states. See
Figure 2.
8
mechanisms) can be described as realising an implicit hierarchical generative model:
a Bayesian hierarchy of ‘hypotheses’ or ‘best guesses’ about the hidden causes of our
sensory states. This ‘Bayesian mechanics’ imposes an upper limit on surprise by
tracking and minimising discrepancies between incoming sensory inputs and top-
down, neuronally encoded predictions (i.e., prediction errors; [20, 31, 53, 54]).
Conditional expectations are encoded by deep pyramidal cells (i.e., representation
units) at each level of the cortical hierarchy that convey predictions downward to
suppress errors at the level below; prediction errors are encoded by superficial
pyramidal cells (i.e., error units) that convey errors forward to revise expectations at
the level above; and neuromodulatory mechanisms regulate the relative influence of
these signals by modifying their precision (see Figure 1b; [31, 49, 50, 55]).
Notably, this scheme has also been leveraged to explain the functional
integration of hierarchically modular neural networks. According to this view, the
brain minimises prediction error by dynamically adjusting the synaptic efficiency of
connections between modules, with backwards connections conveying predictions to
lower levels and forward connections delivering prediction errors to higher ones [20].
Here, cognition is described in terms of the global integration of local neuronal
operations via hierarchical (error minimising) message passing between cortical
regions, a process that is facilitated by a hierarchically nested network architecture
[20].
Under the predictive processing formulation outlined here, prediction errors
quantify the organism’s variational free-energy (and by extension, its surprise). This
predictive process allows us to minimise surprise by updating our internal models
(i.e., through perception, learning, and phenotypic plasticity). Alternatively, we can
also minimise surprise by selectively sampling sensory data that confirms our
expectations, to ensure that our predictions are self-fulfilling (i.e., action). Figure 2
illustrates these two interdependent surprise-resolving processes in terms of action,
which minimises (a bound on) surprise; and perception, which reduces the divergence
between inferred and true states of the world (given some sensory data).
The ensuing perspective on brain dynamics that resolve free energy through
loops of action and perception is called active inference [56-58]. Put simply, this
suggests that action and perception operate synergistically to maintain homeostasis
and optimise the organism’s generative model [15, 59]. In other words, every
organism seeks to maximise sensory evidence for its own existence; it is essentially
9
‘self-evidencing’ [60]. Quite literally, then, the FEP alludes to Maslow’s [61]
‘hierarchy of needs’it suggests that the meaning of life is to self-actualise.
Although surprise cannot be directly evaluated by living systems, it can be
minimised vicariously by minimising a bound on, or proxy for, this quantity:
variational free-energy [15, 16, 52]. Because surprise is mathematically equivalent to
the negative log probability of an outcome (also known as Bayesian model evidence
in machine learning), minimising free-energy compels us to make Bayesian
inferences about our eco-niche. Under the FEP, over time and on average, our actions
will tend to infer or reflect the statistical structure of the environment to which they
are coupled. This explains the intentionality or purposiveness of living systems by
appealing to dynamical principles drawn from complexity science and information
theory in physics, thereby providing a mathematical account of actions guided by our
beliefs. Correspondingly, the FEP supplies a formally expressible and
neurobiologically plausible physics of the mind [3, 4, 62].
An important corollary of this view is that our generative models are
optimised by evolution, neurodevelopment and learning [3, 63, 64]. If each individual
is adapted to their own eco-niche either through natural selection, development, or
learning then the expectations of each of us must differ. Clearly, though, some part
of these expectations must also be inherited, since the characteristic, phenotypical
features of a given species’ generative model are conserved across generations (e.g.,
the basic wiring of the human brain). This segues nicely into the role of our
(Bayesian) prior beliefs about the ways we expect the world to unfold.
By way of explanation, the FEP proffers an elegant, formally expressible
explanation of human neural dynamics across spatiotemporal scales it can be used
to formulate mathematical models of the influence of natural selection acting on
human phenotypes over time [3, 4, 63]. The brain only labels a sensory state as
valuable or unsurprising if it leads to another valuable state, and selection ensures that
an organism progresses through a succession of probable states with adaptive
(homeostatic) value intrinsic, phenotypic states that are unsurprising [15, 16, 52].
Under this view, natural selection reduces surprise by specifying the value of sensory
states through (epi)genetic mechanisms, prescribing a small number of attractive
states with innate value. These states are sought out by living systems because they
minimise surprise by conforming to both their internal states and eco-niche [15]. With
these distinctions in mind, species-typical patterns of cognition and behaviour can be
10
explained as inherited adaptive priors that have been shaped by selection to guide
action-perception cycles towards unsurprising states (e.g., “I will keep moving until I
am rewarded”; [15, 52, 56]); also see [65-67]. In other words, natural selection is
nature’s way of performing Bayesian model selection to minimise the variational free-
energy of our phenotypes (i.e., hierarchical generative models); also see [63]. The
upshot of all this is that the brain does not just contain a hierarchical generative model
of the world, its dynamics also instantiate one – its form and function reflect a
physical transcription of causal regularities in the environment that has been
optimised by evolution within and across nested spatiotemporal scales.
Indeed, central to the architectural claim of the HMM is the evolution of
hierarchical neural connections that reflect lawful statistical regularities in the
environment. Take, for example, the statistical independence between the identity and
location of objects in the visual world knowing what an object is does not tell us
where it is. Strikingly, this statistical independence is reflected in the anatomical
dissociation between the ventral and dorsal streams in the cortical hierarchy, which
encode models or representations of the ‘what’ and ‘where’ attributes of visual
precepts, respectively [68]. This suggests that the structure of the brain recapitulates
the structure of the world in which it is embedded: environmental causes that are
statistically independent are encoded in functionally and anatomically segregated
neuronal structures. Similarly, the hierarchical organisation of the brain mirrors the
hierarchically nested structure of causal regularities in the environment. This
hierarchical nesting marries the hierarchy of temporal scales at which representations
evolve with the hierarchy of temporal scales at which biological phenomena unfold
the lower, more peripheral layers of the neural hierarchy encode rapid environmental
fluctuations associated with sensorimotor processing and stochastic effects; its higher,
more central layers encode increasingly slower regularities related to contextual
changes [69-72].
The idea that the brain instantiates a generative model based on hierarchical
temporal dynamics in the environment makes intuitive sense, given that the content of
our sensorium changes more rapidly than its context [73]. Moreover, the actions of an
organism clearly possess a temporally nested structure (e.g., an arm movement is
composed of smaller elemental movements); it thus makes sense that the organ
responsible for the evaluation and selection of actions mirrors the statistical structure
of the policies to be selected. The temporal structure of neural dynamics has also been
11
demonstrated empirically, both by simulations of perceptual inference and motor
behaviour [71, 74], and studies of the human and primate brain [72, 75, 76]. Finally,
there is good reason to suppose that a temporal neural hierarchy is likely to have been
favoured by selection: it optimises perceptual inference by allowing the organism to
accumulate evidence across different timescales to derive the best explanation for
sensory data [77], and it facilitates adaptive behaviour by enabling top-down,
cognitive control to achieve distal goals [78]. Other selection pressures are also likely
to have been involved. For example, it has been suggested that during co-evolution
with conspecifics, a temporal hierarchy would have been favoured by selection
because it enables the organism to generate and recognise communicative behaviour
that unfolds over multiple, nested timescales [71, 73, 79]. As we discuss later, a
hierarchical structure is also thought to confer other evolutionary advantages. On a
more general note, it is almost self-evident that if the brain entails a generative model
of (the causal structure of) its sensorium and this sensorium is generated by deep
(i.e., hierarchical) temporal processes (e.g., by conspecifics) then neuroanatomy and
neurophysiology must reflect this deep architecture. In this sense, the brain’s
hierarchical and modular anatomy exemplifies the good regulator theorem in
cybernetics, which maintains that any system that can control its environment must be
a good model of that environment [69, 80].
So far, then, we have described non-substantive principles that can be
generalised to any species with a brain not to mention other biological dynamics
(e.g., single-celled organisms [52]; morphogenesis [81, 82]; and plant life [83]).
Every species is equipped with naturally selected Bayesian priors that emerge from
species-typical eco-niches and influence morphology, cognition, and behaviour in
adaptive (i.e., valuable) ways – different organisms instantiate unique ‘embodied
models’ of their specific biological needs and eco-niches [3, 58, 84, 85]. Nevertheless,
in order to explain the human brain – and its relation to our cognition and behaviour
we need to draw upon substantive (ultimate or evolutionary) explanations that can
account for the particular adaptive solutions that have produced the hierarchical
generative models of Homo sapiens [31, 63, 86, 87]. With this in mind, we argue that
the FEP demands recourse to psychology, because it sheds direct light on the
complex, multiscale processes that govern human biobehavioural dynamics in
particular. The HMM does this by synthesising the FEP with an EST of psychology.
12
3. The Evolutionary Systems Theory of Psychology
In psychology, evolutionary systems approaches have traditionally focused on
the complex interplay between evolutionary and developmental processes (e.g., [88-
97]). This approach has since been developed into an integrative EST of human
cognition and behaviour that has the potential to unify major paradigms in the
discipline [1]. The EST in question recasts Tinbergen’s [98] seminal four questions in
ethology in terms of a temporal hierarchy of biological dynamics that extend across
all Homo sapiens: those that produce species-typical ‘functional adaptations’ to the
environment over evolutionary time (e.g., natural selection); intergenerational,
phylogenetic’ mechanisms that introduce evolutionary change by producing heritable
differences between groups (e.g., epigenetic inheritance); ‘ontogenetic’ processes that
unfold over an individual’s lifetime (e.g., gene-environment interactions); and the
proximate ‘mechanisms’ that drive psychology and behaviour in real-time (i.e.,
biopsychosocial dynamics) (see Figure 3). 6 These dynamics are arguably
recapitulated by different research programs in psychology, which concentrate
differentially on four complementary levels of explanation: ultimate explanations for
adaptive, species-typical traits (i.e., evolutionary psychology); epigenetic and
exogenetic explanations for intergenerational, between-group differences (i.e.,
evolutionary developmental biology and psychology); ontogenetic explanations for
individual similarities and differences (i.e., developmental psychology); and
proximate, mechanistic explanations for real-time phenomena (i.e., psychological
6 It should be recognised that this is only one interpretation of Tinbergen’s questions, which continue to
attract debate. For example, in this context, phylogeny is used to refer to the intergenerational
processes responsible for producing evolutionary change within a species, not the evolutionary
outcomes of such processes (e.g., our species’ position on a phylogenetic tree). There is also
considerable debate about whether Tinbergen’s survival valueshould be equated with adaptation,
functionor current utility[99] Bateson P, Laland KN. Tinbergen's four questions: an appreciation
and an update. Trends in ecology & evolution. 2013;28:712-8, [100] Bateson P, Laland KN. On current
utility and adaptive significance: a response to Nesse. Trends in ecology & evolution. 2013;28:682-3,
[101] Nesse RM. Tinbergen's four questions, organized: a response to Bateson and Laland. Trends in
Ecology & Evolution. 2013;28:681-2. We have settled upon ‘functional adaptation’ here because of its
clear evolutionary connotations. As we have discussed elsewhere, how one applies Tinbergen’s
questions will also vary depending on the systemic and temporal scales under scrutiny [3] Ramstead
MJD, Badcock PB, Friston KJ. Answering Schrödinger's question: A free-energy formulation. Physics
of Life Reviews. 2018;24:1-16, [4] Ramstead MJ, Badcock PB, Friston KJ. Variational neuroethology:
Answering further questions: Reply to comments on “Answering Schrödinger's question: A free-energy
formulation”. Physics of life reviews. 2018;24:59-66, [62] Ramstead MJ, Constant A, Badcock PB,
Friston K. Variational ecology and the physics of minds. Physics of life Reviews. Accepted, this issue.
Leaving such terminological issues aside for the purposes of the current discussion, it suffices to say
that human phenotypes emerge from the dynamic interplay between evolutionary, intergenerational,
developmental, and real-time mechanisms, which roughly correspond to different paradigms in
psychology.
13
sub-disciplines; e.g., cognitive, social and clinical psychology) [1]. An important
implication of this view is that in order to explain a given trait, one should seek to
incorporate theories and evidence drawn from each of these levels of inquiry.
So how does this EST of psychology relate to the HMM? Put simply, the
HMM is a first-order hypothesis derived from the synthesis of this broader meta-
theory with the FEP it explains the hierarchical dynamics of the embodied, situated
human brain in terms of (natural and general) selection and self-organisation
interacting at and across evolutionary, intergenerational, developmental, and real-time
spatiotemporal dynamics. The HMM is broadly consonant with other dynamical
theories of biophysics, in that it casts adaptive biobehavioural patterns as the
historical product of reliably recurrent developmental resources reborn in each
generation. These resources are themselves the result of circular interactions between
mechanisms of selection; intergenerational and developmental processes; and the
engagement of humans with their species-typical environments in real-time [25, 89,
93, 102-108].7 By extension, the HMM means that the most effective method to
explain the multiscale dynamics of our brains and behaviour should leverage
multilevel models in psychology that are able to explain both why neurocognitive
mechanisms are adaptive in Homo sapiens, along with how these mechanisms emerge
from circular interactions between evolutionary, intergenerational, developmental,
and real-time processes [1-3].
There is ample evidence to suggest that this temporal hierarchy of biological
dynamics manifests in the development and morphology of the brain. Indeed, both
comparative work and studies on humans suggest that the evolutionary history of the
human brain is reflected across nested levels of neural organisation, extending from
the genes inherited from our hominid ancestors; to the epigenetic transcription factors
that shape gene expression; to the synaptic epigenesis of neural networks over the
course of development; through to the highly distributed and integrated long-range
connections that underwrite conscious awareness [110]. Longitudinal imaging studies
7 Also note that a particularly important constraint that extends across all of these timescales is the
sociocultural environment, since human survival depends on our ability to leverage cultural
information and immersively participate in normative, culturally adapted practices [3] Ramstead MJD,
Badcock PB, Friston KJ. Answering Schrödinger's question: A free-energy formulation. Physics of
Life Reviews. 2018;24:1-16, [87] Ramstead MJ, Veissière SP, Kirmayer LJ. Cultural affordances:
scaffolding local worlds through shared intentionality and regimes of attention. Frontiers in
Psychology. 2016;7, [109] Gallagher S. Enactivist interventions: Rethinking the mind: Oxford
University Press; 2017.
14
examining the maturation of neural networks throughout childhood and adolescence
have also found that the development of the human cortex mirrors our phylogenetic
history the standard developmental sequence sees the maturation of
phylogenetically older, canonical sensorimotor hierarchies that are common among
all mammals, through to the evolutionarily recent, highly integrated association
cortices enjoyed by humans (e.g., [111-113]). This highlights the complementarity of
selection and self-organisation: natural selection ensures the emergence and retention
of highly specialised or segregated sensorimotor networks in infancy, which function
as ‘neurodevelopmental anchors’ that permit the progressive self-organisation of
widely distributed ‘domain-general’ association regions throughout ontogeny that
enhance evolvability by allowing us to respond fluidly to a constantly changing
environment [25, 113-115]. Consistent with this, both evolutionary and
developmental psychologists have long maintained that the brain instantiates a nested
hierarchy of neuronal processing mechanisms that vary in degrees of functional
segregation and integration [22, 23, 89, 90, 93-95, 116-123].
Crucially, this idea is backed by extensive empirical support, ranging from
large meta-analyses of neuroimaging data that provide evidence for functionally
diverse ‘domain-general’ neural subsystems [124-126], through to studies of cross-
and multi-modal context effects in early sensory processing that show that even at the
level of the sensorium, highly segregated ‘domain-specific’ systems exchange data in
a bidirectional fashion [127, 128]. On the other hand, high resolution network-based
analyses have recently provided evidence that different neural ‘modules’ perform
discrete cognitive functions, while highly distributed ‘connector’ regions allow for
their functional integration by coordinating connectivity across ‘modules’ [129, 130].
Importantly, comparative work has further shown that a hierarchical architecture is a
ubiquitous feature of the mammalian brain, progressing from highly segregated
sensorimotor hierarchies found in all mammals through to the higher cortical
association areas that confer the adaptive advantage of heightened cognitive control
among primates [113, 115].
Of particular relevance, the brain’s hierarchical organisation also resonates
with EST. A hallmark feature of complex adaptive systems is that aggregates of
interacting units (e.g., modules) are organized in a hierarchically nested manner; and
that intra-component (e.g., within-module) connections tend to be stronger than inter-
component (e.g., between-module) connections, with neighbouring components
15
showing stronger connections than distal ones [11, 131]. As we alluded to earlier,
there is broad agreement in the life sciences that this sort of structure confers
significant selective advantages. First, it enhances evolvability because deleterious
changes to single components of the system are less likely to lead to total system
failure. Similarly, a hierarchical structure enables the emergence of evolutionary
novelties (e.g., exaptations) without threatening global functioning [21].8 Spatially
compact, functionally connected modules that are relatively sparsely connected to
other modules also conserve the (spatial, processing, and metabolic) cost of neural
connections; preserve specialised kinds of neural processing that unfold over multiple
timescales; and support complex brain dynamics that optimise information processing
[21, 37]. Consistent with this, fine-grained functional connectivity studies suggest that
a hierarchical structure allows cortical networks to optimise the balance between
local, specialised processing and global integration [129, 130]. Interestingly,
computer simulations of evolving networks have also shown that even in the absence
of modularity, a hierarchical structure improves evolvability by adapting faster to new
environments than non-hierarchical structures, because such a structure allows the
system to solve problems by recursively combining solutions to sub-problems [133].
Finally, the hierarchical organisation of the brain promotes self-organised
criticality (colloquially, the ‘edge of chaos’; [12]). This is a fundamental property of
complex adaptive systems that refers to a dynamical state that occupies the
intersection between highly ordered, stable structures and cycles of activity (e.g.,
lattice structures); and highly stochastic, rapidly fluctuating ones (e.g., gaseous
states). This state is known to optimise evolvability by allowing small, extrinsic
changes to create and channel large-scale systemic reorganisations [12, 134, 135].
Recent empirical work has shown that the hierarchical segregation of neural networks
into local neurocognitive mechanisms effectively stretches the parameter range for
self-organised criticality [46]. The nested hierarchy of the brain means that the system
can maintain different degrees of randomness; it is able to entertain subcritical and
8 As noted by a reviewer, these explanations have also been leveraged to argue that evolution favours
redundancy in neural networks. Although we agree that redundancy is important, as discussed here, a
nested hierarchy also confers additional evolutionary advantages. We would also note that it is
important to distinguish between redundancy, where the same function is performed by different
neuronal structures; and degeneracy, which entails many-to-one structure-function relationships.
Specifically, while degeneracy facilitates evolution by conferring systemic robustness, the brain will
seek to minimise redundancy over the course of development because it negatively affects efficiency;
e.g., due to increased metabolic and other costs [132] Friston KJ, Price CJ. Degeneracy and redundancy
in cognitive anatomy. Trends in cognitive sciences. 2003;7:151-2.
16
supercritical dynamics simultaneously, because they can co-exist at different levels of
the hierarchy [136]. Given the selective advantages of being poised at the edge of
chaos, it is unsurprising that a hierarchical structure, which extends this critical
region, has been observed empirically [137].
In closing, it is worth noting that the picture of the brain that we have sketched
above could be readily applied to all primates, not to mention other species [3, 62]. In
this sense, the HMM might be put forward as a theory of embodied, situated brains in
general. Although we certainly encourage such efforts, the reason we have explicitly
coupled this theory with paradigms in psychology is because they are ideally
positioned to cast direct light on the highly sophisticated patterns of free-energy
minimisation unique to Homo sapiens [87]. This is important, because it underscores
the need to synthesise the FEP with substantive research that concentrates on the
cognitive and behavioural dynamics particular to the species in question [4]. To
address this, the HMM weds a generalisable model of the embodied, hierarchical
brain with a clearly articulated meta-theory of different levels of explanation in
psychology which, in principle, encapsulates the disciplinary content knowledge
accumulated by psychologists to date. On the one hand, this model requires
psychologists to explore how the FEP applies to their own research avenues; on the
other, it requires cognitive and behavioural scientists to develop in a bottom-up,
evidence-driven fashion multilevel hypotheses about human neural and behavioural
dynamics (i.e., process theories) that are substantiated by extant findings in
psychology.9 In this sense, the HMM is as much a heuristic for theorising and
research as it is a theory of the brainan issue we will return to shortly. Before we
do, however, it is first important to clearly establish how this conceptual treatment of
the brain directly relates to the mathematics of the FEP. With this in mind, we will
now operationalise the HMM by leveraging the FEP to formally model the dynamics
of the embodied human brain across all four levels of explanation in psychology.
9 This is not to say, of course, that the HMM should be restricted to psychology aloneunpacking the
complexities of the human phenotype clearly requires inter-disciplinary efforts, particularly those
within the cognitive and social sciences, along with comparative and phylogenetic approaches [138]
Daunizeau J. A plea for “variational neuroethology”: Comment on “Answering Schrödinger's question:
A free-energy formulation” by MJ Desormeau Ramstead et al. Physics of life reviews. 2018, [139]
Kirmayer LJ. Ontologies of life: From thermodynamics to teleonomics. Comment on “Answering
Schrödinger’s question: A free-energy formulation” by Maxwell James Désormeau Ramstead et al.
Ibid., [140] Veissière S. Cultural Markov blankets? Mind the other minds gap!: Comment on
“Answering Schrödinger's question: A free-energy formulation” by Maxwell James Désormeau
Ramstead et al. Ibid.
17
4. The HMM Defined
Beyond the fact that they are both ESTs that explain the adaptive, hierarchical
dynamics of the embodied brain, the FEP converges with the EST of psychology in
two pivotal ways. First, we noted in Section 2 that although every organism is adapted
to its specific eco-niche, each generation inherits the adaptive priors of the previous
generation (i.e., species-typical phenotypic traits). This means that we need to
consider the systemic dimension of these phenomena; i.e., the multiscale dynamics
that extend from all Homo sapiens to specific individuals that operate in real-time.
Second, both the FEP and EST of psychology rest on the notion of recursive, causal
interactions between dynamics at different temporal scales. Figure 4 shows how to
express this process formally at each of the timescales over which free-energy
minimisation optimises the state (i.e., perception), configuration (i.e., action),
connectivity (i.e., learning and attention), anatomy (neurodevelopment), and
phenotype (i.e., neural evolution) of living agents that belong to a given class (e.g.,
Homo sapiens) [3]; see Figure 4.
To recapitulate, the HMM syntheses a multi-level EST of human psychology
with the variational formulation of the FEP to provide both a substantive and formally
expressible theory of the brain, mind and behaviour [1, 3]. More precisely, this
hypothesis defines the human brain as: an embodied, complex adaptive control system
that actively minimises the variational free-energy (and, implicitly, the entropy) of
(far from equilibrium) phenotypic states via self-fulfilling action-perception cycles,
which are mediated by recursive interactions between hierarchically organised
(functionally differentiated and differentially integrated) neurocognitive processes.
These ‘mechanics’ instantiate adaptive priors, which have emerged from selection
and self-organisation co-acting upon human phenotypes across different timescales.
Having now defined the HMM, we will close by focusing on its implications for
theorising and research.
5. Using the HMM as a Research Heuristic in Neuroscience and Psychology
At this juncture, we have described both a formal and substantive theory of the
human brain that unifies major paradigms spanning neuroscience and psychology
one that affords both an ultimate and proximate theory of our cognition and
biobehaviour (i.e., adaptive free-energy minimisation); and explains the psyche in
18
terms of hierarchical neural dynamics that work vicariously to minimise surprise. This
model should be seen as a first-order hypothesis derived from the psychological meta-
theory of EST and the FEP. In other words, the HMM shares the same epistemic
status as other influential theories of the brain, such as predictive processing theory
[31, 49, 50] and the massive modularity hypothesis [22, 141, 142], because it provides
cognitive scientists with a systematic heuristic to pose far-reaching questions and
engineer unique, substantive hypotheses from which specific, testable predictions can
be derived [4]. Crucially, some of these predictions make the HMM amenable to
falsification. In particular, the HMM relies on the directly testable second-order
hypothesis that the brain minimises prediction error via hierarchical message passing
in the brain (i.e., predictive coding; [31]), which has already been demonstrated
experimentally by studies of visual processing (e.g., [16, 31, 55]).
For cognitive neuroscientists, the HMM encourages a range of research
avenues already advocated elsewhere. First, it requires finer maps of effective neural
connectivity informed by multiscale structural connectivity findings, along with
empirically informed biophysical and computational models of spatiotemporal
patterns of network activity that capture the ways in which our unique predictive
capacities manifest in particular patterns of hierarchical neural activity across
different contexts [20, 143, 144]. It also appeals to multiscale network approaches that
measure neural activity across timescales, coupled with complementary methods that
map rapid fluctuations in neural patterns in real-time (e.g., EEG and MEG),
maturational changes over developmental time (e.g., diffusion tensor imaging), and
neural mechanisms conserved by evolution (e.g., comparative studies) [138, 145-
147]. Finally, the HMM resonates with approaches in embodied cognition and
neuroethology, which both explore how action-perception cycles emerge from
adaptive brain-body-environment relations [3, 4, 148-151]. Collectively, these
approaches suggest that our understanding of hierarchical neural dynamics depends
on the cumulative weight of empirical methods that are differentially suited for
diverse, synergistic ends.
More generally, the HMM forges a dialectical relationship between
neuroscience and psychology – it favours mutual enlightenment and cross-fertilization
within and between both of these disciplines, allowing insights gleaned from the one
to inform and constrain theorising and research in the other [116, 152-154]. For
neuroscientists, this requires approaches that can isolate the specific psychological
19
factors that govern different patterns of hierarchical neural activity across different
contexts. Methodologies conducive to this kind of research include meta-analyses of
task-based fMRI activation studies to characterise the functional fingerprints of
particular neural regions across different tasks [124, 125], along with empirical work
on the development of ‘cognitive ontologies’ that systematically map the distinct
relationships between well-defined cognitive functions and particular patterns of
neural dynamics [155, 156]. Furthermore, methods from developmental psychology
can allow neuroscientists to better evaluate the ways in which different developmental
trajectories lead to distinctive styles of behaviour and temperament (which, under the
FEP, correspond to different kinds of error-minimising policies, which can vary a
great deal between individuals). In this vein, longitudinal designs that combine
neuroimaging studies on human brain maturation with carefully chosen biological,
psychological and social measures might be used to explore how different
developmental contexts engender stable individual differences in perceptual biases
and active inference [157]. Comparative, cross-cultural, computational, and
dynamical approaches stemming from evolutionary psychology also allow us to study
in great detail the (epi)genetic mechanisms responsible for the emergence,
transmission, and acquisition of our species-typical adaptive priors [138, 152].
Finally, computational models and simulation studies enable us to model how
different levels of dynamical activity interact [3, 149, 158-160], allowing
neuroscientists to explore how the biobehavioural phenomena described and studied
by psychologists reflect adaptive free-energy minimisation within and across different
(evolutionary, intergenerational, developmental, and real-time) contexts. The
outcomes of such analyses can then be confirmed through real-world observations and
experimental work [4, 138].
On the other hand, the FEP offers both a biologically plausible and empirically
tractable formal theory of the human brain, mind, and behaviour to psychologists.
Traditionally, proponents have relied chiefly on computer simulations, fMRI and
EEG to apply dynamic causal models of interactions between hierarchically organised
cortical areas in order to explain perception (e.g., [79]), action (e.g., [57]), attention
(e.g., [161]), and learning (e.g., [162]). More recently, though, others have taken up
the FEP to explain a wide range of psychological phenomena (see [49, 50]), including
anxiety [163], autism [164-166], emotion [167-169], meta-cognition [170], and both
self- and other-representations [171, 172]. It also lends itself to methods that are
20
highly familiar to psychologists, such as the P300 an event-related potential that
may be used experimentally as a non-invasive, temporally sensitive proxy for surprise
[173, 174]. As discussed elsewhere, the FEP can also be exploited to explain large-
scale sociocultural phenomena, including the hierarchical dynamics of scientific
theorising itself [1, 3, 87, 140, 175].
Indeed, by incorporating the FEP, the HMM proffers a new way to explain
cognition and behaviour that can be readily applied to all levels of psychological
inquiry [1]. Although the mathematical apparatus that underwrites it may seem
inaccessible, active inference can be reduced to a simple rubric that can be leveraged
by researchers across psychology’s sub-disciplines cognition and behaviour
function together to minimise surprise. In other words, our lives are a self-fulfilling
prophecy of sorts: everything we think and do stems from the biological imperative to
optimise our predictions about causal regularities in our eco-niche, and to behave in
ways that confirm them. Like others before us (e.g., [31, 50, 60, 163, 166]), we
believe this elegant idea offers a common language to synthesise and explain diverse
findings across the discipline. More particularly, the HMM calls for integrative
hypotheses rallied around four complementary research questions: What is the
adaptive function of a particular trait? What are the intergenerational,
developmental, and real-time mechanisms that produce it? How does it manifest in
beliefs, expectations and predictions that drive self-fulfilling action-perception
cycles? And how does it emerge from particular patterns of hierarchical neural
dynamics?
As we mentioned at the outset, this modelling approach has already been used
to develop an evidence-based EST of the human capacity for depression. Combining
previous applications of the FEP with research spanning all four levels of explanation
in psychology, this second-order hypothesis suggests that depression typically reflects
an evolved biobehavioural strategy that responds adaptively to noxious social
conditions (e.g., exclusion) by minimising the likelihood of unpredictable
interpersonal exchanges. According to this view, normative depressed mood states
instantiate a risk-averse adaptive prior that reduces the likelihood of deleterious
social outcomes by causing adaptive changes in perception (e.g., heightened
sensitivity to social risks) and action (e.g., risk-averse interpersonal behaviours) when
sensory cues indicate a high degree of socio-environmental volatility ([2]; see Figure
5). As discussed elsewhere, this is a neurobiologically plausible scheme that has
21
important implications for diagnosis and treatment in clinical psychology, which can
also be leveraged by cognitive and behavioural scientists to derive more specific,
testable predictions [2].
6. Conclusion
In this article, we have proposed an EST of the human brain predicated on
neuroscience and psychology alike. Although we believe that the HMM offers a
unifying theory of the brain, cognition and behaviour that has the potential to benefit
both of these disciplines by demanding their integration, its explanatory power clearly
rests on the cumulative weight of the second-order hypotheses and empirical evidence
that it generates [4, 138]. Naturally, whether our model inspires consequential
research remains to be seen. If it does, however, it will require sophisticated,
collaborative efforts to elucidate how dynamical interactions between evolutionary,
intergenerational, developmental and real-time processes govern particular patterns of
cognition and biobehaviour; the ways in which the FEP explains such phenomena;
and the hierarchical neural dynamics responsible for producing it. Although
developing such multilevel models is fraught with complexity, a desire for simplicity
should not obstruct our pursuit of veracity. Ultimately, the Devil lies in the details.
Acknowledgments
This work is dedicated to Lucy Morrish for her essential contribution to these ideas.
We are also deeply indebted to Nicholas Allen, Luke Badcock, Axel Constant, Casper
Hesp, Jakob Hohwy, Annemie Ploeger, and Samuel Veissière for their valuable
discussions and comments on earlier drafts. K. Friston is supported by the Wellcome
Trust (Ref: 088130/Z/09/Z) and M.J.D. Ramstead by the Social Sciences and
Humanities Research Council of Canada.
22
References
[1] Badcock PB. Evolutionary systems theory: a unifying meta-theory of
psychological science. Review of General Psychology. 2012;16:10-23.
[2] Badcock PB, Davey CG, Whittle S, Allen NB, Friston KJ. The depressed brain:
An evolutionary systems theory. Trends in cognitive sciences. 2017;21:182-
94.
[3] Ramstead MJD, Badcock PB, Friston KJ. Answering Schrödinger's question: A
free-energy formulation. Physics of Life Reviews. 2018;24:1-16.
[4] Ramstead MJ, Badcock PB, Friston KJ. Variational neuroethology: Answering
further questions: Reply to comments on “Answering Schrödinger's question:
A free-energy formulation”. Physics of life reviews. 2018;24:59-66.
[5] Schrödinger E. What is life? Cambridge: Cambridge University Press; 1944.
[6] Levin S. Complex adaptive systems: exploring the known, the unknown and the
unknowable. Bulletin of the American Mathematical Society. 2003;40:3-19.
[7] Haken H. Principles of brain functioning: a synergetic approach to brain activity,
behaviour and cognition. Berlin: Springer-Verlag; 1996.
[8] Kelso JS. Dynamic patterns: the self-organization of brain and behavior.
Cambridge, MA: MIT Press; 1995.
[9] Depew DJ, Weber BH. Darwinism evolving: systems dynamics and the genealogy
of natural selection. Cambridge, MA: MIT Press; 1995.
[10] Eigen M, Schuster P. The hypercycle: a principle of natural self-organisation.
Berlin: Springer-Verlag; 1979.
[11] Holland JH. Hidden order: how adaptation builds complexity. New York: Basic
Books; 1995.
[12] Kauffman SA. The origins of order: self-organization and selection in evolution.
Oxford: Oxford University Press; 1993.
[13] Ao P. Emerging of stochastic dynamical equalities and steady state
thermodynamics from Darwinian dynamics. Communications in theoretical
physics. 2008;49.
[14] Friston K. A theory of cortical responses. Philosophical transactions of the Royal
Society B: Biological sciences. 2005;360:815-36.
[15] Friston KJ. The free-energy principle: a unified brain theory? Nature Reviews
Neuroscience. 2010;11:127-38.
23
[16] Friston KJ, Kilner J, Harrison L. A free energy principle for the brain. Journal of
Physiology-Paris. 2006;100:70-87.
[17] Breakspear M, Stam CJ. Dynamics of a neural system with a multiscale
architecture. Philosophical Transactions of the Royal Society of London B:
Biological Sciences. 2005;360:1051-74.
[18] Kaiser M, Hilgetag CC, Kötter R. Hierarchy and dynamics of neural networks.
Frontiers in Neuroinformatics. 2010;4:4-6.
[19] Power JD, Cohen AL, Nelson SM, Wig GS, Barnes KA, Church JA, et al.
Functional network organization of the human brain. Neuron. 2011;72:665-78.
[20] Park H-J, Friston K. Structural and functional brain networks: from connections
to cognition. Science. 2013;342.
[21] Sporns O, Betzel RF. Modular brain networks. Annual review of psychology.
2016;67:613-40.
[22] Barrett HC, Kurzban R. Modularity in cognition: framing the debate.
Psychological review. 2006;113:628.
[23] Barrett HC. A hierarchical model of the evolution of human brain
specializations. Proceedings of the national Academy of Sciences.
2012;109:10733-40.
[24] Fodor JA. The modularity of mind: An essay on faculty psychology. Cambridge,
MA: MIT press; 1983.
[25] Anderson ML, Finlay BL. Allocating structure to function: the strong links
between neuroplasticity and natural selection. Frontiers in human
neuroscience. 2014;7.
[26] Chiappe D, Gardner R. The modularity debate in evolutionary psychology.
Theory & Psychology. 2012;22:669-82.
[27] Colombo M. Moving forward (and beyond) the modularity debate: A network
perspective. Philosophy of Science. 2013;80:356-77.
[28] Frankenhuis WE, Ploeger A. Evolutionary psychology versus Fodor: Arguments
for and against the massive modularity hypothesis. Philosophical Psychology.
2007;20:687-710.
[29] Zerilli J. Against the “system” module. Philosophical Psychology. 2017;30:235-
50.
[30] Friston K. Learning and inference in the brain. Neural Networks. 2003;16:1325-
52.
24
[31] Clark A. Whatever next? Predictive brains, situated agents, and the future of
cognitive science. Behavioral and Brain Sciences. 2013;36:181-204.
[32] Fotopoulou A. Time to get rid of the ‘Modular’ in neuropsychology: A unified
theory of anosognosia as aberrant predictive coding. Journal of
neuropsychology. 2014;8:1-19.
[33] Markov NT, Kennedy H. The importance of being hierarchical. Current opinion
in neurobiology. 2013;23:187-94.
[34] Mesulam M-M. From sensation to cognition. Brain: a journal of neurology.
1998;121:1013-52.
[35] Mesulam M. The evolving landscape of human cortical connectivity: facts and
inferences. Neuroimage. 2012;62:2182-9.
[36] Meunier D, Lambiotte R, Fornito A, Ersche K, Bullmore ET. Hierarchical
modularity in human brain functional networks. Frontiers in neuroinformatics.
2009;3:37.
[37] Meunier D, Lambiotte R, Bullmore ET. Modular and hierarchically modular
organization of brain networks. Frontiers in neuroscience. 2010;4:200.
[38] Mumford D. On the computational architecture of the neocortex II. Biol Cybern.
1992;66:251-.
[39] Price CJ, Friston KJ. Degeneracy and cognitive anatomy. Trends in cognitive
sciences. 2002;6:416-21.
[40] Sporns O. Small-world connectivity, motif composition, and complexity of
fractal neuronal connections. Biosystems. 2006;85:55-64.
[41] Baars BJ. Global workspace theory of consciousness: toward a cognitive
neuroscience of human experience. Progress in brain research. 2005;150:45-
53.
[42] Dehaene S, Changeux J-P. Experimental and theoretical approaches to conscious
processing. Neuron. 2011;70:200-27.
[43] Evans JSB. In two minds: dual-process accounts of reasoning. Trends in
cognitive sciences. 2003;7:454-9.
[44] Stanovich KE, West RF. Individual differences in reasoning: Implications for the
rationality debate? Behavioral and brain sciences. 2000;23:645-65.
[45] Kaiser M. A tutorial in connectome analysis: topological and spatial features of
brain networks. Neuroimage. 2011;57:892-907.
25
[46] Moretti P, Muñoz MA. Griffiths phases and the stretching of criticality in brain
networks. Nature communications. 2013;4:2521.
[47] Wang Z, Dai Z, Gong G, Zhou C, He Y. Understanding structural-functional
relationships in the human brain: a large-scale network perspective. The
Neuroscientist. 2015;21:290-305.
[48] Hilgetag CC, Goulas A. Is the brain really a small-world network? Brain
Structure and Function. 2016;221:2361-6.
[49] Clark A. Surfing uncertainty: prediction, action, and the embodied mind. Oxford:
Oxford University Press; 2015.
[50] Hohwy J. The predictive mind. Oxford: Oxford University Press; 2014.
[51] Sengupta B, Stemmler MB, Friston KJ. Information and efficiency in the nervous
system—a synthesis. PLoS computational biology. 2013;9:e1003157.
[52] Friston KJ. Life as we know it. Journal of the Royal Society Interface. 2013;10.
[53] Lee TS, Mumford D. Hierarchical Bayesian inference in the visual cortex. JOSA
A. 2003;20:1434-48.
[54] Rao RP, Ballard DH. Predictive coding in the visual cortex: a functional
interpretation of some extra-classical receptive-field effects. Nature
neuroscience. 1999;2:79-87.
[55] Bastos AM, Usrey WM, Adams RA, Mangun GR, Fries P, Friston KJ. Canonical
microcircuits for predictive coding. Neuron. 2012;76:695-711.
[56] Friston KJ, Daunizeau J, Kiebel SJ. Reinforcement learning or active inference?
PloS one. 2009;4:e6421.
[57] Friston KJ, Daunizeau J, Kilner J, Kiebel SJ. Action and behavior: a free-energy
formulation. Biological cybernetics. 2010;102:227-60.
[58] Friston K. Embodied inference: or ‘‘I think therefore I am, if I am what I think’’.
In: Tschacher W, Bergomi C, editors. The implications of embodiment:
Cognition and communication. Exeter, UK: Imprint Academic; 2011. p. 89-
125.
[59] Friston K, Breakspear M, Deco G. Perception and self-organized instability.
Frontiers in computational neuroscience. 2012;6:44.
[60] Hohwy J. The selfevidencing brain. Noûs. 2016;50:259-85.
[61] Maslow AH. A theory of human motivation. Psychological Review.
1943;50:370-96.
26
[62] Ramstead MJ, Constant A, Badcock PB, Friston K. Variational ecology and the
physics of minds. Physics of life Reviews. Accepted, this issue.
[63] Campbell JO. Universal Darwinism as a process of Bayesian inference. Frontiers
in Systems Neuroscience. 2016;10.
[64] Campbell JO. Towards a unification of evolutionary dynamics: Comment on
“Answering Schrödinger's question: A free-energy formulation” by Maxwell
James Désormeau Ramstead et al. Physics of life reviews. 2017.
[65] McNamara JM, Houston AI. Integrating function and mechanism. Trends in
ecology & evolution. 2009;24:670-5.
[66] McNamara JM, Green RF, Olsson O. Bayes’ theorem and its applications in
animal behaviour. Oikos. 2006;112:243-51.
[67] Stamps JA, Frankenhuis WE. Bayesian models of development. Trends in
ecology & evolution. 2016;31:260-8.
[68] Ungerleider LG, Mishkin M. Two cortical visual systems. In: Goodale MA, Ingle
DJ, Mansfield RJ, editors. Analysis of visual behavior. Cambridge, MA: MIT
Press; 1982.
[69] Friston K, Buzsáki G. The functional anatomy of time: what and when in the
brain. Trends in cognitive sciences. 2016;20:500-11.
[70] Haken H. Synergetics: an introduction. Non-equilibrium phase transition and
self-selforganisation in physics, chemistry and biology. Berlin: Springer-
Verlag; 1983.
[71] Kiebel SJ, Daunizeau J, Friston KJ. A hierarchy of time-scales and the brain.
PLOS Computational Biology. 2008;4.
[72] Murray JD, Bernacchia A, Freedman DJ, Romo R, Wallis JD, Cai X, et al. A
hierarchy of intrinsic timescales across primate cortex. Nature neuroscience.
2014;17:1661.
[73] Kiebel SJ, Daunizeau J, Friston KJ. Perception and hierarchical dynamics.
Frontiers in neuroinformatics. 2009;3:20.
[74] Perdikis D, Huys R, Jirsa VK. Time scale hierarchies in the functional
organization of complex behaviors. PLoS computational biology.
2011;7:e1002198.
[75] Chaudhuri R, Knoblauch K, Gariel M-A, Kennedy H, Wang X-J. A large-scale
circuit mechanism for hierarchical dynamical processing in the primate cortex.
Neuron. 2015;88:419-31.
27
[76] Lerner Y, Honey CJ, Silbert LJ, Hasson U. Topographic mapping of a hierarchy
of temporal receptive windows using a narrated story. Journal of
Neuroscience. 2011;31:2906-15.
[77] Friston KJ, Rosch R, Parr T, Price C, Bowman H. Deep temporal models and
active inference. Neuroscience & Biobehavioral Reviews. 2018;90:486-501.
[78] Pezzulo G, Rigoli F, Friston KJ. Hierarchical active inference: a theory of
motivated control. Trends in cognitive sciences. 2018.
[79] Kiebel SJ, Von Kriegstein K, Daunizeau J, Friston KJ. Recognizing sequences of
sequences. PLoS computational biology. 2009;5:e1000464.
[80] Conant RC, Ross Ashby W. Every good regulator of a system must be a model
of that system. International journal of systems science. 1970;1:89-97.
[81] Kirchhoff M, Parr T, Palacios E, Friston K, Kiverstein J. The Markov blankets of
life: autonomy, active inference and the free energy principle. Journal of The
Royal Society Interface. 2018;15:20170792.
[82] Friston KJ, Levin M, Sengupta B, Pezzulo G. Knowing one's place: a free-energy
approach to pattern regulation. Journal of the Royal Society Interface.
2015;12.
[83] Calvo P, Friston K. Predicting green: really radical (plant) predictive processing.
Journal of The Royal Society Interface. 2017;14:20170096.
[84] Allen M, Friston KJ. From cognitivism to autopoiesis: towards a computational
framework for the embodied mind. Synthese. 2016:1-24.
[85] Gallagher S, Allen M. Active inference, enactivism and the hermeneutics of
social cognition. Synthese. 2016:1-22.
[86] Friston K, Thornton C, Clark A. Free-energy minimization and the dark-room
problem. Frontiers in psychology. 2012;3.
[87] Ramstead MJ, Veissière SP, Kirmayer LJ. Cultural affordances: scaffolding local
worlds through shared intentionality and regimes of attention. Frontiers in
Psychology. 2016;7.
[88] Bjorklund DF, Pellegrini AD. The origins of human nature: Evolutionary
developmental psychology. Washington, DC: American Psychological
Association; 2002.
[89] Caporael LR. Evolutionary psychology: Toward a unifying theory and a hybrid
science. Annual review of psychology. 2001;52:607-28.
28
[90] Geary DC, Bjorklund DF. Evolutionary developmental psychology. Child
development. 2000;71:57-65.
[91] Kenrick DT. Evolutionary psychology, cognitive science, and dynamical
systems: Building an integrative paradigm. Current Directions in
Psychological Science. 2001;10:13-7.
[92] Kenrick DT, Maner JK, Butner J, Li NP, Becker DV, Schaller M. Dynamical
evolutionary psychology: Mapping the domains of the new interactionist
paradigm. Personality and Social Psychology Review. 2002;6:347-56.
[93] Lickliter R, Honeycutt H. Developmental dynamics: toward a biologically
plausible evolutionary psychology. Psychological bulletin. 2003;129:819.
[94] Ploeger A, van der Maas HL, Raijmakers ME. Is Evolutionary Developmental
Biology a Viable Approach to the Study of the Human Mind? 2008.
[95] Ploeger A, van der Maas HL, Raijmakers ME. Is evolutionary psychology a
metatheory for psychology? A discussion of four major issues in psychology
from an evolutionary developmental perspective. Psychological Inquiry.
2008;19:1-18.
[96] Sameroff A. A unified theory of development: A dialectic integration of nature
and nurture. Child development. 2010;81:6-22.
[97] Scott-Phillips TC, Dickins TE, West SA. Evolutionary theory and the ultimate
proximate distinction in the human behavioral sciences. Perspectives on
Psychological Science. 2011;6:38-47.
[98] Tinbergen N. On aims and methods of ethology. Zeitschrift für Tierpsychologie.
1963;20:410-33.
[99] Bateson P, Laland KN. Tinbergen's four questions: an appreciation and an
update. Trends in ecology & evolution. 2013;28:712-8.
[100] Bateson P, Laland KN. On current utility and adaptive significance: a response
to Nesse. Trends in ecology & evolution. 2013;28:682-3.
[101] Nesse RM. Tinbergen's four questions, organized: a response to Bateson and
Laland. Trends in Ecology & Evolution. 2013;28:681-2.
[102] Hendriks-Jansen H. Catching ourselves in the act. Cambridge, MA: MIT Press;
1996.
[103] Jablonka E, Lamb MJ. Evolution in four dimensions: Genetic, epigenetic,
behavioural, and symbolic variation in the history of life. Cambridge, MA
MIT Press; 2005.
29
[104] Laland KN, Sterelny K, Odling-Smee J, Hoppitt W, Uller T. Cause and effect in
biology revisited: is Mayr’s proximate-ultimate dichotomy still useful?
science. 2011;334:1512-6.
[105] Moore DS. The Dependent Gene: The Fallacy of Nature vs Nurture. New York,
NY: Freeman; 2002.
[106] Oyama S. The ontogeny of information: Developmental systems and evolution:
Duke University Press; 2000.
[107] Pigliucci M. Do we need an extended evolutionary synthesis? Evolution.
2007;61:2743-9.
[108] Striedter GF. Stepping into the same river twice: homologues as recurring
attractors in epigenetic landscapes. Brain, Behavior and Evolution.
1998;52:218-31.
[109] Gallagher S. Enactivist interventions: Rethinking the mind: Oxford University
Press; 2017.
[110] Changeux J-P. Climbing brain levels of organisation from genes to
consciousness. Trends in Cognitive Sciences. 2017;21:168-81.
[111] Gogtay N, Giedd JN, Lusk L, Hayashi KM, Greenstein D, Vaituzis AC, et al.
Dynamic mapping of human cortical development during childhood through
early adulthood. Proceedings of the National academy of Sciences of the
United States of America. 2004;101:8174-9.
[112] Gu S, Satterthwaite TD, Medaglia JD, Yang M, Gur RE, Gur RC, et al.
Emergence of system roles in normative neurodevelopment. Proceedings of
the National Academy of Sciences. 2015;112:13681-6.
[113] Buckner RL, Krienen FM. The evolution of distributed association networks in
the human brain. Trends in Cognitive Sciences. 2013;17:648-65.
[114] Badre D. Cognitive control, hierarchy, and the rostro–caudal organization of the
frontal lobes. Trends in cognitive sciences. 2008;12:193-200.
[115] Finlay BL, Uchiyama R. Developmental mechanisms channeling cortical
evolution. Trends in neurosciences. 2015;38:69-76.
[116] Barrett LF. The future of psychology: Connecting mind to brain. Perspectives
on psychological science. 2009;4:326-39.
[117] Eidelson RJ. Complex adaptive systems in the behavioral and social sciences.
Review of General Psychology. 1997;1:42.
30
[118] Geary DC. The origin of mind: Evolution of brain, cognition, and general
intelligence. Washington, DC: American Psychological Association; 2005.
[119] Geary DC, Huffman KJ. Brain and cognitive evolution: forms of modularity
and functions of mind. Psychological bulletin. 2002;128:667.
[120] Karmiloff-Smith A. Beyond modularity: A developmental perspective on
cognitive science. Cambridge, MA: MIT Press; 1992.
[121] Karmiloff-Smith A. Development itself is the key to understanding
developmental disorders. Trends in cognitive sciences. 1998;2:389-98.
[122] Merritt M. Nativism and neurobiology: Representations, representing, and the
continuum of cognition. Review of General Psychology. 2008;12:181.
[123] Quartz SR. The constructivist brain. Trends in cognitive sciences. 1999;3:48-
57.
[124] Anderson ML. After phrenology : neural reuse and the interactive brain.
Cambridge, MA; London, England: The MIT Press; 2014.
[125] Anderson ML. Précis of after phrenology: neural reuse and the interactive brain.
Behavioral and Brain Sciences. 2016;39.
[126] Anderson ML, Kinnison J, Pessoa L. Describing functional diversity of brain
regions and brain networks. Neuroimage. 2013;73:50-8.
[127] Giard MH, Peronnet F. Auditory-visual integration during multimodal object
recognition in humans: a behavioral and electrophysiological study. Journal of
cognitive neuroscience. 1999;11:473-90.
[128] Spence C. Crossmodal correspondences: A tutorial review. Attention,
Perception, & Psychophysics. 2011;73:971-95.
[129] Bertolero MA, Yeo BT, D’Esposito M. The modular and integrative functional
architecture of the human brain. Proceedings of the National Academy of
Sciences. 2015;112:E6798-E807.
[130] Taylor PN, Wang Y, Kaiser M. Within brain area tractography suggests local
modularity using high resolution connectomics. Scientific reports.
2017;7:39859.
[131] Simon HA. The architecture of complexity. Proceedings of the American
Philosophical Society. 1965;106:63-76.
[132] Friston KJ, Price CJ. Degeneracy and redundancy in cognitive anatomy. Trends
in cognitive sciences. 2003;7:151-2.
31
[133] Mengistu H, Huizinga J, Mouret J-B, Clune J. The evolutionary origins of
hierarchy. PLOS Computational Biology. 2016;12.
[134] Bak P, Chen K. Self-organized criticality. Scientific American. 1991;264:46-53.
[135] Kauffman S. At home in the universe: The search for the laws of self-
organization and complexity: Oxford university press; 1996.
[136] Hilgetag CC, Hütt M-T. Hierarchical modular brain connectivity is a stretch for
criticality. Trends in cognitive sciences. 2014;18:114-5.
[137] Hesse J, Gross T. Self-organized criticality as a fundamental property of neural
systems. Frontiers in systems neuroscience. 2014;8:46-59.
[138] Daunizeau J. A plea for “variational neuroethology”: Comment on “Answering
Schrödinger's question: A free-energy formulation” by MJ Desormeau
Ramstead et al. Physics of life reviews. 2018.
[139] Kirmayer LJ. Ontologies of life: From thermodynamics to teleonomics.
Comment on “Answering Schrödinger’s question: A free-energy formulation”
by Maxwell James Désormeau Ramstead et al. Physics of life reviews. 2018.
[140] Veissière S. Cultural Markov blankets? Mind the other minds gap!: Comment
on “Answering Schrödinger's question: A free-energy formulation” by
Maxwell James Désormeau Ramstead et al. Physics of life reviews. 2018.
[141] Buss DM. Evolutionary psychology: A new paradigm for psychological
science. Psychological inquiry. 1995;6:1-30.
[142] Pinker S. How the mind works. New York, NY: Norton; 1997.
[143] Rabinovich MI, Simmons AN, Varona P. Dynamical bridge between brain and
mind. Trends in cognitive sciences. 2015;19:453-61.
[144] Shipp S. Neural elements for predictive coding. Frontiers in psychology.
2016;7.
[145] Sporns O. Network attributes for segregation and integration in the human
brain. Current opinion in neurobiology. 2013;23:162-71.
[146] Sporns O. Contributions and challenges for network models in cognitive
neuroscience. Nature neuroscience. 2014;17:652.
[147] Blakemore S-J. Imaging brain development: the adolescent brain. Neuroimage.
2012;61:397-406.
[148] Beer RD. Intelligence as adaptive behaviour: an experiment in computational
neuroethology. San Diego, CA: Academic Press; 1990.
32
[149] Chiel HJ, Beer RD. The brain has a body: adaptive behavior emerges from
interactions of nervous system, body and environment. Trends in
neurosciences. 1997;20:553-7.
[150] Clark A. An embodied cognitive science? Trends in cognitive sciences.
1999;3:345-51.
[151] Wilson M. Six views of embodied cognition. Psychonomic bulletin & review.
2002;9:625-36.
[152] Badcock PB, Ploeger A, Allen NB. After phrenology: time for a paradigm shift
in cognitive science. Behavioral and Brain Sciences. 2016;39.
[153] Pfeifer JH, Allen NB. Arrested development? Reconsidering dual-systems
models of brain function in adolescence and disorders. Trends in cognitive
sciences. 2012;16:322-9.
[154] Crone EA, Ridderinkhof KR. The developing brain: from theory to
neuroimaging and back. Developmental Cognitive Neuroscience. 2011;1:101-
9.
[155] Poldrack RA. Mapping mental function to brain structure: how can cognitive
neuroimaging succeed? Perspectives on psychological science. 2010;5:753-61.
[156] Price CJ, Friston KJ. Functional ontologies for cognition: The systematic
definition of structure and function. Cognitive Neuropsychology.
2005;22:262-75.
[157] Huys QJ, Dayan P. A Bayesian formulation of behavioral control. Cognition.
2009;113:314-28.
[158] Frankenhuis WE, Panchanathan K, Clark Barrett H. Bridging developmental
systems theory and evolutionary psychology using dynamic optimization.
Developmental Science. 2013;16:584-98.
[159] Friston KJ, Stephan KE, Montague R, Dolan RJ. Computational psychiatry: the
brain as a phantastic organ. The Lancet Psychiatry. 2014;1:148-58.
[160] Holland JH. Emergence: From chaos to order. . Redwood City, CA: Addison
Wesley; 1998.
[161] Feldman H, Friston K. Attention, uncertainty, and free-energy. Frontiers in
human neuroscience. 2010;4:215.
[162] Friston K. Hierarchical models in the brain. PLoS computational biology.
2008;4:e1000211.
33
[163] Hirsh JB, Mar RA, Peterson JB. Psychological entropy: a framework for
understanding uncertainty-related anxiety. Psychological review.
2012;119:304-20.
[164] Palmer CJ, Lawson RP, Hohwy J. Bayesian approaches to autism: Towards
volatility, action, and behavior. Psychological bulletin. 2017;143:521.
[165] Constant A, Bervoets J, Hens K, Van de Cruys S. Precise worlds for certain
minds: An ecological perspective on the relational self in autism. Topoi. In
press.
[166] Van de Cruys S, Evers K, Van der Hallen R, Van Eylen L, Boets B, de-Wit L,
et al. Precise minds in uncertain worlds: predictive coding in autism.
Psychological Review. 2014;121:649-75.
[167] Joffily M, Coricelli G. Emotional valence and the free-energy principle. PLOS
Computational Biology. 2013;9.
[168] Seth AK. Interoceptive inference, emotion, and the embodied self. Trends in
cognitive sciences. 2013;17:565-73.
[169] Barrett LF, Simmons WK. Interoceptive predictions in the brain. Nature
Reviews Neuroscience. 2015;16:419.
[170] Moulin C, Souchay C. An active inference and epistemic value view of
metacognition. Cognitive neuroscience. 2015;6:221-2.
[171] Moutoussis M, Fearon P, El-Deredy W, Dolan RJ, Friston KJ. Bayesian
inferences about the self (and others): A review. Consciousness and cognition.
2014;25:67-76.
[172] Moutoussis M, Trujillo-Barreto NJ, El-Deredy W, Dolan R, Friston K. A formal
model of interpersonal inference. Frontiers in human neuroscience.
2014;8:160.
[173] Kolossa A, Fingscheidt T, Wessel K, Kopp B. A model-based approach to trial-
by-trial P300 amplitude fluctuations. Frontiers in human neuroscience.
2013;6:359.
[174] Mars RB, Debener S, Gladwin TE, Harrison LM, Haggard P, Rothwell JC, et al.
Trial-by-trial fluctuations in the event-related electroencephalogram reflect
dynamic changes in the degree of surprise. Journal of Neuroscience.
2008;28:12539-45.
34
[175] Leydesdorff L. Lifting the Markov blankets of socio-cultural evolution: A
comment on “Answering Schrödinger's question: A free-energy formulation”
by Maxwell James Désormeau Ramstead et al. Physics of life reviews. 2018.
35
FIGURE 1
36
Figure 1. The Hierarchical Organization of Neural Networks. Global brain function
(i.e., cognition) can be described as the global integration of local (i.e., segregated)
neuronal operations that underpin hierarchical message passing among cortical
regions. Global integration is greatly facilitated by the hierarchical organisation of
neural networks into (relatively modular) neurocognitive mechanisms. In network
neuroscience, a neural network is modelled in terms of nodes and their connections,
which are called edges. A node is defined as an integrated unit within a network. In a
fractal or modular hierarchy, each node also comprises a smaller network of nodes
that interact among themselves at a lower nested level. In the brain, this fractal,
encapsulated hierarchy extends from neurons and macrocolumns, through to
macroscopic brain regions and distributed neural networks. According to predictive
coding theory, superficial pyramidal cells compare expectations at each level with
top-down predictions from deep pyramidal cells at higher levels, while
neuromodulatory gating or gain control of superficial pyramidal cells determines their
influence on the implicit belief updating in higher hierarchical levels. Reproduced
from [28].
37
FIGURE 2
38
Figure 2. The Free-Energy Principle. (A) The quantities that define variational free-
energy. These quantities reflect a partition of the system into its internal states,
µ
,
(e.g., states of the brain) and the quantities that describe its exchanges with the
environment; namely, sensory input,
(,)sg a
ηω
= +
, and action,
a
, which alters the
ways in which the organism samples its environment. The environment itself is
specified by equations of motion,
(,)fa
ηη ω
= +
, which describe the dynamics of
(hidden) states of the world,
η
. The term
ω
denotes random fluctuations. Both
internal and active states change synergistically to minimise variational free-energy.
This free-energy is a function of sensory input and a probabilistic representation of
hidden environmental causes (i.e., variational density),
, which is encoded by
the system’s internal states. (B): Alternative expressions for variational free-energy,
which show what its minimisation entails. With respect to action, free-energy can
only be suppressed by increasing the accuracy of sensory data (i.e., selectively
sampling data that are predicted). Conversely, the optimisation of internal states
makes the representation (i.e., variational density) an approximate conditional density
over the causes of sensory input (i.e., perception, which minimises the divergence
between the variational and true posterior density). This optimisation allows the
variational free-energy to impose a tighter bound on surprise and enables the system
to act upon the world to avoid surprising sensory and physiological states.
Reproduced from [3].
39
FIGURE 3
LEVEL OF ANALYSIS
____________
PARADIGM
____________
RELATED DISCIPLINES
DOMAIN OF INQUIRY
____________
META-THEORY
____________
EXEMPLARY HYPOTHESES
TINBERGEN’S QUESTION
____________
TEMPORAL DIMENSION
____________
SYSTEMIC DIMENSION
IV
____________
Psychological Sub-Disciplines
____________
Biology, Chemistry, Computer science,
Medicine, Pharmacology, Physics,
Other cognitive, behavioral & social
sciences
Phenotype x Environment
______________
EST
______________
Biopsychosocial models; Domain-specific
hypotheses; Dynamic systems models; Top-
down & bottom-up processes
Mechanism
______________
Real-time
______________
The Individual in context
III
____________
Developmental Psychology
____________
Biology, Chemistry, Computer science,
Medicine, Pharmacology, Physics,
Other cognitive, behavioral and social
sciences
Genotype x Environment
______________
EST
______________
Biopsychosocial models; Developmental
systems theories; Domain-specific
hypotheses; Epigenesis; Plasticity
Ontogeny
______________
Developmental time
______________
The individual
II
____________
Evolutionary Developmental
Biology/Psychology
____________
Biology, Botany, Computer science,
Ethology, Paleontology, Other
cognitive & behavioral sciences,
Zoology
Group x Environment
______________
EST
______________
Co-evolution; Epigenetic inheritance;
Exogenetic inheritance; Inclusive fitness;
Multilevel, sociality & systems models;
Mutation-selection balance; Natural
selection; Plasticity; Pleiotropy
Phylogeny
______________
Intergenerational time
______________
Groups (e.g., kin)
I
____________
Evolutionary Psychology
____________
Anthropology, Biology, Computer
Science, Ethology, Paleoanthropology,
Sociobiology, Other cognitive &
behavioral sciences, Zoology
Species x Environment
______________
EST
______________
Genetic inheritance; Inclusive fitness;
Modularity; Multilevel, sociality & systems
models; Natural selection; Social
intelligence
Adaptation
______________
Evolutionary time
______________
Homo sapiens
Note: Adapted from Badcock (2012)
INFORMATIONAL EXCHANGE
40
Figure 3. The Evolutionary Systems Theory of Psychology. Human phenotypes,
cognition and behaviour emerge from circular interactions between (general and
natural) selection and self-organization operating within and across Tinbergen’s four
domains of biological dynamics (i.e., adaptation, phylogeny, ontogeny, and
mechanism). The various fields of psychological inquiry explain this process by
formulating models of human phenomena according to four intersecting levels of
analysis: evolutionary hypotheses to explain species-typical, adaptive traits (i.e.,
evolutionary psychology); explanations for intergenerational, between-group
differences (i.e., evolutionary developmental biology and psychology); ontogenetic
explanations for individual similarities and differences (i.e., developmental
psychology); and mechanistic explanations for real-time biobehavioural phenomena
(i.e., the sub-disciplines). These levels of analysis are commensurate and
complementary: evolutionary theories tackle the ultimate questions of psychology by
explaining the adaptive properties of human cognition and biobehaviour; dynamic
systems approaches address its proximate questions by shedding light on the
intergenerational, developmental, and real-time mechanisms responsible for
producing such phenomena. This perspective encapsulates and synthesises the various
paradigms and sub-disciplines of psychology: the recursive informational exchange
between different fields of inquiry allows researchers in each subfield to constrain
their research in light of advances in others, and to integrate findings across different
levels of psychological analysis to develop unique, substantive hypotheses.
Importantly, the non-substantive meta-theory of EST, which formalises the interaction
between (both general and natural) selection and self-organisation, permeates all four
explanatory levels and imposes distinct inclusion criteria upon any derivative of the
meta-theory itself: any multi-level hypothesis derived from this EST must conform to
these two fundamental principles. Adapted from [1].
41
FIGURE 4
Level of inquiry
____________
Temporal scale
Process
____________
Systemic dimension
Free energy formulation
____________
Psychological paradigm(s)
Mechanism
(real-time)
Neurocognition
______________
Perception & action
+
Learning & attention
______________
The individual in context
()=arg min (),()()
()=arg min (),()()
()=arg min  (),()()
()=arg min  (),()()
______________
Psychological subdisciplines
Ontogeny
(developmental
time)
Neurodevelopment
______________
The individual
()=arg
min

(),()()
______________
Developmental psychology
Phylogeny
(intergenerational
time)
Neural microevolution
______________
Groups (e.g., kin)
=arg
min

()
(),()()
______________
Evolutionary developmental biology
and psychology
Adaptation
(evolutionary time)
Neural evolution
______________
Homo sapiens
=arg
min

()
(),()()
_____________
Evolutionary psychology
Informational exchange
42
Figure 4. The Hierarchically Mechanistic Mind.
()
() ( )
( )
,|
ii
Fsa m
µ
denotes the
variational free-energy of sensory data (and its temporal derivatives), , as well as
the states,
µ
, of an agent,
( )
i
ms
, that belongs to a subgroup,
sc
, of a given class,
c
. Action,
a
, regulates the sampling of sensory data; while the internal states of the
organism,
µ
, encode expectations and predictions (i.e., Bayesian beliefs) about the
mean of a probability distribution. Under this formalism, neurocognition entails two
dynamically coupled processes. The first optimises neuronal and effector dynamics
(i.e., perception and action) to attune the organism to its environment by
minimising prediction errors (resp. free-energy) based on a generative model of the
hidden causes of sensory data. The second process optimises synaptic strength and
efficacy over seconds to hours – to encode causal structure in the sensorium and the
precision of prediction errors (i.e., learning and attention). Neurodevelopment
optimises human generative models through activity-dependent pruning and the
maintenance of neural structures and connections, which are transmitted
epigenetically. Neural microevolution optimises average free-energy over generations
of individuals belonging to a subgroup (e.g., kin) of a given class (i.e., conspecifics)
via the (exo- and epi-)genetic transmission of generative models. Neural evolution
optimises average free-energy over time and individuals of a given class (i.e.,
conspecifics) through the effects of selective pressure on their generative models or
priors. Reproduced from [3].
( )
sa
43
FIGURE 5
44
Figure 5. The Evolutionary Systems Model of Depression. Under active inference,
motor and autonomic reflexes mediate action and are driven by descending
(proprioceptive and interoceptive) prediction errors (e.g., reflexes that resolve sensory
prediction errors). Action entails the attenuation of ascending prediction errors (i.e.,
the down-regulation of precision). Prediction errors cannot always be resolved
through action; in which case, the attenuation of sensory precision is suspended. This
suspension enables ascending prediction errors to revise posterior beliefs, which
improves the accuracy of top-down predictions. Here we apply active inference to
depressed mood states. Under this model, when depression is adaptive, it engenders
an increase in the precision of (bottom-up) social (interoceptive and affiliative)
prediction errors when an individual is faced with the threat of aversive interpersonal
outcomes (e.g., exclusion). This increased precision improves perceptual inference
and learning about the probable causes of social stimuli: it heightens sensitivity and
directs attention to socio-environmental cues, while reducing confidence in (top-
down) social predictions. Cognitively, this is reflected by the inhibition or suspension
of goal directed behaviour (e.g., anhedonia), along with an attentional bias toward
social cues and increased rumination about self-other relations. However, depression
becomes pathological when there is a pervasive failure of sensory attenuation, which
induces aberrant beliefs about the likelihood of social rewards and engenders negative
expectations about interactions with others (e.g., pessimism, low self-esteem). These
expectations of negative social outcomes can become self-fulfilling, because they can
lead the individual to search for sensory evidence that social rewards are improbable
and suppress exploratory or acquisitive interpersonal behaviours (i.e., those with
uncertain outcomes). Behaviourally, both adaptive and pathological depressed states
reduce uncertainty within the social world by down-regulating reward-approach
behaviours (e.g., anhedonia, social withdrawal), and by generating signalling
behaviours that elicit interpersonal support (e.g., reassurance seeking) and defuse
potential conflict (e.g., submissive behaviours). Reproduced from [2].
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Genetic and neurobiological research is reviewed as related to controversy over the extent to which neocortical organization and associated cognitive functions are genetically constrained or emerge through patterns of developmental experience. An evolutionary framework that accommodates genetic constraint and experiential modification of brain organization and cognitive function is then proposed. The authors argue that 4 forms of modularity and 3 forms of neural and cognitive plasticity define the relation between genetic constraint and the influence of developmental experience. For humans, the result is the ontogenetic emergence of functional modules in the domains of folk psychology, folk biology, and folk physics. The authors present a taxonomy of these modules and review associated research relating to brain and cognitive plasticity in these domains.
Article
Full-text available
The free-energy principle (FEP) is a formal model of neuronal processes that is widely recognised in neuroscience as a unifying theory of the brain and biobehaviour. More recently, however, it has been extended beyond the brain to explain the dynamics of living systems, and their unique capacity to avoid decay. The aim of this review is to synthesise these advances with a meta-theoretical ontology of biological systems called variational neuroethology, which integrates the FEP with Tinbergen's four research questions to explain biological systems across spatial and temporal scales. We exemplify this framework by applying it to Homo sapiens, before translating variational neuroethology into a systematic research heuristic that supplies the biological, cognitive, and social sciences with a computationally tractable guide to discovery.
Article
Full-text available
How do we navigate a deeply structured world? Why are you reading this sentence first - and did you actually look at the fifth word? This review offers some answers by appealing to active inference based on deep temporal models. It builds on previous formulations of active inference to simulate behavioural and electrophysiological responses under hierarchical generative models of state transitions. Inverting these models corresponds to sequential inference, such that the state at any hierarchical level entails a sequence of transitions in the level below. The deep temporal aspect of these models means that evidence is accumulated over nested time scales, enabling inferences about narratives (i.e., temporal scenes). We illustrate this behaviour with Bayesian belief updating - and neuronal process theories - to simulate the epistemic foraging seen in reading. These simulations reproduce perisaccadic delay period activity and local field potentials seen empirically. Finally, we exploit the deep structure of these models to simulate responses to local (e.g., font type) and global (e.g., semantic) violations; reproducing mismatch negativity and P300 responses respectively.
Article
Full-text available
Autism Spectrum Condition (ASC) presents a challenge to social and relational accounts of the self, precisely because it is broadly seen as a disorder impacting social relationships. Many influential theories argue that social deficits and impairments of the self are the core problems in ASC. Predictive processing approaches address these based on general purpose neurocognitive mechanisms that are expressed atypically. Here we use the High, Inflexible Precision of Prediction Errors in Autism (HIPPEA) approach in the context of cultural niche construction to explain atypicalities of the relational self, specifically its minimal, extended, and intersubjective aspects. We contend that the social self in ASC should not be seen as impaired, but rather as an outcome of atypical niche construction. We unpack the scientific, ethical, and practical consequences of this view, and discuss implications for how the challenges that autistic persons face should be approached.
Article
Full-text available
Motivated control refers to the coordination of behaviour to achieve affectively valenced outcomes or goals. The study of motivated control traditionally assumes a distinction between control and motivational processes, which map to distinct (dorsolateral versus ventromedial) brain systems. However, the respective roles and interactions between these processes remain controversial. We offer a novel perspective that casts control and motivational processes as complementary aspects - goal propagation and prioritization, respectively - of active inference and hierarchical goal processing under deep generative models. We propose that the control hierarchy propagates prior preferences or goals, but their precision is informed by the motivational context, inferred at different levels of the motivational hierarchy. The ensuing integration of control and motivational processes underwrites action and policy selection and, ultimately, motivated behaviour, by enabling deep inference to prioritize goals in a context-sensitive way.
Article
Full-text available
This work addresses the autonomous organization of biological systems. It does so by considering the boundaries of biological systems, from individual cells to Home sapiens, in terms of the presence of Markov blankets under the active inference scheme-a corollary of the free energy principle. A Markov blanket defines the boundaries of a system in a statistical sense. Here we consider how a collective of Markov blankets can self-assemble into a global system that itself has a Markov blanket; thereby providing an illustration of how autonomous systems can be understood as having layers of nested and self-sustaining boundaries. This allows us to show that: (i) any living system is a Markov blanketed system and (ii) the boundaries of such systems need not be co-extensive with the biophysical boundaries of a living organism. In other words, autonomous systems are hierarchically composed of Markov blankets of Markov blankets-all the way down to individual cells, all the way up to you and me, and all the way out to include elements of the local environment.
Book
While Darwinism has successfully resisted reduction to physics, the authors point out that it has from the outset developed and applied its core explanatory concept, natural selection, by borrowing models from dynamics, a branch of physics. The recent development of complex systems dynamics may afford Darwinism yet another occasion to expand its explanatory power. Darwinism's use of dynamical models has received insufficient attention from biologists, historians, and philosophers who have concentrated instead on how evolutionary biology has maintained its autonomy from physics. Yet, as Depew and Weber observe, it is only by recovering Darwin's own relationship to Newtonian models of systems dynamics, and genetical Darwinism's relationship to statistical mechanics and probability theory, that insight can be gained into how Darwinism can successfully meet the challenges it is currently facing. Drawing on recent scholarship in the history of biology, Depew and Weber bring the dynamical perspective to bear on a number of important episodes in the history of the Darwinian research tradition: Darwin's "Newtonian" Darwinism, the rise of "developmentalist" evolutionary theories and the eclipse of Darwinism at the turn of the century, Darwinism's struggles to incorporate genetics, its eventual regeneration in the modern evolutionary synthesis, challenges to that synthesis that have been posed in recent decades by molecular genetics, and recent proposals for meeting those challenges. Bradford Books imprint