Downloaded from http://rstb.royalsocietypublishing.

org/ on August 4, 2017

What does the interactive brain

hypothesis mean for social neuroscience?
rstb.royalsocietypublishing.org A dialogue
Hanne De Jaegher1,2, Ezequiel Di Paolo1,2,3 and Ralph Adolphs4,5,6
1
Department of Logic and Philosophy of Science, IAS-Research Centre for Life, Mind, and Society, University of
Research the Basque Country, Av. De Tolosa 70, 20018 San Sebastián, Spain
2
Department of Informatics, Centre for Computational Neuroscience and Robotics, and Centre for Research in
Cite this article: De Jaegher H, Di Paolo E, Cognitive Science, University of Sussex, Brighton, UK
3
Adolphs R. 2016 What does the interactive Ikerbasque, Basque Foundation for Science, Bilbao, Spain
4
brain hypothesis mean for social neuroscience? Computation and Neural Systems, 5Division of Humanities and Social Sciences, and 6Division of Biology and
Biological Engineering, California Institute of Technology, Pasadena, CA 91125, USA
A dialogue. Phil. Trans. R. Soc. B 371:
20150379. HDJ, 0000-0001-7273-6410
http://dx.doi.org/10.1098/rstb.2015.0379
A recent framework inspired by phenomenological philosophy, dynamical
systems theory, embodied cognition and robotics has proposed the interac-
Accepted: 6 January 2016 tive brain hypothesis (IBH). Whereas mainstream social neuroscience
views social cognition as arising solely from events in the brain, the IBH
argues that social cognition requires, in addition, causal relations between
One contribution of 15 to a theme issue
the brain and the social environment. We discuss, in turn, the foundational
‘Attending to and neglecting people’.
claims for the IBH in its strongest form; classical views of cognition that can
be raised against the IBH; a defence of the IBH in the light of these argu-
Subject Areas: ments; and a response to this. Our goal is to initiate a dialogue between
neuroscience, cognition cognitive neuroscience and enactive views of social cognition. We conclude
by suggesting some new directions and emphases that social neuroscience
might take.
Keywords:
social neuroscience, interactive brain
hypothesis, social interaction, social cognition,
causality, participatory sense-making 1. Introduction
In the context of recent advances in social neuroscience, particularly the avail-
Author for correspondence: ability of methods for investigating brain activity in complex situations,
Hanne De Jaegher including live interaction, novel research questions emerge. Do brains in inter-
action function the same way as in non-interactive situations? Are social
e-mail: [email protected]
interactions simply more complex scenarios involving more dynamical kinds
of processing, but not essentially different from non-interactive cases? Or do
live social engagements between people engender novel phenomena, which
prompt us to reconsider brain function?
A recent proposal strongly vouches for the last option: brains work differ-
ently in social interactive situations. And, moreover, the dynamics of the
interaction itself play important roles in cognitive function [1].
We investigate some implications of this view by raising critical questions.
This will take the form of a dialogue between the authors of this paper—not
to settle the issues or to iron out wider conceptual disagreements once and
for all, but to progress, if not to a final common ground, then hopefully to
some useful inroads into it.

2. The interactive brain hypothesis and how social

neuroscientists should view it (H.D.J. and E.D.P.)
Two of us (H.D.J. and E.D.P.) argue that embodiment and interaction are partly,
but fundamentally, constitutive of social cognition. This view is captured by the
interactive brain hypothesis (IBH): ‘The IBH . . . proposes that social interaction pro-
cesses play enabling and constitutive roles in the development and in the ongoing

& 2016 The Author(s) Published by the Royal Society. All rights reserved.
 Downloaded from http://rstb.royalsocietypublishing.org/ on August 4, 2017

shadow of sensor of participant 1 2

participant 1 participant 1

rstb.royalsocietypublishing.org
tactile
feedback

Phil. Trans. R. Soc. B 371: 20150379

fixed object sensed
participant 2 by participant 1

Figure 1. Set-up for perceptual crossing experiment. Both participants are isolated, each controlling the position of a sensor along a shared virtual one-dimensional
line using a computer mouse. The squares on each side of the line represent the objects that can be sensed by each participant, respectively. Objects are identical in
size. When the sensor touches an object the participant gets a tactile feedback on the finger (green circle). Each participant can sense only three objects, a static one
(black square), the sensor of the other participant (red square) and a ‘shadow’ object that copies exactly the movement of the other’s sensor at a fixed distance (blue
square). (Copyright & 2010 De Jaegher et al. [2]. Licenced under Creative Commons Attribution 3.0 Unported, http://creativecommons.org/licences/by/3.0).

operation of brain mechanisms involved in social cognition, brain, in isolation, make a determining contribution. It is
whether the person is engaged in an interactive situation or not’ ([1], the relations between such external events, the body, and
p. 2). We use here the terminology introduced in [2]: an enabling the brain that matter, or rather, it is only within these
factor is causally necessary for a phenomenon to occur, while a relations, which are not merely contextual, that we can
constitutive factor is part of what makes the phenomenon make sense of brain function during most social cognition.
what it is. While a hypothesis rather than a scientific claim, the In such cases, there is no factoring out of such extra-brain
IBH stimulates a novel perspective on how social neuroscientists elements without removing at the same time something
should construe information processing that generates social be- essential to social cognition as such.
haviour. The unit of analysis is no longer delimited to the brain, What are those instances of social cognition where we
but broadened to include aspects of the social environment with expect extra-neural relational patterns to play a constitutive
which the brain interacts or has interacted. role? Certainly, those involving direct interaction with
There is a range of hypotheses about the role that embodi- others. But also those instances involving the presence of
ment and social context play in social cognition. The weakest others to various degrees ( physical, virtual, etc.) and which
claim is that social interaction is methodologically useful in predispose us to engage interactively, even if we do not or
social neuroscience, as it provides ecological validity and cannot actualize such dispositions.
engages research participants. Another weak claim is that Consider an example where, we argue, social interaction
social interaction needs to be considered as providing important plays a constitutive role in the performance of a social task.
contextual modulation of the social brain. To our knowledge, This is the perceptual crossing experiment by Auvray et al. [4]
nobody disagrees with these claims. They are now actively (figure 1). Two blindfolded participants are told to freely
pursued in social neuroscience, as evidenced in the other move a sensor (a computer mouse) along a shared virtual line.
contributions in this issue; we do not further treat them here. In this virtual ‘world’, each participant can encounter two differ-
A stronger claim is that social interaction facilitates particu- ent kinds of objects, one fixed in space and two moving objects.
lar kinds of brain processes: that is, there is a strong enabling One of the moving objects corresponds to the scanning sensor of
role for social interaction [2]. For instance, it is well known the other participant, and the other is attached to this sensor at a
that development in the absence of social interaction (severe fixed distance, like a ‘shadow’. In terms of their trajectories, the
social deprivation, in humans or other mammals) results in a moving objects are indistinguishable. Whenever her sensor
highly abnormal brain with highly abnormal cognition [3]. encounters an object on this line, the participant receives an
This fact also suggests important constraints on the design of on/off tactile stimulus—a tap on the finger, which is the same
artificial cognitive systems. To our knowledge, this claim is for each kind of object. This situation is symmetrical for both
also uncontentious. We do not discuss this enabling aspect of participants. Note that when a participant’s sensor encounters
social cognition here either. the shadow of the other participant, only the first participant
The hypothesis we discuss here concerns an occurrent will receive a tactile stimulus. When the two sensors meet,
instance of social cognition. We claim that a normal adult both participants receive a stimulus simultaneously.
human brain in isolation is insufficient for a typical instance of Participants are instructed to click the mouse button when-
social cognition. Of course, we acknowledge that the brain ever they judge they are in contact with the other participant.
plays a large, and probably major, role in social cognition.1 As a result, statistically, mouse clicks tend to concentrate on
But processes occurring in that brain at the time of an each other’s sensors (65.9% of clicks) and not on the identically
instance of social cognition are, in the typical case, not fully moving shadow objects (23%). This means that participants
constitutive of social cognition: additional events are also can find each other by ‘perceptually crossing’ their scanning
required. Those additional events involve relations between activities. However, the authors find that the probability of
the brain and ( parts of ) the rest of the world. It is important clicking following a stimulus is approximately the same
to note that we are not claiming that events external to the whether this stimulation comes from the other’s sensor or its
 Downloaded from http://rstb.royalsocietypublishing.org/ on August 4, 2017

shadow. Auvray et al. explain this result in terms of the collec- there is a more or less continuous stream of causal interaction 3
tive dynamics of the interactive configuration. Each participant between brain and the world, and, in the case of the social

rstb.royalsocietypublishing.org
attempts to re-scan many times an object that is apparently world, ‘outputs’ from a brain influence ‘inputs’ (i.e. one
moving, by making back and forth mouse movements. When person influences another) in such a tightly coupled way that
this scanning involves an encounter between both sensors, it becomes impossible to distinguish input from output.
the participants tend to continue this activity for a long time. Instead, say advocates of the IBH, one should treat the whole
By contrast, when one of them is scanning the other’s system (two or more people interacting) as a single, dynami-
shadow, which is moving independently of this scanning cally coupled system. Cognition is constituted by the events
movement, this ‘encounter’ is short-lived. This means that in my brain, the events in the other person’s brain, and the
the dyadic system is organized such that sensor–sensor causal relations between them: the whole system matters.
encounters are more frequent than sensor–shadow encounters, I am sympathetic with what motivates the IBH, and of
which explains why participants can ‘find each other’ in spite course I agree that the classical computer metaphor is

Phil. Trans. R. Soc. B 371: 20150379

of the sensory ambiguity [2]. For this reason, we consider inadequate. Indeed, modern cognitive neuroscience acknowl-
that the interaction process here plays a constitutive role in edges that much of cognition is ‘active’: we continuously
how the social task is organized and performed. move our eyes to redirect visual input [7], we continuously
As this and other examples show, we cannot always assume shift attention to redirect what information is processed, we
that social interactions are mere inputs to cognitive systems, but continuously interact with the environment, especially in
rather, in general, we need to consider as relevant both individ- the case of a social encounter [8,9]. Current enthusiasm
ual and interactive mechanisms from the start [2]. The IBH about Bayesian or predictive coding approaches [10,11]
proposes that social interactions are at the basis of social skills reflects this acknowledgement. But I am confused about
more often than usually assumed. Therefore, the point of the how and why one would need to adopt the IBH, as opposed
IBH is to provide a research guide to specify which social to one of its weaker forms, to incorporate these facts. To make
events and social relations, as well as what kind of brain my confusion more transparent, consider three properties of a
activity, matter and how, to particular instances of social cogni- person that we might want to understand: their observable
tion. As methodologies for investigating brain activity during behaviour, their cognition and their conscious experience.
free interactions continue to develop, it is necessary to theorize Let’s consider these in the light of an experiment that IBH
about these questions, especially as we approach everyday advocates have mustered: the experiment by Auvray et al.
situations involving emergent collective patterns, jointly [4], discussed above (cf. figure 1). In this study, participants
authored actions, and multiple brains and bodies in coordi- find each other’s sensors, even though they themselves
nation. With the IBH we question the tacit assumption that appear unaware of whether they are finding a sensor or a
the best way to approach this challenge is to try to understand shadow. One could quibble about various aspects of this
brain function in isolation by assigning to all extra-neural vari- example as a good example of the IBH in action (it is not par-
ation a role exclusively as inputs or outputs. Instead, we ticularly ‘ecologically valid’; the fact that the subjects cannot
propose to leave open to investigation the conditions under explicitly distinguish sensor from shadow does not show that
which this assumption may be valid as a limiting case. their brains are not representing this distinction, just uncon-
sciously; etc.), but let us take it as an example nonetheless.
Now the question is: what exactly does this experiment
3. Arguments against the interactive brain show? It shows that coupled causal interactions between
two people are required to explain something—what is
hypothesis (R.A.) that? As far as I can tell, it is only a certain aspect of behav-
The other one of us (R.A.) disagrees with the above view iour. Yes, the behaviour of the system cannot be explained
because it seems to disregard a natural partition to causal only by events in individual brains. Me riding a bicycle
interactions in the world. That is, there is a much more also cannot be explained only by events in my brain. Much
direct and dense set of causal interactions internal to the of our behaviour comes about through complex causal
brain, than between brain and external environment. Dis- interactions between our brains and the world, and social
regarding this fact leads to a concern that the IBH renders behaviour is no exception.
unclear the specific role of social neuroscience (as opposed Now, to see the limits of this example, ask yourself what
to social psychology or sociology or behavioural studies of the answer would be with respect to conscious experience. Is
crowd behaviour) in explaining social behaviour. the coupled system of two people interacting supposed to be
In order to understand cognition, we need to partition aware of the distinction between sensors and shadows?
cognitive systems. In particular, we need to partition them Surely not. One good reason is that whatever it is about the
into those parts that should be analysed as inputs to the system that is generating the behaviour of the system under
system, those that are the outputs from the system, and consideration here seems far too meager an example of pro-
those that are actually implementing the cognition. We do cessing to count as cognition. The two people’s brains in
the same thing with computers running programs: there is the experiment are each processing information so as to gen-
a causal interaction with the world that can be treated as erate cognition and conscious experience. The entire system
input, there is processing internal to the computer, and generates a unique behaviour, but that is it. There is not in
there is causal interaction with the world that can be treated addition any kind of collective ‘cognition’ generated for the
as output. same reason that there is not in addition any kind of collec-
The IBH follows the more dynamical systems view that tive consciousness generated (intuitions here may of course
much of situated cognition has adopted [5,6], and claims that diverge; see [12–14]). The reason is that the causal inter-
this partitioning does not reflect how cognition actually takes actions at the systems level that explain the behaviour are
place in the world. Unlike the classical computer metaphor, far too thin to constitute cognition. Cognition requires an
 Downloaded from http://rstb.royalsocietypublishing.org/ on August 4, 2017

extraordinarily dense and complex set of causal interactions that might ultimately ground what it is that the brain’s 4
that are part of an extensive processing architecture. What- representations are about; see the concluding section).

rstb.royalsocietypublishing.org
ever exactly one’s view on what cognition is, it is far more It may well be that ‘causal density’ as I have described it
than a reflex, far more than a fixed action pattern, and instead above is just a proxy for another property that is more fundamen-
is a highly inferential, context-dependent and flexible form of tal to cognition. Perhaps it is computational complexity. Perhaps
information processing. So far as we know, only the brains it is something that requires much more clarification, like ‘own-
of certain animals can generate examples of it. The causal ership’ for a person. Perhaps it is something like ‘manipulability’
interactions between those brains and the rest of the world that could ground our concept of causation; one could imagine
are simply too ‘thin’. manipulability as experimental manipulability by us, or as bio-
This brings me to my core objection against the IBH as a logical manipulability in terms of what is accessible to
hypothesis about cognition: in widening the causal base of evolution or development. Much more debate will be needed
cognition, it negates a distinction that is critical to understand to develop arguments for any of these, but for present purposes

Phil. Trans. R. Soc. B 371: 20150379

cognition, the distinction between those causal events causal density serves as a simple and intuitive metric.
internal to the brain, and those constituting a brain’s relations There is no question that there are collective social phenom-
with the rest of the world. This distinction is huge. A brain’s ena that emerge from causal relations between multiple people
80 billion or so neurons, or a much larger number of compart- and their shared environment. Group behaviours, politics and
ments of those neurons (opinions vary on what to consider the stock market are all examples of this. Each has its proper
the basic processing units in the brain) all causally interact, domain of study required to explain phenomena that emerge
at multiple time scales. A single cortical neuron gets input at those macroscopic levels. Disciplines like political psy-
from perhaps 10 000 other cells and participates in networks chology and economics tackle that. None of these truisms,
at local and global scales. Needless to say, we do not under- however, challenge the neuroscience of social cognition: the
stand exactly how information is processed in the brain, but proper domain of study to explain social cognition is the indi-
clearly it depends on very dense, very complex, sets of bio- vidual brain. The social neuroscientist does not also need to be
logical causal interactions between networks of cells in the studying the stock market. Even if the stock market were a cog-
brain. By comparison, the path of causal inputs to the brain nitive system (unlikely in my view), then this still does not
(or outputs from it to the world) is extremely sparse. There undermine the study of individual brains to understand
are only a million axons from the eye going into the brain. social cognition. If eventually we engineer a computer so
There are many more axons between processing stages advanced that it has cognition, we would not also need to
deeper in the brain (indeed, there are more axons from understand the cognition happening inside the brains of the
higher-level brain regions back down to lower-level regions, people who built that computer. The reasons for the distinc-
such as the ‘feedback’ from visual cortex to thalamus, than tions in all these examples are the same: they are just separate
in the opposite ‘feed-forward’ direction). In the auditory systems. Perhaps there is stock market cognition, AI and
system, there are only about 3000 cells that transduce sound human cognition. I can study them individually, and add
in each ear. Yet from this, through considerably more com- chimpanzee cognition and pet dog cognition. What is impor-
plex processing internal to the brain, we can hear music. tant is to partition the world into systems, the internal
Inside the brain there is cognition and conscious experience. constituents of which interact in ways that do not require also
Outside the brain there are causal relations, and indeed knowing how they interact with the rest of the world. Our
some of those causal relations can be fairly complex and reci- understanding of the world requires such partitioning, and
procal. But they are not part of the brain’s computations and the disciplines that have arisen to explain how the world
they are also not constitutive of cognition. works reflect those partitions. The cognitive holism that the
Indeed, cognition does not require concurrent causal IBH envisions erases real distinctions and, if carried through
interaction with the world at all: we can think, calculate the all the way, would make understanding cognition intractable
product of two numbers and generate images with our eyes because it is everywhere.
closed in a quiet room, or while dreaming. Moreover, a lot
of such internal cognition is social: we think and dream
about other people all the time. All of the occurrent causal 4. In defence of the interactive brain hypothesis
events that constitute such examples of cognition must be
limited to what happens inside our skulls (or perhaps also (E.D.P. and H.D.J.)
our bodies). The IBH deals with this problem by including Let us consider the IBH at its most radical: the claim that the
in its substrate for cognition not only those causal relations dynamics of social interaction play constitutive roles in social
occurring between brain and the world at the time of the cog- cognition. The developmental version seems less controver-
nition, but also relations between brain and the world that sial, although its implications are not trivial (see e.g. [15]).
happened in the past. This has always struck me as a rather In fact, for any environmental factor to developmentally
desperate move that brings us back to how we began this shape the function of brain processes, it cannot be systemati-
section. Yes, of course, cognition depends on the history of cally just an informational input. To play an informational
causal interactions with the world. Had my causal history role strictly requires the stationary functional context of the
been very different, my cognition would also be very differ- system for which a signal serves as an input. Hence, the
ent. But the reason for this difference should be apparent: developmental IBH also necessitates the possibility of inter-
the only mechanism by which my cognition could be chan- action dynamics playing more than just informational roles.
ged in the light of a different causal history is through the Turning to the constitutional version of the IBH, we first
brain. Change my causal history, you change the brain and must stress what the claim is. We defend that the dynamical
hence cognition. All this shows is that causal history is one processes involved in social interactions, which implicate not
particular kind of ‘input’ to the brain over time (albeit one just extra-neural processes but also relational processes
 Downloaded from http://rstb.royalsocietypublishing.org/ on August 4, 2017

between participants (and their surroundings) can be a consti- distinct brains at frequencies more than one order of magnitude 5
tutive part of the processes of social cognition as they are faster than the interactive movements?

rstb.royalsocietypublishing.org
enacted by the individual participants involved [1,2]. The Leaving aside the question of what role (if any) might be
strong version of the IBH simply hypothesizes that this played by such cross-scale synchronization, the evidence
possibility is, in general, a widespread plausibility. This can suggests that interaction patterns produce an entanglement
be criticized in two ways: the extension from possibility to between the brains of the participants. Internally, the wave
plausibility is not empirically warranted, or the very claim of influence across various temporal and spatial scales may
of possibility is wrong. The criticisms of the previous section travel from low to high frequencies via variations in neuronal
are centred on the second option. If this possibility claim is excitability [22 –25]. These top-down effects, evidenced also
wrong, then the constitutive version of IBH falls with it and in arrhythmic cross-frequency couplings [26], have been
only the developmental version remains. associated with different cognitive phenomena, notably
We discuss three aspects in support of the constitutive with the control of visual attention [27–29]. From here it is

Phil. Trans. R. Soc. B 371: 20150379

claim: (i) the non-decomposability of neural and extra- not a big leap to suggest that inter-brain synchronization at
neural processes during interaction, (ii) the functional role high frequencies [21,30,31] is due to high-to-low frequency
of interaction dynamics, (iii) the irreducibility of interactive integration and low-to-high frequency enslavement, with
phenomena such as meaning generated in social interaction the difference that, instead of slow neural oscillations, the
and the co-authorship of interactive acts. processes ‘at the top of the hierarchy’ are the emergent
rhythms of social interaction. This seems the simplest
interpretation of the data, not the only possible one. But
(a) Entanglement until disproven, it is not a bad idea to follow Occam’s advice.
The brain-internal causal density argument discussed above This interpretation is in line with calls to investigate the
seems compelling only if we assume that brains are ‘nearly braided coordination of neural, behavioural and social pro-
decomposable’ systems [16] with respect to body and environ- cesses [32,33]. It also coheres with cumulative evidence of the
ment. Nearly decomposable systems interact with other brain–body as an interaction-dominant system (the opposite
systems without losing their functionality or altering signifi- of a nearly decomposable one), based on findings of corre-
cantly their internal causal relations. Considering the brain lations of neural and behavioural variability across a wide
in this way means to treat its couplings with body and envi- range of time scales [34,35]. Interaction-dominant systems
ronment as inputs. There are solid arguments against the are characterized by the causal inextricability of the various pro-
disposability of body and environment for normal brain cesses involved, as well as the unpredictability of the behaviour
function. Some are based on the abundant evidence of the of the whole from knowledge of the parts in isolation. Evidence
entangled neural, body and environmental dynamics in a of interaction-dominance has also been found to involve extra-
wide range of cognitive performance [17]. A more conceptual neural factors, e.g. in agent–tool systems [36] and during social
argument is Thompson and Cosmelli’s critique of the brain- interaction [37–40].
in-the-vat thought experiment [18]. They argue that it is In view of this evidence, our suggested explanation of
inconceivable for a brain to retain its functionality if separated multiscale inter-brain synchronization engendered by emer-
from body and the world. gent interaction patterns seems plausible. This allows us to
We could assume that the causal support given by body and make two points. The first, which is negative, is that this evi-
environment does not constrain neural function and so, at least dence casts doubt on the causal density argument against the
functionally, the brain could be considered independent. But IBH. Indeed, it would seem that at least under some con-
even in such a case, we cannot infer near-decomposability ditions, brain, body and interactive activity are under
from the evidence of inner causal density alone. We must also mutual causal influence, despite (or thanks to) the density
demonstrate that inner processes are not dominated, shaped of causal linkages in the brain. The second, neutral point
or regulated in their function by external processes, i.e. that raised by entanglement is that if social interaction can have
coupling with the world does not involve nonlinear interactions such an influence on brain activity, then it is clearly possible
across a significant range of timescales. In short, the inner com- that the interactive influence on brain dynamics during
plexity of the brain, which is of course undisputed, is not a instances of social cognition is of a functional kind. To this
deciding factor between the two interpretations discussed positive possibility we turn next.
here: interactional processes as input versus interactional
processes as constitutive of social cognitive function.
Consider the evidence of the entanglement of brain (b) Functional roles for social interaction
and interaction dynamics observed in dual-scanning Evidence of entanglement suggests that we should discard
experiments [19]. According to Simon ([20], p. 204) a nearly the view of interaction patterns as mere inputs to compart-
decomposable system ‘[separates] the high-frequency dyna- mentalized brain processes. But it does not yet say whether
mics of a hierarchy—involving the internal structure of the this more complex picture is sufficient to warrant the interpret-
components—from the low-frequency dynamics—involving ation that interaction dynamics can be constitutive of the
interactions among components’. But this precisely is not the functional aspects of social cognition. What kind of cognitive
case during inter-brain synchronization in live interactions. ‘work’ could be done by social interaction? This question
Using dual electroencephalogram (EEG) scanning during an imi- cannot be answered in general terms. Each case will merit its
tation task with interactors visibly moving their hands freely and own response. But at least in some cases we can provide a
allowing spontaneous synchrony and turn-taking, Dumas et al. story. This is the importance of experiments like perceptual
[21] found inter-brain phase synchronization in the alpha–mu crossing, mentioned before [4]. In it the ecological situation is
(8–12 Hz), beta (13–30 Hz) and gamma (31–48 Hz) bands. maximally simplified without eliminating a key factor: the
How can social interaction affect neural oscillation phase in two free control of the social interaction dynamics by the
 Downloaded from http://rstb.royalsocietypublishing.org/ on August 4, 2017

participants. We do not think that this is an example of ‘just properties of either agent on its own, or to a linear aggregation 6
behaviour’, if by this is meant that no sense-making is of these. The task is transformed from type II to type I—it is

rstb.royalsocietypublishing.org
involved. It is a powerful exemplar that, thanks to its simpli- solved—by the interaction process. There is no need for the par-
city, can help us think differently about individual and ticipants’ brains to represent the distinction between sensor
interactive processes in more complex cases. and shadow at all to solve the task. The participants reap the
The perceptual crossing task is anything but simple. benefits and deal with quasi-disambiguated, type I stimuli:
It is only after the performance has been explained that it ‘if it moves but stays nearby (repeated crossings), then click’.
appears so. In fact, described in strict computational terms, If a process instantiates the solution to a cognitive problem
it is a highly ambiguous, type II problem [41], i.e. a problem it constitutes an instance of cognition. This is what social
where stimuli must be actively discriminated spatially and interaction does in perceptual crossing.
qualitatively using only temporal and proprioceptive cues Further empirical confirmation that social interaction can
(all ‘objects’ found in the virtual space produce the exact play constitutive roles in social cognition is provided by a vari-

Phil. Trans. R. Soc. B 371: 20150379

same on– off tactile stimulation). The task set to the partici- ation of the perceptual crossing experiment by Froese et al. [42].
pants is no less complex than typical discrimination tasks. This variation involves a more sophisticated social cognitive
In fact, it is untypically difficult, because the two moving faculty, that of recognizing the other as an agent. The authors
objects that the participant can encounter (the other partici- found that if they instructed the participants in a perceptual
pant’s sensor and shadow) move identically. Distinguishing crossing task to cooperate as a team in finding each other,
them would require, from an individual perspective, not through several repeated interactions, the probability of click-
only a complex strategy for testing socially contingent reac- ing on the other’s sensor grew to twice as much as that of
tions in these objects but also measuring these reactions in clicking on the shadow object (in the original experiment
a highly ambiguous sensory space. these probabilities are approximately the same; the difference
The fact that this computationally tough problem can be in absolute clicks is given by the interactively skewed
resolved with relative ease in the presence of interactive probability of encountering each object). This means that par-
dynamics does not make this too meager an example of ticipants develop a better way of ‘telling’ if they are in
social cognition. That its difficulty deflates dramatically contact with another agent, for instance, by using prototypical,
once we understand the collective dynamics is precisely the co-authored regularities in the interaction patterns, which in
theoretically pregnant point of the experiment. turn would confirm the direct co-presence of the participants.
The type II regularities in the sensory signals that could help Some pairs developed clear turn-taking patterns. As the
distinguish sensors from shadows are statistically invisible in authors say, these co-authored patterns turned ‘the individual
the absence of a systematic sampling strategy. One way to epistemic task of agency detection into a social pragmatic task
solve the task is to implement a strategy that successfully trans- aimed at mutual coordination’ ([42], p. 4). As mutual recog-
forms type II signals into type I data, i.e. into non-relational and nition is a fundamental aspect of a wide range of cases of
unambiguous inputs [41]. A type I signal by itself contains social cognition, its social constitution in as simple a situation
enough information to determine the next course of action as perceptual crossing is suggestive of an interactive sharing
towards the resolution of the task. This route towards solving of socio-cognitive processes in other cases.
the task involves a biased sampling of the raw sensory streams,
such that the task is rid of its ambiguities. Were this biased
sampling to be implemented in the participants’ brains, we (c) Irreducibility
would not hesitate to acknowledge that the processes involved The examples of entanglement and cognitive functionality evi-
are responsible for the core cognitive workload required to denced in at least some cases of social interaction are indicative
solve the problem. In other words, to solve the perceptual cross- of phenomena that cannot be fully determined by what goes
ing task using this strategy is to find the way of biasing the on in the individual participants’ brains and bodies. But
sampling of sensory inputs so as to transform them from type there is also an important sense in which the acts and meanings
II into type I. that are cognized about in social cognition are themselves part
Now this sampling bias is precisely what is achieved by of emergent interactive phenomena, and not simply a sum-
the collective dynamics, i.e. by the interactive combination mation of individual attributes (such as moods, intentions,
of individual strategies. As shown by Auvray et al. [4], the etc.). To cognize socially, in the enactive understanding of the
interaction process biases the statistical presentation of sen- term, is to skillfully engage in the multiple demands and
sory stimulus towards much more frequent encounters with possibilities of the social world, many of which are directly
the other participant’s sensor, and not the shadow. Mutual or indirectly emergent from social interactions. During interac-
scanning of sensors produces mutual sensory feedback and tive encounters, this skillful engagement does not in general
a permanence in the shared spatial region. This is more necessitate tracking evidence that allows us to infer the
stable than one participant unidirectionally scanning the mental states of others. Often such mental states do not directly
shadow of the other, who is unaware of this scanning and impact on what is immediately required at the present
continues the search in other areas; thus, the scanned moment, or they are directly evident in the acts and responses
shadow object quickly disappears. This is not done con- of the others. Crucially, in such situations of interactive engage-
sciously by the participants but by the relation between their ment, it is not individual cognizing and behaviour that
correlated movements. This cognitive work is neither given sufficiently determines the relevant phenomena: both social
externally (in which case, we would be right in attributing acts and meanings are constituted socially and during the
the solution of the problem to a third party) nor is it generated interactive encounter—think of a handshake, or the act of
internally within the participants’ brains. It is produced by the giving/receiving an object. The interactive constitution
self-organized collective dynamics in which they participate of social acts and meanings is a joint cognitive process that
but whose properties do not correspond to individual necessitates, but is under-determined by, individual cognition;
 Downloaded from http://rstb.royalsocietypublishing.org/ on August 4, 2017

the remainder of determination is given by the relational one that is more amenable to scientific observation and 7
dynamics of the interactive encounter. We call this process experimental manipulation.

rstb.royalsocietypublishing.org
participatory sense-making [43].
Consider escalation as a simple example of what we
mean by irreducibility in the case of interactive phenomena.
Typically, escalation involves an antagonistic pattern of 5. Response (R.A.)
interaction, sustained in time, increasing in intensity, and My co-authors are correct in taking me to reject the possi-
potentially spiralling out of control. Past conflictive inter- bility, not just the plausibility, of the IHB in what I wrote in
actions can predispose the onset of escalation even when §3. However, this depends on three concepts, whose relations
the interaction partners do not individually intend to were argued for only in the vaguest terms; let me say a bit
engage in an antagonistic exchange (see for instance, [44]). more about them here in responding to the arguments of
Sometimes escalation arises spontaneously as a result of §4. The three concepts are cognition, causation and explana-

Phil. Trans. R. Soc. B 371: 20150379

interactive patterns. An example is given by Shergill et al. tory domain (for a discipline). Very roughly, my idea was
[45]. The interaction is quite minimal and involves partici- that features of causation (causal density) put constraints
pants applying a downward push with a finger on the on cognition, and that this put constraints on what social
other one’s hand; an operation that is then repeated, alternat- neuroscience can study in order to understand cognition.
ing roles. Participants are instructed to apply the same force None of us has defined what we mean by cognition,
as the perceived external force applied to them. As the although I alluded to two other concepts as perhaps provid-
turns alternate, the absolute amount of force escalates. ing some reference: computation and consciousness. If
The suggestion is that participants tend to underestimate cognition is taken to be processing that could at least poten-
the force they apply: self-generated force is perceived as tially contribute to the contents of our conscious experience
weaker than externally generated force. Participants compen- [50], then I found it implausible that the coupled system of
sate by increasing force in the next round, resulting in the Auvray experiment had cognition. H.D.J. and E.D.P. do
escalation. Simple as this explanation is, it provides a good not seem to have the same concept of cognition, and instead
model for more general situations: escalation can originate theirs may rely more on how coordination (of behaviour, of
unintentionally by a reciprocal configuration in which a per- meaning) arises from social interactions. I am unclear on
ceived mismatch between one’s own ‘moves’ and those that what my co-authors mean by cognition; but I am also unclear
we are subject to by external action. on what I myself mean. So I think this is one obvious way
The full explanation in this case combines individual forward in our discussion: insofar as all of us are vague on
and relational factors: a tendency to underestimate one’s what we mean by cognition, it opens the way for a revised
own force and the configuration of the alternating interaction understanding of this concept that might reconcile our
pattern. Remove either factor and the explanation fails. apparent differences.
Moreover, the explanation does not involve any high-level How would one go about revising a concept of cognition?
awareness of the escalating pattern or deliberate inten- One place to begin would be by taking the term as relative to
tion to initiate escalation. Like the perceptual crossing a discipline. This brings us to the topic of ‘explanatory
situation, the onset of escalation just happens as part of the domain’. The points made in the previous section all argue
collective dynamics. that the study of the brain alone is insufficient to understand
This simple case exemplifies how interactional dynamics the kind of coupling we see in social interactions. I found the
are not fully under the control of the participants. There is example of the Auvray experiment too detached from what
no escalation module in the brain or an individual intention happens in the brain, but H.D.J. and E.D.P. argue it is not
to escalate in general. It also shows that an important simple, not an atypical example, and not reducible to
aspect of social meaning can relate to these emergent pat- events in the brain and events outside the brain. In short,
terns. Escalation is often associated with the generation of the suggestion of §4 is that social neuroscience could gain
negative affect, which undoubtedly relates to interactional more traction on how it uses the concept of cognition to
history, but as we can see, can emerge as novel social mean- explain behaviour, if it incorporated relations with extra-
ing due to the interaction itself, and not to any individual neural events into its domain of study. This is an empirical
intentions. Similar processes where social meaning is gener- suggestion: social neuroscientists should try to take this
ated by interactional patterns were already described by stance, and see how far they get with it. Will it be helpful in
Gregory Bateson in terms of schismogenesis and feedback explaining human social behaviour, or will it create compli-
[46], and taken up in psychotherapeutic contexts (see e.g. cations if we widen the discipline of social neuroscience in
[47 –49]). The objects of social meaning are themselves this way? This seems like a reasonable practical position. If cog-
interactively generated as well as apprehended. nition is somewhat relativized to a discipline in this way, shifts
Social interaction processes can be very hard to disentan- in the explanatory domain of the discipline would result in
gle from individual brain and body dynamics. They can also corresponding shifts in the concept of cognition.
play specific functional roles in the solution of a cognitive The final issue concerns causation: I felt that this was
task. And they can give rise to objects, meanings and actions much ‘denser’ in the brain than between brain and environ-
that are irreducibly interactive. These complex realities in no ment, but the only metric I offered were sheer numbers of
way eliminate the possibility of scientific inquiry. On the con- axons. H.D.J. and E.D.P. argue that this is not the right
trary, in some cases they result in simpler explanations than metric, because even very small physical connections can
those that are unduly constrained to be skull-bound, as we result in profound influences. I think they are right. This
witness in the case of escalation and perceptual crossing. then leaves me to retreat to something other than causal den-
Far from making social cognition fuzzier and mysterious, sity as the distinguishing feature that delimits processing in
the IBH in fact seeks to provide a more objective foundation, the brain from processing involving events outside the
 Downloaded from http://rstb.royalsocietypublishing.org/ on August 4, 2017

brain. The only other good metric that comes to mind is itself can show emergent dynamical patterns that do not 8
something like ‘evolvability’ or ‘manipulability by evolution’. reduce to the activity of the mice. This conclusion is conso-

rstb.royalsocietypublishing.org
That is, there is a strong intuition that evolution can direct nant with Cosmelli & Thompson’s suggestions about the
changes in cognition through changes in the brain, but not brain-in-a-vat thought experiment [18]. The upshot is that
changes in the physical environment. Unfortunately, this we may be able to describe some aspects of cognition reason-
intuition only works for the non-social environment. For ably well as input –output transformations by the brain,
the social environment, there is instead a strong plausibility whereas others cannot be so described. Social cognition
that brains co-evolved, and so cognition indeed could typically may be of the latter kind.
evolve through changes across multiple brains. This leads us to consider the practical issue of which
To summarize: my current concept of cognition, however are the criteria for delineating the systems under study.
ill-defined, is squarely centred on the brain. But I have not One possible reading of the IBH could be that everything
made a serious attempt to revise this, and it is possible that matters, and so there is no right decomposition into causal

Phil. Trans. R. Soc. B 371: 20150379

such a revision would result in a concept with more utility. systems that could illuminate a particular instance of social
The argument that causal density specifies nature’s joints cognition. This is certainly not the intended reading.2 To
for a cognitive system is problematic because actual physical highlight the embeddedness of brain activity within a behav-
density of connections is probably not the right metric. These ing and interacting body is not to render social neuroscience a
considerations lead to the conclusions of the next section, on hopeless endeavour. It is to raise awareness that certain
which all three of us agree. assumptions, such as the assumption of decomposability,
can be problematic if formulated uncritically. There exist
many experimental approaches that would allow the simul-
taneous study and manipulation of neural and interactive
6. Suggestions for social neuroscience dynamics, as we have mentioned.
In writing this article, all three of us acknowledge that under- Besides these practical claims, the IBH can also be seen as a
standing the brain and cognition is incredibly hard. All foundational claim about what grounds social cognition. From
approaches should have the provisional status of ‘hypoth- input–output transformations alone, meaning cannot emerge;
esis’—something the IBH explicitly does have, but standard for what do the inputs and outputs stand for? The IBH assigns
information processing views usually do not. We agree that meaning from the wider perspective of (i) social interaction and
historical views of cognition as computation over represen- (ii) the wider enactive theory about cognition as sense-making
tations are unlikely to adequately describe how brains that it forms part of (e.g. [55]). From an enactive perspective,
work. Rather than defining cognition as that kind of infor- meaning emerges in virtue of historical and concurrent pat-
mation processing that is unique to brains, we would prefer terns of interaction. That something like this must happen,
to think of cognition as involving brains in some way yet to for example, as an infant learns her very first words seems
be fully understood, possibly including causal relations uncontentious. Associating the word ‘book’ with seeing a
between brains, bodies, other people and even the non- book presumably works by the infant and another person
social environment. Evolution has made use of whatever both looking at the book and saying and hearing the word
substrates are available to generate flexible behaviour, and book. But why does she learn the word book? Why is it mean-
we simply do not know yet what those substrates are. ingful to her at all? Here, the wider framework of participatory
The IBH can be seen as making a practical claim for scien- sense-making out of which the IBH grew makes claims about
tists, namely that there is a more compact explanation of how cognizers encounter the world as meaningful (which is
human social behaviour if we adopt the interactive stance how enactivists define cognition) [43,55]. Certainly, there is
than if we stick with the classical input –output stance. Con- room for rich further debate here. The key future challenge
sider how far one could push the classical brain-in-a-vat for the enactive approach is to develop further concepts and
thought experiment. There have been recent experiments hypotheses, and to continue to articulate them in ways that
using optogenetics in mice that manipulate brain activity so make contact with other frameworks. In this way, concepts
precisely that they literally reconstitute the pattern of neur- like ‘participatory sense-making’ can be articulated into
onal activity that would have been evoked by encoding an domain-specific claims, hypotheses and explanations that
actual sensory stimulus [51]. Such experimentally created pat- relate to conceptions of meaning, as they vary between
terns of activity in the brain cause the mouse to behave as if it relevant disciplines. Formulating the IBH is an attempt to do
remembered an actual stimulus. While this experiment seems precisely this for the field of social neuroscience.
to show that we can understand cognition and behaviour as A closing question is where to find a home for social
divorced from the environment, it actually points to the value neuroscience in all of this. The difficulty arises when social
of the IBH as a framework to understand what is happening. neuroscience attempts to study that which underlies behav-
Suppose the experiment attempted to recreate the pattern of iour and cognition, when it attempts to explain how
activity involved in an actual, reciprocal, social encounter. It meaning is generated, and why social cognition and social
quickly becomes apparent that to do so would require behaviour exhibit particular forms and features. After all,
mimicking the other animal as a social stimulus. But as the social cognition shows substantial differences if we compare
other animal responds to our experimental mouse, this is a dog, a chimpanzee or, for that matter, people from different
not a fixed input, but rather a complex, time-varying input cultures or at different ages. How can we explain these differ-
embedded in causal loops with the very behaviour we wish ences—differences that render social behaviour meaningful
to experimentally control. Our surrogate ‘input pattern’ for individuals of each species, culture and epoch, but less
ends up being not only extremely complex, but in fact meaningful as we cross between these. Perhaps the largest
cannot be specified in the absence of an analysis of the first contribution of the IBH to social neuroscience is to show
mouse as involved in a socially coupled interaction, which that it is impossible to answer these questions if the only
 Downloaded from http://rstb.royalsocietypublishing.org/ on August 4, 2017

data we entertain are from a single adult brain in one species.

That is, we need to consider species-typical social interactions
Endnotes 9
1
This is not to say that we think that understanding the brain is suffi-

rstb.royalsocietypublishing.org
not just in the context of meaningful situations, but also in the cient for understanding cognition. From an enactive perspective,
context of evolution and development. Echoing the well- what matters is the embodied subject in relation to her world or
known ethological refrain that ‘nothing makes sense in meaningful environment. Interestingly, the same kind of argument
biology except in the light of evolution’ ([56], p. 125), we in favour of the IBH presented here in the context of social neuro-
science could be made in the context of many approaches to
would urge that social neuroscience should incorporate at
embodied cognition that remain methodologically individualistic.
least comparative neuroscience, developmental neuroscience, For these, the body is crucial for the mind, meaning the individual
and input from sciences that study social interaction into its body and not its engagement in social interactions. From the enactive
domain of study (see e.g. [57]). perspective, by contrast, both body and social engagement are
primordial [1,2].
Authors’ contributions. E.D.P. and H.D.J. are the primary authors of §§2 2
The situation is not unlike other debates in biology. Arguments for

Phil. Trans. R. Soc. B 371: 20150379

and 4; R.A. is the primary author of §§3 and 5; all authors contributed the importance of non-genetic factors in evolution and development
equally to §§1 and 6, and all authors provided substantial input to all (e.g. [52,53]) are often met with the criticism that one cannot study
sections of the paper. every conceivable causal factor scientifically. But there are positive
Competing interests. We have no competing interests. counter-proposals that distinguish between different causal roles.
Funding. H.D.J. is funded by a Ramón y Cajal Fellowship, RYC-2013- For instance, Woodward [54] suggests that one can discriminate
14583. R.A. was supported in part by a Conte Center grant from between causal factors according to different criteria such as their
the National Institute of Mental Health (USA). stability or non-contingency, specificity and appropriateness for the
level of explanation. Manipulability could be another factor for this
Acknowledgements. We thank Guillaume Dumas, Frederick Eberhardt,
kind of consideration.
Riitta Hari and the reviewers for their comments on the manuscript.

References
1. Di Paolo EA, De Jaegher H. 2012 The interactive 13. Clark A, Chalmers DJ. 1998 The extended mind. 25. Le Van Quyen M. 2011 The brainweb of cross-scale
brain hypothesis. Front. Hum. Neurosci. 6, 1–16. Analysis 58, 7–19. (doi:10.1093/analys/58.1.7) interactions. New Ideas Psychol. 9, 57 –63. (doi:10.
(doi:10.3389/fnhum.2012.00163) 14. Rowlands M. 2003 Externalism: putting mind and 1016/j.newideapsych.2010.11.001)
2. De Jaegher H, Di Paolo EA, Gallagher S. 2010 Can social world back together again. Georgetown, DC: 26. He BJ, Zempel JM, Snyder AZ, Raichle ME. 2010 The
interaction constitute social cognition? Trends Cogn. Sci. Acumen/McGill-Queens University Press. temporal structures and functional significance of
14, 441–447. (doi:10.1016/j.tics.2010.06.009) 15. Byrge L, Sporns O, Smith LB. 2014 Developmental scale-free brain activity. Neuron 66, 353–369.
3. Harlow HF, Harlow MK. 1962 Social deprivation in process emerges from extended brain –body– (doi:10.1016/j.neuron.2010.04.020)
monkeys. Sci. Am. 207, 136–146. (doi:10.1038/ behavior networks. Trends Cogn. Sci. 18, 395–403. 27. Buschman TJ, Miller EK. 2007 Top-down versus
scientificamerican1162-136) (doi:10.1016/j.tics.2014.04.010) bottom-up control of attention in the prefrontal and
4. Auvray M, Lenay C, Stewart J. 2009 Perceptual 16. Simon HA. 1962 The architecture of complexity. posterior parietal cortices. Science 315, 1860–1862.
interactions in a minimalist virtual environment. Proc. Am. Philos. Soc. 106, 467–482. (doi:10.1126/science.1138071)
New Ideas Psychol. 27, 32 –47. (doi:10.1016/j. 17. Anderson ML, Richardson MJ, Chemero A. 2012 28. Saalmann YBP, Pigarev IN, Vidyasagar TR. 2007
newideapsych.2007.12.002) Eroding the boundaries of cognition: implications of Neural mechanisms of visual attention: how top-
5. Clark A. 1997 Being there: putting brain, body and embodiment. Top. Cogn. Sci. 4, 717 –730. (doi:10. down feedback highlights relevant locations. Science
world together again. Cambridge, MA: MIT Press. 1111/j.1756-8765.2012.01211.x) 316, 1612– 1615. (doi:10.1126/science.1139140)
6. Kelso JAS. 1995 Dynamic patterns: the self- 18. Cosmelli D, Thompson E. 2011 Brain in a vat or 29. Lakatos P, Karmos G, Mehta AD, Ulbert I, Schroeder
organization of brain and behaviour. Cambridge, body in a world? Brainbound versus enactive views CE. 2008 Entrainment of neuronal oscillations as a
MA: MIT Press. of experience. Philos. Top. 39, 163– 180. (doi:10. mechanism of attentional selection. Science 320,
7. Noë A. 2004 Action in perception. Cambridge, MA: 5840/philtopics201139119) 110–113. (doi:10.1126/science.1154735)
MIT Press. 19. Babiloni F, Astolfi L. 2014 Social neuroscience and 30. Astolfi L, Toppi J, de Vico Fallani F, Vecchiato G,
8. Findlay JM, Gilchrist ID. 2003 Active vision: the hyperscanning techniques: past, present and future. Salinari S, Mattia D, Cincotti F, Babiloni F. 2010
psychology of looking and seeing. Oxford, UK: Oxford Neurosci. Biobehav. Rev. 44, 76 –93. (doi:10.1016/j. Neuroelectrical hyperscanning measures
University Press. neubiorev.2012.07.006) simultaneous brain activity in humans. Brain
9. Przyrembel M, Smallwood J, Pauen M, Singer T. 20. Simon HA. 1969/1996 The sciences of the artificial. Topogr. 23, 243–256. (doi:10.1007/s10548-010-
2012 Illuminating the dark matter of social Cambridge, MA: MIT Press. 0147-9)
neuroscience: considering the problem of social 21. Dumas G, Nadel J, Soussignan R, Martinerie J, 31. Astolfi L et al. 2011 Imaging the social brain by
interaction from philosophical, psychological, and Garnero L. 2010 Inter-brain synchronization during simultaneous hyperscanning during subject
neuroscientific perspectives. Front. Hum. Neurosci. 6, social interaction. PLoS ONE 5, e12166. (doi:10. interaction. IEEE Intell. Syst. 26, 38 –45. (doi:10.
190. (doi:10.3389/fnhum.2012.00190) 1371/journal.pone.0012166) 1109/MIS.2011.61)
10. Koster-Hale J, Saxe R. 2013 Theory of mind: a 22. Canolty RT et al. 2006 High gamma power is phase- 32. Dumas G, Kelso JAS, Nadel J. 2014 Tackling the
neural prediction problem. Neuron 79, 836. (doi:10. locked to theta oscillations in human neocortex. social cognition paradox through multi-scale
1016/j.neuron.2013.08.020) Science 313, 1626–1629. (doi:10.1126/science. approaches. Frontiers in Psychology 5, 1 –4. (doi:10.
11. Clark A. 2013 Whatever next? Predictive brains, 1128115) 3389/fpsyg.2014.00882)
situated agents, and the future of cognitive science. 23. Womelsdorf T et al. 2007 Modulation of neuronal 33. Hari R, Kujala MV. 2009 Brain basis of human social
Behav. Brain Sci. 36, 181–204. (doi:10.1017/ interactions through neuronal synchronization. Science interaction: From concepts to brain imaging. Physiol.
S0140525X12000477) 316, 1609–1612. (doi:10.1126/science.1139597) Rev. 89, 453 –479. (doi:10.1152/physrev.00041.
12. Noë A. 2010 Out of our heads: why you are not your 24. Buzsáki G, Draguhn A. 2004 Neuronal oscillations in 2007)
brain, and other lessons from the biology of cortical networks. Science 304, 1926– 1929. (doi:10. 34. Kelso JAS, Dumas G, Tognoli E. 2013 Outline of a
consciousness. London, UK: Macmillan Press. 1126/science.1099745) general theory of behavior and brain coordination.
 Downloaded from http://rstb.royalsocietypublishing.org/ on August 4, 2017

Neural Netw. 37, 120–131. (doi:10.1016/j.neunet. 42. Froese T, Iizuka H, Ikegami T. 2014 Embodied social 49. Fogel A. 1993 Developing through relationships: 10
2012.09.003) interaction constitutes social cognition in pairs of origins of communication, self and culture. London,

rstb.royalsocietypublishing.org
35. Van Orden GC, Holden JG, Turvey MT. 2003 Self- humans: a minimalist virtual reality experiment. Sci. UK: Harvester Wheatsheaf.
organization of cognitive performance. J. Exp. Rep. 4, 1–10. (doi:10.1038/srep03672) 50. Searle J. 1990 Consciousness, explanatory inversion,
Psychol. 132, 331–350. (doi:10.1037/0096-3445. 43. De Jaegher H, Di Paolo E. 2007 Participatory sense- and cognitive science. Behav. Brain Sci. 13, 585 –
132.3.331) making: an enactive approach to social cognition. 642. (doi:10.1017/S0140525X 00080304)
36. Dotov DG, Nie L, Chemero A. 2010 A demonstration Phenomenol. Cogn. Sci. 6, 485 –507. (doi:10.1007/ 51. Ramirez S et al. 2013 Creating a false memory in
of the transition from ready-to-hand to unready-to- s11097-007-9076-9) the hippocampus. Science 341, 387–391. (doi:10.
hand. PLoS ONE 5, e9433. (doi:10.1371/journal. 44. Granic I. 2000 The self-organization of parent-child 1126/science.1239073)
pone.0009433) relations: beyond bidirectional models. In Emotion, 52. Oyama S. 1985 The ontogeny of information:
37. Shockley K, Santana M-V, Fowler CA. 2003 Mutual Development, and self-organization. dynamic systems developmental systems and evolution. Cambridge,
interpersonal postural constraints are involved in approaches to emotional development (eds MD UK: Cambridge University Press.

Phil. Trans. R. Soc. B 371: 20150379

cooperative conversation. J. Exp. Psychol. Hum. Lewis, I Granic), pp. 267–297. Cambridge, UK: 53. Jablonka E, Lamb MJ. 2005 Evolution in four
Percept. Perform. 29, 326 –332. (doi:10.1037/0096- Cambridge University Press. dimensions: genetic, epigenetic, behavioral, and
1523.29.2.326) 45. Shergill SS, Bays PM, Frith CD, Wolpert DM. 2003 symbolic variation in the history of life.
38. Richardson MJ, Marsh KL, Schmidt RC. 2010 Two eyes for an eye: the neuroscience of force Cambridge, MA: MIT Press.
Challenging egocentric notions of perceiving, acting, escalation. Science 301, 187. (doi:10.1126/science. 54. Woodward J. 2010 Causation in biology: stability,
and knowing. New York, NY: Guildford. 1085327) specificity, and the choice of levels of explanation.
39. Riley MA, Richardson M, Shockley K, Ramenzoni VC. 46. Bateson G. 1936 Naven: a survey of the Biol. Philos. 25, 287– 318. (doi:10.1007/s10539-
2011 Interpersonal synergies. Front. Psychol. 2, problems suggested by a composite picture of the 010-9200-z)
1– 7. (doi:10.3389/fpsyg.2011.00038) culture of a New Guinea tribe drawn from three 55. Thompson E. 2007 Mind in life: biology,
40. Bedia MG, Aguilera M, Gómez T, Larrode DG, Seron points of view. Stanford, CA: Stanford University phenomenology, and the sciences of mind.
F. 2014 Quantifying long-range correlations and 1/f Press. Cambridge, MA: Harvard University Press.
patterns in a minimal experiment of social 47. Watzlawick P, Helmick Beavin J, Jackson DD. 1967 56. Dobzhansky T. 1973 Nothing in biology makes sense
interaction. Front. Psychol. 5, 1 –12. (doi:10.3389/ Pragmatics of human communication. A study of except in the light of evolution. Am. Biol. Teach. 35,
fpsyg.2014.01281) interactional patterns, pathologies, and paradoxes, 125–129. (doi:10.2307/4444260)
41. Clark A, Thornton C. 1997 Trading spaces: p. 294. New York, NY: Norton. 57. De Jaegher H, Peräkylä A, Stevanovic M. 2016 The
computation, representation, and the limits of 48. Stern DN. 1977/2002 The first relationship: infant co-creation of meaningful action: bridging enaction
uninformed learning. Behav. Brain Sci. 20, 57– 66. and mother, 2nd edn. London, UK: Harvard and interactional sociology. Phil. Trans. R. Soc. B
(doi:10.1017/S0140525X97000022) University Press. 371, 20150378. (doi:10.1098/rstb.2015.0378)