1 s2.0 S0149763417301732 Main
1 s2.0 S0149763417301732 Main
1 s2.0 S0149763417301732 Main
Review article
A R T I C L E I N F O A B S T R A C T
Keywords: Within research into neurodevelopmental disorders, little is known about the mechanisms underpinning changes
Autism Spectrum Disorder in symptom severity across development. When the behavioural presentation of a condition improves/symptoms
Neurodevelopmental disorders lessen, this may be because core underlying atypicalities in cognition/neural function have ameliorated. An
Compensation alternative possibility is ‘compensation’; that the behavioural presentation appears improved, despite persisting
Compensatory mechanisms
deficits at cognitive and/or neurobiological levels. There is, however, currently no agreed technical definition of
Adaptation
compensation or its behavioural, cognitive and neural characteristics. Furthermore, its workings in neurode-
Camouflaging
Cognitive phenotype velopmental disorders have not been studied directly. Here, we review current evidence for compensation in
Behavioural phenotype neurodevelopmental disorders, using Autism Spectrum Disorder as an example, in order to move towards a
Remediation better conceptualisation of the construct. We propose a transdiagnostic framework, where compensation re-
Theory of mind presents the processes responsible for an observed mismatch between behaviour and underlying cognition in a
Executive function neurodevelopmental disorder, at any point in development. Further, we explore potential cognitive and neural
Good outcome mechanisms driving compensation and discuss the broader relevance of the concept within research and clinical
Under-diagnosis
settings.
Late diagnosis
Female presentation
Unaffected siblings
⁎
Corresponding author.
E-mail addresses: lucy.livingston@kcl.ac.uk (L.A. Livingston), francesca.happe@kcl.ac.uk (F. Happé).
http://dx.doi.org/10.1016/j.neubiorev.2017.06.005
Received 1 March 2017; Received in revised form 25 May 2017; Accepted 13 June 2017
Available online 19 June 2017
0149-7634/ © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
2017), developmentally-relevant mechanisms for improvement, in- There is, to our knowledge, only one review paper on the topic (Ullman
cluding compensation, should be an important focus of research. and Pullman, 2015), which explores the specific compensatory function
Second, compensation could be a useful way to unpick some of the of the declarative memory system in five neurodevelopmental dis-
heterogeneity amongst neurodevelopmental disorders, which is fre- orders, including ASD. Crucially, the authors’ review relies on a defi-
quently proposed to be one of the greatest challenges to understanding nition of compensation that is reminiscent of that described in the aging
these conditions (Thapar et al., 2017). Finally, studying the mechan- and neuropsychological literature; that compensation reflects how an
isms underlying compensation could be fundamental to informing early intact neurocognitive process/system might take over, or compensate
intervention research, whose principle aim is to improve long-term for, the functioning of a defective process/system in order to maintain
prognosis. And yet, in order to directly investigate compensation, we typical behaviour and/or cognitive task performance. Indeed, Ullman
must have a reasonable definition to guide our measurements and de- and Pullman (2015) suggest that in ASD, where socio-cognitive func-
rive testable hypotheses. tioning is compromised, intact declarative memory ability may scaffold
In this paper, we aim to create a working definition of compensation social behaviour; for example, the ability to recall previously learned
relevant to neurodevelopmental disorders and use this definition to i) social rules may replace intuitive understanding of social cues, thereby
review evidence for compensation, garnering examples from ASD, ii) contributing to an appropriate social response.
propose a preliminary framework of the workings of compensation, iii) There is, however, good reason to question whether a definition
explore its potential cognitive and neural underpinnings, and finally, derived from the study of individuals who have acquired their deficits
iv) discuss the research and clinical implications of studying neurode- (e.g., brain-damaged individuals/aging adults), necessarily extends to
velopmental disorders within this compensation framework. neurodevelopmental populations (Johnson, 2017; Thomas and
Karmiloff-Smith, 2002). For example, Johnson (2017) has highlighted
2. Defining and measuring compensation in neurodevelopmental how focal brain damage during the pre/perinatal period may be com-
disorders pensated for by early reallocation of function to intact brain regions,
but that in the case of conditions such as ASD, where more wide-spread
2.1. Compensation in the psychological literature early brain disturbance is observed (or postulated, e.g., general synapse
dysfunction), an alternative explanation of compensation may be re-
Within psychological research in general, the term ‘compensation’ quired. Additionally, brain injury in healthy adults may trigger a host of
has been widely used. In instances where participants, who are ex- compensatory processes (e.g., enhanced connectivity from damaged to
pected to be limited in a particular set of resources (be this due to a frontal regions; Sharp et al., 2014) that are not necessarily comparable
psychiatric condition, old age, or an experimental manipulation to cases where a cascade of atypical neural function has existed from
amongst healthy participants), perform better than expected on a psy- very early in development.
chological task, the possibility that they have in some way compen- In our endeavor to find a definition of compensation drawn from
sated, is often speculated upon by authors. This compensatory hy- observations in neurodevelopmental phenotypes, we take inspiration
pothesis is generally backed up by evidence showing that ‘compensated’ from research into a developmental condition that has a relatively
participants have achieved this ‘typical’ performance with the recruit- circumscribed cognitive deficit and has received some preliminary
ment of additional resources, be these neurobiological, cognitive, or discussion with regards to compensation; namely developmental dys-
genetic. For example, in the literature on aging, researchers have used lexia.
the term to describe how older adults can demonstrate atypical acti-
vation (enhanced or decreased) of task-relevant brain areas or activa- 2.3. Lessons from developmental dyslexia
tion of additional regions not typically recruited by younger adults, in
order to perform a task just as well as their younger counterparts (see Developmental dyslexia is characterised by a specific impairment in
Grady, 2012). Equally, within research into neuropsychological pa- reading, not otherwise accounted for by intellectual or visual abilities
tients, the term ‘compensation’ refers to the brain’s ability to rely on (American Psychiatric Association, 2013). The condition is proposed to
alternative neural routes after typical routes have been compromised by be underpinned by a core deficit in phonological processing (Snowling,
brain damage, in order for patients to make improvements in beha- 2014), which contributes to an array of behavioural symptoms amongst
viour/cognitive abilities (e.g., Price and Friston, 2002). The term dyslexics (e.g., spelling errors and slow reading and word recognition;
‘partial compensation’ is used to describe how an attempt to counteract Thambirajah, 2010). Critically, although the majority of children with
limited resources may not always be efficient enough to support wholly dyslexia experience these difficulties persistently (Hatcher et al., 2002;
‘typical’ behaviour or cognitive task performance. Maughan et al., 2009), a subset of individuals, referred to as ‘com-
Despite frequent use of the term ‘compensation’, there is no tech- pensated dyslexics’ (Lefly and Pennington, 1991), eventually establish
nical or universal definition. The precise interpretation of its meaning typical reading skills by the time they enter adulthood (Callens et al.,
or the meaning of ‘compensatory brain activity’ is specific to the par- 2012; Gallagher et al., 1996).
ticular task and participant population in question. The literature also In principle, there are at least three possible ways in which dyslexic
suggests that the process of compensation could actually exist and op- individuals’ primary symptoms could lessen. First, the phonological
erate at multiple levels, from molecular and/or genetic pathways (for processing deficit at the cognitive level may genuinely remit, thus
instance, synaptic plasticity in order to counteract atypical con- supporting good reading ability. Second, phonological processing may
nectivity; e.g., in ASD, Bourgeron, 2015) to broader cognitive systems be delayed rather than deficient in these children, so that there is
and behaviour (for instance, atypical neural functioning to support eventually developmental ‘catch up’. Third, good reading ability may
typical cognitive task performance; e.g., in ASD, White et al., 2014). In be facilitated by alternative neurocognitive pathways that are in-
this paper, due to the complexity and novelty of the phenomenon, we dependent of phonological routes. On inspection of the literature, the
will focus on conceptualising compensation across levels of behaviour, first two possibilities do not appear to hold up empirically. Amongst
cognition and whole neural networks only, solely within the scope of highly ‘compensated’ dyslexics, who do not exhibit measurable spelling
neurodevelopmental disorders. or reading difficulties (e.g., those in higher education), significant
phonological processing deficits are revealed when tapped with sensi-
2.2. Compensation in the literature on neurodevelopmental disorders tive enough cognitive probes, such as rapid picture naming (Gallagher
et al., 1996; Ingvar et al., 2002; Parrila et al., 2007; Swanson, 2012).
With no agreed definition, it is unsurprising that compensation in Further, these individuals’ reading abilities are not necessarily com-
neurodevelopmental disorders has received little empirical attention. parable to those of typically developing individuals under certain
730
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
Hypothetical distinctions between different mechanisms that could promote improved outcome; Compensation (shallow), Compensation (deep), Genuine remediation, Delayed maturation. Social skills at the behavioural level and theory of mind
contexts; for example, their reading speeds are significantly slower than
(ToM) at the cognitive level in ASD are used as an example here, but could be substituted with another behaviour, underlying cognitive ability and neurodevelopmental disorder. These predictions require empirical testing.
achieved via one or more atypical routes that bypass phonological
processing. These alternative processes may be sufficient in certain
Delayed maturation
ting the orthographic forms of unfamiliar words to familiar ones where
compensators.
spelling of the two is similar (Cavalli et al., 2016; Parrila et al., 2007)
integrated.
and relying upon visual imagery (Bacon and Handley, 2014) and the
semantic context of words (Corkett and Parrila, 2008; Pugh et al.,
2001). This notion of non-phonological routes to ‘typical’ reading is
further corroborated by evidence at the neural level, such as findings of
Genuine remediation
everyday situations.
2011; Shaywitz et al., 2004). Greater functional connectivity between
task challenges).
posterior visual regions has also been reported (Koyama et al., 2013),
compensators.
perhaps reflecting the potential compensatory role of visual strategies
(for an exhaustive review of compensatory neural pathways, see
Pammer, 2014).
negatively impacted.
degree of observable symptoms), and actual ability, exhibited in un-
derlying cognitive and/or neural function. Accordingly, our working
definition of compensation is as follows: the processes contributing to
situations.
nections with typical sensory inputs may not wire in the usual way.
However, in contrast to compensation, we would expect those in-
Behavioural
Genetic
731
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
732
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
relationships (Billstedt et al., 2005; Eaves and Ho, 2008; Farley et al., has demonstrated that a substantial proportion (20%) of these in-
2009). We do note, however, that the majority of studies into adult dividuals will also go on to follow an autism trajectory (Elsabbagh and
outcome thus far rely heavily on definitions of outcome imposed by Johnson, 2010). Of particular interest to the notion of compensation,
neurotypicals to neurotypical standards, and may not necessarily reflect are those at-risk siblings who, despite displaying some early ‘red flags’
the best outcomes as perceived by autistic individuals themselves. for autism at 12 months of age (for review, see Jones et al., 2014), are
These studies to date consider ‘good outcome’ to have occurred, when indistinguishable from typically developing individuals in terms of so-
autistic individuals appear, at least on the surface, to be experiencing a cial and communicative behaviour by 3 years of age. For example,
less impairing version of the condition. A similar pattern is seen Macari et al. (2012) found this trajectory amongst 9% of their sample of
amongst other neurodevelopmental disorders, including developmental at-risk siblings. To the extent that these early behavioural markers re-
dyslexia, as discussed earlier, and ADHD. For example, a third of chil- liably predict later ASD diagnosis, rather than broader developmental
dren with ADHD no longer exhibit clinically impairing symptoms by issues (e.g., ADHD, cognitive impairment), these individuals might re-
adulthood (Faraone et al., 2006). Crucially, if this ‘good outcome’ has present some form of early, unconscious compensation, which buffers
arisen despite persistence in underlying core cognitive deficits, rather against the expression of ASD in behaviour.
than because these deficits have genuinely resolved, these individuals Johnson et al. (2015b) have proposed that, in the face of substantial
represent a subgroup that fit our aforementioned definition of com- risk for ASD in early development, a series of adaptive/compensatory
pensation. neural processes might take place, thereby reducing the severity of
There is good reason to suspect that autistic adults with ‘good autistic symptoms manifested in behaviour later on. This may be to the
outcome’, who typically show good intellectual and language abilities, extent that certain individuals bypass an autism diagnosis altogether.
continue to have characteristically autistic cognitive profiles, despite Equally, the most severe symptom presentations might signal a very
displaying “[social] behaviour that more and more shades into nor- limited capacity for this early neural compensation, which Johnson
mality” (Frith, 1991, p. 31). Socially adapted behaviour might be (2012) suggests is rooted in the availability of prefrontal cortex (PFC)
achieved via overt and conscious strategies, despite core socio-cognitive based executive functions. This standpoint poses an intriguing theore-
abilities, such as theory of mind, remaining impaired. For example, an tical possibility concerning the broader autism phenotype (BAP), which
individual may be taught to make eye contact without necessarily being is the sub-clinical expression of autistic traits, often seen in relatives of
able to extract mental state information from that eye contact. Further, those with ASD (Sucksmith et al., 2011). Traditionally, BAP has been
there is evidence to suggest that logical ‘hacking’ and intact verbal explained in terms of unaffected relatives having a milder ‘hit’ for ASD
abilities might facilitate success on explicit theory of mind tasks that (e.g., milder theory of mind deficit) compared to their affected relative,
directly ask participants to reason about mental states, in both children which keeps them below the diagnostic threshold. However, reframing
(e.g., Cantio et al., 2016; Happé, 1995; Peterson et al., 2007; Scheeren this in light of compensation, opens up the possibility that in some
et al., 2013) and adults (Lever and Geurts, 2016a; Senju et al., 2009). cases, unaffected relatives may not necessarily possess a milder cogni-
Such strategies, however, cannot support performance on more sensi- tive burden for ASD per se, but instead, possess a greater capacity to
tive, implicit measures of theory of mind (e.g., anticipatory eye gaze compensate for this cognitive burden (Skuse, 2007), resulting in an
based on attribution of a false belief), in which persistent difficulties are apparently milder behavioural presentation. This does not exclude the
revealed (Senju et al., 2009; Schneider et al., 2013). This notion of a possibility, which is indeed the prevailing view, that many relatives
fundamental and enduring difficulty in understanding others’ minds is express a milder phenotype due to genuinely milder underlying autistic
further highlighted by the fact that in situations where learned strate- difficulties.
gies may not be sufficient to fully compensate, such as novel, fast-paced In order to assess the extent to which unaffected siblings might in
everyday social interactions, even ‘good outcome’ individuals report some cases be exceptional at compensating, one could investigate a
difficulty ‘keeping up’. cognitive probe or biological marker shared amongst individuals with
Evidence for a substantial mismatch between observable social be- ASD and their siblings (i.e., an endophenotype), which signals some
haviour and underlying theory of mind, as described above, amongst kind of aetiological load or ‘hit’ for ASD (here we have focused on
autistic individuals who otherwise appear socially adapted casts doubt theory of mind impairment, but other examples could include atten-
on claims that some autistic people may genuinely transition off the tional abnormalities or EEG markers; see Jones et al., 2014). Com-
autism spectrum; for instance, in research describing children who pensation could be inferred when so called ‘unaffected’ siblings are
reach ‘optimal outcome’ after early behavioural intervention (Fein demonstrating a typical behavioural profile, in spite of the presence of a
et al., 2013; Helt et al., 2008). These studies report on individuals who cognitive deficit or biological marker characteristic of those with di-
had a reliable diagnosis of ASD in childhood yet by young adulthood agnosed ASD. By contrast, unaffected siblings who appear typical in
are indistinguishable from their typically developing peers, even in the terms of the cognitive characteristic/biological marker of interest,
absence of any additional support. The extent to which these in- would likely not be compensating, but rather, showing a typical be-
dividuals are showing compensation in the face of continuing under- havioural profile due to a genuine lack of ‘hit’ for ASD. Methods that
lying difficulties, rather than a genuine ‘remission’ of ASD, is yet to be aim to quantify aetiological load for ASD within the general population
explored. Crucially, the underlying socio-cognitive profile of these in- could help clarify this distinction between unaffected siblings who have
dividuals has not been empirically investigated. However, reports of a lesser ‘hit’ for ASD versus those who are compensating in the face of a
continued social awkwardness and reduced quality of friendships substiantal ‘hit’; for example, siblings that are compensating should
(Orinstein et al., 2015) suggest that these individuals continue to ex- demonstrate high common genetic risk for ASD. This could, for ex-
perience significant social difficulties. This points to a broader issue ample, be estimated using polygenic risk scores, which are based on
within research into outcome in ASD, which has predominantly relied single nucleotide polymorphisms associated with the disorder in ques-
upon behavioural measures. If we are to gain a better understanding of tion. Furthermore, it will be useful to determine if siblings can com-
compensation amongst these highly able individuals, a closer, more in- pensate for rare genetic risk (e.g., de novo mutations) by investigating
depth look at the discrepancy between behavioural and cognitive out- the genetic patterns of all affected/unaffected members of a family,
comes will be required. alongside cognitive/biological markers for ASD.
In the case of theory of mind, highly sensitive tasks will be required
3.2. At-risk individuals transitioning away from the autism phenotype to reveal subtle difficulties with mentalising in unaffected siblings.
Currently, there are only a few theory of mind tasks that are able to
The prospective study of infants at high genetic and/or environ- capture individual differences in typically developing individuals (e.g.,
mental risk for ASD (those with an older sibling with an ASD diagnosis) Movie for the Assessment of Social Cognition, Dziobek et al., 2006;
733
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
Strange Stories Film Task, Murray et al., 2017; Adult-Theory of Mind, to show clinically impairing symptoms due to a superior ability to
Brewer et al., 2017). Continued longitudinal investigation of siblings compensate (Dworzynski et al., 2012). Support for this notion comes
will also be required to clarify whether those who demonstrate early from cross-sectional observations that ASD symptom severity is
autistic-like features that eventually disappear are genuinely compen- equivalent between males and females in childhood, yet by adulthood,
sating for an underlying liability for ASD specifically, which might be females display fewer social symptoms (Lai et al., 2011; Rynkiewicz
revealed with a cognitive or biological probe, or whether these ‘red et al., 2016).
flags’ in fact signal reversible social atypicalities, for example, due to Clinical observations and self-reports also suggest that females may
developmental delay. be particularly motivated and/or skilled in ‘camouflaging’ their social
difficulties (Mandy and Tchanturia, 2015; Simone, 2010), although
3.3. Late diagnosis these difficulties may continue to be experienced as impairing. Of in-
terest here, a recent qualitative investigation (Hull et al., 2017) into
Since the earliest reported cases over 70 years ago (Asperger, 1944 ‘camouflaging’ in autistic adults distinguishes masking and compensa-
[translated in Frith, 1991]; Kanner, 1943), autism has been con- tion as two sub-components of camouflaging; the former reflecting
ceptualised as a condition with childhood-onset. However, many in- suppression of autistic behaviours and the latter reflecting active stra-
dividuals come for an ASD diagnosis for the first time in adulthood tegies to support appearing ‘neurotypical’. Contrastingly, the frame-
(Geurts and Jansen, 2012; Happé and Charlton, 2012; Lai and Baron- work presented in this paper would encompass both of these instances
Cohen, 2015). This is thought in part to mirror increasing public as compensation as they both fit our definition of a discrepancy be-
awareness of ASD and a substantial widening of diagnostic criteria tween behaviour and underlying cognitive ability. Instead, we distin-
(Hansen et al., 2015; Rutter, 2005). Additionally, the latest addition of guish between simpler and fairly inflexible methods (e.g., suppressing a
the Diagnostic and Statistical Manual of Mental Disorders, DSM-5 behaviour) with more sophisticated ones (e.g., explicitly developed
(American Psychiatric Association, 2013), has for the first time re- strategies to reason about mental states), in terms of the depth of
cognised that autistic symptoms “may not become fully manifest until compensation (See Section 4.1).
social demands exceed limited capacities” (p. 50). Hence, the under- Claims for advanced compensation in the female versus male autism
lying assumption regarding first diagnosis in adulthood, is that core phenotype would be best supported by evidence for less severe beha-
autistic difficulties have always existed for these individuals, but the vioural presentation in females versus males, despite equally severe
ability to compensate in some way has supported a relatively ‘neuro- ASD-related cognitive and/or brain atypicalities. Additionally, if this
typical’ presentation, at least up until adulthood, where social demands male-female disparity was found to be more distinct in older in-
have exceeded compensatory ability. Alternatively, these individuals dividuals, this would imply enhanced compensation in females devel-
diagnosed in adulthood may have exhibited some autistic behaviours oping across the lifetime. There has to our knowledge been only one
earlier in life, but these symptoms were not sufficiently impairing for a preliminary investigation of how the discrepancy between observed
diagnosis. For example, certain individuals may construct a niche for behaviour and underlying cognition might differ by sex. Lai et al.
themselves that lessens the burden of their social difficulties (e.g., (2016) explored the difference between standardised scores of ‘internal’
having a partner to act as their ‘social brain’, an employment en- versus ‘external’ measures; theory of mind ability (inferred from per-
vironment that is low in social demands) and only when this niche is formance on the Reading the Mind in the Eyes task; Baron-Cohen et al.,
disturbed do their autistic behaviours become impairing enough to seek 2001a) and self-rated Autism-Spectrum Quotient (AQ; Baron-Cohen
and warrant a diagnosis (see Section 4.2). et al., 2001b) scores, versus observer-rated symptoms (Autism Diag-
Demographically, this group of late diagnosed individuals are fairly nostic Observation Schedule [ADOS]; Lord et al., 2000). They con-
reminiscent of those ‘good outcome’ individuals first diagnosed in cluded that this difference, which they refer to as an ‘operationalised
childhood. Studies from adult ASD clinics suggest up to 50% are em- camouflaging measure’, was greater for diagnosed females than males.
ployed or in higher education (Happé et al., 2016; Hofvander et al., Further, a greater difference between internal and external levels, i.e., a
2009) and 33–50% live independently (Hofvander et al., 2009; greater ‘camouflaging’ score, was associated with medial-temporal
Lehnhardt et al., 2016). Ascertaining whether a behaviour-cognition based volume differences in the female brain only. The data showed
mismatch truly characterised these individuals in earlier life, i.e. that that there was no sex difference on the Reading the Mind in the Eyes
‘neurotypical’ behaviour masked core autistic difficulties, is challenging task and instead, the male-female difference in ‘camouflaging’ scores
as there are no studies to our knowledge that address, even retro- was driven by higher self-rated symptoms in females on the AQ, and
spectively, what these adults looked like as children. However, anec- higher observer-rated symptoms in males on the ADOS. This supports
dotal and qualitative reports from these individuals themselves suggest the notion that autistic females might appear less behaviourally severe
that underlying difficulties were present long before the time of diag- than males, despite equally severe doses of cognitive deficit. We do,
nosis; for example, in expressions of having felt different or mis- however, draw attention to a potential issue with the use of the AQ as a
understood throughout their lives (e.g., Bargiela et al., 2016; Hickey proxy of underlying cognitive difficulties, as higher AQ scores in au-
et al., 2017). Accurately quantifying compensation in this late diag- tistic females versus males may simply reflect greater awareness rather
nosed group will also require sensitive theory of mind tasks, where task than greater experience of underlying difficulties.
performance cannot rely on the very strategies that these individuals Studying the female versus male autism phenotype may be a useful
may have developed to ‘cope’ without a diagnosis up until adulthood. way to gain insight into the mechanisms involved in compensation. So
far, research has been limited (e.g., Lai et al., 2016) by the fact that
3.4. Female autism presentation females in the clinic are more likely to have accompanying intellectual
and/or behavioural difficulties (Lai et al., 2015; van Wijngaarden-
Although ASD has historically been a predominantly male condition Cremers et al., 2014). Furthermore, our diagnostic tools, derived pri-
(Fombonne, 2009), there is mounting evidence to suggest that ASD may marily from the study of autistic males, may be less sensitive to the
be more prevalent in females than previously assumed (Lai et al., 2015; female autism phenotype (Rutter et al., 2003), thereby increasing the
Robinson et al., 2013). Females in general get diagnosed significantly chances of seriously under-estimating females’ behavioural impair-
later than their male counterparts (Begeer et al., 2013; Rutherford ments. There is therefore a limited amount that could be discovered
et al., 2016) and are more likely to receive an alternative diagnosis about enhanced compensation in females by studying diagnosed in-
before ASD is confirmed (Bargiela et al., 2016; Begeer et al., 2013). One dividuals, who may be less likely to have compensatory resources due
potential explanation (of several, see Lai et al., 2016) for later diagnosis to additional impairments. The most informative approach to identi-
amongst females is that females with high autistic traits are less likely fying successful female compensators may be through population-based
734
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
studies (e.g., Brunsdon et al., 2015) of females who report high autistic functioning in adulthood.
traits and demonstrate poor socio-cognitive abilities, but nevertheless The possibility does remain that for certain autistic individuals,
compensate sufficiently for these difficulties so that they sit below the compensation extends deeper. That is to say that genuinely sophisti-
diagnostic threshold for ASD. It is also possible that internalising pro- cated alternative routes to good theory of mind performance might
blems experienced by autistic females may overshadow their autistic exist. These routes are unlikely to use the usual computational ma-
symptoms in a clinical assessment, thus delaying the time it takes to chinery that allows effortless representation of mental states in typically
gain a diagnosis. This is particularly problematic if internalising developing individuals from the second year of life, but may still pro-
symptoms have arisen from the effort of deliberately compensating in vide a sufficient, albeit slower, route to mental state attribution. For
the first place. instance, autistic individuals may rely on particular cognitive strengths,
such as detail-focused perception and exact memory (see Happé and
4. Key characteristics of compensation Frith, 2006), to process and analyse social information, although no
empirical research has been conducted on this to date. We predict that
Consideration of the ASD literature has isolated four ASD-related deep compensation should be relatively more flexible and resistant to
phenomena, as described in Section 3, which fit our working definition break down under stress and/or fatigue than shallow compensation.
of compensation. This process has simultaneously highlighted other Quantifying shallow compensation will require accurate measure-
potentially important characteristics of compensation that may con- ment of both autistic behaviour and underlying social cognition and
tribute to theory development. Here, we propose three hypothetical perhaps, exploration of how the severity of symptoms alters with the
features of compensation, with the aim of being able to unite fairly changing demands of the environment. Ideally, numerous measures of
discrete examples of compensation (good outcome, at-risk individuals autistic behaviours across different contexts should be taken into ac-
transitioning away from the autism phenotype, late diagnosis and fe- count. We might expect shallow-compensated individuals to show a
male autism presentation) under one umbrella term, not only within particularly uneven behavioural profile across multiple contexts (e.g.,
one specific neurodevelopmental disorder group, such as ASD, but also research setting, home/work setting, novel scenarios). If shallow com-
in a transdiagnostic manner. pensation exists, then we should observe break down in certain con-
texts, when demands exceed compensatory ability. Quantifying deep
4.1. Compensation may be shallow or deep compensation, however, will prove particularly challenging, especially
if the supporting processes are sophisticated enough to support mental
Compensation is unlikely to be spread evenly, meaning that there state attribution, even in novel scenarios or relatively difficult theory of
will be certain difficulties more easily compensated for than others mind tasks. Instead, investigating the underlying neural signatures
(Ullman and Pullman, 2015). For instance, in the case of ASD, learning (e.g., using fMRI or EEG) associated with good task performance might
to laugh at a joke that relies on mental state understanding when others prove particularly useful by shedding light on the nature of the alter-
do is much simpler than working out why the joke was funny in order to native route to good task performance.
formulate an appropriate social response. It is therefore conceivable
that the compensation sitting beneath the surface of ‘typical’ behaviour 4.2. Compensation is modulated by the environment
could either be fairly shallow or instead could extend much deeper (see
Table 1). We suggest that ‘shallow compensation’ is akin to the use of a The immediate environment may serve either to facilitate or impede
white stick by the visually impaired. It enables one to avoid obstacles, the workings of compensation. That is to say that the environment
but does not go further in mimicking vision. By contrast, we suggest might directly modulate the extent to which cognitive difficulties are
that echolocation would be ‘deep compensation’ as it allows the for- manifest in behaviour. This could be regarded as a form of external
mation of a rich spatial representation (the end point of vision), albeit compensation (derived from external sources rather than those internal
via a different route. to the individual), which we here define as environmental scaffolding.
In ASD, examples of both shallow and deep compensation can be For example, in the case of developmental dyslexia, the orthography
identified. In terms of the former, there are strategies that allow one to you have to acquire has a major effect on the visibility of the phono-
navigate the social world superficially (e.g., making deliberate eye logical processing difficulties. In languages with regular phoneme-gra-
contact, imitating others, inhibiting undesirable social behaviours). pheme pairings (like Italian), few children are slow to learn to read,
These strategies are inflexible, do not work well in novel situations or although dyslexics can still be identified using specific cognitive tasks
when social cues are especially ambiguous. Importantly, these strate- (e.g., rapid picture naming). In the case of ASD, we can speculate that
gies do not support underlying socio-cognitive processes, i.e., they environments where social rules are very explicit will make it easier for
would be unlikely to support intact performance on sensitive measures an autistic person to know how to behave socially, thereby facilitating
of theory of mind. Moreover, we predict that shallow compensation is compensation. Equally, the environment may serve to impede or dis-
fragile and can rapidly break down, particularly when compensatory rupt compensation; for example, increasing demands in the environ-
resources are limited, such as under stress, anxiety or mental fatigue. ment may outstrip compensatory capacity at particular points in de-
Shallow compensation might explain why even those autistic in- velopment, thereby revealing autistic traits in behaviour. The child who
dividuals with ‘good outcome’, who score well on assessments of au- had a predictable home environment tailored to their needs might be
tistic behaviours, often say they still struggle with day-to-day social increasingly required to interact with others at school and into higher
scenarios. The Autism Diagnostic Observation Schedule (ADOS; Lord education and/or work. For those seeking diagnosis in adulthood, the
et al., 2000), a standardised one-to-one assessment, may in fact provide task of compensating may have been achievable in childhood, up until a
an optimal environment for compensation, since it involves structured critical point when environmental demands (on e.g., social in-
interaction with a trained researcher in a quiet space. However, across dependence, self-care) increased substantially.
multiple real-life social scenarios, the most difficult aspect to compen- The potential for the environment to modulate compensatory pro-
sate for, i.e., a fundamental difficulty with intuitively understanding cesses and thereby, the expression of symptoms, should be dis-
other minds, will continue to present challenges for many autistic in- tinguished from the ability of the environment to promote ‘good out-
dividuals with so-called ‘good outcome’, as demonstrated in persistent come’ for the individual, without the behavioural manifestation of
difficulties with social relationships and meeting new people. Shallow symptoms necessarily lessening. We refer to this phenomenon as en-
compensation may also partially explain why those receiving a late vironmental accommodation. This may be particularly apparent as in-
diagnosis sometimes describe having ‘burnt out’, as shallow compen- dividuals begin to self-select their environments across development
satory strategies become inefficient to support day-to-day social (e.g., particular academic and/or occupational settings) in order to
735
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
match their abilities (e.g., strengths in detail processing, difficulties in something important about the cost of compensation to those who
social interaction), a phenomenon that Johnson et al. (2015b) have transition away from the autism phenotype early in life, is an inter-
referred to as ‘niche construction’. Certain individuals may increasingly esting avenue for future research. Finally, it can be noted that residual
select environments that hold non-social abilities (i.e., autistic psychiatric difficulties have been reported even amongst those ‘optimal
strengths) to a higher standard than social skills (i.e., autistic difficul- outcome’ individuals who no longer meet diagnostic criteria for ASD
ties). Therefore, although these individuals may continue to display (Fein et al., 2013; Mukaddes et al., 2017).
typically autistic behaviours, their immediate environment or niche is The exact nature of the relationship between compensation and
much more accommodating of these (or actually positively embraces mental health difficulties, such as anxiety, depression and suicidal
these), such that their subjective quality of life and mental wellbeing ideation, requires empirical clarification and in particular, a long-
are less likely to be negatively affected. There may be particular so- itudinal approach. Importantly, distinguishing between shallow and
cieties/cultures where environmental accommodation is more likely to deep compensation should help to further our understanding of the
take place. For example, in an environment or society where an in- specific types of compensation, if any, that genuinely come at a cost to
dividual does not have to read many other minds, there would be fewer the individual. For example, there is reason to suspect that those in-
opportunities for that individual’s poor theory of mind ability to have a dividuals engaging in deep compensation, which permits relatively
detrimental effect on their quality of life. Future research should aim to flexible social understanding, might not necessarily be vulnerable to
highlight the different ways in which particular environments may i) subsequent mental health problems. This may particularly be the case
directly interact with or modulate compensation, as well as ii) promote when individuals seek niche environments for themselves that com-
good outcome, independent of compensation. plement their array of cognitive abilities/deficits (i.e., when environ-
mental accomodation takes place; see Section 4.2).
4.3. Compensation may come at a cost
5. Potential neurocognitive mechanisms
The very idea of compensation implies that a secondary, perhaps
less well-suited set of cognitive resources/machinery is being used in The neurocognitive means by which compensation can be achieved
the absence of resources/machinery that would typically serve the is relatively unexplored. However, investigating those cognitive factors
purpose. For example, domain-general resources may be used in the that have previously been associated with improved prognosis, i.e.
absence of domain-specific (e.g., socio-cognitive) ones. If these domain- ‘good outcome’, might be the most promising line of investigation.
general resources are finite (and shared with other tasks), they will Additionally, we would expect the ability to compensate to be in-
rapidly exhaust after continuous use for compensation and equally, if dependent from the core cognitive burden for the disorder itself
they are preferentially allocated to compensation, this will be at the (Morton and Frith, 1994), if we are to account for the wide hetero-
expense of other tasks dependent upon these domain-general resources. geneity in the degree to which behavioural symptoms either persist or
This hypothesis is in line with reports from autistic individuals them- improve. This is corroborated by studies from the literature on ASD
selves indicating that the task of ‘pretending to be normal’ during social demonstrating that the severity of autistic symptoms in childhood is a
interaction is mentally tiring and stressful (Bargiela et al., 2016; Hull poor predictor of the extent to which behavioural improvements will be
et al., 2017). One young man with ASD, Russell Lehmann, describes the made (Fountain et al., 2012; Levy and Perry, 2011). Instead, it should
improvements in his own autistic symptoms from childhood to adult- be informative to explore those cognitive factors that show a substantial
hood as “stemming from nothing but grueling, demanding, exhausting degree of variance amongst individuals within the disorder group, yet
work” (personal communication). remain intact for a proportion of individuals. It is possible that the
The incidence of mental health problems amongst ‘compensated’ compensatory role of these cognitive factors could be relevant trans-
individuals might speak to the demanding and taxing nature of com- diagnostically (see Ullman and Pullman, 2015) and we therefore in-
pensation. For instance, additional mental health difficulties amongst clude mention of other neurodevelopmental disorders in the examples
late diagnosed adults are the rule rather than the exception (Happé below.
et al., 2016; Lever and Geurts, 2016b). Geurts and Jansen (2012) found
that 53% of individuals coming for first diagnosis in adulthood and 5.1. Intellectual ability
receiving an ASD diagnosis had previous contact with a mental health
clinic and exhibited high levels of depression and anxiety symptoms. Neurodevelopmental disorders such as ASD are not defined by a
Additionally, Cassidy et al. (2014) reported extremely high prevalence particular level of intellectual ability, spanning the whole IQ range,
of suicidal ideation (66%) amongst individuals attending an adult ASD from profound intellectual disability (IQ < 50) to average or above
clinic, which substantially exceeds estimated figures amongst adults average intelligence (IQ > 115; Charman et al., 2011). ADHD is also
diagnosed in childhood or adolescence (e.g., Balfe and Tanta, 2010). suggested to be aetiologically distinct from intellectual disability, such
There are a number of possible explanations for this association be- that diagnoses can be seen amongst populations with low, average or
tween late diagnosis and heightened mental health problems. First, the high IQ (Katusic et al., 2011; Wood et al., 2011). In spite of the fact that
effort of compensating across the lifetime might have a downstream IQ and severity of symptoms are not intrinsically linked (i.e., a severe
detrimental impact on mental health. Second, those individuals who are symptom profile can exist alongside extremely high IQ), IQ might play a
more likely to compensate in the first place might be those with addi- crucial role in the developmental unfolding of these disorders across the
tional mental health problems; for example, a more anxious individual lifetime, i.e., changes in symptoms. Indeed, higher IQ in childhood is
might be especially driven to ‘appear normal’. Lastly, but equally pos- one of the strongest predictors of ‘good outcome’ in later life for autistic
sible, additional mental health difficulties may have over-shadowed the individuals (Billstedt et al., 2005; Blumberg et al., 2016; Howlin et al.,
presentation of autistic symptoms to the extent that individuals do not 2004; Farley et al., 2009; Fein et al., 2013; Magiati et al., 2014). In-
receive a diagnosis of ASD until after a lifetime of mis/partial diag- terestingly, higher childhood IQ also characterises those that remit from
noses. ADHD symptoms by young adulthood, compared to those who persist
Interestingly, additional mental health difficulties (e.g., anxiety) are (Cheung et al., 2016). IQ, therefore, may represent a key feature of the
frequently reported amongst unaffected siblings of those with ASD propensity/ability to compensate. Indeed, in ASD, there is evidence to
(Hallett et al., 2013; Shephard et al., 2017). Whether these heightened suggest that greater IQ, in particular verbal IQ (Durrleman and Franck,
mental health problems reflect a core characteristic of the BAP, an 2015; Fisher et al., 2005; Happé, 1995), may help certain individuals to
aetiological link between ASD and other psychiatric problems, the bootstrap their limited theory of mind ability. The exact nature of the
stressful experience of growing up with a sibling with ASD or instead, relationship between IQ and compensation, however, requires
736
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
clarification. It is equally possible that the propensity to compensate behaviours, to plan behaviours before and throughout social interac-
early in life might drive IQ throughout development (Anderson, 2008). tion, and to be flexible in potentially unpredictable social scenarios
For example, some early forms of compensation might facilitate social might help to reduce observable social symptoms. There is also evi-
development, thereby providing an improved learning environment for dence to suggest that intact executive function might be important to
acquiring the domain-general skills that are tapped in an IQ test. Fi- those subgroups of individuals showing apparently good compensation.
nally, whether high IQ could support compensation across all settings is For example, Troyb et al. (2014) have reported comparable executive
unknown. For example, IQ might best facilitate compensation in a function ability across a range of tasks measuring inhibition, set-
shallow manner, within structured and predictable settings (e.g., an shifting, planning and working memory between ‘optimal outcome’ and
ADOS assessment). It may or may not be sufficient to support deeper typically developing individuals. Additionally, Lehnhardt et al. (2016)
compensation, in the fast-paced and unpredictable social settings of found that late diagnosed females had significantly greater executive
everyday life. function ability on tasks involving cognitive flexibility and processing
Good intellectual ability is also implicated in the previously dis- speed, compared with late diagnosed males, suggesting that females
cussed examples of compensation in ASD. For example, an fMRI study who have ‘coped’ without a diagnosis until adulthood may have done so
investigating neural correlates during a task of social exclusion in au- because of their good ability to monitor and regulate their social be-
tistic adolescents, their unaffected siblings and typical controls, found haviour. Future exploration of how executive function and/or in-
that in the unaffected siblings only, higher IQ was associated with a tellectual abilities might differ between early and late diagnosed groups
more ‘typical’ neural activation pattern (Bolling et al., 2015). The au- will serve to clarify the degree to which these cognitive abilities do
thors suggest that in the face of substantial risk for ASD, higher IQ indeed specifically facilitate compensation amongst females and those
might steer certain siblings towards typical social cognition. In terms of at risk of late diagnosis. Moreover, the extent to which executive
the female autism phenotype, evidence suggests that those who do meet function plays a specific role in compensation over and above general
diagnostic criteria are more likely to have poorer IQ, which tips them intellectual ability remains to be seen. One study (Pugliese et al., 2015)
over the diagnostic threshold, compared to other females with equally suggests that executive function ability is a stronger predictor of ‘good
high autistic traits (Dworzynski et al., 2012). Whether females are outcome’ (in terms of better adaptive living skills) for autistic children
simply less likely to come to clinical attention in the absence of addi- and young adults, compared to general intelligence.
tional intellectual difficulties due to the insensitivity of our diagnostic
instruments, or because higher IQ genuinely plays a specific role in 5.3. Neural mechanisms
compensating for underlying difficulties is a crucial distinction to be
made by future research (e.g., by examining the impact on mental Investigation at the neural level might be informative in de-
health and quality of life). termining the extent to which good behavioural, and possibly cognitive,
Higher IQ also appears to be an important feature of the late diag- performance is underpinned by engagement of the systems used by
nosed population. For example, Lehnhardt et al.’s (2016) study found typically developing individuals, suggesting genuine remediation or
that males and females diagnosed in adulthood had exceptionally high delayed maturation (see Table 1), or instead, alternative pathways,
verbal and non-verbal IQ. Further, females had a significantly greater suggesting compensation. There are numerous ways in which com-
processing speed compared to males, as reflected in the Digit-Symbol- pensatory processing might be reflected in an atypical neural signature.
Coding subtest of the WAIS IQ test, which the authors interpreted as a First, neural compensation may be evident in extra ‘neural effort’ re-
particular compensatory advantage for females when processing and quired from the same neural network used by neurotypicals. For ex-
analysing fast social cues. Finally, cross-sectional studies of adults ample, in the ASD literature, there are examples of fMRI studies de-
seeking a diagnosis in adulthood suggest that increasing age is asso- monstrating hyper-activation of the so-called ‘theory of mind’ network
ciated with greater intellectual ability (Happé et al., 2016) and visual (medial prefrontal cortex, posterior cingulate and lateral temporal
memory (Lever and Geurts, 2016b). One potential explanation of these cortices) when performing theory of mind tasks, implying that ‘social’
findings is that older adults who have compensated for longer in their tasks are still solved via social means, albeit, atypically. For example,
lifetime until seeking clinical support may have done so with the aid of White et al. (2014) found that even autistic adolescents who con-
enhanced intellectual abilities. sistently passed a battery of theory of mind tasks (i.e., exhibited ‘ty-
pical’ cognitive performance) demonstrated an atypical pattern of ac-
5.2. Executive function tivation of the ‘theory of mind’ network similar to those individuals
who consistently performed poorly. This latter finding highlights the
Executive function refers to a constellation of higher-order cognitive strength of combining cognitive and neuroimaging methods when in-
abilities in planning, inhibition and cognitive flexibility, which may be vestigating compensation, in particular deep compensation, where
fractionable from one another (Joseph and Tager-Flusberg, 2004). Al- cognitive task performance might remain intact.
though executive dysfunction has previously been proposed to play an Neural compensation may also be reflected in the recruitment of
aetiological role in the development of autistic symptoms (Ozonoff alternative networks that either replace or support the functioning of
et al., 1991; Russell, 1997), particularly non-social symptoms (e.g. the dysfunctional network. Whether an alternative system could ever
Yerys et al., 2009), there is increasing evidence to suggest that it is not a fully substitute for the functioning of the ‘theory of mind’ system is yet
universal feature of ASD (Brunsdon et al., 2015; Cantio et al., 2016; to be established. There is, however, evidence to support Johnson
Geurts et al., 2014; Wallace et al., 2016). This has led to the suggestion et al.’s (2015b) proposal that the PFC is a candidate brain region for
that executive dysfunction should be conceived as an individual dif- compensatory processing, given its fundamental role in top-down co-
ference amongst the autistic population, rather than as a core feature of ordination of other cortical regions. For instance, Kaiser et al. (2010)
ASD itself (Johnson, 2012). Instead, executive function might remain used fMRI whilst individuals with ASD, their unaffected siblings and
relatively intact for a proportion of autistic individuals and can there- controls, watched biological motion. Of particular interest, they found a
fore be recruited to support compensation. Indeed, Johnson et al. unique pattern of activity, only for the unaffected siblings, in the su-
(2015b) speculate that those with a double hit of ASD and executive perior temporal sulcus and most intriguingly, the PFC, suggesting that
dysfunction might be completely stripped of compensatory resources PFC-dependent brain activity might be characteristic of individuals at
and thus will be the most likely to exhibit a severe behavioural profile. risk of ASD, who bypass an ASD trajectory via compensation. Evidence
One can speculate about the mechanisms by which executive for recruitment of additional brain areas not otherwise used by typically
function ability might facilitate compensation, particularly shallow developing individuals is also found in the literature on ADHD (for a
compensation. For example, greater ability to inhibit undesirable social review, see Fassbender and Schweitzer, 2009) and developmental
737
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
dyslexia (see Section 2.3). cognitive (Brunsdon and Happé, 2014) and socio-cognitive (Happé
The alternative possibility is that, at a neural level, compensation et al., 2017) atypicalities associated with ASD. Whilst we await im-
will be represented rather idiosyncratically, such that there is great proved characterisation of these cognitive deficits, as well as potential
variability from one ‘compensated’ individual to another, reflecting the biological markers, ascertaining compensation in the clinic may be best
numerous possible pathways to compensation. Interestingly, evidence done through observation of behaviour across multiple contexts that
for this may be disorder-specific. For example, in the case of ASD, differ in social demands.
studies measuring neural activation during theory of mind tasks con- The study of compensation may also provide insight into supporting
sistently find either hyper-activation (see references above) or hypo- those on the autism spectrum. For example, we may hope to boost
activation (Kana et al., 2015; Lombardo et al., 2011; O’Nions et al., compensatory resources early on in individuals at risk for ASD
2014) of the ‘theory of mind’ network, in both children and adults with (Johnson, 2012). Additionally, individuals that are limited in their
ASD, rather than any evidence of the idiosyncratic neural patterns that compensatory ability, for instance, those with both low IQ and/or poor
might have been predicted a priori (Happé and Frith, 2014). Finally, it executive function, may represent a particularly vulnerable subgroup of
should be noted that compensation could be reflected in suppressed individuals. And yet, we have here challenged the notion that com-
brain activity, as suggested in the aging literature (see Grady, 2012). pensation is a universally desirable process. It may be that those sub-
It remains to be investigated by future research which of the po- groups suspected to be engaging in good compensation, particularly
tential neural mechanisms outlined above would be the most successful shallow compensation, are also at risk for mental health problems.
form of compensatory processing. That is to say, which would be the Therefore, one should perhaps be cautious in universally promoting the
most efficient and effective for supporting, for example, ‘typical’ social use of compensatory mechanisms as a means of fostering genuinely
behaviour in an ADOS assessment (i.e. shallow compensation), and ‘good’ outcome amongst autistic individuals. Indeed, there is evidence
which would extend to supporting theory of mind task performance and to suggest that our objective, neurotypical definitions of ‘good out-
everyday, flexible social interaction (i.e. deep compensation). We might come’, which dominate the literature, do not necessarily coincide with
suspect that those individuals engaging in shallow compensation will autistic individual’s self-rated quality of life (van Heijst and Geurts,
continue to exhibit persistent brain atypicalities that are characteristic 2015). More research is needed to identify neurodiverse concepts of
of ASD. However, the possibility remains that those engaging in deep ‘good outcome’, and thinking about neurodevelopmental disorders
compensation may in fact demonstrate fewer ASD-relevant brain aty- through a compensation lens may be helpful in this.
picalities compared to other autistic individuals, particularly with in-
creasing age, as compensatory strategies are refined. 7. Limitations and outstanding questions
6. Implications of a compensation framework The impetus for defining and subsequently describing compensation
in this paper was to synthesise the limited evidence so far for this
Thinking about neurodevelopmental disorders within a compensa- phenomenon. For the purpose of demonstration, we have focused on
tion framework could have a range of implications for research and one single cognitive deficit (theory of mind) within one condition
clinical practice. First, compensation might begin to explain why chil- (ASD), in order to explore how compensation might operate at beha-
dren who appear behaviourally similar in childhood (e.g., Fountain vioural and cognitive levels. In reality, there are likely multiple cog-
et al., 2012) can follow divergent pathways to outcome, thus demon- nitive deficits that might be compensated for in different ways, across
strating great heterogeneity in the degree to which their symptoms numerous conditions. This latter point feeds into the issue of hetero-
improve, persist or worsen. Second, evidence so far suggests that geneity, which is a defining feature of ASD. There may truly exist sub-
compensation might play a role in under- or mis-diagnosis in the clinic. groups of individuals who i) genuinely remit from ASD and its cognitive
Diagnostic procedures for neurodevelopmental disorders, such as ASD, characteristics (as suggested by Fein et al., 2013), or ii) follow a more
continue to rely on the severity of behavioural symptoms presented to benign trajectory of the condition, thereby promoting good outcome via
the clinical observer. And yet, if there exists a substantial behaviour- some mechanism other than compensation (e.g., delayed maturation).
cognition mismatch for certain subgroups, including some females and Further research will be required to detect the subtle but qualitative
late diagnosed adults, the extent of an individual’s underlying socio- differences between these possible subgroups and their aetiological
cognitive difficulties would not necessarily be apparent in the clinic. An bases/biomarkers. In Table 1 we have outlined hypothetical distinc-
individual with very limited understanding of others’ minds may on the tions between compensation (both shallow and deep) and the other
surface appear not sufficiently ‘impaired’ to warrant diagnosis, espe- mechanisms that might promote improved outcome. Most importantly,
cially if compensation has supported not only improved behavioural alternative explanations to compensation will only be plausible if there
presentation but also some level of independence (e.g., educational or is good evidence to suggest that stable underlying difficulties no longer
occupational success). Furthermore, even autistic individuals who exist, which, as outlined earlier, will require sensitive cognitive probes.
eventually achieve ‘good outcome’ and perhaps no longer meet diag- Ultimately the most robust method to address this question will be a
nostic criteria for ASD, but endure persistent difficulties at the cognitive longitudinal developmental approach.
level, will continue to have important support needs that may otherwise The notion of compensation also poses intriguing questions about
be overlooked. Finally, those individuals that compensate sufficiently to mechanisms underlying various interventions for neurodevelopmental
sit just below the diagnostic threshold, but experience socio-cognitive conditions. For example, we can ask whether such interventions are
difficulties comparable to other diagnosed individuals, not only risk attempting to genuinely alleviate core difficulties or whether they are in
mis-diagnosis (Rutter, 2011) but may also be the most vulnerable to fact trying to promote a layer of compensation, which supports the
mental health problems. individual to reach a more neurotypical outcome. In the case of inter-
Ideally, joining cognitive tasks (e.g., theory of mind tasks) and ventions aimed at improving social skills during childhood/adoles-
biological markers (e.g., genetic and neural markers for ASD) with the cence, it seems that these may be targeting shallow compensation; for
assessment of behaviour in the clinic would be the optimal way to example, explicitly teaching individuals rules for how to make and
determine if individuals are compensating and as such, avoid under- or maintain eye contact, initiate conversation and behave appropriately in
mis-diagnosis. However, at present, this is far from straightforward. The particular social scenarios. This form of training is likely inflexible and
precise cognitive features that underpin conditions such as ASD are not there is evidence for failure to generalise (for a review, see Fletcher-
unequivocally understood. In this paper, for the purpose of illustration, Watson et al., 2014). Early intensive interventions, however, (e.g.,
we have discussed theory of mind impairment as it is a widely reported Early Start Denver Model, Dawson et al., 2010; Focused Playtime In-
core deficit in ASD (Happé, 2015), but there is an array of potential tervention, Kasari et al., 2014), where there is an intention to rebuild
738
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
739
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
P., Kuntsi, J., 2016. Cognitive and neurophysiological markers of ADHD persistence Hansen, S.N., Schendel, D.E., Parner, E.T., 2015. Explaining the increase in the pre-
and remission. Br. J. Psychiatry 208, 548–555. http://dx.doi.org/10.1192/bjp.bp. valance of autism spectrum disorders: the proportion attributable to changes in re-
114.145185. porting practices. JAMA Pediatr. 169, 56–62. http://dx.doi.org/10.1001/
Corkett, J.K., Parrila, R., 2008. Use of context in the word recognition process by adults jamapediatrics.2014.1893.
with a significant history of reading difficulties. Ann. Dyslexia 58, 139–161. http:// Happé, F., 1994. An advanced test of theory of mind: understanding of story characters’
dx.doi.org/10.1007/s11881-008-0018-1. thoughts and feelings by able autistic, mentally handicapped, and normal children
Dawson, G., Rogers, S., Munson, J., Smith, M., Winter, J., Greenson, J., Donaldson, A., and adults. J. Autism Dev. Disord. 24, 129–154. http://dx.doi.org/10.1007/
Varley, J., 2010. Randomized, controlled trial of an intervention for toddlers with BF02172093.
autism: the Early Start Denver Model. Pediatrics 125, e17–e23. http://dx.doi.org/10. Happé, F., 1995. The role of age and verbal ability in the theory of mind task performance
1542/peds.2009-0958. of subjects with autism. Child Dev. 66, 843–855. http://dx.doi.org/10.1111/j.1467-
Durrleman, S., Franck, J., 2015. Exploring links between language and cognition in 8624.1995.tb00909.x.
autism spectrum disorders: complement sentences, false belief, and executive func- Happé, F., 2015. Autism as a neurodevelopmental disorder of mind-reading. J. Br. Acad.
tioning. J. Commun. Disord. 54, 15–31. http://dx.doi.org/10.1016/j.jcomdis.2014. 3, 197–209. http://dx.doi.org/10.5871/jba/003.197.
12.001. Happé, F., Charlton, R.A., 2012. Aging in autism spectrum disorders: a mini-review.
Dworzynski, K., Ronald, A., Bolton, P., Happé, F., 2012. How different are girls and boys Gerontology 58, 70–78. http://dx.doi.org/10.1159/000329720.
above and below the diagnostic threshold for autism spectrum disorders? J. Am. Happé, F., Frith, U., 2006. The weak coherence account: detail-focused cognitive style in
Acad. Child Adolesc. Psychiatry 51, 788–797. http://dx.doi.org/10.1016/j.jaac. autism spectrum disorders. J. Autism Dev. Disord. 36, 5–25. http://dx.doi.org/10.
2012.05.018. 1007/s10803-005-0039-0.
Dziobek, I., Fleck, S., Kalbe, E., Rogers, K., Hassenstab, J., Brand, M., Kessler, J., Woike, Happé, F., Frith, U., 2014. Annual research review: Towards a developmental neu-
J.K., Wolf, O.T., Convit, A., 2006. Introducing MASC: a movie for the assessment of roscience of atypical social cognition. J. Child Psychol. Psychiatry 55, 553–557.
social cognition. J. Autism Dev. Disord. 36, 623–636. http://dx.doi.org/10.1007/ http://dx.doi.org/10.1111/jcpp.12162.
s10803-006-0107-0. Happé, F., Ronald, A., Plomin, R., 2006. Time to give up on a single explanation for
Eaves, L.C., Ho, H.H., 2008. Young adult outcome of autism spectrum disorders. J. Autism autism. Nat. Neurosci. 9, 1218–1220. http://dx.doi.org/10.1038/nn1770.
Dev. Disord. 38, 739–747. http://dx.doi.org/10.1007/s10803-007-0441-x. Happé, F., Mansour, H., Barrett, P., Brown, T., Abbott, P., Charlton, R.A., 2016.
Eigsti, I.-M., Stevens, M.C., Schultz, R.T., Barton, M., Kelley, E., Naigles, L., Orinstein, A., Demographic and cognitive profile of individuals seeking a diagnosis of autism
Troyb, E., Fein, D.A., 2016. Language comprehension and brain function in in- spectrum disorder in adulthood. J. Autism Dev. Disord. 46, 3469–3480. http://dx.
dividuals with an optimal outcome from autism. NeuroImage Clin. 10, 182–191. doi.org/10.1007/s10803-016-2886-2.
http://dx.doi.org/10.1016/j.nicl.2015.11.014. Happé, F., Cook, J.L., Bird, G., 2017. The structure of social cognition: in(ter)dependence
Elsabbagh, M., Johnson, M.H., 2010. Getting answers from babies about autism. Trends of sociocognitive processes. Annu. Rev. Psychol. 68, 243–267. http://dx.doi.org/10.
Cogn. Sci. 14, 81–87. http://dx.doi.org/10.1016/j.tics.2009.12.005. 1146/annurev-psych-010416-044046.
Faraone, S.V., Biederman, J., Mick, E., 2006. The age-dependent decline of attention Hatcher, J., Snowling, M.J., Griffiths, Y.M., 2002. Cognitive assessment of dyslexic stu-
deficit hyperactivity disorder: a meta-analysis of follow-up studies. Psychol. Med. 36, dents in higher education. Br. J. Educ. Psychol. 72, 119–133. http://dx.doi.org/10.
159–165. http://dx.doi.org/10.1017/S003329170500471X. 1348/000709902158801.
Farley, M.A., McMahon, W.M., Fombonne, E., Jenson, W.R., Miller, J., Gardner, M., Helles, A., Gillberg, C.I., Gillberg, C., Billstedt, E., 2015. Asperger syndrome in males over
Block, H., Pingree, C.B., Ritvo, E.R., Ritvo, R.A., Coon, H., 2009. Twenty-year out- two decades: stability and predictors of diagnosis. J. Child Psychol. Psychiatry 56,
come for individuals with autism and average or near-average cognitive abilities. 711–718. http://dx.doi.org/10.1111/jcpp.12334.
Autism Res. 2, 109–118. http://dx.doi.org/10.1002/aur.69. Helt, M., Kelley, E., Kinsbourne, M., Pandey, J., Boorstein, H., Herbert, M., Fein, D., 2008.
Fassbender, C., Schweitzer, J.B., 2009. Is there evidence for neural compensation in at- Can children with autism recover? If so, how? Neuropsychol. Rev. 18, 339–366.
tention deficit hyperactivity disorder? A review of the functional neuroimaging lit- http://dx.doi.org/10.1007/s11065-008-9075-9.
erature. Clin. Psychol. Rev. 26, 445–465. http://dx.doi.org/10.1016/j.cpr.2006.01. Hickey, A., Crabtree, J., Stott, J., 2017. ‘Suddenly the first fifty years of my life made
003. sense’: experiences of older people with autism. Autism. http://dx.doi.org/10.1177/
Fein, D., Barton, M., Eigsti, I.-M., Kelley, E., Naigles, L., Schultz, R.T., Stevens, M., Helt, 1362361316680914.
M., Orinstien, A., Rosenthal, M., Troyb, E., Tyson, K., 2013. Optimal outcome in Hoeft, F., Meyler, A., Hernandez, A., Juel, C., Taylor-Hill, H., Martindale, J.L., McMillion,
individuals with a history of autism. J. Child Psychol. Psychiatry 54, 195–205. http:// G., Kolchugina, G., Black, J.M., Faizi, A., Deutsch, G.K., Siok, W.T., Reiss, A.L.,
dx.doi.org/10.1111/jcpp.12037. Whitfield-Gabrieli, S., Gabrieli, J.D.E., 2007. Functional and morphometric brain
Fisher, N., Happé, F., Dunn, J., 2005. The relationship between vocabulary, grammar, and dissociation between dyslexia and reading ability. PNAS 104, 4234–4239. http://dx.
false belief task performance in children with autistic spectrum disorders and chil- doi.org/10.1073/pnas.0609399104.
dren with moderate learning difficulties. J. Child Psychol. Psychiatry 46, 409–419. Hoeft, F., McCandliss, B.D., Black, J.M., Gantman, A., Zakerani, N., Hulme, C., Lyytinen,
http://dx.doi.org/10.1111/j.1469-7610.2004.00371.x. H., Whitfield-Gabrieli, S., Glover, G.H., Reiss, A.L., Gabrieli, J.D.E., 2011. Neural
Fletcher-Watson, S., McConnell, F., Manola, E., McConachie, H., 2014. Interventions systems predicting long-term outcome in dyslexia. PNAS 108, 361–366. http://dx.
based on the theory of mind cognitive model for autism spectrum disorders (ASD). doi.org/10.1073/pnas.1008950108.
Cochrane Database Syst. Rev.. http://dx.doi.org/10.1002/14651858.cd008785. Hofvander, B., Delorme, R., Chaste, P., Nydén, A., Wentz, E., Ståhlberg, O., Herbrecht, E.,
pub2. CD008785. Stopin, A., Anckarsäter, H., Gillberg, C., Råstam, M., Leboyer, M., 2009. Psychiatric
Fombonne, E., 2009. Epidemiology of pervasive developmental disorders. Pediatr. Res. and psychosocial problems in adults with normal-intelligence autism spectrum dis-
65, 591–598. http://dx.doi.org/10.1203/PDR.0b013e31819e7203. orders. BMC Psychiatry 9, 35. http://dx.doi.org/10.1186/1471-244X-9-35.
Fountain, C., Winter, A.S., Bearman, P.S., 2012. Six developmental trajectories char- Howlin, P., Goode, S., Hutton, J., Rutter, M., 2004. Adult outcome for children with
acterize children with autism. Pediatrics 129, e1112–e1120. http://dx.doi.org/10. autism. J. Child Psychol. Psychiatry 45, 212–229. http://dx.doi.org/10.1111/j.1469-
1542/peds.2011-1601. 7610.2004.00215.x.
Frith, U., 1991. Asperger and his syndrome. In: Frith, U. (Ed.), Autism and Asperger Hull, L., Petrides, K.V., Allison, C., Smith, P., Baron-Cohen, S., Lai, M.-C., Mandy, W.,
Syndrome. Cambridge University Press, Cambridge, pp. 1–39. 2017. “Putting on my best normal”: social camouflaging in adults with autism
Frith, U., 2004. Emanuel Miller lecture: confusions and controversies about Asperger spectrum conditions. J. Autism Dev. Disord. http://dx.doi.org/10.1007/s10803-017-
syndrome. J. Child Psychol. Psychiatry 45, 672–686. http://dx.doi.org/10.1111/j. 3166-5.
1469-7610.2004.00262.x. Ingvar, M., af Trampe, P., Greitz, T., Eriksson, L., Stone-Elander, S., von Euler, C., 2002.
Frith, U., 2012. Why we need cognitive explanations of autism. Q. J. Exp. Psychol. 65, Residual differences in language processing in compensated dyslexics revealed in
2073–2092. http://dx.doi.org/10.1080/17470218.2012.697178. simple word reading tasks. Brain Lang. 83, 249–267. http://dx.doi.org/10.1016/
Frith, U., 2013. Autism and dyslexia: a glance over 25 years of research. Perspect. S0093-934X(02)00055-X.
Psychol. Sci. 8, 670–672. http://dx.doi.org/10.1177/1745691613507457. Johnson, M.H., 2012. Executive function and developmental disorders: the flip side of the
Gallagher, A.M., Laxon, V., Armstrong, E., Frith, U., 1996. Phonological difficulties in coin. Trends Cogn. Sci. 16, 454–457. http://dx.doi.org/10.1016/j.tics.2012.07.001.
high-functioning dyslexics. Read. Writ. 8, 499–509. http://dx.doi.org/10.1007/ Johnson, M.H., 2017. Autism as an adaptive common variant pathway for human brain
BF00577025. development. Dev. Cogn. Neurosci. 25, 5–11. http://dx.doi.org/10.1016/j.dcn.2017.
Georgiades, S., Bishop, S.L., Frazier, T., 2017. Editorial perspective: Longitudinal research 02.004.
in autism−introducing the concept of ‘chronogeneity’. J. Child Psychol. Psychiatry Johnson, M.H., Gilga, T., Jones, E., Charman, T., 2015a. Annual research review: Infant
58, 634–636. http://dx.doi.org/10.1111/jcpp.12690. development, autism, and ADHD–early pathways to emerging disorders. J. Child
Geurts, H.M., Jansen, M.D., 2012. A retrospective chart study: the pathway to a diagnosis Psychol. Psychiatry 56, 228–247. http://dx.doi.org/10.1111/jcpp.12328.
for adults referred for ASD assessment. Autism 16, 299–305. http://dx.doi.org/10. Johnson, M.H., Jones, E.J., Gilga, T., 2015b. Brain adaptation and alternative develop-
1177/1362361311421775. mental trajectories. Dev. Psychopathol. 27, 425–442. http://dx.doi.org/10.1017/
Geurts, H.M., Sinzig, J., Booth, R., Happé, F., 2014. Neuropsychological heterogeneity in S0954579415000073.
executive functioning in autism spectrum disorders. Int. J. Dev. Disabil. 60, 155–162. Jones, E.J., Gilga, T., Bedford, R., Charman, T., Johnson, M.H., 2014. Developmental
http://dx.doi.org/10.1179/2047387714Y.0000000047. pathways to autism: a review of prospective studies of infants at risk. Neurosci.
Grady, C., 2012. The cognitive neuroscience of ageing. Nat. Rev. Neurosci. 13, 491–505. Biobehav. Rev. 39, 1–33. http://dx.doi.org/10.1016/j.neubiorev.2013.12.001.
http://dx.doi.org/10.1038/nrn3256. Joseph, R.M., Tager-Flusberg, H., 2004. The relationship of theory of mind and executive
Hallett, V., Ronald, A., Colvert, E., Ames, C., Woodhouse, E., Lietz, S., Garnett, T., Gillan, functions to symptom type and severity in children with autism. Dev. Psychopathol.
N., Rijsdijk, F., Scahill, L., Bolton, P., Happé, F., 2013. Exploring anxiety symptoms in 16, 137–155. http://dx.doi.org/10.1017/S095457940404444X.
a large-scale twin study of children with autism spectrum disorders, their co-twins Kaiser, M.D., Hudac, C.M., Shultz, S., Lee, S.M., Cheung, C., Berken, A.M., Deen, B.,
and controls. J. Child Psychol. Psychiatry 54, 1176–1185. http://dx.doi.org/10. Pitskel, N.B., Sugrue, D.R., Voos, A.C., Saulnier, C.A., Ventola, P., Wolf, J.M., Klin, A.,
1111/jcpp.12068. Vander Wyk, B.C., Pelphrey, K.A., 2010. Neural signatures of autism. Proc. Natl.
740
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
741
L.A. Livingston, F. Happé Neuroscience and Biobehavioral Reviews 80 (2017) 729–742
Skuse, D.H., 2007. Rethinking the nature of genetic vulnerability to autistic spectrum analyis. Autism 19, 158–167. http://dx.doi.org/10.1177/1362361313517053.
disorders. Trends Genet. 23, 387–395. http://dx.doi.org/10.1016/j.tig.2007.06.003. van Wijngaarden-Cremers, P.J., van Eeten, E., Groen, W.B., van Deurzen, P.A., Oosterling,
Snowling, M.J., 2014. Dyslexia: a language learning impairment. J. Brit. Acad. 2, 43–58. I.J., van der Gaag, R.J., 2014. Gender and age differences in the core triad of im-
http://dx.doi.org/10.5871/jba/002.043. pairments in autism spectrum disorders: a systematic review and meta-analysis. J.
Steinhausen, H.C., 2009. The heterogeneity of causes and courses of attention-deficit/ Autism Dev. Disord. 44, 627–635. http://dx.doi.org/10.1007/s10803-013-1913-9.
hyperactivity disorder. Acta Psychiatr. Scand. 120, 392–399. http://dx.doi.org/10. Varnet, L., Meunier, F., Trollé, G., Hoen, M., 2016. Direct viewing of dyslexics’ com-
1111/j.1600-0447.2009.01446.x. pensatory strategies in speech in noise using auditory classification images. PLoS One
Sucksmith, E., Roth, I., Hoekstra, R.A., 2011. Autistic traits below the clinical threshold: 11, e0153781. http://dx.doi.org/10.1371/journal.pone.0153781.
re-examining the broader autism phenotype in the 21st century. Neuropsychol. Rev. Wallace, G.L., Yerys, B.E., Peng, C., Dlugi, E., Anthony, L.G., Kenworthy, L., 2016.
21, 360–389. http://dx.doi.org/10.1007/s11065-011-9183-9. Assessment and treatment of executive function impairments in autism spectrum
Swanson, H.L., 2012. Adults with reading disabilities: converting a meta-analysis to disorder: an update. Int. Rev. Res. Dev. Disabil. 51, 85–122. http://dx.doi.org/10.
practice. J. Learn. Disabil. 45, 17–30. http://dx.doi.org/10.1177/ 1016/bs.irrdd.2016.07.004.
0022219411426856. White, S.J., Frith, U., Rellecke, J., Al-Noor, Z., Gilbert, S.J., 2014. Autistic adolescents
Thambirajah, M.S., 2010. Developmental dyslexia: clinical aspects. Adv. Psychiatr. Treat. show atypical activation of the brain’s mentalizing system even without a prior
16, 380–387. http://dx.doi.org/10.1192/apt.bp.108.006239. history of mentalizing problems. Neuropsychologia 56, 17–25. http://dx.doi.org/10.
Thapar, A., Cooper, M., Rutter, M., 2017. Neurodevelopmental disorders. Lancet 1016/j.neuropsychologia.2013.12.013.
Psychiatry 4, 339–346. http://dx.doi.org/10.1016/S2215-0366(16)30376-5. Wood, A.C., Rijsdijk, F., Johnson, K.A., Andreou, P., Albrecht, B., Arias-Vasquez, A.,
Thomas, M., Karmiloff-Smith, A., 2002. Are developmental disorders like cases of adult Buitlaar, J.K., McLoughlin, G., Rommelse, N.N.J., Sergeant, J.A., Sonuga-Barke,
brain damage? Implications from connectionist modelling. Behav. Brain Sci. 25, E.J.S., Uebel, H., van der Meere, J.J., Banaschewski, T., Gill, M., Manor, I., Miranda,
727–788. http://dx.doi.org/10.1017/S0140525x02000134. A., Mulas, F., Oades, R.D., Roeyers, H., Rothenberger, A., Steinhausen, H.C., Faraone,
Troyb, E., Rosenthal, M., Eigsti, I.-M., Kelley, E., Tyson, K., Orinstein, A., Barton, M., Fein, S.V., Asherson, P., Kuntsi, J., 2011. The relationship between ADHD and key cog-
D., 2014. Executive functioning in individuals with a history of ASDs who have nitive phenotypes is not mediated by shared familial effects with IQ. Psychol. Med.
achieved optimal outcomes. Child Neuropsychol. 20, 378–397. http://dx.doi.org/10. 41, 861–871. http://dx.doi.org/10.1017/S003329171000108X.
1080/09297049.2013.799644. Yerys, B.E., Wallace, G.L., Sokoloff, J.L., Shook, D.A., James, J.D., Kenworthy, L., 2009.
Ullman, M.T., Pullman, M.Y., 2015. A compensatory role for declarative memory in Attention deficit/hyperactivity disorder symptoms moderate cognition and behavior
neurodevelopmental disorders. Neurosci. Biobehav. Rev. 51, 205–222. http://dx.doi. in children with autism spectrum disorders. Autism Res. 2, 322–333. http://dx.doi.
org/10.1016/j.neubiorev.2015.01.008. org/10.1002/aur.103.
van Heijst, B.F., Geurts, H.M., 2015. Quality of life in autism across the lifespan: a meta-
742