Habituation and novelty detection fNIRS brain responses in 5‐ and 8‐month‐old infants: The Gambia and UK

Abstract The first 1,000 days of life are a critical window of vulnerability to exposure to socioeconomic and health challenges (i.e. poverty/undernutrition). The Brain Imaging for Global Health (BRIGHT) project has been established to deliver longitudinal measures of brain development from 0 to 24 months in UK and Gambian infants and to assess the impact of early adversity. Here results from the Habituation‐Novelty Detection (HaND) functional near‐infrared spectroscopy (fNIRS) task at 5 and 8 months are presented (N = 62 UK; N = 115 Gambia). In the UK cohort distinct patterns of habituation and recovery of response to novelty are seen, becoming more robust from 5 to 8 months of age. In The Gambia, an attenuated habituation response is evident: a larger number of trials are required before the response sufficiently suppresses relative to the response during the first presented trials. Furthermore, recovery of response to novelty is not evident at 5 or 8 months of age. As this longitudinal study continues in The Gambia, the parallel collection of socioeconomic, caregiving, health and nutrition data will allow us to stratify how individual trajectories of habituation and recovery of response to novelty associate with different risk factors and adaptive mechanisms in greater depth. Given the increasing interest in the use of neuroimaging methods within global neurocognitive developmental studies, this study provides a novel cross‐culturally appropriate paradigm for the study of brain responses associated with attention and learning mechanisms across early development.

This means that over 80 million children in LMICs fail to develop a core set of age-appropriate skills that allow them to maintain attention, understand and follow simple directions, communicate and cooperate with others, control aggression, and solve complex problems. Compromised development of these skills may have a significant impact on their subsequent academic achievement, mental health and economic status, and consequently their potential to lead full and productive lives and support future generations.
The first 1,000 days of life (from conception to 2 years of age) represents a critical window for brain and nervous system development in humans and have been described "as a foundation and catalyst of human development in the balance of the life course" (Bornstein, 2014). While only a fraction of our lifespan, it is characterized by prodigious physiological, psychological and physical change.
From the first moments of life, the as-yet physically immature neonate manifests stimulus specific physiological, behavioural and cortical responses across the senses: vision (Farroni et al., 2013;Farroni, Csibra, Simion, & Johnson, 2002), audition (DeCasper & Fifer, 1980;Ockleford, Vince, Layton, & Reader, 1988), smell (Bartocci et al., 2000) and touch (Lejeune et al., 2010). In the weeks and months that follow, infants assimilate information through multiple interactions with their world; developing and honing physical, expressive and receptive skills. The brain is at its most plastic during this time, and these changes are the outward manifestation of increasing cortical specialization to its particular environment (Johnson, 2011). Over the course of an infant's individual development the brain adapts both to the general features of the environment shared by others, and to the individual circumstances into which it is born (Johnson, Jones, & Gliga, 2015).
How the individual brain adapts during a lifespan in relation to early environmental adversity, can be thought of in different ways.
First, as related to "resilience," or rather the ability of the individual to avoid deviation from a typical developmental trajectory in the face of environmental challenges (Karatsoreos & McEwen, 2013).
Second as related to "ontogenic adaptation" whereby an individual develops in such a way that allow an optimal fit between environmental demands and brain functioning. Rather than making the assumption that one trajectory is favourable over the other, this notion allows for adaptive advantages to be experienced in different contexts (Johnson et al., 2015). This can be conceptualized along different time scales.
Adaptability of the brain may be essential for normal function and brain development in response to events ranging from the acute (e.g. premature birth), to the medium-term (e.g. response to acute periods of infection leading to undernutrition/sleep disruption at one age point) to the chronic (e.g. continued presence of early adversity such as undernutrition or poverty throughout childhood). Furthermore, these factors may work synergistically whereby an infant born prematurely may be at greater risk of infectious illnesses and may consequently experience prolonged undernutrition (Nelson, 2017). In addition factors such as poverty can be extrinsically bound to further psychosocial risk factors (Jensen, Berens, & Nelson, 2017), such as child maltreatment, domestic violence or stress (Giovanelli, Reynolds, Mondi, & Ou, 2016). With perhaps more widespread consequences, poverty will also require caregivers to allot an increased allocation of time to income generation and household duties and thus decreased opportunities for infant-caregiver interactions and decreased access to family resources (Milteer, Ginsburg, Council On Communications And Media, & Committee On Psychosocial Aspects Of Child And Family Health, 2012; Worku et al., 2018).
While many studies suggest that the presence of these risk factors in infancy has a lasting impact throughout the life course (Hackman & Farah, 2009;Martorell et al., 2010;Victora et al., 2008), very little is known about the neural bases of these early deficits, particularly in low-and middle-income countries. This may be partly a consequence of limited availability of child-friendly portable brain imaging methods (Isaacs, 2013;Lloyd-Fox et al., 2017). Therefore the investigation of the developing brain in remote rural or urban "field based" settings has until recently been broadly limited to behavioural assessments (Georgieff, 2007;Sabanathan, Wills, & Gladstone, 2015).
Importantly, while behavioural responses can be measured from birth (i.e. visual attention, sucking modulation) early in development these responses may be more variable or subtle and therefore less reliable than a neural measure. Furthermore, observable changes in an individual's behaviour may not be measurable in pre-verbal infants and/or become predictive of developmental outcome until much later in development (i.e. toddlerhood)-despite significantly earlier changes in brain functional and anatomical specialization-thereby missing a window of vulnerability for initiating an intervention. Neural measures may allow us to access this window and provide earlier diagnostic tools.
To contribute to our understanding, it is therefore imperative neurocognitive development is studied longitudinally and from an

RESEARCH HIGHLIGHTS
• Global health research has been limited to the use of behavioural assessments to measure the effect of exposure to early adversity (socioeconomic and health challenges) • Functional near-infrared spectroscopy (fNIRS) is a portable, non-invasive, low-cost imaging technique that provides a new avenue for measuring brain function in low-resource settings.
• A novel cross-culturally appropriate paradigm demonstrated habituation and novelty detection brain responses in 5-and 8-month-old infants in the UK.
• In the Gambia, habituation brain responses were attenuated, and a recovery of response to novelty absent, at the group level, at 5-8 months of age. early age, taking contemporaneous measurements of brain function in parallel with measurements of exposure to environmental challenges. The alternative-assessing impairment at school age and attempting to retrospectively decode the cascading effects of early environmental insults-is unlikely to reveal clear and reliable targets for intervention.
The Gambia, the smallest country on the African continent, has, for over 70 years, partnered with the UK Medical Research Council (MRC) to conduct research to further understand the impact of undernutrition and poverty and provide interventions to promote healthy growth in the population. This research focus is partly due to the contrasting weather patterns which alternate between 6 months where rain occurs and six months of extreme dryness, which directly affects the availability of key nutrients (Moore et al., 1997). The majority (60%) of the roughly 2 million inhabitants live in the coastal regions surrounding the capital, whereas the remainder of the population live rurally, often supporting themselves through subsistence farming (Hennig et al., 2017). The Gambia is one of the lowest ranking countries with regard to gross national income, years of schooling and life expectancy, with over half of adults in The Gambia never having received formal education (Hennig et al., 2017). For the last decade, education level has risen as there is now free universal education which 97% of children attend to primary level (The Gambia Government National Education Statistics by Gender, 2016). Marriages are commonly polygamous with over half of wives living with one or two cowives (Hennig et al., 2017). Over the past decades infant and child mortality has decreased, birth spacing has increased, and overall family size has gone down (Nabwera, Fulford, Moore, & Prentice, 2017 ac.uk). One of their rural research sites is located in the West Kiang District, which is 145 km inland from the capital of The Gambia. The field station where the study took place is reached by 4 × 4 vehicles on unmade roads and located within a remote inland village, Keneba.
Due to its isolation the field station in the village maintains all facilities necessary for research and clinical care (generator powered electricity, water supplied by bore hole, satellite communication).
The community in this region relies predominantly on subsistence farming and, thus, eating patterns and income vary greatly between the annual wet and dry seasons (Hennig et al., 2017;van der Merwe et al., 2013). Additionally, there is a high prevalence of infectious disease (van der Merwe et al., 2013). The combination of prenatal growth retardation, poor-quality and often contaminated foods and high levels of infection cause moderate to severe growth faltering (failure to thrive) in height and weight gain from around three months of age in the local population of West Kiang (Lunn, 2000;Lunn, Northrop-Clewes, & Downes, 1991;van der Merwe et al., 2013). Growth faltering is most severe beyond the first six-months of life, when non-breast milk feeds are introduced (Eriksen et al., 2017).
Over the last 6 years the use of infant friendly neuroimaging techniques has been pioneered at this MRCG at LSHTM field station in The Gambia (Begus et al., 2016;Lloyd-Fox et al., 2017, 2014Papademetriou et al., 2014). These proof-of-concept studies have led to the establishment of a prospective longitudinal study to measure early neurocognitive development during the first 2 years of life at two parallel sites, in the UK and The Gambia (see Supplementary information Data S1 for demographical information on The Gambia).
This project aims to implement brain imaging measures (functional ) and family-caregiving assessments (caregiver-infant interaction videos and questionnaires) into a framework of regular collection of biological, socioeconomic, parental health and nutritional data at both sites. The data presented in this study are from the full UK cohort (N = 62) and the first 115 infants studied in The Gambia 1 and longitudinal data collection is ongoing.
The longitudinal design includes nine data collection phases; antenatal (32-36 weeks gestation), and postnatal-1-3, 7-14 days, 1, 5, 8, 12, 18 and 24 months of age. In the full-scale BRIGHT study, the cohort size in The Gambia was selected-to provide adequate power to support within-cohort comparisons-on the assumption that approximately 25%-30% of the cohort will be stunted (z-score of length-for-age <2 standard deviations below the WHO reference) by 2 years of age (Nabwera et al., 2017). While a second cohort within The Gambia (i.e. infants from an urban coastal where undernutrition is less prevalent) was originally considered, there are considerably varying environmental and cultural factors to take into account (which are more rapidly changing relative to rural areas due to economic development, and the associated health and nutrition transition). Therefore a larger cohort was purposefully selected within the rural district. The purpose of the BRIGHT study is not to draw comparisons between the UK and Gambia cohorts. This design will firstly allow the modelling of longitudinal changes in brain function, cognitive development and growth within the rural Gambian population. Second, through the collection of parallel behavioural and environmental data it will allow the identification of critical developmental moderators, mediators and markers of resilience. Third, given that neuroimaging data provide the backbone of this study, it was essential that a UK cohort was established to measure the longitudinal developmental trajectories of the different fNIRS and EEG paradigms (some of which have not been studied at these ages before). This was chosen to broadly match the context of previously acquired developmental neuroimaging data, given that to date the vast majority of research of this kind has been undertaken in Western, Educated, Industrialised, Rich and Democratic (WEIRD) populations (Henrich, Heine, & Norenzayan, 2010). fNIRS data are acquired at each visit from 1 to 24 months of age (6 time-points): the design includes measures of visual and auditory social selectivity, selective attention through measures of habituation and recovery of response to novelty, working memory development and functional connectivity. This broad range of assessments will enable typical pathways of brain development to be mapped across a range of cognitive domains within this rural Gambian population.
Secondarily, this project will investigate whether key environmental factors (i.e. undernutrition, poverty, parent-infant interaction) have an adverse or positive impact on brain development within this cohort across multiple cognitive functions or whether there are crucial periods of sensitivity for particular sub-domains of early processing and specialization.
Over the last 20 years research has shifted focus to try and understand whether poverty associated risk factors are associated with differences in specific neurocognitive systems such as memory, language or attention (Farah et al., 2006;Mezzacappa, 2004;Noble, Norman, & Farah, 2005) While largely restricted to the study of low SES within high-income countries, this behavioural research with school-aged children has led authors to propose that early development of these foundational skills may have cascading effects on later development. Motivated by evidence of behavioural deficits in these neurocognitive domains, several recent EEG studies have reported evidence of differences in neural mechanisms of attention in children (D'Angiulli, Herdman, Stapells, & Hertzman, 2008;Stevens, Lauinger, & Neville, 2009) and younger infants (Tomalski et al., 2013) living within low SES settings.
This study implements a Habituation and Novelty detection (HaND) paradigm at 1 to 24 months of age to allow the investigation of the development of discriminatory neural responses associated with-attention, learning and memory mechanisms-across infancy and early toddlerhood. Here, data are presented from two of the longitudinal time points (5 and 8 months) across both research sites (the UK and The Gambia). The HaND paradigm is designed to allow the measurement of habituation (or repetition suppression) responses, and recovery of response and subsequent novelty detection to a change in stimuli. Such effects have been observed in single cell recordings and adult neuroimaging studies (Desimone & Duncan, 1995;Grill-Spector, Henson, & Martin, 2006). In adult studies neural repetition suppression has been associated with cognitive processes of discrimination or learning, however, this is less well understood in infancy. Following on from earlier behavioural work (DeCasper & Fifer, 1980;DeCasper & Spence, 1986;Mehler et al., 1988), which evidenced newborns' ability to detect a change in speaker during passive listening to spoken sentences, several recent fNIRS studies have shown that repeated exposure to identical stimuli, either visual or auditory, produces neural habituation in very young infants (Benavides-Varela et al., 2011;Bouchon, Nazzi, & Gervain, 2015;Nakano, Watanabe, Homae, & Taga, 2009), and a recovery of response to novelty (e.g. a change in speaker) (Benavides-Varela et al., 2011;Nakano et al., 2009). For this study, a novel fNIRS paradigm was developed, designed to be objective and appropriate for use across different populations, to extend the applicability to older, awake infants and enable the investigation of developmental trajectories of habituation (repetition suppression) and novelty detection. All these previous fNIRS studies on habituation have been undertaken with newborns (one study at 3 months of age), while they are sleeping, under laboratory conditions. In this study infants were awake, and this paradigm was presented as part of a battery of other fNIRS tasks. Therefore the design of previous studies (Benavides-Varela et al., 2011;Nakano et al., 2009) was adapted with the intention of (a) increasing data retention, (b) utilizing a situation more similar to everyday learning situations , (c) creating a paradigm that would be more appropriate for application in a noisier and variable environment such as that often encountered in low-resource global health research contexts (Sabanathan et al., 2015) and (c) allowing the investigation of responses in participants across a broad age range from 0 to 24 months of age. The paradigm employed a naturalistically presented and engaging sentence, spoken by a culturally appropriate adult (i.e. UK or Gambian), within an environment containing naturalistic engaging (and potentially interfering) additional visual stimuli. Infants listened to a repeating spoken sentence during familiarization trials, which switched to a novel speaker during novelty trials, then returned to the familiar speaker at the end of the session. The rationale for using a change in voice was that infants in both cultures should be capable of discriminating pitch changes induced by a change in gender of the speaker; these are salient features of the infants' environment. This rapid repetition suppression (habituation) model was adopted rather than a paradigm with a delay between adaptor and target as it has been suggested that these are less demanding in terms of memory capacity and may be more sensitive to developmental changes within infant populations (Nordt, Hoehl, & Weigelt, 2016). Furthermore, the feasibility of using sentence-level adaptation neuroimaging paradigms has been previously shown in adults (Dehaene-Lambertz et al., 2006), and differential haemodynamic responses to a change in speaker from male/female has been evidenced from very early in life in infants born preterm at 28-32 weeks gestation (Mahmoudzadeh et al., 2013). The fNIRS arrays were designed so that responses could be recorded over auditory associative brain regions including the inferior frontal gyrus (IFG), middle and superior temporal regions and extending to the temporo-parietal junction across both cohorts . These regions have been previously linked with auditory repetition suppression and novelty detection in early infancy (Benavides-Varela et al., 2011;Bouchon et al., 2015;Mahmoudzadeh et al., 2013;Nakano et al., 2009) and adulthood (Belin & Zatorre, 2003;Dehaene-Lambertz et al., 2006), and subsequent recovery of response to novel speech (Benavides-Varela et al., 2011, 2017Bouchon et al., 2015;Dehaene-Lambertz et al., 2006).
In this study, infants were tested at 5 and 8 months of age at two sites (The UK and The Gambia). It was hypothesized that initial responses would manifest over auditory associative regions, diminish following repetition and recover as the novel stimulus was detected. Furthermore, it was hypothesized that increasingly rapid habituation (and therefore more mature and specialized neural processing) would be evident as the infants become older. An increasing number of studies are reporting adverse effects of poverty associated risk factors on the developing brain (D'Angiulli et al., 2008;Stevens et al., 2009;Tomalski et al., 2013;Xie et al., 2018). While current analyses cannot assess the direct relation to these risk factors as data collection is not complete for longitudinal growth and risk modelling, the diverse exposure to environmental challenges (i.e. poverty, undernutrition) present in the Gambian cohort allowed the investigation of these developmental brain responses within a novel rural low-resource population in which it was hypothesized neural discriminatory responses could be affected.

| Participants
Participants were recruited from two sites: Cohort 1 in the UK and  (Hennig et al., 2017).
Ethical approval was given by the joint Gambia Government-MRCG Ethics Committee, and informed consent was obtained in writing, or via thumbprint if individuals were unable to write, from all parents/ carers prior to participation.
To be included in the present fNIRS study, infants from both cohorts must have been born at term (37-42 weeks' gestation) and within the UK at normal birth weight (>2.5kg). In the West Kiang District in The Gambia, a combination of prenatal growth retardation, poor-quality and often contaminated foods and high levels of infection cause moderate to severe growth faltering in height and weight gain from around 3 months of age in the local population (Lunn et al., 1991;Lunn, 2000;van der Merwe et al., 2013). While infants with an indication of severe growth faltering (weight-for-height (WHZ) or head circumference (HCZ) z-score less than −3 according to WHO standards) were not excluded, growth measures were recorded at each time point (see Table 1). Overall, four infants had z-scores falling below −3 in WHZ or HCZ, of which only one had scores below −3 at more than one age point. Furthermore, while there are multiple ethnic groups in the country, mostly associated with their own language and cultural customs, the largest proportion of the population belongs to the Mandinka ethnic group (Jukes & Grigorenko, 2010).
To avoid confounds in the sample due to translation of stimuli and questionnaires into multiple languages, it was decided to only enrol families of the Mandinka group into this study.
This study presents Habituation-Novelty detection (HaND) data from the first two postnatal longitudinal time points (5 and 8 months) to be conducted while the infants were awake (as at 1 months they were testing while asleep). Following each session, infant's data could be excluded for the following reasons; due to (a) motion artefact in after the beginning of the study, and several of these infants were referred to the clinic as they were unwell. Therefore this exclusion criterion was higher than in the UK.

| Experimental procedures
Infants wore custom-built fNIRS headgear consisting of two ar- Psychtoolbox (Brainard, 1997;Kleiner et al., 2007;Pelli, 1997 In addition to the fNIRS study, anthropometric measures were performed. Infant lengths and weights were measured by using a Harpenden Infantometer length board (Holtain Ltd) and electronic baby scale (model 336; Seca), to a precision of 0.1 cm and 0.01 kg respectively. Mid upper arm circumference (MUAC) was measured by using a paper measuring tape to a precision of 0.1 cm. Head circumference, as a proxy for brain size, was measured to the nearest 0.1 cm with a stretch-proof measuring tape (model201; Seca) around the maximum circumference of the head (forehead to occiput).

| Data processing and analysis
Within each optical array, the NIR light measured by the detectors will have travelled from the sources through the skin & connective tissue, skull and underlying brain tissue layers. The NIRS system measured the light attenuation from each source detector pair.
These light attenuation measures were used to calculate changes in the concentration of oxy-haemoglobin (HbO 2 ) and deoxy-haemoglobin (HHb) in µMol which were used as haemodynamic indicators of cortical neural activity (Obrig & Villringer, 2003). The procedure of analysis, using in-house programmes developed in MATLAB®, followed a similar protocol to previous infant research (Cristia et al., 2013;Gervain et al., 2011;Lloyd-Fox et al., 2010). Data pre-processing included a channel pruning step with several pre-defined thresholds. The first threshold established a minimum DC value of attenuation reading of 3e-4. This threshold is based on previous experience using the NTS system, and will exclude channels where there is not enough light from the source reaching the corresponding detector (e.g. due to hair blocking either optode, or one of the optodes being unclipped from the array). The second threshold sets the maximum acceptable difference between the coefficients of variation in the attenuation readings for the two wavelengths per channel (set at 0.2), therefore this criteria will exclude channels where the noise characteristics per wavelength are significantly different. Finally, a power spectrum density analysis of the raw signal will discard channels with strong frequency components unrelated to the experiment. For each infant, the raw intensity data were inspected using automated quality control scripts that included the above criteria, and channels with poor signal readings, excess variability in the data measured with the coefficient of variation, large baseline drifts or interference from the Eye Tracker (detected with a frequency analysis of the data) were excluded for further analysis.
Infants with less than 60% of valid channels ( these can be found in Supplementary Information. To resolve statistical problems of multiple comparisons for these group analyses the false discovery rate (FDR) correction (Benjamini & Hochberg, 1995) was applied.
To use a more data driven, yet anatomically informed approach, these t-values were then entered into the main analyses using a Cluster based permutation approach (Maris & Oostenveld, 2007) to select the region of interest (i.e. the cluster of nearest-neighbour channels with the strongest collective response to Fam1) for each Cohort at each age point. This approach provided a pathway to guide ROI selection, where no prior studies were available, and to confirm the t test results. Furthermore, the cluster-based permutation analysis is a non-parametric statistical test that provides a solution to the multiple comparisons problem for data collected simultaneously from multiple collection points close to each other in space (Maris & Oostenveld, 2007). This method has been pre-

| UK cohort
For the infant cohort in the UK at 5-month of age, whilst responses are seen to follow a pattern of habituation and recovery to novelty at this age point (see Figure 4), overall the main effect of epoch was

| D ISCUSS I ON
This study successfully implemented a novel habituation and novelty detection fNIRS infancy paradigm into a longitudinal study in The Gambia and the UK.

| UK cohort
In accord with the hypotheses, in the UK cohort the analyses revealed three main findings. First, infants demonstrated a habituation (or repetition suppression) response to repeated auditory stimuli (spoken sentence), localized to middle and superior temporal regions of the cortex. Second, infants evidenced a novelty detection (or dishabituation) response to the presentation of the spoken sentence by a novel speaker in these same regions. Third, between 5 and 8 months habituation to repeated stimuli and recovery of response to novelty became increasingly evident, as the infants became older, indicative of developmental specialization.
In contrast to previous fNIRS research (Benavides-Varela et al., 2011;Bouchon et al., 2015;Nakano et al., 2009)-where habituation and novelty detection responses were found in temporal and inferior frontal regions in infants 0-3 months of age-the main cluster permutation analyses and subsequent repeated measure ANOVAs did not demonstrate significant responses until 8 months of age. The predicted pattern of responses at five months of age was weaker with differential responses-indicative of habituationevident at a borderline level during the ROI analyses and significant during the channel level preliminary analyses. The omnibus linear mixed model, however, did not find age-related differences in activation patterns suggesting that overall the differences between the 5-and 8-month-old responses were not substantial.
While the relative contribution to the current findings, of key differences between this study and previous research, cannot be fully teased apart, several factors can be considered in turn. First, and of import here, state of alertness (sleep vs. awake) may have significantly impacted the results. All previous research on habituation and novelty detection of brain responses using fNIRS has been conducted while infants sleep in contrast to this study which was undertaken while infants were awake. Interestingly, prior to their five and eight month visits infants also undertook the same study at 1 month of age while asleep. Preliminary analyses show that habituation responses were evident in this cohort at this earlier time point, similarly to previous research (Benavides-Varela et al., 2011;Bouchon et al., 2015). Therefore the presentation of stimuli during sleep may alter the speed at which infants' habituate to external stimuli (i.e. faster during sleep, no competing visual input interference). Second, it is likely that during the course of development the attention, learning and memory mechanisms associated with these responses will be impacted by the complexity and context of the study. While previous fNIRS studies have reported detection of novelty in the neonatal period (Benavides-Varela et al., 2011;Bouchon et al., 2015;Mahmoudzadeh et al., 2013) here the response was found to be weaker at the younger age point (5 months relative to 8 months). The increased complexity in this study-including stimulus complexity; a change in the identity of a speaker during the presentation of a full sentence (compared to a novel word, speaker or tone), and environmental complexity; a naturalistic multimodal context with interfering external stimuli-may have therefore altered the age at which novelty responses in this context are seen. Third, different effects may have been found with a different arrangement of fNIRS channels across additional cortical regions-see Limitations section-though note that the fNIRS arrays used in this study cover the majority of prefrontal and temporal regions found to be selective regions in two previous fNIRS studies on novelty detection and habituation (Benavides-Varela et al., 2011;Nakano et al., 2009).
Importantly, the overall study is designed to assess developmental trajectories across multiple longitudinal assessments during the first 2 years of life, rather than to establish whether infants can, or cannot, habituate/detect novelty to particular stimuli by a particular age per se. Therefore, as data continue to be collected across additional time points (12, 18 and 24 months), this will allow the investigation of developmental changes in repetition suppression and novelty detection across each cohort and the mapping of developmental trajectories within individual infants.  (Henrich et al., 2010). As this study continues in The Gambia, the increased sample sizes and age points in the Sub-Saharan cohort will enable causal pathways of developmental specialization to be mapped in the presence of the diverse range of environmental and early adversity factors to which these infants are exposed. The collection of this data from as early in life as possible should allow the investigation of how different factors (such as maternal or infant undernutrition, family poverty) compound early development. Furthermore, the project aims to further understand how other factors (i.e. caregiving practices, day to day interactions) may compensate for the effects of, or allow adaptation to, these early risk factors. Therefore as this study continues the relationship between the development of repetition suppression and recovery of response to novelty and, to take one example, the complexity of the home environment (i.e. frequency of adult/child speech, environmental sounds and caregivers that the infant is exposed to) will be explored through the concurrent measurement of home caregiving interviews and assessments using automated language environment analysis (i.e. LENA). Furthermore, the project will draw together neurocognitive correlates of brain function from across different methodologies (EEG, attentional measures in eye-tracking tasks, fNIRS) to further understand how trajectories of brain development observed within the habituation and novelty detection responses relate to other patterns of brain specialization across social, attention, memory, language and functional connectivity paradigms. This will allow us to further understand whether risk factors such as under nutrition differentially effect the development of emerging functional networks and regional hubs across the brain.

| LI M ITATI O N S
Localization of responses: Previous research has found, in addition to the temporal region identified in this study, that right lateralized prefrontal areas were recruited during novelty detection (Benavides-Varela, Hochmann, Macagno, Nespor, & Mehler, 2012;Nakano et al., 2009). Familiarization and novelty responses were not localized to prefrontal regions of the fNIRS arrays during analyses. However, in future work it could be important to extend coverage of the fNIRS arrays to explore novelty effects to a change in speaker further as the current arrays did not extend fully into dorsolateral areas of the frontal cortex.
Importantly, this analytical approach allowed a data-driven approach to isolate regions of interest. Therefore, in lieu of individual MRIs (which are not practically feasible in this context), this allowed the location of functional brain responses within the UK and The Gambia cohorts to be isolated to appropriate channel clusters separately. In future work, this could be independently run for different anatomical regions (i.e. temporal/frontal). Furthermore, as the relation between developmental trajectories of growth and brain function within The Gambian cohort are explored further the same approach can be used to identify ROIs at an individual level to accommodate differences in brain size and anatomical structure due to severe undernutrition.
Mapping of co-variables: A limitation of the analysis presented in this paper is that it is currently limited to two time points (5 and 8 months of age) from a longitudinal study following children to 2 years of age. A primary objective of the BRIGHT study in The Gambia will be to assess the impact of longitudinal growth against brain development within this cohort. At this stage, there is not sufficient data to model growth across the period of key interest (birth to 2 years of age).
UK-Gambia comparisons: The aim of this study was to assess developmental specialization of brain function within cohorts, rather than draw direct comparisons between the UK and The Gambia. The UK cohort was included in this study to provide additional data regarding fNIRS longitudinal measures of the developmental specialization of the brain-in a more comprehensively researched population in the UK-so as to broadly match the context of previous research which has largely been conducted in "WEIRD" populations. As such the UK data set has allowed the successful measurement of age-related changes in habituation and recovery to novelty brain responses in this new paradigm. In turn this will allow the investigation of the variability in responses within the Gambian cohort. Furthermore, through the continuation of data collection at 12, 18 and 24 months of age, individual trajectories of habituation and recovery of response to novelty will be stratified within The Gambian cohort across the first 2 years of life.

| CON CLUSORY REMARK S
Importantly, neuroimaging measures such as fNIRS allows the identification of markers of atypical/adaptive function from a far earlier age (i.e. from birth) than behavioural assessments are typically able to (i.e. from 1 to 2 years onwards). The presented findings form part of a newly emerging field of research, which will in the future hopefully yield valuable insights into global neurocognitive development.
Fundamentally by increasing understanding of longitudinal brain specialization in the UK, and developmental trajectories of brain function in The Gambia, this project will assess how infants' react to their environment in ways allowing for an adaptive developmental outcome, thus resulting in a cortical structure and functioning most fit for their experienced circumstances (Johnson et al., 2015).
The long-term aim of this research is to establish fNIRS as a universal assessment tool for the investigation of the impact of adversity on cognitive development, identify individuals at greatest risk and target interventions from an early age before critical developmental milestones have been affected.

ACK N OWLED G EM ENTS
We thank the parents and infants who took part in this study without whom this work would not have been possible. We also thank the broader team of staff at the MRC Keneba Field Station for supporting us in the collection of this data. This study was supported The full sample has also now been recruited into the study in The Gambia: N = 223.