Abstract

Introduction:

Emerging evidence suggests that neurocognitive processes, notably executive function (EF), should be considered important contributors to the observed variability in language outcome in both normally hearing (NH) and deaf children. Neural correlates of EF have potential clinical value, especially for infants and young children, in whom behavioural assessments can be unreliable.

Methods:

In the current study, we aimed to examine cortical correlates of EF in NH children aged 4–6 years old using functional near-infrared spectroscopy (fNIRS), a non-invasive, neuroimaging technique. In 23 NH children we measured activity in superior temporal and prefrontal cortex bilaterally whilst participants performed a behavioural EF task.

Results:

A significant association between the amplitude of EF-evoked cortical activation overlying the left dorsolateral prefrontal cortex and language performance was observed in NH children. To demonstrate that fNIRS is feasibly capable of measuring cortical responses specifically to behavioural EF in paediatric cochlear implant (CI) recipients, we also present the data for one deaf child aged 4.3 years with (CIs) who underwent the same simultaneous behavioural EF tasks and fNIRS imaging as the NH children.

Discussion:

The current results not only highlight the importance of higher-order top-down EF processes in language development, but they also demonstrate potential for fNIRS to provide cortical correlates of EF performance. Future applications of this technology could not only help explain variability in language outcome in both NH children and deaf children with CIs but also facilitate earlier intervention, such as EF rehabilitation strategies, where required at an earlier stage in a child’s life.

1 Introduction

Executive function describes the cognitive processes required to achieve planned and goal-directed actions (Diamond, 2013) and is generally accepted to consist of three main elements: (1) working memory (WM), temporarily holding and processing information in immediate memory to guide reasoning and decision-making; (2) inhibition control (IC), the ability to suppress specific thoughts or behaviour and; (3) cognitive flexibility (CF), the ability to switch between a mental or behavioural task in response to changing demands (Miyake et al., 2000). Despite not being described in Miyake et al. (2000) model, many researchers consider attention to be fundamental to EF development (Garon et al., 2008). In the typically developing child, EF begins to develop during infancy with rapid changes occurring during the preschool years that continue to develop through adolescence and into adulthood (Anderson, 2002; Espy et al., 2001; Zelazo et al., 2013). An association between EF and language processing is well established (Fuhs and Day, 2011; Kuhn et al., 2016), with the current literature commonly describing a reciprocal and bidirectional relationship between EF and language in both NH and hearing impaired children, particularly in early childhood (Kronenberger et al., 2020; Singer and Bashir, 1999). Executive function also regulates listening effort that, subsequently, increases listening proficiency (Kronenberger et al., 2020). This increased listening proficiency is proposed to allow for more effective language processing, an important requirement for deaf children with CIs who need to encode the degraded input from their hearing device (Pichora-Fuller et al., 2016).

A plethora of neuroimaging studies in children with typical brain development demonstrate a relationship between behavioural measures of EF and activity within the prefrontal cortex (PFC) (Aron et al., 2014; Konishi et al., 1998; Laird et al., 2005; Monchi et al., 2001; Wager and Smith, 2003). With respect to the auditory-deprived brain, evidence suggests that congenital deafness affects the development of neuronal connections between primary auditory (i.e., superior temporal cortex) and higher-order cortical areas such as the PFC (Sharma et al., 2009). Since both prefrontal and temporal brain regions are reportedly involved in WM processing (Baddeley, 2003), it is perhaps not that surprising that pre-lingually deaf children with CIs are two-to-five-times more likely than NH children to have deficient EF skills (Kronenberger et al., 2014a). Variable performance in EF has been observed in paediatric CI recipients (who are encouraged to use oral rather than visual/sign language) as early as during the preschool years (Beer et al., 2014), a period during which the development of EF and language skills are potentially related (Pichora-Fuller et al., 2016). This evidence contributes to the hypothesis that the observed variability in language outcome amongst paediatric CI recipients may be accounted for by differences in neurocognitive processing (Conway et al., 2011; Geers et al., 2011). Interestingly however, a study by Hall et al. (2018) showed that deaf native signers who had access to American Sign Language from birth showed no evidence of problems with performance-related EF compared with NH children. This suggests that deafness itself does not account for poor EF, rather a lack of early access to language whether that be in sign or speech, has a stronger impact. Hence, when analysing the relationship between EF and language development in deaf children, the mode of communication is a vital factor. For deaf children who have received a cochlear implant, individual communication experiences both pre- and post-implantation need to be considered.

EF in preschool children with NH has been shown to improve with behavioural interventions (Blair and Raver, 2014; Diamond and Lee, 2011), hence there may be value in integrating such treatments into existing speech and language therapy regimens for children with and without hearing loss (Kronenberger et al., 2020). Currently, assessments of EF abilities are reliant on behavioural measures. However, since very young children are often difficult to assess with behavioural techniques, there is often a delay of many years before it is possible to detect impaired EF development that delays any subsequent treatment.

Functional near-infrared spectroscopy (fNIRS) is an increasingly popular, non-invasive optical imaging technique that can be used safely and repeatedly for studying cortical function (Saliba et al., 2016). It has many advantages over other neuroimaging techniques, which include being: (1) acoustically silent; (2) portable, allowing for subjects to be imaged in both clinical and research environments; (3) relatively resistant to head movements so that infants and children can be scanned whilst awake and sitting on a parent’s knee; (4) compatible with hearing devices, including CIs. Indeed, in NH individuals and deaf CI users alike, fNIRS has reliably measured cortical responses in the paediatric population (Blasi et al., 2014; Lawrence et al., 2020; Mushtaq et al., 2020; Sevy et al., 2010). fNIRS images the haemodynamic response to neuronal activity in the brain via the use of near-infrared light (Boas et al., 2014). Low-power near-infrared light is directed through the scalp and into the cortex; the intensity of the light returning to the surface of the scalp is then detected. Changes in the concentration of oxygenated haemoglobin (HbO) and deoxygenated haemoglobin (HbR) can be measured, which are then subsequently interpreted as an indirect reflection of neuronal activity.

The overall aim of this work was to evaluate whether behavioural EF task-evoked cortical activation correlates with language performance in NH children aged 4–6 years. We use fNIRS imaging to principally target prefrontal and superior temporal brain regions in NH children to examine cortical activation during visual WM and IC behavioural EF tasks. We also include one deaf child with CIs to evaluate whether fNIRS is capable of successfully measuring cortical responses, specifically to behavioural EF, in paediatric CI recipients. It was our understanding at the time of conducting the current study, that although fNIRS had been utilised to examine behvaioural EF-evoked cortical activation in NH children (Smith et al., 2017; Mehnert et al., 2013), it had never previously been used to investigate cortical responses to behavioural EF in children with CIs. Where comparisons are made between NH children and one deaf child with CIs, this is on a qualitative and descriptive case study type basis only.

2 Materials and methods2.1 Participants and ethical approval

The design was approved by the Health and Social Care Research Ethics Committee A (HSC RECA) (REC reference: 19/NI/0121) and sponsored by Nottingham University Hospitals NHS Trust (Research & Innovation reference: 19ET008). Written informed consent was obtained from the accompanying parents or guardians of all participants. All NH participants were native English speakers, able to talk in simple sentences, had normal or corrected-to-normal vision and were aged between 4 years, 0 months and 6 years, 11 months. Any known cognitive, motor, or language disorder were defined as exclusion criteria. Children were recruited into the NH group if their parent/guardian declared no known problems with their hearing. The deaf child had been diagnosed with a bilateral profound hearing loss at 2 years of age and underwent bilateral cochlear implantation at 3 years of age. With respect to communication mode, the deaf participant was born into a NH family and was encouraged to develop spoken English language from birth with no regular exposure to visual/sign language. Although a formal cognitive assessment was not performed, there was no known cognitive or medical disorders apart from profound sensorineural deafness.

2.2 Study design

Twenty-four NH children were recruited to this experiment, with a mean age of 5.3 years (range 4.0–6.8 years, [SD] = 0.8 years). Table 1 details individual ages of the recruited participants. One deaf child with CIs was also recruited (male, aged 4.3 years). However, one NH child refused to wear the fNIRS equipment during the Go/NoGo behavioural experiment and refused to perform the N-back task. Therefore, cortical activation data acquired from 23 NH participants and one deaf child with CIs was subjected to analysis. Behavioural EF data was acquired for all 24 NH children and the deaf children with CIs, except for the one NH child who refused to perform the N-back task.

Participant IDAge in years14.425.434.245.354.864.976.384.096.0105.9114.6126.8134.9146.0156.0166.2175.2185.0195.6205.4216.0224.8234.4245.8

A table to demonstrate the individual age of the 24 NH participants.

Participant 1 was excluded from fNIRS analysis since they refused to perform the N-back task and wear the fNIRS equipment during the Go/NoGo task. N.B. Mean age 5.3 years (SD 0.8).

2.2.1 Equipment

Testing was conducted in a sound-attenuated room with the lighting dimmed. Participants were seated approximately 75 cm from a visual display unit. Brain activity was non-invasively measured using a Hitachi (Tokyo, Japan) ETG-4000 continuous-wave fNIRS system. The ETG-4000 measures simultaneously at wavelengths of 695 nm and 830 nm (sampling rate 10 Hz) and uses frequency modulation to minimise crosstalk between channels and wavelengths (Scholkmann et al., 2014). A dense sound-absorbing screen was placed between the fNIRS equipment and the listening position, resulting in a steady ambient noise level of 38 dB SPL (A-weighted). During the main fNIRS task, participants entered their responses using an “RTbox” button box (Li et al., 2010). The experiment was implemented in MATLAB (MathWorks, Natick, MA, USA) using the Psychtoolbox-3 extensions (Brainard, 1997; Pelli, 1997).

2.2.2 Behavioural stimuli

The Go/NoGo and N-back tasks were employed to examine behavioural EF-evoked fNIRS responses since these tasks have been commonly employed in neuroimaging studies examining IC and visual WM (Smith et al., 2017). The methodology and stimulation paradigms for the Go/NoGo (visual IC task) and N-back (visual WM) tasks that were utilised in the Smith et al. (2017) fNIRS study were implemented in current experiments.

2.2.2.1 Go/NoGo behavioural task

In this task, yellow and purple spaceships appeared sequentially on the visual display unit (VDU). Participants were instructed to press a button on the “RTbox” response box when they saw a yellow spaceship, and not when they saw a purple spaceship (Figure 1). A mandatory practice session with both yellow and purple spaceships was completed and repeated, if necessary, until the child attained a score of ≥70% correct. Subsequently, each participant was presented with four Go blocks (control condition) alternating with four NoGo blocks (inhibition condition) with 10 s of rest between each block. The Go blocks included only yellow spaceships, accounting for basic visual processing and motor response. During the rest block a fixation cross was presented on a uniform background. The NoGo blocks included both yellow and purple spaceships at a ratio of 10:4. For all blocks, 10 spaceships appeared for 500 milliseconds (ms) with an inter-stimulus interval of 1,000 ms, for a total of 15 s per block, with 10 s of rest between blocks.

The response inhibition control Go/No-Go task. Adapted with permission from Smith et al. (2017).

2.2.2.2 N-back behavioural task

The N-back task consisted of a WM condition (1-back) and a control condition (0-back). For the 1-back condition, various black shapes (square, circle, triangle, star) on a uniform background were presented to participants who were instructed to push a button if they saw the same shape twice in a row (Figure 2). Stimuli were presented for 1,500 ms each with an inter-stimulus interval of 500 ms. Practice blocks were completed until a score of ≥70% correct was obtained. The participant then completed four blocks of 16 s each with a target/non-target ratio of 3:5. During the 0-back blocks, participants were told to press a button whenever they saw a circle (an example circle was shown prior to onset of the block). Four 0-back blocks were completed, each block lasting 16 s with a target/non-target ratio of 3:5. A 10 s rest period occurred between all blocks.

The working memory N-back task. Adapted with permission from Smith et al. (2017).

As per Smith et al. (2017), behavioural analyses for the Go/NoGo and N-back tasks involved calculating task accuracy. Overall task accuracy was determined using d’prime, which was calculated using the following formula:

where z is the z score for the ith participant, omission error rate (OER) represents not pressing the response button when seeing a yellow spaceship and commission error rate (CER) reflects pressing when seeing a purple spaceship. d’prime is known to represent task accuracy while accounting for different response styles (Macmillan and Creelman, 2004).

2.2.3 fNIRS task procedure

The study design used was based on previous studies performed in our laboratory (Lawrence et al., 2018; Lawrence et al., 2020; Wijayasiri et al., 2017) by simultaneously performing fNIRS imaging along with the behavioural tasks.

2.2.4 fNIRS measurements and definition of regions of interest (ROIs)

fNIRS measurements were made with a total of 33 optodes arranged in a 3 × 11 array within an elasticated fabric head cap. The array comprised 17 emitter and 16 detector optodes with a fixed inter-optode distance of 30 mm, providing a penetration depth into the brain of approximately 15 mm (Gibson et al., 2005; Strangman et al., 2014). This resulted in a total of 52 measurement channels, primarily covering the prefrontal and superior temporal cortex (STC) bilaterally.

The international 10–20 system allowed correlation of defined external anatomical landmarks with underlying cortical areas (Jasper, 1958). The array was positioned on each participant’s forehead such that the emitter and detector optodes were centred horizontally at Fpz, a rationale commonly used in established fNIRS studies that have specifically examined EF-evoked cortical activation (Perlman et al., 2016; Smith et al., 2017). The bottom middle detector was carefully placed at location Fpz and the bottom row of optodes on each side aligned with the trajectory from Fpz and Oz (Figure 3). Although the array was constantly centred at Fpz, the inter-optode distance was fixed and hence the location of all other optodes was approximate and depended on the size and shape of each participant’s head. To evaluate variation in head size between individuals, the following measurements in centimetres were made on all participants: (1) head circumference, with the measurement taken along the trajectory of Fpz to Oz bilaterally; (2) the nasion to inion distance, and; (3) the distance between pre-auricular points.

Placement of the fNIRS measurement array. (a) Photographic illustration of optode array placement with the emitter optode locations in red and detector optode locations in blue. The bottom centre optode (indicated by yellow circle) is placed at Fpz, which is located at 10% of the nasion to inion distance away from the nasion (indicated by the yellow dot); (b) The yellow dashed line indicates the trajectory from Fpz to Oz, which was used to guide placement of the lowermost row of optodes.

Specifically, we aimed to primarily measure cortical activation in the PFC and STC bilaterally. Since increased activation overlying Brodmann areas (BAs) 10 and 46 (located within the PFC) has been observed in multiple fNIRS studies examining responses to WM and IC tasks (Perlman et al., 2016; Smith et al., 2017), these BAs were targeted for the EF regions of interest (ROIs) bilaterally. Although the primary auditory cortex is located medially within the lateral sulcus (Penhune et al., 1996), secondary auditory regions important for speech processing are located more laterally within the STC (Friederici, 2011). These secondary auditory areas, which include BA 22 and 42, were targeted as auditory ROIs bilaterally. The definition of both EF and auditory ROIs was enabled by the international 10–20 system combined with measurements of participants’ head size. As the inter-optode distance within the array remained fixed, extrapolated measurements of the distance of each optode from defined landmarks of the 10–20 system made it possible to determine over which BA each fNIRS recording channel was approximately located in participants at the group level. Figure 4 shows the typical coverage of the fNIRS optode array in addition to mean distance of head measurements for the 24 participants that were included in the fNIRS data analysis (NH children, n = 23; deaf child with CIs, n = 1). Interestingly, the SD for all these measurements did not exceed the fixed inter-optode distance of 3 cm. The EF ROIs in the PFC were subsequently defined as right BA 10 (Ch#26, 36, 47), left BA 10 (Ch#27, 38, 48), right BA 46 (Ch#25, 35, 46) and left BA 46 (Ch#28, 39, 49). The ROIs in the STC were defined as the right and (Ch#32, 33, 43) left (Ch#41, 42, 52) auditory ROIs.

Typical coverage by the fNIRS optode array for all participants with respect to the international 10–20 system and corresponding Brodmann areas. The area of the brain typically covered by the optode array is indicated by the bold dashed line in relation to the positions of the 10–20 system in schematic (a) and the corresponding Brodmann areas in schematic (b). The optode array typically spanned all cortical areas within the bold dashed line, hence good coverage of the prefrontal and superior temporal areas was achieved. The mean distance of specific measurements for 24 participants (NH children, n = 23; deaf child with CIs, n = 1) along with the standard deviations are shown in table (c). The mean distances in this table, in combination with the landmarking system of the international 10–20 system, were used to devise the bold dashed line in both schematics (a,b).

Once the position of the optode array was completed, a photograph was taken of the final placement for reference purposes. During testing, participants were instructed to remain still and keep head movements to a minimum to reduce motion artefacts in the recorded data.

2.2.5 Analysis of fNIRS data

Analysis of the fNIRS data was performed in MATLAB (MathWorks, Natick, MA, USA) using functions provided in the HOMER2 package (Huppert et al., 2009) together with custom scripts. The analysis was performed in a similar manner to previous studies conducted in our laboratory (Anderson et al., 2017; Lawrence et al., 2018; Lawrence et al., 2020; Mushtaq et al., 2019, 2020; Wiggins et al., 2016; Wiggins and Hartley, 2015; Wijayasiri et al., 2017). The following steps were performed:

Exclusion of channels with poor signal quality: we used the scalp coupling index (SCI) to identify and exclude channels suffering from poor optode—scalp contact (Pollonini et al., 2014). We excluded channels with SCI < 0.3 chosen to exclude only the worst 5% of channels in our dataset. Experience from previous studies in our laboratory showed that excluding only the worst 5% of channels strikes an appropriate balance between the desire to include only high-quality channels in the analysis versus the risk of committing type II (false negative) statistical errors if the effective sample size is too heavily reduced due to extensive channel exclusions.

Conversion to optical density: the measured light intensity levels were converted to optical density using the HOMER2 hmrIntensity2OD function, a standard step in fNIRS data analysis (Huppert et al., 2009).

Motion-artefact correction: after the raw fNIRS intensity signals had been converted, a wavelet filtering technique was conducted using the HOMER2 hmrMotionCorrelationWavelet function, a technique described by Molavi and Dumont (2012). This enabled correction of motion artefact by eliminating outlying wavelength coefficients which are assumed to be artefacts by implementing a probability threshold. We excluded wavelet coefficients lying further than 0.719 times the interquartile range below the first or above the third quartiles. By avoiding the need to reject contaminated trials from the data and instead processing the signals to suppress artefacts, wavelet filtering can enhance data yield (Molavi and Dumont, 2012).

Bandpass filtering: the optical density signals were band-pass filtered between 0.02 and 0.5 Hz to attenuate low-frequency drift and cardiac oscillations.

Conversion to estimated changes in haemoglobin concentrations: optical density signals were then converted to estimated changes in the concentration of HbO and HbR through application of the modified Beer–Lambert Law (Huppert et al., 2009). A default value of 6 was used for the differential path-length factor at both wavelengths. Note that the continuous-wave fNIRS system used in the present study allows for the estimation only of relative changes in haemoglobin concentrations across conditions and not absolute concentrations.

Isolation of the functional haemodynamic response: we applied the haemodynamic modality separation (HMS) algorithm described by Yamada et al. (2012) to isolate the functional component of the haemodynamic signal and suppress systemic physiological interference (Yamada et al., 2012). This algorithm attempts to separate functional and systemic signals based on the assumption that the correlation between HbO and HbR will be different in each case. Although this approach does not accurately account for all statistical properties of the noise typically found in fNIRS data (Huppert, 2016), in previous studies we have found application of this algorithm to be beneficial to the detection of auditory cortical activation (Lawrence et al., 2018; Lawrence et al., 2020; Wiggins et al., 2016; Wijayasiri et al., 2017); in particular, application of the HMS algorithm was shown to substantially improve the test–retest reliability of auditory fNIRS measurements (Wiggins et al., 2016).

Quantification of response amplitude: to quantify the level of behavioural EF task-evoked cortical response, the pre-processed fNIRS signal was subjected to a general linear model (GLM) approach previously described in Wijayasiri et al. (2017) and Lawrence et al. (2018, 2020). The GLM was applied to the continuous data collected over the duration of the imaging session. The design matrix included a set of three regressors (corresponding to the canonical haemodynamic response plus its first two temporal derivatives) for each experimental condition within each behavioural task, plus a further set for the rest trials within each task. Each trial in each behavioural task was modelled as a short epoch corresponding to the actual duration stimulation for that trial. Within each condition, the canonical and temporal-derivative regressors were serially orthogonalized with respect to one another (Calhoun et al., 2004). Model estimation was performed using the two-stage ordinary least squares procedure described by Plichta et al. (2007), which incorporates a correction for serial correlation (Cochrane and Orcutt, 1949). The ‘derivative-boost’ technique (Calhoun et al., 2004) was used to estimate response amplitude: this technique calculates an amplitude value that is a function of both the canonical (nonderivative) and the derivative terms of the model; the resulting amplitude estimates are less affected by any systematic differences in latency or dispersion between conditions, compared to if the amplitude is estimated from the canonical term alone.

The beta weights of the canonical haemodynamic response function (HRF) were extracted at each measurement channel, for each stimulation condition within each behavioural task, and for all participants. The employed haemodynamic signal separation method (Yamada et al., 2012) assumes a fixed linear relationship between HbO and HbR in the functional response. Therefore, the results of all statistical analyses are identical regardless of whether conducted on the beta weights extracted for the HbO or HbR parameter. Only results corresponding to the beta estimates of the HbO parameter of the functional component are presented here. These beta weights were used to quantify the amplitude of behavioural EF task-evoked cortical activation and were subjected to further statistical analyses as specified in the relevant sections of this manuscript.

2.2.6 Assessment of language performance

When originally designing the fNIRS experiment, a face-to-face assessment of language performance by a National Health Service (NHS) Speech and Language Therapist using the Preschool Language Scale—Fifth Edition (PLS-5UK®) was planned for all participants. Unfortunately, the COVID-19 pandemic commenced mid-way through the participant recruitment schedule. The unprecedented restrictions and demands of the pandemic on NHS clinical services prevented the NHS Speech and Language Therapists using the PLS-5UK® assessment for participants recruited after the onset of COVID-19. Due to the inability to continue with the PLS-5UK® assessment, the CCC-2® was administered to parents of all participants recruited to the fNIRS experiment after the pandemic commenced.

With respect to the PLS-5UK® assessment, this is an individually administered test that measures children’s receptive and expressive language skills (Zimmerman et al., 2014). It has been developed for use in children from birth to 7 years and 11 months and consists of two standardised scales which are auditory comprehension (AC) and expressive communication (EC). A total language score (TLS), a combination of the AC and EC score, may also be calculated. PLS-5UK® provides norm-referenced information due to being standardised on a UK population. It is therefore able to evaluate how a particular child is functioning in comparison to NH peers with both substantial reliability and validity (Zimmerman et al., 2014). The PLS-5UK® assessment has also been used by UK auditory implant programmes for ongoing manual monitoring of language development in CI recipients for many years after device insertion as part of their routine NHS care pathway.

The CCC-2® is a checklist that can be completed by a parent of a child aged 4 years and over who speaks in simple sentences (Bishop, 2003). The General Communication Composite (GCC) score calculated by the CCC-2® tool may be used to identify children who are likely to have clinically significant communication problems. Standardised GCC scores are based on normative data from a sample of children from the United Kingdom (UK). The CCC-2® has been demonstrated as a reliable and valid tool through data acquired from NH clinical samples (Bishop, 2013) and has also been shown to identify communication difficulties when administered to children using CIs (Ramirez-Inscoe and Moore, 2011).

2.3 Statistical analyses2.3.1 Behavioural data during fNIRS task

Bivariate linear regression analysis was employed to assess for any correlation between behavioural performance (d’prime task accuracy) on the Go/NoGo and N-back EF tasks and language ability in NH children (either PLS-5UK® TLS or CCC-2® GCC score). Twelve out of the 24 NH children performed a PLS-5UK® assessment prior to the onset of COVID-19, whereas the other 12 NH children were assessed with the CCC-2® tool. However, one child that was assessed with the PLS-5UK® tool refused to perform the N-back task. The mean age for children in the PLS-5UK® and CCC-2® groups was 5.2 (SD 0.89) and 5.0 (SD 0.71) years, respectively with no statistically significant difference in age between these two groups.

2.3.2 fNIRS data

Our statistical analysis aimed to establish the relationship between language performance and the amplitude of behavioural EF task-evoked cortical activation in NH children measured with fNIRS on a channel-wise basis. One NH child refused to wear the fNIRS equipment whilst performing the Go/NoGo task and refused to perform the N-back task. Therefore, cortical activation data acquired from 23 NH children and one deaf child with CIs during the Go/NoGo and N-back tasks was subjected to analysis. With respect to the statistical analyses that examined the relationship between cortical correlates of EF and language performance, the 23 NH children were allocated to subgroups according to whether the PLS-5UK® (n = 11) or CCC-2® (n = 12) was performed. Bivariate correlation analysis was performed using the value for the difference in the beta weight between the EF and control condition, i.e., (NoGo vs. Go and 1 back vs. 0-back) and the PLS-5UK®/CCC-2® score.

To examine cortical responses during the behavioural Go/NoGo and N-back tasks, beta weights representing the amplitude of cortical activation for each condition within each task were extracted for each measurement channel across the optode array. For each condition within a task (i.e., the Go and NoGo conditions within the Go/NoGo task and the 0-back and 1-back conditions within the N-back task), single sample, one-tailed t-tests were conducted on the beta weights to detect measurement channels that displayed significant levels of cortical activation compared to rest. The same analysis was also performed to detect which measurement channels detected significantly increased cortical activation from the control to the EF condition (i.e., from the Go to the NoGo condition and from the 0-back to the 1-back condition). However, by testing for significant activation in all 52 individual measurement channels, multiple comparisons are performed and the risk of Type I error is increased. These multiple comparisons were accounted for by applying the false discovery rate (FDR method) across channels (Benjamini and Hochberg 1995). The original formulation of the FDR procedure, which assumes independence or slight positive dependency across tests, was used in line with recommendations for fNIRS data analysis (Singh and Dan, 2006). Statistical significance was assessed against an FDR-corrected threshold of q < 0.05.

For each condition within the Go/NoGo and N-back tasks, the time course of cortical activation within the predefined ROIs was plotted. This enabled visualisation of the behavioural EF-evoked haemodynamic responses to determine whether plausible, artefact-free, HRFs had been obtained in each task condition. For each measurement channel, the time course of HbO and HbR concentration changes were block-averaged using the HOMER2 hmrBlockAvg function (Huppert et al., 2009). For each task condition these were then averaged over the relevant measurement channels to produce separate time courses for each ROI.

Following the use of a GLM approach, the resultant beta weights representing the amplitude of cortical activation were averaged across the ROI measurement channels for each participant. To investigate cortical responses to the Go/NoGo and N-back tasks in NH children, beta weights for each condition were subjected to statistical testing using a Linear Mixed Model [LMM] (West et al., 2007). Analyses were conducted using IBM SPSS Statistics for Windows Version 26.0 software (IBM Corporation, Armonk, NY, USA). Unfortunately, the sample size for deaf children with CIs prevented statistical analysis with LMMs for this group. However, to investigate EF-evoked cortical activation in NH children during both the Go/NoGo and N-back tasks, beta weights were analysed separately for each ROI. Two LMMs were performed for each ROI. The first LMM included two fixed factors of ‘condition’ and ‘age group’ to estimate the fixed effect of task condition and age on behavioural EF-evoked cortical activation within each ROI. In addition, a ‘condition – age group’ interaction term was specified to understand whether an effect of condition on cortical activation differed between younger and older children. Age was represented as a continuous variable, but a median split of 4.9 years was used to allocate participants to either a ‘younger’ or ‘older’ age group so that the effect of age could be evaluated as a categorical variable within the model. To evaluate the effect of task performance (d’prime) on behavioural EF-evoked cortical activation, the covariate ‘performance (d’prime)’ was entered into the second LMM as a fixed effect. To account for variance related to age, ‘age’ was also specified as a fixed covariate in this LMM. In all LMMs, a random intercept for ‘participant’ was included with ‘scaled identity’ selected as the covariance type to account for the correlation between beta weights within a participant. A restricted maximum likelihood estimation method was adopted for all LMMs.

3 Results3.1 fNIRS experiment3.1.1 The relationship between behavioural EF task performance and language ability

Performance accuracy (d’prime) on the Go/NoGo task was not associated with the PLS-5UK® TLS (r = −0.24, p = 0.45; n = 12) or CCC-2® GCC score (Tb = 0.22, p = 0.34; n = 12) in NH children (both 2-tailed). Similar findings were observed for the N-back task with no association between d’prime and PLS-5UK® TLS (r = 0.08, p = 0.83; n = 11) or d’prime and CCC-2® GCC score (Tb = 0.11, p = 0.63; n = 12) in NH children (both 2-tailed).

3.1.2 Channel-wise analysis of relationshi