Predictors of executive function among 2 year olds from a Thai birth cohort

Abstract

Executive function (EF) is a critical skill for academic achievement. Research on the psychosocial and environmental predictors of EF, particularly among Southeast Asian, agricultural, and low income/rural populations, is limited. Our longitudinal study explored the influence of agricultural environmental, psychosocial, and temperamental factors on children’s emerging EF. Three-hundred and nine farm worker women were recruited during the first trimester of pregnancy. We evaluated the effects of prenatal insecticide exposure and psychosocial factors on “cool” (i.e., cognitive: A-not-B task, looking version) and “hot” EF (i.e., affective, response inhibition) measures of emerging EF. Maternal urine samples were collected monthly during pregnancy, composited, and analyzed for dialkylphosphate (DAP) metabolites of organophosphate insecticides. Psychosocial factors included socioeconomic status, maternal psychological factors, and quality of mother-child behavioral interactions. Backward stepwise regressions evaluated predictors of children’s EF at 12 (N=288), 18 (N= 277) and 24 (N=280) months of age. We observed different predictive models for cool EF, as measured by A-not-B task, vs. hot EF, as measured by response inhibition tasks. Report of housing quality as a surrogate for income was a significant predictor of emerging EF. However, these variables had opposite effects for cool vs. hot EF. More financial resources predicted better cool EF performance but poorer hot EF performance. Qualitative findings indicate that homes with fewer resources were in tribal areas where children must remain close to an adult for safety reasons. This finding suggests that challenging physical environments (e.g., an elevated bamboo home with no electricity or running water), may contribute to development of higher levels of response inhibition through parental socialization methods that emphasize compliance. Children who tended to show more arousal and excitability, and joy reactivity as young infants in the laboratory setting had better cognitive performance. In contrast, maternal emotional availability was a significant predictor of hot EF. As expected, increased maternal exposure to pesticides during pregnancy was associated with worse cognitive performance but was not associated with inhibitory control. Identifying risk factors contributing to the differential developmental pathways of cool and hot EF will inform prevention strategies to promote healthy development in this and other unstudied rural, low income Southeast Asian farming communities.

1. Introduction

Executive function (EF) refers to a set of cognitive skills that involve effective problem-solving, coordinated thoughts and actions, adaptation to change, and sustained goal-directed behaviors. These skills significantly contribute to academic achievement, school readiness, and prosocial behavior, while challenges with EF skills could be linked to issues in behavior, health, and finances (Diamond, 2013). Developmental trajectories and underlying mechanisms of EF have been widely studied for over two decades among global north countries, (Carlson, Zelazo, & Faja, 2013; Haft & Hoeft, 2017; Zelazo & Müller, 2002). Historically, researchers believed that EF was a unitary process that did not develop until the preschool years, perhaps due to the challenges of assessing EF in young children (Hendry, Jones, & Charman, 2016; Zelazo & Müller, 2002). Recent research suggests a hierarchical framework in which simpler neurodevelopmental abilities emerge earlier during infancy as the foundation of the more complex EF skills in older childhood and that continue developing through adolescence and adulthood, in parallel with the development of the prefrontal cortex (PFC) (Cuevas, Rajan, & Bryant, 2017; Hendry, Jones, & Charman, 2016). There has been a growing interest in studying EF during birth to 24 months, as the brain is highly responsive and adaptable to experiences and holds significant potential for impacting lifelong development (Black et al., 2017). Researchers are now investigating the emergence of EF skills, shedding light on factors influencing EF skills in infancy and their developmental trajectory. For example, some researchers have adopted a biopsychosocial perspective, considering both intrinsic (e.g., biological) and extrinsic (e.g., psychosocial) factors, that contribute to EF development (Cuevas, Rajan, & Bryant, 2017), while others propose an information processing approach (e.g., Zelazo & Müller, 2002).

While research on EF has been particularly prominent in Global North countries, efforts are underway to explore development of EF in diverse populations and cultural contexts. In addition, cultural variation (e.g., social norms), and environmental factors (e.g., housing, neurotoxicants) may alter brain structure and function relevant for EF development (Grantham-McGregor et al., 2007; Haft & Hoeft, 2017; Sternberg, 2014). Thus, exploring influential factors of EF in specific populations and cultures is crucial for understanding the underlying mechanisms and designing interventions that benefit each culture’s specific needs (Sternberg, 2014).

This study explored the intrinsic and extrinsic influences that play roles in the development of EF among infants who live in rural/hill-tribal agriculture families in the northern region of Thailand. The infants and families in this study were from a cohort study to investigate the impact of prenatal insecticide exposure on child neurodevelopment. Subsets of our participants also had some environmental vulnerabilities such as low income and parental education and challenging home environments. For example, while some participants lived in a small, rural town in concrete or wooden homes with electricity and running water, hill-tribal housing, isolated in the mountains of Thailand, consists of bamboo huts elevated approximately 1.5 meters from the ground, without electricity or running water (Moonpanane et al., 2022; Tongdhamachart & Tommanon, 2020).

In sum, conducting research on influences of early EF development in the understudied rural/hill-tribal agriculture populations has the potential to generate insights into the diversity of EF development and contribute more information to the design of targeted early intervention strategies.

1.1. Literature Review

1.1.1. EF Developmental Framework in Infancy

The common principal skills of EF, derived from studies in older children and adults, include working memory, inhibitory control, and cognitive flexibility, which are associated with activation of the PFC (Diamond, 2016). In brief, working memory (WM) allows children to maintain and manipulate goals, rules, and information, facilitate accurate responses, and organizing behaviors to achieve objectives. Inhibitory control (IC) helps children regulate impulsive behaviors or autonomic responses, which enables them to perform more appropriate behaviors. Inhibitory control may involve attentional control as well as motor inhibition (Colombo & Cheatham, 2006; Welsh et al., 2010). Lastly, cognitive flexibility (CF) helps enabling children to ignore distractions and focus on essential stimuli and shifting their attention, thoughts, or actions adaptively to accommodate new information or rules (Diamond, 2013; Raver & Blair, 2016). However, EF has been delineated through various conceptual frameworks. Studies in early childhood (3-5 years old) have indeed supported a unitary construct of EF (Wiebe, Espy, & Charak, 2008; Wiebe et al., 2011). However, when exploring infancy (1-2 years old), the prevailing approach has been the adoption of the two-core EF framework (Castillo & Lopez, 2022). This framework encompasses working memory (WM) and inhibitory control (IC) features, based on assessments focused on infants’ abilities to retain and manipulate information, as well as inhibit dominant responses to achieve goals (Cuevas, Rajan, & Bryant, 2017).

Furthermore, a two-core concept of EF, described as cool and hot EF (Zelazo & Müller, 2002), differentiates “cool” and “hof” EF based on brain injury studies and subsequent neuroimaging studies (e.g., Moriguchi, 2022). That is, cool EF is associated with the dorsolateral prefrontal cortex (DL-PFC), which supports abstract, decontextualized problem solving while hot EF is associated with orbitofrontal cortex (OFC) and is primarily associated with affective/motivational aspects of problem solving. Although there are no definitive tests that isolate cool or hot EF, some tests rely more heavily on cool or hot aspects of EF (Zelazo & Müller, 2002).

Moreover, there has been ongoing discussion about the multifaceted nature of IC (Castillo & Lopez, 2022; Diamond, 2016; Hendry et al., 2022). Diamond (2016) described two types of IC as “cook” and “hot” IC, based on the level of temptation present in the task. Cool IC or so called inhibition of attention (Diamond, 2016) or competitive inhibition (Hendry et al., 2022), refers to inhibitory control in situations where emotional arousal is minimal and requires the ability to inhibit a dominant or previously-established competitive response in the presence of competing alternatives (e.g., A-not-B task). In addition, it has been shown that cool IC is more relevant to WM than hot IC, even though WM and IC are inter-correlated and therefore not completely distinct skills (Diamond, 2016; Hendry et al., 2022). Hot IC, also called response inhibition, inhibition of action (Diamond, 2016), or directed global inhibition (Hendry et al., 2022), refers to inhibitory control in situations where the temptation or emotional arousal is elevated and requires the child to inhibit a response towards an attractive object and to wait until a particular signal is given (e.g., delay of a snack).

Despite findings regarding the unitary construct of EF in early childhood, we have chosen to adopt the cool and hot framework to elucidate the distinct types of motivation observed during infancy assessments. We defined “cool EF” as tasks in which infants must inhibit their attention towards the familiar location of a toy to find the toy at a new location (i.e., A-not-B), and “hot EF” tasks as those in which infants must inhibit their responses towards an affective object (i.e., Snack Delay, Crayon Delay, and Prohibited Toy Task). Due to time constraints of our study protocol, we could not conduct both cool and hot EF during the same session. We chose to measure A-not-B at 12 months to align with previous research employing tasks to assess cool EF (Castillo & Lopez, 2022) and at 18 months to capture development in late infancy. We measured response inhibition tasks as an indicator of hot EF at 24 months, considering existing literature and the ability to internalize and comply with requests (Castillo & Lopez, 2022; Kochanska, Murray, & Harlan, 2000).

1.1.2. Intrinsic factors

Our study focused on two intrinsic factors: early neurobehavioral tendencies and infant temperament.

1.1.2.1. Neurobehavior

Neurobehavioral tendencies refer to the individual differences in patterns of reflexive, neurophysiological behavior and responsiveness that emerge shortly after birth. These tendencies include reflexes, attention, arousal levels, reactivity to stimuli, and self-regulation abilities (Lester & Tronick, 2004). Studies have found that poorer neurobehavioral performance during the newborn period (e.g., suboptimal reflexes, high excitability, poor self-regulation) was longitudinally associated with poorer child developmental outcomes (e.g., cognitive, language, and motor) at 2 years (Spittle et al., 2017) and during preschool years (Meether et al., 2021). However, we found no study investigating the associations between infant neurobehavior and the distinct aspects of cool and hot EF during infancy (1-2 years).

1.1.2.2. Temperament

Temperament refers to individual differences in emotional and behavioral responses to specific situations (Rothbart, 2007). Rothbart and colleagues proposed three broad temperament constructs: (1) surgency/extraversion or a strong tendency to express positive emotionality, approach impulses, high-intensity pleasure, and high activity level; (2) negative affectivity or a tendency to express fear, sadness, distress to limitations, as well as difficulty in recovery from distress, and (3) effortful control or attentional and inhibitory control, low-intensity pleasure and high soothability (Putnam, Rothbart, & Gartstein, 2008). Rothbart and Derryberry (2013) described surgency/extraversion and negative affectivity as the reactive dimensions of temperament, and effortful control as a regulatory dimension. Noticeably, the effortful control dimension and some EF skills, particularly inhibitory control, overlap with each other (Cuevas, Rajan, & Bryant, 2017). Children with surgency/extraversion temperaments tend to show a greater tendency to seek out opportunities for exploring new and challenging tasks, including problem-solving activities and willingness to engage in novel experiences (Suor et al., 2019). However, it can also make children prone to impulsivity, and distractability. In fact, surgency has been negatively associated with academic performance (Liu et al., 2018). On the contrary, children with high negative affectivity may experience heightened sensitivity, expressed as fear, sadness, or anger, and may become easily overwhelmed and difficult to soothe when experiencing challenging or novel situations. Negative affectivity may, therefore, interfere with the ability to perform cognitive or cool EF tasks (Blair, 2002; Liu et al., 2018).

Studies of EF during infancy reveal inconsistent findings of the relations between temperament and EF, particularly among demographically diverse samples. For example, Frick et al. (2018) reported that parent-report of high surgency in 10-month-old children was correlated with cool EF performance at 18 months in a high-SES Swedish sample, but not with hot EF. However, this significant association was not found in a low-to-middle SES Latin American sample (Gago Galvagno et al., 2019), nor in a high-SES Chinese sample (Zhang & Wang, 2022).The reported associations between negative affectivity and early EF are also inconsistent. Some studies report no correlation (Frick et al., 2018; Gago Galvagno et al., 2019; Zhang & Wang, 2022), but others report negative associations between negative affectivity and hot EF (Gagne & Goldsmith, 2011; Leve et al., 2013). Furthermore, Gagne and Goldsmith (2011) found opposite correlations between child anger, .an indicator of negative affectivity, and inhibitory control when using different assessment methods for hot IC (i.e., laboratory observation vs. parent report). Altogether, the findings from different demographic samples suggest that research using more diverse groups of children across cultures will be important in understanding EF development globally. The inconsistent findings from different studies using different types of measurements also need further exploration. Therefore, we expanded our assessment of child temperament by using both parent report and laboratory observation.

1.1.3. Extrinsic factors

Several studies have suggested that children from disadvantaged psychosocial backgrounds are more likely to be at risk of poor EF. Research among the developing countries, such as India, Nepal, or Dominica has shown that children, particularly from low-income families and rural areas, are at risk of less mature cognitive developmental outcomes due to, for example, limited resources, stressful environments, and lower quality of parent-child relationships (Grantham-McGregor et al., 2007; Haft & Hoeft, 2017). Moreover, several epidemiological and public health studies found adverse effects of toxicant exposure on child EF. In addition, toxicant exposure during critical periods of brain development can disrupt the normal functioning of neural systems involved in EF (Davis et al., 2019; Liu & Lewis, 2014).

1.1.3.1. Housing quality and Parenting

Research has shown that experiencing poverty and negative parenting can induce stress hormones, particularly cortisol, in early childhood, and thus can adversely impact the developing frontal cortex and cognitive abilities (Blankenship et al., 2019; Doom & Gunnar, 2013). Additionally, Feola et al. (2020) reported that an adverse association between salivary cortisol levels and school-age childhood EF was mediated by different frontal cortical thickness. Gago Galvagno et al. (2019) suggested that poverty measurements should assess the distal context (e.g., physical characteristics of the child’s environment) rather than focusing on income exclusively. Several investigators have observed that lower levels of parental education and quality of housing (e.g., lack of an interior toilet and safe drinking water) also have an adverse impact on the development of EF (Blair, Raver, & Berry, 2014; Korucu et al., 2019). Unfortunately, the investigation of associations between home environment and emerging EF during infancy is very limited, especially in predominantly low-income groups (Ursache et al., 2013). Keller et al. (2004) compared hot EF performance between 18- to 20-month-old urban middle-SES and rural/agricultural low-SES children and found that rural/agricultural children performed better on hot EF tasks than the urban middle-SES children. They suggested that rural/agricultural, low-SES psychosocial and environmental conditions might facilitate earlier development of inhibition of action, despite more disadvantaged home environments (e.g., mud brick houses without water supply) (Lamm et al., 2018).

Whereas unfavorable home environments adversely influence EF, high quality of parenting positively influences EF. Several studies from a range of countries have shown that high quality of mother-child interactions is essential for the subsequent development of EF skills in children (Bernier, Carlson, & Whipple, 2010; Cheng et al., 2018; Lehman et al., 2002). For example, Chinese children whose mothers had higher parenting quality when their child was 1 year old, tended to develop better EF by 2 and 3 years of age (Cheng et al., 2018). The result is consistent with a US study which reported positive associations between quality of parenting during infancy and children’s later cool EF but not hot EF (Bernier, Carlson, & Whipple, 2010). Our study assessed quality of living (i.e., physical housing quality, ownership of equipment/appliances), and parenting to explore the impact of the home’s environmental and psychosocial factors on emerging hot and cool EF.

1.1.3.2. Exposure to toxicants

Prenatal exposure to neurotoxicants has been shown to impact child emotional and cognitive development. Neurotoxicants may include heavy metals (e.g., lead, mercury) (Shah-Kulkarni et al., 2020; Valeri et al., 2017), air pollutants (Gonzalez-Casanova et al., 2018), and insecticides (Eskenazi et al., 2007; Sagiv et al., 2021). Our study is a part of (Baumert et al., 2022b), a study designed to examine the impact of prenatal exposure to organophosphate (OP) insecticides on child neurodevelopment. Our previous report (Baumert et al., 2022a) demonstrated that pregnant women employed on farms had higher concentrations of urinary dialkylphosphate (DAP) metabolites of OP insecticides than US cohorts. Several previous studies reported adverse effects of prenatal exposure to OP insecticides on child neurodevelopmental outcomes, including EFs (Binter et al., 2020; Furlong et al., 2017; Sagiv et al., 2021). Moreover, Rauh et al. (2012) found associations between prenatal OP exposure and structural changes in older children (6-11 years old), such as inward deformations in the dorsal and mesial surfaces of the left superior frontal gyrus, a region pertinent for the development of executive function. In addition, prenatal OP insecticide exposure was associated with poorer performance on working memory tasks (Sagiv et al., 2021; Thistle et al., 2022), although results were somewhat inconsistent for hot EF tasks (Binter et al., 2020; Thistle et al., 2022). In sum, we included prenatal insecticide maternal biomarkers indicative of insecticide exposure (i.e., DAP metabolites) as an extrinsic predictor of emerging EF in the current study.

1.4. Present study

The aims of our study were to examine the influences of emerging EF during the first 2 years of life in a unique, understudied population of agricultural low-income rural families and hill tribe minority groups in Thailand. We explore the associations between intrinsic (e.g., newborn neurobehavior and temperament) and extrinsic factors (e.g., quality of housing, parenting, and insecticide exposure) on emerging EF based on a biopsychological perspective. Considering previous findings suggesting distinct aspects of cool and hot EF, we assessed separate multiple regression models to examine the significant influences on cool and hot EF. We hypothesized that suboptimal infant neurobehavior will be negatively associated with infant cool and hot EF. Either of the reactive temperament dimensions may be negatively associated with cool EF skills, with surgency perhaps showing more variable relations depending on task demands. Quality of parenting will be positively associated with cool and hot EF, but quality of the home environment may have a paradoxical effect on hot but not cool EF. Lastly, high prenatal exposure to OPs will be adversely associated with cool and hot EF.

2. METHODS

2.1. Participants

Participants were 323 mother-child dyads recruited during their first antenatal care appointment at hospitals in Chom Thong and Fang districts, Chiang Mai province, Thailand. Healthy women, ages 18- 40 with a singleton pregnancy, consuming < 2 alcoholic drinks per day, no illegal drug use, and employed as farm workers were included. Informed consent was obtained from all study participants prior to enrollment. This study was reviewed and approved by the Institutional Review Board at Emory University (Rutgers University reliance) and the Ethical Review Board at Chiang Mai University (Chulalongkorn University reliance) (Baumert et al., 2022b; Sittiwang et al., 2022). Infants > gestational age of 32 weeks and birth weight > 1500 grams with no birth anomalies continued in the study. Fourteen participants dropped out or were lost to follow-up. The final sample consisted of 309 mother-child dyads,

2.2. Measures

2.2.1. Maternal Interview, Questionnaires, and Medical Record Abstraction

Trained nurses recruited pregnant women and administered an intake questionnaire to determine study eligibility. After informed consent, trained research assistants administered a structured questionnaire to collect demographic, medical, and occupational information (Baumert et al., 2022b). The Test of Non-verbal Intelligence (TONI-IV) was completed by participants as a language-free assessment of intellectual ability (Ritter et al., 2011). Maternal depression was assessed using the Depression Anxiety Stress Scale (DASS) to indicate maternal psychological state (Brown et al., 1997; Oei et al., 2013). All questionnaires used in this study were translated to Thai, back translated, then reviewed by the study’s principal investigators (PS, NF) before use.

2.2.2. Child Executive Function Measures

2.2.2.1. Cool EF Tasks

2.2.2.1.1. A not B, looking version (Bell & Adams, 1999; Marcovitch et al., 2016)

In brief, we adapted the looking A- not- B task with 0 and 2s delay intervals to assess cool EF. Two opaque plastic cups were separately placed in front of each child. In the first phase (same- side phase) , an attractive animal toy was hidden under a cup in location A, (randomly assigned to the left (L) or right (R) side of the child). In the reversal phase, the toy was hidden in location B, opposite location A. After hiding the toy, the experimenter called the child’ s name to bring the child’ s attention to the midline and asked, “Where is the toy?” The experimenter waited 5s to observe the child’s first eye-movement response. If the child’s gaze moved toward and fixated briefly on the correct cup, experimenter revealed the toy, noted a correct response, and then continued to the next same- side trial. If the child gazed toward the incorrect cup or made no attempt to find it within 5s, the experimenter revealed the correct location and noted an incorrect response. Each child needed 2 consecutive correct responses in the same- side phases before moving on to the reversal phase ( e. g. , L-L- R pattern) . The test was discontinued if the child had 4 consecutive incorrect responses for the same- side phase. After the reversal trial, the second round of same- side and reversal phase was repeated with opposite pattern (e.g., R-R-L), and continued to the third round (e.g., L-L-R). Only children who had 2 corrects out of 3 reversal trials were then tested with 2s delay interval. We calculated percent correct trials from the response in reversal trials for both non-delay and 2s delay, then summed to create a composite score. Three experimenters reviewed video records for 5-6 cases/month to determine inter-rater reliability (κ = .90 - 1.00).

2.2.2.2. Hot EF Tasks

We assessed hot EF performance with Crayon-Delay, Snack-Delay, and Prohibited-Toys tasks in fixed order and video recorded each infant’s performance. Mothers were informed not to interact with the child during the tasks or to interact minimally, such as telling the child to stay seated or return to the task site.

2.2.2.2.1. Crayon-Delay Task (Joyce et al., 2016)

Each child was seated at a child-size table opposite the experimenter. To minimize mother child interaction, the mother answered a questionnaire while remaining out of the child’s sight. The experimenter put an opened box of crayons on the table, leaving a couple of crayons on the table and a paper within the child’s reach, while explaining, “I need to go outside to find some things for a new game. Please do not touch the paper or crayons until I come back. ” After this instruction, the experimenter began timing the experiment and left the room, leaving the child with crayons and paper for 60s. Upon return, the experimenter allowed the child to color with the crayons for an additional 5 min.

From videos, we coded the latency from prohibition to the time the child touched the crayon or the box. A higher score indicated the child delayed his/her response longer. Three experimenters scored latencies of response for 5–6 cases/month to assess inter-rater reliability (ICC = .95 – 1.00).

2.2.2.2.2. Snack Delay Task (Kochanska, Coy, & Murray, 2001)

This task consists of 4 trials with four delay times (10, 15, 20, and 30s). At the start of the 10s-trial, each child was seated at a child- size table opposite the experimenter. The experimenter instructed the child: “Keep your hands on the table and wait to get the snack after the bell rings.” The experimenter put a snack plate covered by a transparent cup on the table within the child’ s reach. The bell was placed on the table out of the child’ s reach and under the experimenter’ s control. Timing began when the experimenter released her hand from the snack plate. The experimenter waited until half of the scheduled delay time elapsed, then lifted the bell without ringing it. When the appropriate delay time for the trial was reached, the experimenter rang the bell and allowed the child to eat the snack. This established procedure was repeated until all 4 delay-time trials were finished. If the child grabbed the snack before the experimenter released her hand from the snack plate, the trial was noted a “ failure”. If the child failed 3 times for the 10s- trial, the task ended with no subsequent delay- time trials administered. Scores ranged from 0 to 9, with a higher score indicating longer delay (Spinrad et al. (2007). We estabished inter-rater reliability among three experimenters via video records for 5– 6 cases/month (κ = . 85 – 1.00).

2.2.2.2.3. Prohibited-Toy Task (Kochanska, Coy, & Murray, 2001)

In the mother- prohibited phase ( 10 minutes) , the experimenter instructed the mother: “There will be a shelf with three toys in the room. The child is not allowed to play with them during the test. You need to prohibit your child from touching the toys on the shelf and encourage him/her to play only with the wooden shapes provided for the first 10 min.” Next, the experimenter let mother and child enter the room while the experimenter remained outside. After 10 min, the experimenter entered the room and instructed the mother to say to the child: “Please play with the wooden shapes while I am answering this questionnaire. Do not touch or play with any of the toys on that shelf”. Then, the mother completed a questionnaire at a table with her back turned to the child. Before leaving the room, the experimenter said to each mother: “Please do not interact or repeat the commands. You can comfort your child with minimum interaction if s/he becomes fussy or cries.” The experimenter added “I will be back after 5 min” and left.

The child- internalization of prohibition phase started when the experimenter completely left the room. In this phase, after 1 minute passed, an unfamiliar female entered the room and played with toys on the shelf for 1 minute, then exited, leaving the child and mother in the room for 3 minutes. When the experimenter returned to the room after the allotted 5 minutes, the child was allowed to play with the prohibited toys for the next 5 minutes. Following Kochanska, Coy and Murray (2001)and Harden et al., 2015, we coded behaviors from video records during the child internalization phase with higher scores indicating greater response inhibition. Three experimenters coded video records for 5-6 cases/month to establish inter-rater reliability (κ = .84 – 1.00).

2.2.3. Measures of intrinsic factors

2. 2. 3. 1. NICU Network Neurobehavioral Scale (NNNS) (Lester, Tronick, & in collaboration with T. Berry Brazelton, 2004)

We used the NICU Network Neurobehavioral Scale (NNNS) to assess newborn neurobehavior. Two study nurses, certified by a Brown University certified trainer and blind to infant exposure, administered the NNNS while the mother observed the infant’ s assessment. To establish inter -rater reliability , nurses met monthly to perform independent assessments for approximately 20% of the total sample. Thirteen dimensions are summarized from NNNS raw scores. Based on latent profile analysis, three NNNS profiles were identified for infants: Profile1 (N=183): low need for handling, high self-regulation, and low arousal/excitability; Profile2 (N=75) : moderate need for handling, moderate levels of arousal, moderate self-regulation and excitability; Profile3 (N=62): higher need for handling, low attention and self- regulation, high arousal and excitability, and greater stress/abstinence (Sittiwang et al., 2022).

2.2.3.2. Infant Behavior Questionnaire – Revised – Short Form (IBQ-r-sf) (Putnam et al., 2014)

We also assessed child temperament by maternal report. At the 12-month appointment, mothers completed the IBQ-r-sf. Mothers reported frequency of child behaviors in several emotion eliciting situations during the previous 2 weeks on a 7-point Likert scale that are summarized in 14 scales.

We performed principal axis factoring with varimax rotation and derived 2 factors, positive affectivity (POS) and negative affectivity (NEG). The eight scales defining POS were activity level, approach, duration of orienting, high intensity pleasure, low intensity pleasure, perceptual sensitivity, smiling and laughter, and vocal reactivity (α = .87). The mean scores of each scale were averaged to create a POS score. The remaining five scales; fear, sadness, falling reactivity, soothability, and cuddliness defined NEG (α = .64). The mean scores of each scale were averaged to create a NEG score. Distress to limitations loaded equally with both factors (POS loading = 0.47; NEG loading = 0.46); therefore, it was interpreted separately from the other 2 factors.

2.2.3.3. Laboratory Temperament Assessment Battery (Lab-TAB) (Goldsmith & Rothbart, 1999)

We used Attractive Toy Placed Behind Barrier and the Puppet Game episodes of Lab-TAB to assess anger and joy reactivity, respectively. We video recorded Lab-Tab sessions with one camera placed to capture facial expressions and a second camera to capture the child’s body posture for scoring. Each child sat on mother’s lap during the test.

Attractive Toy Placed Behind Barrier:

This anger episode was divided into 3 trials. Each trial consisted of a 15s- play phase and a 30s- observation phase. The experimenter sat with the child at a table, approximately 1 meter to the child’ s left. The experimenter informed mothers to remain quiet, maintain a neutral facial expression, and keep maternal soothing at a minimum.

Each episode began when the experimenter demonstrated playing with a colored rattle by shaking it, then letting the child play freely with the rattle for 15s.Next, the experimenter placed a clear plexiglas barrier on the table in front of the child and within the child’s reach. The rattle was taken from the child and placed behind the plexiglas for 30s (observation). Finally, the rattle was returned to the child to relieve any negative emotion that occurred.

We modified Goldsmith and Rothbart (1999)’s coding to code the intensity of struggling approach (i.e., attempts to get toy from behind the barrier) and struggling withdrawal (i. e. , attempts to withdraw from the situation).. We used Izard’s AFFEX system (Goldsmith & Rothbart, 1999) to code facial anger.

We adapted the data reduction approach of Planalp et al. ( 2017) to create a composite score for child anger reactivity from the following mean and peak scores across 18 epochs: intensity of facial anger, intensity of distress vocalization, intensity of struggling approach, and intensity of struggling withdrawal (Cronbach’s alpha= 0.77). The higher score indicated more intense and frequent occurrences of anger reactivity.

2.2.3.3.2. Joy

Puppet Game:

The joy episode was divided into 5 trials. The first trial consisted of a rabbit and a giraffe puppet greeting each other and the child. The second to fourth trial consisted of brief conversation for the puppets when tickling the child, using a standard dialogue provided by Goldsmith and Rothbart (1999). In the fifth trial, both puppets were placed in front of the child and the child was allowed to play freely with them for 30s.

We used the coding system by Goldsmith and Rothbart ( 1999) for joy reactivity and data reduction by Planalp et al. ( 2017) to summarize mean and peak scores for the following: intensity of smiling, presence of laughter, presence of positive vocalizations, and presence of positive motor acts, across all epochs to create one composite score (Cronbach’ s alpha= 0.82) . Higher scores indicated more intense and frequent occurrences of joy reactivity.

2.2.4. Prenatal Environmental factors

2.2.4.1. OP Insecticide Exposure Assessment across Pregnancy

Urine samples were obtained up to 6 times during pregnancy at each antenatal care visit. Details about the composite scheme and DAP metabolite descriptive statistics in the samples can be found in Baumert et al. (2022a). Only ΣDAP (the total of 6 metabolites) was used in data analysis.

Composited urine samples were analyzed using a previously validated method which was also cross-validated by gas chromatography-mass spectrometry. Two blank samples, four QC samples and calibrants were analyzed concurrently with unknown samples in each analytical run. Relative standard deviations of the QC pools ranged from 4.5-12.6% (Prapamontol et al., 2014). The laboratory also participated in the proficiency testing program administered by the German External Quality Assessment Scheme. Although there is no universally accepted measure of urinary dilution, we evaluated creatinine for correction of urinary dilution in our models. This approach allows comparison and consistency with other epidemiological studies of pesticide effects on neurodevelopment (Eskenazi et al., 2007; O’Brien, Upson, & Buckley, 2017). Creatinine was measured by diluting urine samples 1000-fold with water after spiking with its isotopically labeled analogue. Diluted samples were analyzed by liquid chromatography tandem mass spectrometry coupled with electrospray ionization (Kwon et al., 2012). During the analysis of creatinine, a certified reference material (SRM) obtained from the National Institute of Standards and Technology (NIST) was included (NIST SRM 3667).

2.2.5. Postnatal Environmental factors

2.2.5.1. Housing quality at 12 months

Because it is difficult to determine income among farm workers in LMIC settings, we developed the Assets Questionnaire as an alternate measure and used factor analysis to derive 3 primary factors: (1) housing quality: items rated to reflect quality of building materials (e.g., scales range from earthen → bamboo → wood/concrete) and utilities (e.g., electricity (Yes/No), flooring (range from dirt → bamboo → tile/polished wood); (2) safety of home and neighborhood: maternal ratings of home and neighborhood spaces and safety, and (3) asset value: total value of appliances, farm equipment, vehicles, and land/home ownership (Howe et al., 2012).

2. 2. 5. 2. The Infant/Toddler Home Observation for Measurement of the Environment inventory ( IT-HOME)(Caldwell & Bradley, 2003)

We used the IT-HOME, administered by trained research assistants and conducted in each participant’s home, to investigate the overall quality of the child’ s home environment. The total IT-HOME score was calculated by summing 6 content domains including parental responsivity, acceptance of child, organization of the environment, play materials, parental involvement, and variety of stimulation. The higher total IT-HOME score indicated better quality of child home environment. (Cronbach’s alpha= 0.78).

2.2.5.3. Maternal Emotional Availability (EAS) (Biringen, 2008)

Two EAS-certified experimenters administered the EAS using the same set of toys across 12 and 24 months. We selected toys common to the local area, suitable for these ages, and likely to promote diverse play. The EAS was separated into 2 free-play sessions, 10 min each. During the first session, we brought the mother and child to the laboratory room to observe and video record the mother-child interaction in an unfamiliar place. The mother was told that she and her child would be left by themselves in the room to play for 10 minutes as they would at home. After 10 minutes, the experimenter returned to the mother and instructed her to ask the child to help return the toys to the basket. The second 10 minute free-play session began with the same instruction and was intended to evaluate how the mother guided the child after becoming familiar with the setting and the experimenter.

We coded videos from 2 free play sessions for maternal sensitivity, structuring, non-intrusiveness, and non-hostility with scores ranging from 7 (optimal) to 1 (non-optimal). We averaged across the maternal scores to create one composite score of maternal emotional availability (Cronbach’s alpha = 0.60) (Salo et al., 2020). We monitored inter-rater reliability via video records across the two certified EAS experimenters for 5-6 cases/month (ICC = 0.78 - 0.98).

2.3. Procedure

Maternal and child participants came to our clinic four times at the following child ages: 5 weeks, 12, 18, and 24 months (Table 1). Prior to each testing session, the experimenter performed a health check with the mother to ascertain that the child was healthy and in appropriate condition to participate (e.g., had enough sleep, not hungry). Participants were compensated 600 Thai baht for each testing visit. The IT-HOME and the assets questionnaire were administered during a home visit at 12 months by the trained research assistants.

Table 1.

Maternal and child measurements in each visit

Visit	Measurements
5-week	1. NNNS
12-month	1. A-not-B 2. IBQ-r-sf 3. Fab-TAB (i.e., Puppet Game and Attractive Toy Placed Behind Barrier) 4. EAS 5. IT-HOME 6. Assets Questionnaire
18-month	1. A-not-B
24-month	1. Hot EF tasks (i.e., Crayon Delay, Snack Delay, Prohibited Toys Task) 2. EAS

NNNS = NICU Network Neurobehavioral Scale, IBQ-r-sf = Infant Behavior Questionnaire revised short form, Lab-TAB = Laboratory Temperament Assessment Battery, EAS = Emotional Availability Scale, IT-HOME = The Infant/Toddler HOME inventory.

2.4. Statistical Analysis

Scores from the response inhibition or hot EF tasks were transformed into z-scores. The correlations among these hot EF tasks ranged from r = 0.32 to 0.37 (p < .001). We used principal component analysis with promax rotation to create a composite score. Factor-loadings ranged from 0.73 - 0.76 and were summarized into one factor, reflecting the average z-score for the 3 hot EF tasks,

Frequencies and percentages for categorical data, means, standard deviations, and histograms for continuous variables were examined for the distribution of variables. Bivariate and partial Spearman correlations measured the association between outcomes and predictors of interest. Partial correlations controlled a priori for child birth weight, paternal years of education, maternal nonverbal intelligence, child sex, and maternal depression measured at 12 months as these variables have been found to be significant factors in child development in past research. They were included as covariates rather than predictors of interest because they have the lowest potential for immediate intervention. Fisher’ s method was used to create p- values for all correlations. ANOVA were used to assess differences of means across levels of the categorical variable (NNNS profiles).

Because of the multicollinearity between predictors and the exploratory nature of the research question, backward stepwise regression with a retention alpha of 0. 10 was applied to assess which variables had more power in predicting the outcomes than others after controlling for the set of covariates used in partial correlations.

Although no single variable had more than 10% missing data, almost 50% of the data were missing values for at least one of the variables used in the analyses. Hence, multiple imputation, with 25 imputed data sets, was used to calculate estimates and p-values for partial correlations, and to generate results from regression analyses.

3. Results

3.1. Descriptive Statistics

The data availability for each measurement, mean, SD, and range of scores are presented in Table 2. The data for each measurement differs based on the participants available for testing ( e. g. , missed appointments, child cooperation). We observed no differences in marital status, maternal education, preterm birth, child’s sex, gestational age, smoking, nor alcohol use between participants who have data for all measurements and those who have missing data. Little’ s test failed to find any patterns of missingness based on the observed data ( p = .096). Further observations during data collection suggest that there is no reason indicating that missingness was correlated with potential values that would have been observed. Thus, we assume data is missing completely at random.

Table 2.

Descriptive statistics for Predictors and EF outcomes (N = 309)

Vanable (n)	n	M(SD)	Min - Max
Predictors
ΣDAP (nmol/L)	308	142.7 (219.9)	42.6 – 2852.6
ΣCreatinine (nmol/L)	256	278.8 (107.7)	42.7 – 630.8
NNNS Profile1	173
NNNS Profile2	71
NNNS Profile3	62
IBQ-r-sf Positivity	287	5.06 (0.75)	2.18 – 6.77
IBQ-r-sf Negativity	287	3.70 (0.68)	1.66 – 5.03
Joy Reactivity	271	5.04 (2.70)	0.00 – 11.20
Anger Reactivity	285	12.80 (4.89)	0.00 – 22.45
Housing Quality	301	−0.00 (0.59)	−1.62 – 2.45
Safety of Home	301	0.01 (0.69)	−2.49 – 0.99
Asset Value	301	−0.00 (0.71)	−1.08 – 3.14
Total HOME	301	32.40 (5.42)	13 – 44
EAS at 12m	248	4.78 (0.96)	2.75 – 7.00
EAS at 24m	267	4.77 (0.86)	2.75 – 7.00
Outcomes
A-not-B at 12m	284	31.22 (51.85)	0 – 200
A-not-B at 18m	266	74.44 (75.83)	0 – 200
HotEF at 24m	275	0.01 (0.79)	−1.56 – 1.31
Covariates
Male	156
Female	153
Child Birthweight (kg.)	307	3.01 (0.40)	1.73 – 4.22
Maternal Nonverbal Intelligence; TONI	303	82.00 (8.97)	60 – 109
Maternal Depression at 12m.	288	3.64 (4.74)	0 – 24
Paternal Education (years)	288	8.02 (4.28)	0 – 16

ΣDAP = summed dialkylphosphate metabolites, NNNS = NICU Network Neurobehavioral Scale, IBQ-r-sf = Infant Behavior Questionnaire revised short form, EAS = Emotional Availability Scale, TONI = The Test of Non-verbal Intelligence; 12m = 12 months, 18m = 18 months, 24m = 24 months, HotEF = composite score of crayon delay, snack delay, and prohibited toy task

Table 3 presents Spearman’ s Rho bivariate correlations among the study variables and covariates. The A-not-B at 12 months was not correlated with any predictors nor later EF performance at 18 and 24 months. Interestingly, we found 167 out of 284 infants (58%) with zero scores on A-not-B task. We conducted further investigation of 12-month object permanence performance (OPP) as a foundational skill for A- not- B task. We categorized the 12-month infants into two OPP groups: 1) the underdeveloped object permanence infants (39%), as indicated by their inability to consistently achieve two consecutive correct responses in same-side trials of A-not-B task, and 2) the developed object permanence infants (61%) who could manage correct responses in 2 consecutive same- side trials. We observed a correlation between the object permanence groups and A- not- B scores at 12 months (r = .64, p < .0001). Although our procedure to assess OPP was exploratory (i.e., same-side trials), it helps clarify that many infants with underdeveloped object permanence at 12 months were likely not capable of performing the A-not-B task. This supports our rationale for not including 12-month A-not-B as a cool EF variable in our regression model.

Table 3.

Bivariate Spearman‘s Rho Correlations for Predictors, EF outcomes, and Covariates.

	AnotB12 m	AnotB18 m	HotEF24 m	ΣDAP	ΣCreatinine	profile1	profile2	profile3	IBQ – r – sf POS	IBQ – r – sf NEG	JoyReactivity	AngerReactivity	HouseQuality	Safetyofhome	Assetvalue	TotalHOME	EAS 12 m	EAS 24 m	ChildSex	Birthweight	TONI	M. Depression
AnotB 12 m	1.00
AnotB 18 m	.04	1.00
HotEF 24 m	−.03	−.22 ***	1.00
ΣDAP	.02	−.17 **	.13 **	1.00
ΣCreatinine	.06	−.02	−.08	.22 **	1.00
NNNSprofile1	−.02	−.06	.07	.17 **	.02	1.00
NNNSprofile2	−.01	−.09	.03	−.02	.04	−.63 ***	1.00
NNNSprofile3	.03	.17 **	−.11	−.19 **	−.06	.58 ***	−.28 ***	1.00
IBQ-r-sf POS	.02	.07	−.19 **	−.17 **	.05	−.08	−.03	.13 *	1.00
IBQ-r-sf NEG	.08	−.04	.11	−.12 *	−.17 **	−.11	.07	.06	.11	1.00
Joy Reactivity	.04	.18 **	−.15 *	−.06	−.08	.03	−.05	.08	.16 **	.00	1.00
Anger Reactivity	.00	−.01	−.06	−.12 *	.09	−.05	.05	.01	.08	.06	−.07	1.00
House Quality	.01	.19 **	−.25 ***	.08	−.04	−.01	−.00	.01	.26 ***	−.18 **	.16 **	−.03	1.00
Safety of Home	−.01	.02	−.07	.08	−.01	−.04	−.01	.06	.18 **	−.08	.05	.04	.12 *	1.00
Asset Value	−.04	.14 *	−.23 ***	−.09	−.04	−.07	−.03	.12 *	.24 ***	.08	.10	.09	.22 **	.07	1.00
Total HOME	−.01	.19 **	−.29 ***	−.27 ***	.06	−.08	−.03	.14 *	.30 ***	.06	.19 **	.08	.21 ***	.12	.32 ***	1.00
EAS 12 m	−.04	.01	.09	−.07	−.10	.05	.00	−.06	.09	−.01	.09	.09	.05	.09	.05	−.02	1.00
EAS 24 m	−.02	−.07	.17 **	−.09	.07	.04	−.03	−.02	.03	.10	.13*	.15 *	−.03	.02	.08	.07	.17 *	1.00
Child Sex	.02	−.07	.01	−.06	−.04	.01	−.02	.01	.13 *	.02	.08	.12 *	.11 *	.00	.07	.05	.10	.07	1.00
Birthweight	.05	. 09	−.01	.05	.06	.10	−.08	−.04	.06	.00	.12 *	.06	.02	.05	.03	−.00		.01	.08	−.08	1.00
TONI	.05	.13 *	−.17 **	−.16 **	.14 *	−.07	−.01	.10	.31 ***	−.06	.17 **	.04	.30 ***	.11	.36 *	.31 ***	.08	.08	.15 *	−.00	1.00
Maternal Depression	.00	.03	−.08	−.19 **	.09	−.11	.03	.10	.14 *	.19 **	.00	−.01	−.13 *	−.12 *	.15 *	.12 *	.04	.05	−.01	−.03	.05	1.00
Paternal Education	.05	.19 **	−.29 ***	−.21 ***	.11	−.12 *	.03	.12 *	.30 ***	.02	.22 ***	.14 *	.17 **	.09	.36 ***	.36 ***	.08	.07	.10	.05	.42***	.14 *

^*p < .05,

^**p < .01,

^***p < .001;

12 m = 12 months, 18 m = 18 months, 24 m = 24 months, HotEF = composite score of crayon delay, snack delay, and prohibited toy task, ΣDAP = summed dialkylphosphate metabolites, NNNS = NICU Network Neurobehavioral Scale, IBQ-r-sf = Infant Behavior Questionnaire revised short form, POS = positivity, NEG = Negativity, House Quality = Housing Quality, EAS = Emotional Availability Scale, TONI = The Test of Non-verbal Intelligence, Child Sex (1 = boys, 2 = girls)

At 18 months, the increased numbers of infants with developed object permanence (82%) as well as a significant correlation between object permanence and A- not-B scores (r = . 55, p < . 0001) demonstrates that when the infants were older, the percent who demonstrated object permanence was markedly increased. Thus, many infants appeared to have a foundational skill and were capable of the A-not-B task at this age, as compared to their earlier performance at 12 months. Therefore, the 18-month A-not-B score was included as a cool EF variable in our model.

Finally, means (SEs) for A-not-B at 18 months and hot EF tasks at 24 months by NNNS profile are presented in Table 5.

Table 5.

Unadjusted and adjusted Means and SE of EF outcomes across NNNS profiles

	Unadjusted Means (SE)		Adjusted Means (SE)
	A-not-B 18m	HotEF 24m	A-not-B 18m	HotEF 24m
NNNS Profile1	68.72 (6.39)	0.069 (0.063)	70.54 (6.33)	0.032 (0.060)
NNNS Profile2	60.12 (10.11)	0.076 (0.102)	60.06 (9.90)	0.104 (0.100)
NNNS Profile3	96.65 (10.28)	−0.138 (0.106)	91.00 (10.25)	−0.074 (0.103)
ANOVA p-value	0.027	0.22	0.075	0.45

18m = 18 months, 24m = 24 months, HotEF = composite score of crayon delay, snack delay, and prohibited toy task, NNNS = NICU Network Neurobehavioral Scale

3.2. Predictors of A-not-B at 18 months

After control for a priori confounders (Table 4), prenatal OP insecticide metabolites were negatively associated with A-not-B performance. Joy reactivity at 12 months, housing quality, and the total IT-HOME environment predicted better A-not-B performance as did the child NNNS Profile3 (high arousal and excitability). We then performed backward stepwise regression analysis to examine which environmental or temperamental variables were most predictive of A-not-B while controlling for covariates. We observed significant associations with housing quality (B = 20.83, SE = 8.28, p = .012) and marginally significant associations with joy reactivity (B = 3.75, SE = 2.02, p = .065) as those variables were most predictive of better A-not-B performance.

Table 4.

Bivariate and partial correlations (controlling for ΣCreatinine, child sex, birthweight, TONI, maternal depression, and paternal education)

	Bivariate Correlations		Partial Correlations
Predictors	A-not-B 18m	HotEF 24m	A-not-B 18m	HotEF 24m
ΣDAP	−.17**	.13*	−.13*	.07
IBQ-r-sf POS	.07	−.19**	.01	−.10
IBQ-r-sf NEG	−.04	.11	−.05	.13*
Joy Reactivity	.18**	−.15*	.14+	−.10
Anger Reactivity	−.01	−.06	−.03	−.02
Housing Quality	.19**	−.25***	.16*	−.22***
Asset Value	.14*	−.23***	−.07	−.14*
Total IT-HOME	.19**	−.29***	.12*	−.20**
EAS 12m	.01	.09	−.00	.12
EAS 24m	−.06	.17**	−.08	.21***

⁺p < .06,

^*p < .05,

^**p < .01,

^***p < .001;

12m = 12 months, 18m = 18 months, 24m = 24 months, HotEF = composite score of crayon delay, snack delay, and prohibited toy task, ΣDAP = summed dialkylphosphate metabolites, IBQ-r-sf = Infant Behavior Questionnaire revised short from, POS = positivity, NEG = Negativity, EAS = Emotional Availability Scale, TONI = The Test of Non-verbal Intelligence

Additionally, we observed that infants with NNNS profile 3 tended to perform better on the A-not-B task compared to the other profiles, but the A-not-B mean differences between profiles were marginally significant after adjusting for covariates (Table 5).

3.3. Predictors of Hot EF at 24 months

In contrast to A-not-B (cool EF) performance, we observed that housing quality and the total IT-HOME environment were negatively correlated with response inhibition (hot EF) after controlling for a priori covariates. Maternal emotional availability at 24 months, however, was positively correlated with inhibitory control (see Table 4). Our backward stepwise regression analysis revealed mostly similar results to the partial correlation analysis. That is, after controlling for confounders, lower housing quality (B = −.21, SE = .08, p = .008), and total IT-HOME scores (B = −.03, SE = .01, p = .002) were significantly associated with better inhibitory control while increased, concurrent maternal emotional availability (B = .20, SE = .50, p < .0001) was a significant predictor of better inhibitory control. Additionally, we found higher mother-report of child negativity on the IBQ (B = .15, SE = .07, p = .035) was significantly associated with better inhibitory control. Lastly, we did not observe any significant differences in the hot EF among the NNNS profiles, as presented in Table 5.

4. Discussion

We observed that extrinsic environmental factors and infant intrinsic temperament predicted emerging EF skills among infants from rural hill tribe families in Thailand. However, differential associations were observed for cool and hot EF. Although housing quality and a more supportive home environment at 12 months were positive predictors of cool EF (i.e., A not B), we observed negative associations of these external factors with tests of hot EF (e.g., snack delay). On the other hand, maternal emotional availability observed in the laboratory at 24 months predicted better response inhibition but was not associated with cool EF. In addition, while maternal biomarkers of OP insecticides, as indicators of prenatal exposure, predicted lower cognitive performance, these biomarkers were not associated with response inhibition or hot EF. Intrinsic characteristics reflecting infant optimal neurobehavior and positive temperament were marginally associated with better cognitive performance. In contrast, higher infant negativity based on maternal report was associated with better hot EF, but newborn neurobehavior did not associate with hot EF. These predictive models suggest pathways for EF development that could lead to different interventions based on the differential predictors for emerging cool and hot EF.

4.1. Associations among housing quality, home environment, and cool vs. hot EF

We found that lower housing quality contributed to poorer cool EF. Our finding is consistent with a previous study in low-income Latin American families in which living in a more vulnerable environment (e.g., low parental education, poor housing materials, no water access) was adversely associated with EF performance (Gago Galvagno et al., 2019). In our study, partial correlations revealed that a better home environment, with particularly high parental responsivity and availability of learning materials, predicted better A-not-B performance. We also observed that children whose parents were more responsive and had more learning materials, showed more joy reactivity and engagement during the puppet game. We hypothesized that children living with more responsive parents, who can provide more learning materials, such as eye-hand coordination toys or muscle activity toys, were more familiar with engaging toys and tasks and thus achieved better performance on the cool EF task.

In contrast, poorer quality of housing and home environment predicted an improved ability to perform hot EF tasks (e.g., waiting longer for snack, and not playing with the attractive toys). Qualitative findings indicated that participants residing in more vulnerable structural housing environments, such as living in a bamboo hut elevated above the ground without appropriate railings, electricity, or water supply may foster parental control over infant actions to ensure their safety. Surprisingly, our study revealed that lower parental responsivity and fewer learning materials observed during our home visit were also associated with better hot EF performance even after controlling for covariates.

Lower parental responsivity indicated that parents were quieter, less responsive to researchers, and exhibited less parent-child interaction during the home visit. Our observations revealed a similar pattern of lower responsivity and fewer learning materials among rural and tribal Thai families reported in Williams et al. (2003). The quietness toward the researchers may be attributed to Thai cultural tendency towards deference to individuals of higher status (e.g., nurses, researchers) (Moonpanane et al., 2022; Williams et al., 2003). These parental deferential characteristics along with lack of stimulating environments potentially contribute to heightened infant inhibition of action in the laboratory. Moreover, we observed that these infants were habitually still and close to their mothers, as they were often wrapped around their mother’s chest with a cloth throughout infancy. These practices may contribute to the development of stronger infant response inhibition through limits on infant movement and compliance with the caregiver’s expectations. Previous research in Greece (Keller et al., 2004) and Germany (Lamm et al., 2018) revealed that children from agricultural, low-income tribal families where hierarchical socialization is valued (e.g., children are expected to comply with adults) performed better on response inhibition tasks than children from urban middle-SES families, where child-centered parenting is valued.

This finding demonstrates specific advantages of growing up in challenging home environments, where parents are required to be more considerate of child safety. This environment seems to be beneficial for the development of an infant’s response inhibition.

4.2. Associations between quality of parenting and cool vs. hot EF

Maternal emotional availability at 24 months was positively associated with better response inhibition at 24 months. That is, mothers who provided more sensitivity and guidance for the infant instead of hostile and intrusive approaches during free-play promoted better response inhibition in their infant. For example, at 24 months, mothers who were more emotionally available guided the infant to sit on a chair and played together with toys on the table. They also talked with their infant and responded to questions during the free-play without interrupting or responding negatively. These characteristics of parenting, observed in the laboratory setting, facilitate infants’ compliance with adult direction. This finding is consistent with a previous study in which maternal sensitivity and maternal non-intrusiveness, measured by parental emotional availability, were associated with parental reports of preschool-age inhibitory control. Higher parental sensitivity improved child inhibitory control, but higher parental intrusiveness was negatively associated with inhibitory control (Geeraerts et al., 2021).

We did not observe significant associations between maternal emotional availability at 12 months and cool EF at 12 or 18 months. Unfortunately, we did not evaluate maternal emotional availability at 18 months, and so we could not examine concurrent associations between maternal emotional availability and cool EF at 18 months. Alternatively, measurement of maternal emotional availability and A-not-B task at 18 months of age could have improved the strength of the association.

We also found no significant relationship between maternal emotional availability and other significant predictors. This suggests that better maternal emotional availability independently promoted infant hot EF and potentially is a strong protective factor to help infants develop the ability to regulate their behavior. However, we observed a negative association between hot EF and observer ratings of maternal responsivity during the home visit. This may be due to different observational settings eliciting different maternal behaviors. The IT-HOME involves ratings of naturalistic maternal behaviors toward the child during the home visit. Thai culture emphasizes politeness toward a visitor in the home which may subdue their typical behaviors. In contrast, maternal emotional availability in the laboratory setting instructed the mother to interact and play with her child in her usual manner. Although the mothers knew that they were being videotaped, there were no observers physically present in the playroom setting.

4.3. Associations between prenatal exposure to OP insecticides and cool vs. hot EF

We confirmed our hypothesis that maternal biomarkers of OP insecticides would adversely affect cool EF but contrary to our hypothesis was not associated with hot EF. Our findings are consistent with previous studies. For example, Thistle et al. (2022) reported a significant association between higher maternal urinary OP insecticide metabolites and poorer cool EF performance (i.e., non-verbal working memory), but no association with hot EF performance (i.e., snack delay task) among preschool children. In addition, birth cohort studies in the US documented an adverse impact of maternal biomarkers of OPs on subsequent infant cognitive and perceptual performance on the Bayley Scales of Infant Development. However, these infant studies emphasized tests consistent with cool rather than hot EF (Engel et al., 2011; Eskenazi et al., 2007). Using parent-report (i.e., Child Behavior Checklist), Eskenazi et al. (2007) reported significant associations between maternal biomarkers of OPs and 24-month infant pervasive developmental problems (e.g., unresponsive to affection, avoids eye contact). However, our study did not assess child emotional/behavioral problems at this age. Future studies could consider inclusion of regulatory control and emotional/behavioral measures to explore the effects of prenatal OPs on the emergence of EF and behavioral disorders.

4.4. Associations between neurobehavior at 5 weeks and cool vs. hot EF

We hypothesized that optimal neurobehavior at 5 weeks would be associated with better cool and hot EF. In contrast with the majority of studies, we found that infants who had greater excitability and arousal but lower attention and self-regulation (i.e., Profile 3) tended to perform better on A-not-B than infants in Profiles 1 or 2. Infants with the characteristics of Profile 3 are typically more difficult to soothe and their emotional reactivity can inhibit their ability to perform well on tasks that require sustained attention. Although Profile 3 infants had the highest levels of excitability and arousal relative to Profiles 1 and 2, their scores were within the normative range observed by Provenzi et al. (2018) in a sample of healthy, full term US infants. Moreover, Profile 3 attention and self-regulation scores were also within one standard deviation of the US normative values. In contrast, the most common profile for Thai infants in this study (N=182) was characterized by extremely low arousal and excitability (e.g., > 50% of subjects with excitability = 0) with attention and self-regulation within the normative US value. Therefore, the excitability and arousal exhibited by infants within Profile 3 may suggest greater environmental responsivity than seen among infants in Profiles 1 and 2 that contributed to their performance on A-not-B. However, mechanisms behind these associations among the participants’ unique environmental factors, neurobehavior, and EF development remains unclear. Further investigations exploring the typical and atypical profiles of neurobehavior with more diverse populations among Thai Infants will enhance our understanding of the study results.

4.5. Associations between temperament at 12 months and cool vs. hot EF

We observed that infants who were more engaged with the puppet by smiling, vocalizing, and approaching, tended to perform better in A-not-B task at 18 months. The correlation between joy reactivity at 12 months and later cool EF performance is consistent with a previous study reporting a positive correlation between surgency temperament (e.g., high pleasure or enjoyment) and better performance on A-not-B (Frick et al., 2018). Even though we did not find significant correlation between maternal-report of infant positive temperament and cool EF, infant positive and enthusiastic reaction to the laboratory task was positively correlated with maternal-report of a positive temperament. The inconsistent finding between laboratory observation and maternal-report may be due to different aspects of infant temperament captured by these measures. The laboratory observation only captured infant reactivity in one situation (i.e., engaging in puppets’ conversations and tickles). The maternal-report captures many situations in the infant’s daily life.

In contrast, our results revealed a significant relation between higher mother-report of infant negativity (e.g., inhibited approaches to novelty, difficult to soothe) and higher response inhibition performance. As infants with more negativity tend to be inhibited in response to novelty, the unfamiliarity of a stranger and situation may facilitate hot EF performance as the infant with a more fearful or anxious response may inhibit themselves because of fear of strangers or situations. If infants were inhibited by the novelty, they may exhibit wary behaviors consistent with response inhibition.

In sum, we found distinct pathways between infant temperament and EF development. Even though only trends, infants who exhibited higher joy reactivity when engaging in with puppets at 12 months tended to perform better on A-not-B. In contrast, maternal report of infant negativity at 12 months was significantly associated with better emerging hot EF at 24 months. These findings suggest that a more positive temperament tends to be a better predictor of cool EF, while negativity tends to be associated with inhibited behavior among the rural/hill-tribe infants.

This study only explored direct associations between infant temperament and the development of EF. Further investigations should examine more complex interactions between infant temperament, environmental factors (e.g., parenting), and the development of EF for designing targeted interventions. For example, Suor et al. (2019) found that associations between child temperament and EF performance were moderated by parenting quality. Children with high surgency benefited more from optimal parental guidance than children rated low in surgency. Additionally, children with low negative emotionality benefited more from effective parental control practices than children with high negative emotionality.

4.6. Relations among cool and hot EF

Our results revealed developmental change in A-not-B performance (cool EF) from 12 to 18 months. In contrast with previous findings found in western middle-class infants (Cuevas & Bell, 2014; Marcovitch et al., 2016), 58% of our 12-month-old participants could not perform the reversal trials of A-not-B, possibly due to a lack of object permanence. Furthermore, when we categorized 12-month-old participants according to OPP, we observed that 39% did not achieve object permanence, because they could not produce 2 consecutive correct responses for same-side trials. This may indicate a natural developmental pattern among the infants in this population where cool EF may not be as well-developed until 18 months. Alternatively, it may be related to disadvantaged environments that delay cognitive development. Interestingly, Clearfield and Niman (2012) found that approximately 60% of infants from low-SES environments showed perseverative errors at 12 months, while high-SES infants showed lower errors at approximately 20%.

Additionally, we did not observe a significant correlation between 12-month object permanence and 18-month A-not-B in our study. However, our study solely assessed infants’ performance in reversal trials to evaluate cool EF without incorporating a specific procedure to assess object permanence. We recommend that future studies should consider incorporating object permanence scores as outlined by Cuevas and Bell (2010) to provide a more comprehensive understanding of early cognitive development. Moreover, future studies should include more diverse populations in Thailand. Thus, we can understand normative EF development among Thai infants.

We found a negative correlation between infants’ performance on cool and hot EF tasks. Infants who performed better on the A-not-B task tended to show poorer performance on response inhibition tasks. The negative correlation between cool and hot EF tasks in this study is incongruous with previous studies which typically reported non-significant correlations between these tasks (Frick et al., 2018; Gago Galvagno et al., 2019; Hendry et al., 2022). These incongruent findings may be explained by several differences in study design and participant characteristics. Previous studies focused on urban middle-to-high SES families in European countries (Frick et al., 2018; Hendry et al., 2022), and mixed low and high SES families in Latin America (Gago Galvagno et al., 2019), in contrast with this study which focused on low-SES agricultural families in Thailand. These differences in SES, housing, cultural norms, and values potentially have a significant influence on individual infant development (Grantham-McGregor et al., 2007; Haft & Hoeft, 2017; Sternberg, 2014). Moreover, since cool and hot EF were not measured at the same age, this may explain the negative correlation. This finding may also suggest some interaction between cool and hot EF during infancy. At least one longitudinal study of cool and hot EF with younger infants also reported a negative, but nonsignificant correlation between A-not-B task at 10 months and response inhibition tasks at 16 months (Hendry et al., 2022).

Finally, environmental and temperamental factors appear to make distinctive and differential contributions to early cool and hot EF. As discussed, better quality of housing, greater newborn arousability and excitability, and more joy reactivity across infancy may facilitate infants’ approach behavior and engagement in a problem-solving task, resulting in better cool EF performance. In contrast, hot EF requires infants to inhibit their action to approach the attractive stimuli and therefore, better performance is associated with environmental factors typically regarded as adverse such as poor housing and a more fearful temperament that contribute to a tendency to inhibit responsivity in novel situations.

4.7. Suggestions and limitations

The strength of our study on emerging EF is the prospective cohort study design in which approximately three-hundred infants were evaluated from birth to 2 years of age. We used a multi-method approach that included direct observational measurements along with maternal questionnaires to explore the role of biological and psychosocial factors on the development of emerging EF. Our findings support the distinct developmental pathways of cool and hot EF during the first two years of life. We suggest further research to assess these domains separately, in order to examine the contributions of intrinsic and extrinsic factors to the emergence of EF in infancy.

Due to the lack of normative developmental data for EF among Thai infants, direct comparisons with the broader population are unfeasible. This absence hinders contextualizing our findings within the local developmental landscape. To address this, future research efforts should focus on establishing tailored normative data for the Thai population, enriching our understanding of executive function development in this context.

Our study revealed that the cool EF development of Thai infants was comparatively delayed compared to infants in western countries. This disparity may be attributed to variances in environmental and parental factors. Notably, we found that external factors such as improved home environments and optimal parenting practices showed stronger associations with the development of cool and hot EF, as opposed to intrinsic factors. Additionally, our findings suggest that infant participants with low arousal, excitability, and joy reactivity may be at higher risk for poorer cool EF development. To foster EF development, it is essential to prioritize enhancing living conditions, improving parenting quality, and focusing on infants exhibiting low reactivity. Further investigation is required to understand the longitudinal implications of cool and hot EF pathways on later childhood outcomes. Additionally, exploring the intricate interplay between extrinsic and intrinsic factors influencing EF development will aid in designing effective early intervention programs for children with diverse EF needs.

Lastly, it is important to note that our findings are from a unique agricultural population. Further study may include more diverse participants to investigate the similarities or differences between different cultures and the degree to which this study’s findings may be generalized to other populations.

5. Conclusion

In sum, the present study explored extrinsic and intrinsic predictors of emerging EF skills during infancy to toddlerhood in an understudied, low-income Thai farming community. We found different developmental predictors of cool and hot EF skills. Quality of housing, home environment characteristics, and parenting are stronger predictors than other environmental risk factors, such as prenatal OP insecticide exposure and low parental education. Empirical evidence on cool and hot EF in infancy will help future researchers better understand the foundations of EF and key contributors to its development. This should assist them in the design of intervention programs for general EF development as well as specific interventions for hot or cool domains of EF that benefit infants and children with different needs.

Highlights.

• Prenatal organophosphates exposure was associated with lower cool executive function.

• Better home environment was associated with better cool, but lower hot executive function.

• Maternal emotional availability was associated with hot executive function but was not associated with cool executive function.

• Surgency was associated with better cool EF, but negativity was associated with better inhibitory control.

Data Availability

Data will be made available on request.