COgnitive Dynamics in Early Childhood

1 INTRODUCTION AND RATIONALE
1.1 Background
This study aims to investigate the nature and implications of cognitive
variability, which refers to the fluctuations in performance that individuals
exhibit across cognitive tasks over time. While cognitive abilities such as
reasoning, memory, and vocabulary have been extensively studied, the focus has
predominantly been on mean performance, neglecting the rich and informative
dimension of variability. Traditionally, cognitive abilities have been regarded
as stable traits with lifelong implications (Deary, 2014). However, this
perspective has overshadowed cognitive variability across diverse contexts and
situations.

Literature has highlighted the substantial implications of cognitive
variability on real-life and educational outcomes (Gottfredson & Deary, 2004).
For instance, individuals exhibiting higher variability in cognitive
performance are more prone to being misallocated into inappropriate educational
settings, leading to long-term consequences (Woodrow, 1932). Moreover,
cognitive variability has been identified as a potential early warning marker
for adverse outcomes and neurodevelopmental disorders like attention deficit
hyperactivity disorder (ADHD) (Fagot et al., 2018; Kofler et al., 2013).

Despite speculations about the importance of cognitive variability
persisting for almost a century (Nesselroade, 1991), empirical research on this
topic has been limited due to logistical challenges and the lack of appropriate
quantitative techniques. However, recent studies have provided evidence that
individuals exhibit variability in cognitive performance across different
trials, hours, days, and even seasons (Kelly & Beltz, 2020; Licher et al.,
2019; Rabbitt et al., 2001; Sievertsen et al., 2016). Notably, this variability
is particularly pronounced during periods of rapid cognitive development, such
as childhood and old age (Galeano Weber et al., 2018; MacDonald et al., 2006;
Siegler, 1994).

Understanding cognitive variability is crucial for several reasons.
Firstly, individuals with greater variability are more likely to be
inaccurately stratified in schools or careers, resulting in lifelong
consequences and emphasizing the need for fair and accurate assessment methods
(Cattell, 1966). Secondly, variability in performance implies that individuals
may spend a significant proportion of time performing below thresholds of
adequate performance, which can have serious implications, particularly in
high-stakes professions such as piloting or surgery (MacDonald et al., 2006).
Additionally, cognitive variability has been linked to neurodevelopmental
disorders like ADHD, and interventions targeting variability have shown promise
in alleviating symptoms (Kofler et al., 2013). Lastly, reducing cognitive
variability holds potential for enhancing day-to-day functioning and reducing
challenges for vulnerable individuals (Li & Lindenberger, 1999; MacDonald et
al., 2006).

In this unique longitudinal design using gamified versions of classic
cognitive domains we will measure variability across a range of tasks at
multiple levels of temporal resolution: months, days, occasions and trials. 600
children (from which 200 will also be in the neuroimaging arm) will be measured
for a period of three years. Once per year they will take part in a burst: A
week where they will be measured three times a day; and up to two other
measurement occasions if classrooms or parents of individual participants
agree. We will use cutting edge methodology to understand the behavioural,
neural and environmental mechanisms of variability, as well as the longitudinal
consequences of variability on cognitive development and the emergence of
mental health symptomatology such as ADHD. By combining the strengths of deep
phenotyping with cutting edge quantitative modeling, we will be able to test
and develop theories of cognitive development, demonstrate the role of brain
structure and function in supporting cognitive dynamics and determine the
effect of cognitive variability on developmental outcomes.

Testing young children is important in the CODEC study because
cognitive variability is particularly pronounced during periods of rapid
cognitive development such as early childhood (Galeano Weber et al., 2018;
MacDonald et al., 2006; Siegler, 1994). By focusing on young children, the
study aims to capture and analyse the dynamic changes and fluctuations in
cognitive performance during this critical developmental stage. Additionally,
understanding cognitive variability in early childhood is essential to identify
potential risk factors and early warning markers for neurodevelopmental
disorders such as ADHD (Fagot et al., 2018; Kofler et al., 2013). By assessing
cognitive variability in young children, the study can contribute to early
detection and intervention strategies. Previous MRI scanning experiences at the
Donders Institute have shown that starting at the age of 8 years, children can
clearly understand what is happening, articulate any questions or concerns, and
yield high quality data. Hence, only children from the age of 8 years on will
be included in the imaging arm. In the behavioural arm, which only has minimal
risk, 7-year-old-children can be included to avoid exclusion of younger
children in the same class as the 8-year-olds. They will be allowed to
participate in the imaging arm from the moment they turn 8. Furthermore, as
previously mentioned, since children change school settings (primary to
secondary) at the age of 11-12 years, we aim to focus recruitment on children
aged 8-10 years at the first measurement to best balance the minimal burden and
scientific insight into a key developmental period, allow for maximum retention
across the three years while they are still attending primary school.

1.2 Measurement and Modelling
To understand the causes and consequences of variability, we must first
measure it, and measure it well (Flake & Fried, 2020). In previous studies, a
wide range of variability metrics have been used. However, they suffer from a
range of practical and theoretical limitations. For instance, simple summary
metrics such as iSD (individual standard deviations) or ICV (coefficient of
variation) can be biased due to neglect of autoregressive structure (de
Haan-Rietdijk et al., 2016), individual differences in mean performance or
trend-like changes over time. Moreover, simplistic measures such as iSD or iSD2
ignore measurement error inherent in variability (Wang & Grimm, 2012). The
challenges of modelling these sources of variation in a principled manner have
hitherto precluded widespread, accurate measurement of variability. To overcome
these challenges, we will use the new, flexible, integrative mathematical
framework of Dynamic Structural Equation Modeling (DSEM) (Asparouhov et al.,
2018; Hamaker et al., 2018; McNeish & Hamaker, 2020 - see 'methodology'). DSEM
allows us to simultaneously estimate all components of the time course.
Moreover, using a tailored multilevel SEM framework and Bayesian estimation,
each of these sources of variability can be estimated as a random effect (i.e.
as varying between individuals). Crucially, variability can then be modelled as
dependent and/or independent variable in a fully integrated manner.
The first measurement occasion will entail a *burst* measurement,
yielding a week of data on N=600 across all tasks. Fitting a DSEM to this wave
of data will yield a rich set of variability estimates. These will be used to
generate a completely novel descriptive measure to capture individual
differences: The Variability Performance Profile (VPP). This VPP is specific to
each individual and will capture a rich description of how variable a child is
across cognitive domains (reasoning, speed, memory, language, exploration) and
temporal scales (trials, occasions, days). Further measuremen

2 OBJECTIVES
2.1 Primary Objective:
To address the existing knowledge gaps and advance our understanding of
cognitive variability, the CODEC study aims to integrate experience sampling
methods, longitudinal designs, deep phenotyping cohorts, and state-of-the-art
statistical methodologies to investigate three core questions:
1. How does cognitive variability differ between individuals?
2. What are the neural, psychological, and environmental mechanisms that
underlie cognitive variability?
3. What are the long-term consequences and outcomes associated with differences
in cognitive variability?

2.2 Secondary Objective(s):
1. Measurement and Modelling:
• To develop a comprehensive and accurate measurement of variability using DSEM
• To explore variability across different domains and temporal resolutions
2. The mechanisms of Variability:
• To explore the role of strategy exploration and exploitation as drivers of
variability.
3. Long-term consequences of variability:
• To investigate variability as an early marker for atypical development,
particularly in neurodevelopmental disorder such as autism and ADHD

Study Design: longitudinal observational cohort study with children aged 7 to
10 at first measurement. We will use a flexible design, meaning that we will
aim to start measuring 8-year-old-children to follow them for the duration of 3
years (while they are still in primary school), although to allow for a big
enough sample size, we will also measure children eager to participate from
ages ranging between 7 and 10 at the first testing session (behavioural part)
and between 8 and 10 at the first testing session for the imaging part.* *
Duration: 3 years; Participants for the behavioural testing only, will
be examined for a *burst period* (1 week of testing for 2-3 times a day) once a
year for the duration of 3 years (3 burst periods in total) and up to two
additional brief (25 minutes per occasion) measurements a year depending on
classroom or parent agreements. Participants also taking part in the imaging
study will also be scanned once at the start of the study (year 1) and once at
the end of the study (year 3).

Setting:
-The majority of data collection will be in classrooms of collaborating schools.
-A subset of individuals may use our individual enrolment route, and thus
perform their tasks at home.
-200 children will take part in 2 imaging sessions, which will take place at
the Donders Institute for Brain, Cognition, and Behaviour at Radboud
University.

Justification of the Design:
The study aims to investigate the cognitive development of children over time
and the impact of cognitive variability on various tasks. A longitudinal cohort
study design is appropriate for this research as it allows for the examination
of changes in variability of cognitive abilities over time. Frequent sampling
is required to separate different temporal resolutions (e.g. trial to trial,
occasion to occasion, day to day), ensure appropriate power, and separate
developmental effects from retest effects.

The risk for participants taking part in this study is negligible. The full
sample will participate in a tablet based cognitive study similar to tasks
already implemented as part of widely used educational platforms. A subgroup of
the participants will also take part in the neuroimaging phenotyping arm. MRI
is a non-invasive technique. MRI has been widely used in*children of similar
(and younger) ages without apparent harmful consequences when inclusion
criteria are followed. As the aim of the study is to understand individual
differences in cognitive variability in early childhood, this study requires
the participation of children. The study will run for the duration of 3 years
and children will participate in behavioural testing several times a year (one
week of high intensity testing - *burst-week*, and up to 2 other measurement
occasions per year if classrooms or parents of individual participants agree).
The subset of 200 children taking part in the MRI study will have an MRI
session in the first year, and a follow up scan after an interval of
approximately 3 years. Since children change school settings (primary to
secondary) at the age of 11-12 years, we focus recruitment on
7-to-10-year-old-children to allow for a maximum retention across the three
years while they are still attending primary school. Based on previous positive
experiences with scanning children at the Donders Institute starting from
8-year-old-children, only children aged 8 or older will be allowed to take part
in imaging sessions.
The burden for the behavioural part will consist of repeated
participation (several times a year: 1 burst-week and up to two other
measurement occasions per year if classrooms and individual participants agree)
on short mood and sleep scales accompanied by cognitive measures, tested in a
playful way on our tablet-based platform (m-Path). For the duration of 1 week
(i.e., burst-week) each year (for 3 years), children will play 3 games for a
duration of 15 minutes (5 minutes per game), 2 or 3 times a day (depending on
the possibilities within the classrooms or at home). Further testing occasions
will consist of children playing all 5 games once for a total duration of 25
minutes (5 minutes per game).
The burden for the imaging arm will consist of 2 on-site visits at the
Donders Institute for parent(s) and child. For the children, the burden
associated with participation comprises two imaging sessions (once at the
beginning of the study and once approximately 3 years after). While
8-11-year-old children do not rate their participation in MRI research as scary
or annoying according to the DISCO-RC questionnaire (DISCOmfort in Research
with Children; Staphorst et al., 2017;
https://vragenlijst.kindenonderzoek.nl/resultaten/mri-scan/), some children
mention feeling tired or bored. During scan sequences which allow it, a silent
animation video will play to decrease boredom, and we will regularly interact
with participants in between scan sequences. The MRI session will take a total
of ~2 hours and will involve a screening and welcome (15 minutes), answering a
set of short questionnaires on mood and tiredness (15 minutes), a mock scan to
familiarise children (30 minutes), the actual MRI session (~55 minutes), and a
thank-you note/debrief with a small gift (5 minutes). In order to further
reduce potential discomfort in the scanner, children will receive hearing
protection suitable for their age and their head will be supported with foam
pillows. During the two task-based fMRI blocks (2*8.5 minutes) we will record
gaze direction and pupil dilation using the Eyelink 1000 Plus system. During
the scan session, children will have continuous contact with the experimenters
and can ask to leave the MRI scanner at any time without consequences.* *
The accompanying adult will be asked to complete a range of
questionnaires at each visit, which will take approximately 35 minutes and/or
to take part in the same 5 cognitive games as the child previously took part in
during the behavioural part (25 minutes). Participation of the parents on the
questionnaires and cognitive tasks will require consent on a different form to
the child*s consent. Parental participation is not required for the child to
take part in the imaging arm.