You are on page 1of 11

Acta Psychologica 209 (2020) 103089

Contents lists available at ScienceDirect

Acta Psychologica
journal homepage: www.elsevier.com/locate/actpsy

Cognitive capacity in self-directed learning: Evidence of middle school T


students’ executive attention to resist distraction
Õnne Uus⁎, Paul Christian Seitlinger, Timo Tobias Ley
School of Educational Sciences, Tallinn University, Narva road 25, 10120 Tallinn, Estonia

ARTICLE INFO ABSTRACT

Keywords: Self-directed learning (SDL) is a rapidly developing trend in schools, although its prerequisites, such as children’s
Self-directed learning skills and abilities to plan and monitor their own learning, have not been investigated in detail. Due to additional
Working memory capacity cognitive load it induces, SDL has been in some cases found to be detrimental for learning, especially for students
Executive functions with a lower cognitive capacity. With this study, we explored some of the causes for the variability in learning
Long term memory
gains. We examined 111 middle school students' self-directed category learning using an exploratory web-task
Cognitive load theory
Concept formation
for autonomous learning, focusing on their information search (browsing a taxonomy of unknown dinosaurs)
Complex learning and their memorization of respective category labels. We were interested to detect whether students’ perfor-
mance in a complex span task (Ospan) was also reflected in their search and learning behavior. Results revealed
different learning gain trajectories in the latter task, where higher WMC students were more confident about
their learning. Also, the students with lower WMC were found to search the taxonomy by repeatedly searching
the same (basic type of) dinosaur exemplar. In line with prior findings about human mental capacity restrictions
and cognitive load theory, the present work evidenced the important role of students’ resistance to distraction,
and its relation to differences in self-directed search and memorizing. The results imply the need to teach me-
tacognitive skills and offer supportive scaffolding in order to avoid cognitive overload in SDL.

1. Introduction Buschkuehl, Jaeggi, & Farkas, 2018; Van Deur & Murray-Harvey,
2005). All those mental acts, however, add a heavy load on young
A critical determinant of modern pedagogy is self-directed learning learners’ cognitive capacity, thus hampering effective memorising of
(SDL) based on the active inquiry of the learner, where students new information (Garner, 2009; Nęcka & Lulewicz, 2016; Sweller, Van
themselves control and make choices concerning their learning, au- Merrienboer, & Paas, 1998; Sweller, van Merriënboer, & Paas, 2019;
tonomously solve problems and form conceptual conclusions (de Bruin Zepeda, Richey, Ronevich, & Nokes-Malach, 2015). Cognitive load
& van Merriënboer, 2017; Thomas, Strage, & Curley, 1988). However, theory (CLT) offers a convincing account of how learning tasks induce
to be successful in SDL one needs specific skills: to set learning goals, processing load that directly affects the learner’s ability to construct
plan what to learn more about, select strategies and also monitor new knowledge in LTM (de Bruin & van Merriënboer, 2017). Particu-
whether the goals are met (Van Deur & Murray-Harvey, 2005). As SDL larly important for high quality learning, especially when it involves
outcome, like any new knowledge acquisition, depend on human in- sequential subtasks, seems to be the interplay of working memory
formation processing constraints (Gyselinck, Jamet, & Dubois, 2008) (WM) supported by executive attention to resist distractions
inextricably linked to individual memory functions: encoding, reten- (Schwaighofer, Bühner, & Fischer, 2017. p. 60). WM is conceptualized
tion, retrieval, that expect a quick interplay of the learner's temporary as an “integral bottleneck” for the memorization system, as it is used for
mental workspace and long-term memory (LTM) storage (Cowan, 2014; registering, manipulating and shortly saving information that is then
Gilhooly & Logie, 2004). As a result, complex SDL requires a much stored in LTM (Baddeley, 1996; Cowan, 2014). Executive attention is
greater number of cognitive operations (compared to traditional responsible for maintaining the present information despite inter-
learning), such as to plan and search for learning items as well as to ference and can be regarded as a higher order control process in making
monitor and evaluate one’s own current learning stage (Rutherford, decisions or for problem-solving (Coolidge & Wynn, 2005). Therefore,

Abbreviations: SDL, self-directed learning; WM, working memory; WMC, working memory capacity; EF, executive function; LTM, long-term memory; CLT, cognitive
load theory

Corresponding author.
E-mail addresses: onne.uus@tlu.ee (Õ. Uus), paul.seitlinger@tlu.ee (P.C. Seitlinger), tobias.ley@tlu.ee (T.T. Ley).

https://doi.org/10.1016/j.actpsy.2020.103089
Received 4 January 2020; Received in revised form 5 May 2020; Accepted 11 May 2020
Available online 03 July 2020
0001-6918/ © 2020 Elsevier B.V. All rights reserved.
Õ. Uus, et al. Acta Psychologica 209 (2020) 103089

any attentional insufficiency can cause errors that can lead to one's (Schwaighofer et al., 2017). Whereas SDL requires from the learner
biased judgment on own current learning state and thereby decision even more efficient focusing (i.e., executive attention), concept-crea-
about what to do next and how to do it (Roebers, 2017; Sprenger et al., tion and assessment of their recently gained knowledge to make pro-
2011). However, especially the latter lays the ground in SDL, and gress (Sweller et al., 1998). However, to our knowledge there is no
substantial differences have been found in children's attention control, systematic research about the two coupled and mental capacity de-
especially in executive aspect, since there exist marked differences in pendent processes to form new concepts and search for new informa-
developmental maturing (Best & Miller, 2010; de Bruin & van tion. In this regard, since cognitive load and abilities to allocate one’s
Merriënboer, 2017; Nęcka & Lulewicz, 2016). This may result in poor own cognitive resources across all the required acts are related – how to
metacognitive judgment on whether one has successfully learned and help learners not to cognitively overburden is a continuing challenge.
render SDL as ineffective (de Bruin & van Merriënboer, 2017; Thomas Therefore, we need firstly to address SDL activities to optimize the
et al., 1988). This restriction manifests especially in unfamiliar topics, learner's intellectual load in order to gain the ultimate phenomena of
where several unknown units interrupt processing in each momentary the learning (Sweller et al., 1998. p 251). Based on the aspects men-
processing step (Kirschner, Sweller, & Clark, 2006; Roebers, 2017). This tioned above, the current study examines complex learning in parti-
is because the human cognitive architecture is severely limited, al- cular conditions: along the entire learning-session led by the learner (to
lowing to include in WM only small amount of elements concurrently gather specific data related to individual cognitive variables) to con-
(Cowan, 2014) to deal with visual, verbal or auditory material and save centrate on the specific attention control perspective of the learner
it in LTM (Baddeley, 1996; Sweller et al., 1998). In other words, the (Sweller et al., 2019) that can either aid or restrict school students'
limited WM capacity restricts an “online” processing to quickly form learning progress in SDL. Our study design differs from prior work
new links when too many new items are entering concurrently (Sweller exploring complex learning mainly by means of self-reporting (Garner,
et al., 1998). At the same time, the multicomponent tasks accom- 2009; Rutherford et al., 2018). In this paper, we therefore follow a
panying SDL might cause cognitive overload that hampers executive different strategy, namely, we have created an environment where
attention control (Best & Miller, 2010; Ecker, Lewandowsky, Oberauer, some of the processes can be directly traced on a more fine-grained
& Chee, 2010; Sprenger et al., 2011), thereby failing to bring relevant level and related directly to cognitive capacity: students select the items
items into WM to link them with prior knowledge and then saving in themselves, keep their attention on relevant info-pieces and integrate
LTM (Sweller et al., 1998). Therefore, executive control and WM–LTM them into LTM concepts. The current task is one example of how
interaction are the most important factors affecting one's mental op- complex learning can be used to study some of those fine-grained
erations in memorizing new information (Nęcka & Lulewicz, 2016; processes which are typical in the majority of complex learning tasks
Sweller et al., 1998; Sweller et al., 2019; Unsworth, Fukuda, Awh, & led by the student.
Vogel, 2014) that, however, is crucial for complex SDL. Schwaighofer
et al. (2017) have evidenced that the percentage of correctly recalled 1.1. Present work
items decreases, as cognitive load increases (see also Sprenger et al.,
2011). Yet, better initial executive control enables more efficient recall We addressed specific feedback on how the learners construct and
later (Shipstead, Harrison, & Engle, 2015). This is because executive memorize (i.e., form their own new concepts) while learning an un-
attention enables to mentally highlight only those items needed for familiar topic. To this end, we developed a novel search-and-learning
efficient solution of the present processing task, so leaving additional task (Dino), where to investigate learners’ step-by-step-measured pat-
time for practicing to be learned items. Hence, students with poor at- tern to allocate their attentional capacity across animal category names
tention control are slower in concurrent processing which is one of the to be memorized within a given 3D-taxonomy. We designed such a
primary reasons for the strong relationship between higher-order cog- research scenario based on the aspect that, as generally in the tradi-
nition and complex learning (Unsworth, Redick, Heitz, Broadway, & tional learning paradigm (e.g., in science disciplines) it is typical that a
Engle, 2009). However, as the mental capacity of school-age children is novice learner gains already developed domain-specific classification
not yet fully matured (Nęcka & Lulewicz, 2016), they get even more systems to-be-learned. However, in SDL, the learners have to be able to
severely confined when dealing with novel information (de Bruin & van form appropriate concepts by themselves, so sequentially constructing
Merriënboer, 2017). Though, to develop full understanding, it is im- their new knowledge. For instance, while acquiring a zoological tax-
perative to direct one’s attention only to task-relevant items, to shift onomy, it means getting to know its phylogenetic dimensions (e.g.,
attention from one item to another and establish links between them locomotion) as well as every dimension’s feature values (e.g., “not able
and prior knowledge (Schwaighofer et al., 2017). Accordingly, as at- to fly”/“able to fly”). Also, to identify the typical combination of spe-
tentional and shifting skills are based on executive control, they lay the cific features (in terms of which combinations belong to a particular set)
ground for one's progress in complex learning (Rutherford et al., 2018). and then learn the respective category name. Doing so, learners form
Still, children's abilities to perform in complex tasks improve until (at their own mental representations via the exemplars (see Fig. 1) based
least) the age 15 (Best & Miller, 2010). on what the respective category members look like. Therefore, one must
The problem is that although the SDL paradigm needs proactive first create (in WM) their own needed links (into LTM) and only then is
students, who are able to flexibly select adaptive strategies (Garner, able to mentally operate with them as associated representations.
2009) to monitor and control their own learning progress (Sweller However, the accompanying cognitive load can seriously influence such
et al., 2019), this specific competency is rarely taught (de Bruin & van initial processing, especially in case of lower mental capacity people
Merriënboer, 2017). However, for basic functions (planning, mon- (e.g., children whose cognitive capability is not yet mature, nor do they
itoring, evaluating), children first need to learn and practice their own always have enough prior linking associations compared to adults).
metacognitive strategies, since without such skills they face overload Therefore, we set our research goal to find out the following: Does the
that leads to insufficient or superficial knowledge (de Bruin & van ecological character of individual cognitive capacity (i.e., the ability to
Merriënboer, 2017; Garner, 2009; Sweller et al., 2019; Zepeda et al., maintain information in the face of distraction) have an impact on
2015). The severe constraints in WM and executive functions (EF) performance in the complex search-and-learning tasks in a classroom
discussed above clearly relate to SDL: Students iterative search for in- setting? The rationale for this originated from the assumption that
formation includes two core aspects which subjects them to a dual- complex learning imposes heavy loads on novice students’ WM to resist
demand task: (a) to create conceptual relations of the learning domain, distraction leading to high need for executive control. Whereas the
(b) to simultaneously memorize gathered information (Sweller et al., latter is not with equal spurt matured (due to a normal developmental
1998. p 267). Hence, individual differences in those skills mainly ex- path as well as in terms of individual differences). Accordingly, we set
plain the variance in cognitive achievements in complex learning our specific research questions as follows:

2
Õ. Uus, et al. Acta Psychologica 209 (2020) 103089

constructions, can diminish the present processing, thus restricting the


later recall of the last ones (Sweller et al., 1998).

2. Materials and methods

In order to detect participants' ability to memorize the needed info-


items while being “distracted” (see Section 2.2), we adapted the web-
based O-span (Unsworth et al., 2014; Unsworth, Heitz, Schrock, &
Engle, 2005) task (ɑ = 0.83). We aimed to compare each participant's
Ospan results to another cognitively demanding task Dino directed by
themselves (see Section 2.1). It is worth noting that the experiment took
place in the format of all students together participating in the same
classroom (to replicate the daily lesson in a usual classroom environ-
ment as realistically as possible). Therefore, the social aspect might
have an impact on the results.

2.1. Exploration task Dino

We implemented the Dino as a simple web-based environment, in


which 3D taxonomy (size/flying ability/skull type) could be explored
Fig. 1. Dino taxonomy used in the present study.
along seven (~2 min) phases. The core difference in Dino (from pre-
Corresponding image (Fig. 2) assumed to “link” those three specific features to
create associations by displayed picture (as typical to this particular category vious studies) is that students self-direct the presentation of the learning
members) by note: “This is how such dinosaur, with those features you lastly items, following a prompt: “The animal I want to see next” (1) “is
selected, look like”. small”/“is large” (2) “can fly”/“can’t fly” (3) “has a crest”/“has no
crest” (see 2.1.1). Doing so, it was possible to reach (corresponding to
one's decided selection of the features) altogether eight dinosaur cate-
(1) Does participants' cognitive capacity relate to their respective
gories. Each binary decision within each selection-triple was in the
confidence growth in complex task progress (i.e., to correctly
data-file automatically recorded either 0 – marking this selection as
memorize features-specific category names – primary task, while
without … (“large”, “flying ability and “crest”) then respectively: (1)
cumulatively forming additional concepts – secondary task)?
"small” (2) “running” and (3) “smooth-head”. Or recorded 1 – marking
(2) What dynamics appear in students' hesitation patterns (“I do not
as with … (together with these three features) then respectively: (1)
know”) in terms of WMC (i.e., primary task: memorization ability in
"large”, (2) “flying” and (3) “skull-crest”. The to-be-learned task here
the face of secondary task distraction: to select further features-
was to find (according to features-selections) the relevant category
combinations)?
names (of dinosaurs). However, there were many different options to
relate those three features (reaching then different category-names of
We proposed the following hypotheses:
respective animals). To track the participant's “learned” pattern, eight
H1. There will emerge different trajectories of confidence growth in judgments (see Section 2.1.2) were implemented: the first at the very
newly acquired knowledge depending on learners’ cognitive capacity. beginning of the task (where all students assumed to be not familiar
with the content) and then respectively one judgment after each study
H2. Higher WMC enables more efficient concepts formation in
phase (that should reveal the learning patterns).
memorizing manifesting in lower hesitation (i.e., “I do not know”).
H3. There will emerge a preference for particular learning-exemplars to 2.1.1. Study phase
study (with specific taxonomy-characteristics). The Dino task consists of 8 learning phases (each of which reaching
the respective selection of a dinosaur category). Starting each cycle, a
Our assumption was that depending on the individual cognitive
learner is asked to make three binary selections (see Table 1) of preg-
capacity, the learner will either benefit from resist distractors
iven features to be made sequentially (one by one on consecutively
(Unsworth & Spillers, 2010) that enables more efficient memorizing,
presented pages).
thus also more efficient recall, hence more steadily reach towards
Corresponding to the current selection from the taxonomy (as
surely learned gain. Or, when distracted to memorize (i.e., less able to
shown in Table 1), the respective path (in Fig. 1) led to an image ap-
resist distractors), poorer in recall manifesting in alternating dynamics
pearing on the screen (see Fig. 1).
towards surely learned. When trying to search, learn and memorize
After that, below the image students were asked to “choose the
several info-items simultaneously (Sweller et al., 2019), human cogni-
correct category-label” from the names presented below (together with
tive capacity limitations in terms of a cumulative overlap of previous
five “distractor” labels), where they should click on the correct one,

Table 1
Taxonomy to make binary decisions (in Dino task).
Dimensions 1 and 2
Locomotion and Body size

Can fly Can’t fly

Dimension 3 Is large Is small Is large Is small

Skull shape Has a crest Pteranadon Scaphognathus Parasaurolophus Rinchenia


Has no crest Istiodactylus Dorygnathus Edmontosaurus Pisanosaurus

3
Õ. Uus, et al. Acta Psychologica 209 (2020) 103089

colour and wrong by red colour. To enhance one's memorizing, only the
correct category name together with the respective picture (one per/
page) were redisplayed (highlighted by green colour). Also, to motivate
learners, the current accuracy scores were shown, like in games: the
number of correct episodes divided by the total number of episodes
already passed. As one's cognitive burden in Dino might arise by the
competing associations interfering one's mental processing (e.g., Pisa-
nosaurus–“small” as well as Rinchenia–“small”), we therefore assumed
by giving constant feedback revealing only the correct label of that
category that it would help the learner to conform the preliminary
concept. That is, repeating this features-combination association to-
gether with the relevant name should enable the student to shape their
mental link between phylogenetic features and the correct category-
label, before starting a new learning-cycle. At the beginning of the Dino
task, the participants’ initial activities in terms of decision-making
cannot be expected to be very systematic yet. By giving continuous
feedback provided in every cycle, the categorization is expected to
become more accurate (not just “playing a game”). According to the
cognitive load theory, we had a reason to expect some differences re-
vealed in learners' behavioural pattern depending on their cognitive
capacity.
Fig. 2. The Dino image example corresponding to the selection of specific
features (“Small”/“Can't fly”/“No skull-crest”).
2.1.2. Testing phase
To get a gradual overview of the participants’ knowledge gain (i.e.,
they learned indeed), one's learning progress was systematically tested.
For this, six picture-label pairs on the screen appeared (one after an-
other), where the students had to judge whether each picture (with
specific feature-sets) and the label below the image matched, an-
swering: “Surely yes,” “Probably yes,” “I do not know,” “Probably no,”
or “Surely no”. The answers were mapped to numbers with “I do not
know” always coded 0. The response “Surely yes” was determined by
value 2 and “Probably yes” by value 1 respectively. Table 2 shows the
coding system. This enabled to analyse the participants’ real knowledge
gain (at that moment). It is worth noting that in the Dino experimental
task an interference (similarly to any complex learning) could arise
along cumulatively selecting alternative feature-triple combinations, as
the overlap of the features could hamper one's processing per/moment
to (i) relate with an appropriate name, and at the same time (ii) com-
pare with all other (previously chosen) feature-triple combinations to
make a “link” to be memorized.

2.2. Complex WM tasks O-span

We adapted the validated O-span (with good internal consistency


Fig. 3. The Dino image example corresponding to the selection of specific 0.78) by Unsworth et al. (2005) that includes the dual paradigm in
features (“Large”/“Flying”/“With skull-crest”). which to-be-remembered items are interspersed with “deliberately
distracting activities” (Unsworth et al., 2014). To-be-remembered items
e.g., (Fig. 2) Pisanosaurus (i.e., Small/No wings/No crest in a skull) and were letters {F, H, J, K, L, N, P, Q, R, S, T, Y} appearing at the centre of
(Fig. 3) Pteranodontia (i.e., Large/Wings/Crest). It was possible to the screen (one after another for 1 sec each). After each letter, a “dis-
reach many other dinosaur images corresponding to the selection of traction” appeared by the math equation (each on a separate page)
other specific features set. After selecting one label of six, respective asking to be checked as true or false (e.g., 3 × 7 = 22). In every trial,
feedback “right” or “wrong” was given. Right was highlighted by green the letters and equations alternated until the last letter was presented,
then students were asked to recall all presented letters in the same

Table 2
Coding scheme to map each student's judgment on numbers (in Dino Task).

Note. Gray-shaded cells indicate correct response patterns

4
Õ. Uus, et al. Acta Psychologica 209 (2020) 103089

order. The scores of (a) correctly judged equations (mean = 91.37%; reasons can have impact on school-students’ mental contribution, hence
standard deviation [SD] = 0.05) and (b) correctly recalled letters possibly not honest if labelling categorically). Instead, we used either
(dependent variable) (ranging between 0 and 75; mean = 54.92; SD = continuous variable (i.e., the memorization proportion in Ospan) to
10.93) were involved into the analysis. Following Ospan authors’ compare participants' performance in another (i.e., self-directed task
(Unsworth et al., 2005; Unsworth et al., 2014) suggestion, we included Dino) learning scenario; or the Ospan's quartiles, then conditionally
only the math achievements with minimum rate of 85% into the ana- comparing "lower WMC" (quartile 1, who kept in mind fewer to-be-
lysis. The latter approach was implemented as a “proving” indicator learned items) and "higher WMC" (quartile 4, who kept in mind more
(i.e., the respective student contributed equally in both) of the dual items). Mean scores were 39.81 (SD = 4.47) for the “lower” and 66.87
demand (e.g., not just repeated only those items to be memorized – (SD = 3.62) for the “higher” WMC group t36.7 = −17.1; P < 0.001).
sequential letters, while not calculating in order to “save energy”) with Secondly, participants´ responses in Dino task (see coding scheme
proper math equations to be judged (Unsworth et al., 2005). Table 2) were analysed in two aspects: (1) expressing the dynamics of
their learning progress (i.e., their self-estimated learned state per/mo-
2.3. Participants and study design ment given either under high confidence “Sure”; or under low con-
fidence “Probably” respectively); (2) their hesitation dynamics (i.e., “I
Participants were Estonian-speaking students of Grades 7 and 8 do not know”) exposing their dynamics of either becoming more con-
(altogether 228 initially starting the experiment tasks) from four dif- fident (i.e., gaining total confidence) or not (in lately acquired knowl-
ferent high schools in Tallinn, Estonia (average age = 15.1 years; SD = edge).
0.61). They were not compensated financially. Their performance was
not disclosed to their teachers. Exclusions were made from the initial 3.1. Hypothesis control
sample on the basis of the following exclusion criteria: (i) In O-span –
not adequate contribution (fostering in only one direction from the H1. There will emerge different trajectories of confidence growth in
dual-task), (ii) In Dino (study phase) – searching for only one feature- newly acquired knowledge depending on learners’ cognitive capacity.
combination, (iii) In Dino (testing phase) – predominantly answering “I We included into the analysis the following self-estimated responses
do not know”, (iv) Identification problem – many participants' data-files that we interpreted respectively: “Sure yes” given under high con-
had to be excluded due to invalid personal identification code (although fidence (i.e., totally confident), and “Probably yes” given under low
they were initially asked to add the same code in both tasks), (v) confidence (i.e., not confident).
Technical obstacles – some of the classroom computers did not display The growth of confidence pattern (Fig. 4) presents that some par-
the images correctly, couple of those unexpectedly just shut down ticipants (i.e., quartile 4 as “higher WMC”) started with their higher
during the tasks (not saving data), (vi) There were also students, who rate of confidence already at the beginning of the task and continued
(although contributing) did not click until the very last page of the task, with their cautious and steady (i.e., more linear) dynamics reaching a
hence their (respectively one or the other) data was not saved auto- considerably higher level of surely confident, than others (e.g., quartile
matically. Additional aspect should be considered here, namely social 1 as “lower WMC”).
noise (i.e., working together with the entire class-set in the same Based on this overview we defined the calculated proportion ratio of
classroom), which could hamper individual mental operations (espe- the “Sure yes” towards the “Probably yes” (see Fig. 5), that enabled us
cially in conditions of already high cognitive load while self-directing to determine the participants' index of self-estimated general con-
their complex learning). Indeed, these were difficult trade-offs between fidence level to examine its dynamics. We see (Fig. 5) that “lower
ecological (classroom setting) and internal validity. Correspondingly, WMC” students' (i.e., quartile 1) general confidence trajectory was ra-
the final number of the students involved was 111. ther intermittent (although, fragmentarily raising, still lowered again,
especially towards the end of the task). While “higher WMC” partici-
2.3.1. Procedure pants' (i.e., quartile 4) general confidence graphs show rather calm
We first asked for permission from school directors and then tea- dynamics (in contrast to the latter, especially raising at the end of the
chers. A letter describing the study, its aim, data management and task).
confidentiality ensurance was sent to the parents, where we asked for We performed a linear regression of the criterion Confidence on Test
their consent (to allow their children to participate in the study). phase (categorical predictor: Phases 1–8) and Ospan performance (con-
Participants provided their personal consent in entering experiment tinuous predictor: nCorrect), explaining about 6 % of the variance in
(after being informed that they could quit at any moment). All sessions the criterion (adjusted R2 =.05, p<.001). As Table 3 shows, both
took place in their everyday class groups administered as daily “les- predictor variables contributed significantly to predicting the criterion,
sons” (duration 45 min) starting with brief instructions, then followed where the impact of time appears to be stronger than the impact of
by O-span and Dino tasks. Every student worked on a personal com- students’ performance in the Ospan task (see the beta coefficients).
puter and was asked to find a place in the classroom (as separate as H2 Higher WMC enables more efficient concepts formation in
possible from others) to ensure individual contribution. Telephones, memorizing manifesting in lower hesitation (i.e., “I do not know”).
pencil/paper and/or other helping tools were not allowed. Instructions Fig. 6 gives an overview of the number of “I do not know” answers
and demonstrations were performed by the first author. Two assisting for each learning phase.
teachers gave personalized instructions (if needed) and ensured that the Average proportion of judgments “I do not know” (Fig. 7) by each
students stayed focused and did not engage in task-irrelevant activities. learning phase shows clear differences:
Students who finished quickly were allowed to leave as quietly as Average of self-estimated judgments “I do not know” by partici-
possible. pants’ WMC quartiles (Fig. 8) presents a clearly different proportion
between the “higher” and “lower” WMC participants. An analysis of
3. Results variance confirmed the existence of a statistically significant difference
between the quartiles F(3.28) = 3.866, p = 0.02. The following Tukey
First of all, participants´ Ospan scores were analysed by the pro- post-hoc test highlighted the different groups: 4 vs 1 and 4 vs 2
portion of memorized capital letters (the first task) in light of math (p<0.05).
calculations (as the second task of the dual demand). We included into H3. There will emerge a preference for particular learning-ex-
examination (to define the proportion of correct memorization) of min. emplars to study (with specific taxonomy-characteristics).
85% appropriate math rate (as determined by the Ospan authors). To We calculated the average occurrence (i.e., images appeared on the
avoid a strict estimation of “High” or “Low” WMC (since many different screen) proportion of each dinosaur within each WMC quartile that

5
Õ. Uus, et al. Acta Psychologica 209 (2020) 103089

Fig. 4. "Total confidence" and "low confidence" trajectories (in Dino task).

revealed differences. A more detailed analysis shows that among all Based on the Dino task preprogrammed 3D characteristics system,
options of dinosaur types, the appearing frequency of category-image we can conclude that in such a different frequency aspect (of the
Pisanosaurus emerge from others, as revealed by ANOVA F(1,90) = Pisanosaurus images) for to reach this particular category-name, those
8,053, (p<0.01). Therefore, since the Dino task environment was participants in “lower WMC” group (quartile 1) consequently had to
programmed to show an equal distribution of image presentation (in prefer respective heuristics more often in the study phase (i.e., con-
the judgment phase) across all animal category names involved in the structing the learning samples) that corresponds to this image ac-
task, it was reasonable to conclude from this result that some of the cording to specific features-combination “Small”/“No wings”/“No
participants (more than others) have chosen a respective combination skull-crest” (as shown Fig. 1 as well as Fig. 2: on the left).
of characteristics that led to this particular image. Kruskal-Wallis test
confirms this image to be of the category of Pisanosaurus, χ2 = 8.9532;
degree of freedom [df = 3]; P < 0.05). The following Tukey post-hoc 4. Discussion
test confirms that in the latter finding exists a statistically significant
difference (see Fig. 9) in the comparison pair between the groups 4 vs 1, The main aim of this study was to examine two basic cognitive
p=0.029 (p<0.05). processes needed in complex SDL in general: information search and
memorizing. We found that cognitive capacity relates to the

Fig. 5. Confidence (“Sure yes” towards “Probably yes”) ratio dynamics (across Dino task).

6
Õ. Uus, et al. Acta Psychologica 209 (2020) 103089

Table 3 they are still developing (Nęcka & Lulewicz, 2016). Nonetheless, in-
Summary of linear regression of Confidence on Test Phase and Ospan perfor- trinsic cognitive load (that involves information complexity as well as
mance the present knowledge of the individual) makes it possible to change
Criterion Predictor variable β SE t value only when changing what needs to be learned or when changing the
learner’s expertise (Sweller et al., 2019). While extraneous cognitive
Confidence Intercept −.20 .247 −0.79 load (that encompasses how the information is presented and what the
Phase 2 .39 .161 2.41⁎
learner should do) can be changed when instructional procedures
Phase 3 .45 .168 2.65⁎⁎
Phase 4 .65 .173 3.76⁎⁎⁎ change to reduce element interactivity (Sweller et al., 2019). Therefore,
Phase 5 .52 .174 3.00⁎⁎ germane cognitive load “redistributes” WM resources (for to get
Phase 6 .63 .175 3.59⁎⁎⁎ learned) from extraneous activities how? to deal with intrinsic mental
Phase 7 .77 .175 4.37⁎⁎⁎
activities via the effort to acquire knowledge to increase one’s expertise to
Phase 8 .90 .174 5.19⁎⁎⁎
Ospan nCorrect .01 .004 2.53⁎
gain solutions to problems (Gyselinck et al., 2008). However, the more
resources a learner devotes to dealing with extraneous cognitive load,
⁎⁎⁎
p<0.00. the fewer remain for managing intrinsic cognitive load, leading to less
⁎⁎
p<0.01. learning, therefore, germane and intrinsic cognitive loads are tightly

p<0.05. interrelated (Sweller et al., 2019). As such, the task presentation as
extraneous load what to do? is especially important in the context of
participants’ confidence growth trajectory of its progress to correctly complex SDL, where intrinsic cognitive load is already high because of
memorize features-specific category names, while cumulatively con- the task novelty and/or lack of pre-knowledge within the domain. As a
structing additional features-set combinations. We conclude that result it will be difficult for the learner to encode and save in parallel,
“higher-WMC” participants more efficiently resisted distractions to because of the problem to divide executive attention at the same time
process current learning-items to be memorized. This result is in line (1) to resist distractions and (2) to control the WM content as updating
with previous findings pointing at the risk of overload especially in (Gyselinck et al., 2008). The latter constraining aspect was also mani-
novel multitasking situations, where must ensure that relevant goal is fested in the present study, instanced here by the dual-task paradigm,
being maintained through resistance towards an interference, that where two tasks involve (or compete for) the same limited cognitive
might interrupt the processing (Cowan, 2014; Nęcka & Lulewicz, 2016; resources: the secondary task performance is affected by the cognitive
Unsworth & Spillers, 2010). The latter directly relates to SDL, where load induced by the primary task (Gyselinck et al., 2008). Which, ac-
one's iterative search of information includes two core aspects: (a) to cording to the second research question about the dynamics in students'
create conceptual relations, (b) to simultaneously memorize sequen- hesitation patterns “I do not know” got an answer that Higher WMC
tially gathered information (Sweller et al., 1998). Hence, as human enables more efficient concept formation (manifested in lower rate of
cognitive capacity lays the ground for one’s conceptual memorization, hesitation). We propose that higher-WMC shielded those students
individual differences in those skills mainly explain the variance in change in their expertise via more efficient binding of learning-items
cognitive achievements in complex learning (Schwaighofer et al., (so constantly lowering their intrinsic cognitive load as their expertise
2017). Especially in terms of cognitive load perspective that is related raised). That is, they did not get distracted, which is needed for bene-
(a) to the difficulty of the task; (b) information complexity; (c) initial ficial memorizing (Sweller et al., 2019) while forming their own con-
knowledge in this field. Whereas, the cognitive load theory (CLT) ex- ceptual links due to which such linked connections were later more
plains how learning tasks can induce processing load, affecting the easily reachable for them (Ecker et al., 2010; Miyake et al., 2000;
learner’s ability to construct their knowledge into LTM, in particular, Shipstead et al., 2015) during recall. The latter seemingly gave them the
when the elements must be maintained simultaneously (de Bruin & van benefit of increasing their sense of confidence with a rather steady
Merriënboer, 2017). While human limited WM “space” can process only trend. That is, germane cognitive load “redistributed” their WM re-
small amounts of information per/time (Cowan, 2014; Kirschner et al., sources from extraneous activities how to do it? to intrinsic effort to
2006; Schwaighofer et al., 2017; Shipstead et al., 2015; Sweller et al., elaborate their own knowledge enabling them to sequentially create
2019). Moreover, children’s cognitive capability is even more limited as quality memory schemas (Gyselinck et al., 2008) as seen by their

Fig. 6. “I do not know” judgments overview (in Dino task).

7
Õ. Uus, et al. Acta Psychologica 209 (2020) 103089

Fig. 7. “I do not know” judgments by learning phases (in Dino task).

judgments towards “Sure yes”. On the other hand, our results revealed when changing one's expertise "understood" that their only option to
(by Hypothesis 4) that lower-WMC participants more often met a par- achieve the goal is to reduce own mental load (i.e., to “redistribute”
ticular type of learning examples. They had seemingly constructed re- WM resources) from extraneous load to deal with intrinsic load (Sweller
spective characteristics leading to a specific image such as "small”/ et al., 2019). As such, they put their effort helping themselves to gain
“running”/“smooth-head” corresponding to the category of Pisano- the knowledge keeping mental load as low as possible (preferring the
saurus. We propose that using such a strategy, lower-WMC students easiest taxonomy-combination "Small"/"Running"/"Smooth-head" as
either assumingly tried to keep their cognitive load as low as possible, without distracting features as possible). Doing so they might hope to
e.g., as Pisanosaurus involves quite basic and more easily remembered increase their expertise to get this task done as solution to problem
features to-be-learned thereby not overburdening own cognitive cap- (Gyselinck et al., 2008). Since the more resources a learner devotes to
ability for one’s concurrent processing. Or, it is also possible that, since dealing with extraneous cognitive load, the fewer remain for managing
such self-helping method is the easiest way to differentiate between the intrinsic cognitive load, leading to less learning (Sweller et al., 2019).
animals within taxonomy, they may repeat just in case the same char- The latter is confirmed also by the finding in this work that lower-WMC
acteristic-construction without daring to explore other alternatives. participants were more hesitating (as seen by their “I do not know”
Understandably, as we cannot claim this but only assume by our results, pattern) than higher-WMC students. This is in line with the prior sug-
this would be an interesting topic for further studies. Hence, we can gestion that, as concept formation (i.e., schema construction) involves
only hypothetically propose that lower-WMC students facing intrinsic two basic functions: (1) the storage organization in LTM and (2) re-
cognitive load and realizing that it is only possible to solve this task duction of the WM load, therefore those two should “constitute the

Fig. 8. Average of "I do not know" judgments by quartiles (in Dino task).

8
Õ. Uus, et al. Acta Psychologica 209 (2020) 103089

Fig. 9. Frequency of Pisanosaurus.

primary role in education as well as training systems” (Sweller et al., 1998. Schwaighofer et al., 2017; Sweller et al., 2019). To be efficient in im-
p 256). Since one’s ability to form new concepts directly depends on plementing the approach of SDL, we should first focus on the learner’s
cognitive load (Schwaighofer et al., 2017), the percentage of correctly WM–LTM interaction (Sweller et al., 2019), especially in terms of the
recalled items decreases as cognitive load increases (Shipstead et al., executive aspect in it (Schwaighofer et al., 2017). Since, the latter en-
2015; Sprenger et al., 2011). Whereas, the most critical aspect in ables the interrelated searching and learning acts, which are central in
complex learning is to maintain information despite interference, yet various SDL scenarios, where the learner has to connect collections of
individual differences in the ability to block distractors (which is highly interacting items (Sweller et al., 2019) that often cannot be learned in
interdependent of executive attention) form the basis of WMC isolation creating an even heavier load on executive functioning until
(Coolidge & Wynn, 2005). Therefore, WMC is a rather domain-free the most suitable solution can be found (Best & Miller, 2010;
process, since any mental task requires the use of at least some domain- Schwaighofer et al., 2017). This is why an expert can quickly re-link the
specific stimuli, while the processing skills work in conjunction with elements, so freeing up the processing “space” to find a solution
WMC. Thus, no single psychological measurement can capture only the (Cowan, 2014), whereas from the novice, it requires high demand on
sole executive section of WMC yet taps the both: the domain-free WMC executive control that leads to a heavy cognitive load (Kirschner et al.,
as well as the required domain-specific skills (Coolidge & Wynn, 2005, 2006; Roebers, 2017; Sweller et al., 2019). Consequently, the capacity
p. 9). However, executive control depends on the cognitive maturity to of the WM is not simply a matter of how much information a person can
enable the EF’s foundational components: WM retention, inhibition, hold temporarily, it is also one’s ability to put that information to use
and shifting (processes needed for goal-directed behaviour) orche- (Shipstead et al., 2015). All those aspects are especially crucial for
strated by the human prefrontal cortex, which development continuing school-age children, whose cognitive mechanism is more limited than
still throughout childhood, adolescence, and even young adulthood in adults (Nęcka & Lulewicz, 2016). Thereby, our empirical findings are
(Best & Miller, 2010; Nęcka & Lulewicz, 2016). Whereas, inhibitory in line with prior work related to one’s cognitive capacity core aspects:
control in particular prevents a person from accepting inappropriate the span to hold current items active and the effectiveness to bind
solutions (Nęcka & Lulewicz, 2016). At the same time, we have many elements into task-relevant structures (Nęcka & Lulewicz, 2016;
students in the classrooms, who are due to different reasons and in spite Shipstead et al., 2015). Once information has been efficiently saved in
of their cognitive ab abilities level, even more vulnerable to high cog- LTM, it can be more easily transferred back to be reused (Sweller et al.,
nitive load because of anxiety, emotional (Visu-Petra, Stanciu, Benga, 2019). In which, as shown by our results, there is a more efficient trend
Miclea, & Cheie, 2014) or motivational issues (Zepeda et al., 2015) of knowledge gain revealed in higher-WMC participants' learning pat-
directly hampering executive control (Schwaighofer et al., 2017). tern. Although our results were not so surprising (prior findings have
Thereby, the inhibitory control, that forms a precondition over the also concluded different outcomes of learning as well as differences in
possible lures, shapes the most crucial aspect for complex learning, cognitive capacity), they shed clear light, especially in the complex SDL
especially for school-age children (Nęcka & Lulewicz, 2016). Conse- context. Moreover, as instanced by our example (of learning and search
quently, as SDL expects specific activities requiring internal application activities), the school students’ (whose higher-order abilities develop
of proper skills and strategies from the learner (Rutherford et al., 2018; with unequal paths) mental interaction between WM and LTM fostered

9
Õ. Uus, et al. Acta Psychologica 209 (2020) 103089

by executive attention can either support or delay the progress in SRL individuals. Consequently, while integrating the learning methods led
gain. The latter is important to foresee before implementing learning by the school-students themselves, we cannot take for granted that all
scenarios directed by school students themselves, as well as in guiding learners perform gaining similar results. Therefore, an assistive meth-
the learning material presentation in a proper way that would optimize odology is required in order to efficiently scaffold the progress of
all learner's cognitive load for his/her efficient intellectual perfor- learning led by oneself, especially for students with lower capability to
mance. optimize their cognitive load. Also, it is needed to help learners to ac-
quire their own multifunctional metacognitive competencies for effec-
5. Limitations and future work tive learning as today students should become critically thinking citi-
zens with proper strategies in order to prevent burnout, especially in
This study had a few limitations. All conclusions here were drawn light of the 21st century's need for quick reorientation in rapidly
on the basis of a sample of data collected using a particular method changing circumstances expecting fast adaptation. When provided with
under specific, controlled conditions. However, the main construct, the more sophisticated preparing, the school students’ benefit will be
moderating role of executive attention additionally depends on the multiple: they should succeed academically in school by resisting cog-
context characteristics, such as social noise in the classroom and time nitive overload and also remain motivated to continue education. So
constraints creating additional cognitive demand. Therefore, our future being up for lifelong learning, this will give them the advantage of
work will focus on different samples, designs, and settings. We had to resilience, which, in turn, will support their long-term overall well-
exclude many participants, yet, this fact shows vividly the different being.
skills or will of students in the same age group. Similar restricting re-
sults were also reported by Nęcka and Lulewicz (2016), where a pro- CRediT authorship contribution statement
portion of the sample did not fulfil the established accuracy criteria. In
the current work by the O-span’s accuracy criterion (Unsworth et al., Õnne Uus: Conceptualization, Methodology, Investigation,
2005; Unsworth et al., 2014), we excluded 15 participants. In addition, Formal analysis, Visualization, Writing - original draft. Paul
in Dino, participants were excluded if their data clearly showed no Christian Seitlinger: Methodology, Resources, Software, Data
contribution. Here, the aspect of motivation (i.e., willingness to con- curation, Writing - review & editing. Timo Tobias Ley: Project ad-
tribute) might have played a major role, where many students mono- ministration, Writing - review & editing, Supervision.
tonously responded only “I do not know”. In a way it should be taken
into account that especially in case of teenagers, the students’ interest, Acknowledgements
mood, and will to contribute are, indeed, hectic.
This study also had a few positive aspects: (1) a setting similar to This work was partially funded by the European Union’s Horizon
real SDL were used. (2) Data which was gathered moment-to-moment 2020 research and innovation program [grant agreement number
allowed us exploration along the entire learning process in order to 669074, project CEITER: http://ceiter.tlu.ee].
analyse and make valid conclusions about the dynamics and trajectories
of learning behaviour. (3) Highlighting the role of WM and executive Ethical approval
attention in higher-order cognition in general, particularly in complex
SDL (de Bruin & van Merriënboer, 2017; Schwaighofer et al., 2017), our Informed consent was obtained for experimentation. All procedures
results are in accordance with earlier findings, especially in terms of performed in the study were in accordance with the ethical standards of
cognitive load theory (Sweller et al., 2019). Hence it inspires us to the institutional research committee. The privacy rights were carefully
further develop suggestions for more specific metacognitive skills in followed, no any harm caused, no additional accompanying aspects
order to facilitate the SDL approach, as specific scaffolding strategies observed.
are highly needed until school students' (as still children's) own skills
are acquired and/or their cognitive mechanism is sufficiently mature. References
In line with Nęcka and Lulewicz (2016) suggestion that particular ex-
ecutive control of inhibitory function prevents one from accepting in- Baddeley, A. (1996). The fractionation of working memory. Proceedings of the National
appropriate solutions, it might be less important for an adult because of Academy of Sciences, 93(24), 13468–13472 References.
Best, J. R., & Miller, P. H. (2010). A Developmental Perspective on Executive Function.
already matured capabilities, but in school-age children, the attention Child Development, 81(6), 1641–1660.
control aspect seems crucial. This study gives value especially in terms de Bruin, A. B., & van Merriënboer, J. J. (2017). Bridging cognitive load and self-regu-
of complex learning as well as SDL paradigm, where learning outcomes lated learning research: A complementary approach to contemporary issues in edu-
cational research. Learning and Instruction, 51, 1–9.
unequivocally depend on one’s active contribution to maintain atten- Coolidge, F., & Wynn, T. (2005). Working Memory, its Executive Functions, and the
tion control. Even more, students with under-performance or low Emergence of Modern Thinking. Cambridge Archaeological Journal, 15(1), 5–26.
achievement might lose their last motivation to learn, if faced with Cowan, N. (2014). Working Memory Underpins Cognitive Development, Learning, and
Education. Educational Psychology Review, 26(2), 197–223.
learning problems for a long time already (i.e., due the difficulties of
Ecker, U. K. H., Lewandowsky, S., Oberauer, K., & Chee, A. E. H. (2010). The Components
overload, even possible dropout in the case of too high cognitive load). of Working Memory Updating: An Experimental Decomposition and Individual
The latter, however, can lead to academic burnout, which, in turn, leads Differences. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36,
170–189.
to other mental health problems. Therefore, children’s complex
Garner, J. (2009). Conceptualizing the relations between executive functions and self-
learning skills should be treated carefully and their limited, often fragile regulated learning. The Journal of Psychology, 143(4), 405–426.
cognition should not be overloaded, because this directly influences Gilhooly, K., & Logie, R. H. (2004). Working memory and thinking: Current issues in thinking
their overall functioning capability. Instead, we need to focus on laying and reasoning. Psychology Presshttps://doi.org/10.4324/9780203346754.
Gyselinck, V., Jamet, E., & Dubois, V. (2008). The role of working memory components in
the groundwork in children mental capacity at an early age, helping multimedia comprehension. Applied Cognitive Psychology: The Official Journal of the
them gain, develop, train and improve their universal metacognitive Society for Applied Research in Memory and Cognition, 22(3), 353–374.
skills that enable to know on “how my cognition and mental – as well as Kirschner, P. A., Sweller, J., & Clark, R. E. (2006). Why Minimal Guidance During
Instruction Does Not Work: An Analysis of the Failure of Constructivist, Discovery,
learning – operations are manageable at all” and are usable in different Problem-Based, Experiential, and Inquiry-Based Teaching. Educational Psychologist,
circumstances. 41(2), 75–86.
Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D.
(2000). The Unity and Diversity of Executive Functions and Their Contributions to
6. Conclusion Complex ““Frontal Lobe” Tasks: A Latent Variable Analysis” Cognitive Psychology, 41,
49–100.
Mental acts, learning and cognitive capability differ among Nęcka, E., & Lulewicz, A. (2016). Capacity, Control, or Both–Which Aspects of Working

10
Õ. Uus, et al. Acta Psychologica 209 (2020) 103089

Memory Contribute to Children’s General Fluid Intelligence? Polish Psychological Unsworth, N., Fukuda, K., Awh, E., & Vogel, E. K. (2014). Working memory and fluid
Bulletin, 47(1), 21–28. intelligence: capacity, attention control, and secondary memory retrieval. Cognitive
Roebers, C. M. (2017). Executive function and metacognition: Towards a unifying fra- Psychology, 71, 1–26.
mework of cognitive self-regulation. Developmental Review, 45, 31–51. Unsworth, N., Heitz, R. P., Schrock, J. C., & Engle, R. W. (2005). An automated version of
Rutherford, T., Buschkuehl, M., Jaeggi, S. M., & Farkas, G. (2018). Links between the operation span task. Behavior research methods, 37(3), 498–505.
achievement, executive functions, and self-regulated learning. Applied Cognitive Unsworth, N., Redick, T. S., Heitz, R. P., Broadway, J. M., & Engle, R. W. (2009). Complex
Psychology, 32(6), 763–774. working memory span tasks and higher-order cognition: A latent-variable analysis of
Schwaighofer, M., Bühner, M., & Fischer, F. (2017). Executive functions in the context of the relationship between processing and storage. Memory, 17(6), 635–654.
complex learning: Malleable moderators? Frontline Learning Research, 5(1), 58–75. Unsworth, N., & Spillers, G. J. (2010). Working memory capacity: Attention control,
Shipstead, Z., Harrison, T. L., & Engle, R. W. (2015). Working memory capacity and the secondary memory, or both? A direct test of the dual-component model. Journal of
scope and control of attention. Attention, Perception, & Psychophysics, 77(6), Memory and Language, 62(4), 392–406.
1863–1880. Van Deur, P., & Murray-Harvey, R. (2005). The Inquiry Nature of Primary Schools and
Sprenger, A. M., Dougherty, M. R., Atkins, S. M., Franco-Watkins, A. M., Thomas, R. P., Students' Self-Directed Learning Knowledge. International Education Journal, 5(5),
Lange, N., & Abbs, B. (2011). Implications of cognitive load for hypothesis generation 166–177.
and probability judgment. Frontiers in Psychology, 2(JUN), 129. Visu-Petra, L., Stanciu, O., Benga, O., Miclea, M., & Cheie, L. (2014). Longitudinal and
Sweller, J., van Merriënboer, J. J., & Paas, F. (2019). Cognitive architecture and in- concurrent links between memory span, anxiety symptoms, and subsequent executive
structional design: 20 years later. Educational Psychology Review, 1–32. functioning in young children. Frontiers in psychology, 5, 443.
Sweller, J., Van Merrienboer, J. J., & Paas, F. G. (1998). Cognitive architecture and in- Zepeda, C. D., Richey, J. E., Ronevich, P., & Nokes-Malach, T. J. (2015). Direct instruction
structional design. Educational psychology review, 10(3), 251–296. of metacognition benefits adolescent science learning, transfer, and motivation: An in
Thomas, J. W., Strage, A., & Curley, R. (1988). Improving students’ self-directed learning: vivo study. Journal of Educational Psychology, 107(4), 954.
Issues and guidelines. The Elementary School Journal, 88(3), 313–326.

11

You might also like