Introduction
The terms Executive Function (EF) is used to refer to a set
of top-down processes that allow to regulate one’s thoughts
and behaviors
(Miyake & Friedman, 2012)
. The abilities
that refer to these processes are inhibitory control, planning,
cognitive flexibility and working memory (WM).
Inhibitory control or Inhibition may be defined as «A
deliberate overriding of a dominant or prepotent response»
(Miyake & Friedman, 2012, p. 2)
even though inhibition is a
multi-componential construct itself and comprehends many
different abilities such as managing impulses and
interferences, both behavioural and cognitive
(Diamond,
2013; Nigg, 2004)
. Due to this complexity, inhibition lately
has been theorized as an ability that more than being a
dimension per se may be a general factor of executive
control that may also influence all the other abilities
(Miyake & Friedman, 2012)
. Working Memory is the ability
to hold in mind useful information while completing a task
when the information is no longer present. It is possible to
distinguish two type of WM, verbal and visuo-spatial WM
(Baddeley, 1986; see Engle, Kane, & Tuholski, 1999 a for a
review).
Diamond (2013)
lately reviewing the literature
theorizes a bi-directional relation between Inhibition and
WM, describing them as separate dimensions yet strongly
related.
1.1 Why is important to have a good Inhibitory
Control?
This ability has an important control and filter role that
involve being able to manage thoughts, actions and, in
general, interferences. Inhibition allows us to keep the
attention focused, to suppress the override of a prepotent
response or an, usually, automatic action in case of need or
danger. Inhibition imply also the ability to regulate behavior
and emotions, without this kind of control we would be at
the mercy of impulses, and in constant danger of taking bad
decision or to not be able to manage complex situations.
1.2 Development of Inhibition
The ability to inhibit responses has been observed in
several studies since early childhood
(Kochanska, Murray &
Coy 1997; Diamond, 2002; Jones, Rothbart & Posner, 2003;
Carlson, 2005; Garon, Bryson & Smith, 2008)
.
Best and
Miller (2010)
report a significant peak of development in
inhibitory processes in preschools years. In effect, in
behavioural response inhibition task (e.g. Go-noGo) and
also in more complex ones that involve also WM (e.g. the
Stroop test), a successful performance is registered already
by the age of four. From five to eight this ability registers a
continuous growth, in particular in task in which inhibition
and WM are combined
(Gerstadt, Hong & Diamond, 1994;
Carlson, 2005)
. Other studies document a significant
increase in these abilities during childhood, in particular
from three to four year of age
(see Clark et al., 2013)
.
Gandolfi, Viterbori, Traverso and Usai (2014) analyzing the
organization of this construct in preschoolers and trying to
investigate the latent structure of cognitive processes
involved in inhibition in children from the 24 to 48 month,
suggest that inhibitory processes are not yet differentiated
before 36 months of age, after which a distinction between
different inhibitory dimensions emerges.
1.3 What we Know about Inhibition and WM during adolescence?
Prencipe et al. (2011)
, in order to evaluate changes and
improvements across ages, analyze the performance of a
large sample of children and mid-adolescents between 8 and
15 years of age at a battery of Inhibition and WM tasks and
dividing the sample into age groups (8 and 9 year-olds, 10
and 11 year-olds, 12- and 13-year-olds, and 14- and
15-yearolds) results in WM capacity show the largest improvements
at backward Digit Span relatively early in adolescence,
between the two youngest age groups. This result is
generally consistent with the findings of Hooper, Luciana,
Conklin & Yarger (2004). In general studies of WM during
adolescence come across variable results: some studies
report level in terms of performance comparable to the one
registered in adults already by early adolescence
(Asato,
Sweeney, & Luna, 2006; Crone, Wendelken, Donohue, van
Leijenhorst & Bunge, 2006)
while others show a more
linear pattern in terms of gradual improvements through late
adolescence
(e.g. Hooper et al., 2004; Huizinga, Dolan &
Van der Molen, 2006; Lamm, Zelazo & Lewis, 2006)
.
Prencipe et al. (2011)
, point out how in backward digit span
10 and 11-year-olds in their study performed as well as
typical adults
(Janssen, Krabbendam, Jolles, & van Os,
2003; Kinsella et al., 1996)
.
1.4 WM and Inhibition which relation?
Cognitive Inhibition is important to support WM: to achieve
a goal is very important to select carefully relevant
information in order to let the goal guide our behavior
decreasing the probability of being driven by an automatic
response that could imply a failure in pursuing the goal.
Moreover to relate ideas or events it is extremely important
to be able to move the focus from one to another without
being stuck only to one and also to be able to elaborate and
combine facts in a creative way, resisting the impulse to
repeat the same patterns in thinking. Lastly, inhibition is
very important to filter internal and external distractions and
to prevent our mental space to be overwhelmed by
information, resisting to proactive interference and deleting
no longer relevant facts or materials
(Hasher & Zacks 1988;
Zacks & Hasher, 2006)
.
2.This Study
The present work is aimed to investigate how inhibition and
WM do organize during adolescence, in literature is
documented how these abilities develop until the later years
of this particular and unique developmental stage. For this
reason the aim of the study is also to try different structural
equations models in order to investigate if WM and
inhibition cluster in a unitary dimension or if they are
recognizable as separate dimensions connected by a strong
relation or inter-relation. We hypothesize that all inhibition
tasks would cluster into one factor and WM tasks into
another, in fact this pattern has already been tested in
adulthood
(Miyake et al., 2000; Miyake & Friedman, 2012)
.
As mentioned is still quite unclear how these aspects do
organize in this great time of changes in the human brain, in
particular we hypothesize that WM and Inhibition may
result strongly related. In order to test these hypothesis we
selected tasks created to investigate different aspects of
inhibition, in agreement with the literature who suggests
that different abilities are related to this construct (Nigg,
2004).
Participants
240 Adolescents, (158 females, 82 males) attending High
school, age range 14 to 19.
Inhibition Task:
Antissacade task
(Adapted from Roberts, Hager & Heron,
1994)
. This task is a common measure of oculomotor
inhibition. Participants practiced on 22 trials and then
received 90 target trials. The dependent measure was a
balanced index obtained by dividing the mean of all reaction
times for each trial and the proportion of correct trials.
Flanker task
(Eriksen,1974)
. This task is a cognitive
measure of the ability of managing interferences.
Congruent, incongruent and neutral trials were shown. A
congruent trial is one in which the flankers are associated
with the same response as the target, whereas in an
incongruent trial, the flankers are associated with a
competing response. The task was composed by a practice
block of eight trials (two congruent, two incongruent and
two neutral) and a test block of 48 trials (16 trials for each
condition). The dependent measure was a balanced index
obtained by dividing the mean of all reaction times for each
trial and the proportion of correct trials.
Go-No Go task
(Donders, 1969)
. The Go-No go is a
classical inhibition paradigm meant to assess the ability to
stop an automatic response. The task was composed by a
practice phase of 20 trials and a test phase of 100 trials.
Participant were asked to answer as fast as they could every
time the “Go” stimulus appeared and received a feedback
every time they pressed reporting their RT so they could
realize how fast/slow they were. The dependent measure
was a balanced index obtained by dividing the mean of all
reaction times for each trial and the proportion of correct
trials.
Stop Signal task
(Logan, 1994)
. The task requires to stop a
motor response while already activated. The experiment
consists of two phases: a practice phase of 32 trials and an
experimental phase of three blocks of 64 trials. In both
phases, each trial starts with the presentation of the fixation
sign, which is replaced by the primary-task stimulus after
250 msec. The dependent measure is the SSD index, which
is an adjusted measure, based on the accuracy and the
consequential adjustment of the delay.
Working Memory task
The two following WM tasks are widely used as complex
measures of WM capacity, while the third and last one is a
more classic visuo-spatial span measure.
Symmetry Span task (SymmSpan)- Kane et al. (2004).
This span is composed by two different visual task
performed at the same time. The first one consists in
recalling a sequence of square that appear on the screen
while the second one consists in judging if some figures are
symmetric or not. Dependent variable was the absolute span
score the SSPAN score: the sum of all perfectly recalled
sets. So, for example, if an individual recalled correctly two
squares in a set size of two, three squares in a set size of
three, and three squares in a set size of four, their SSPAN
score would be five (2 + 3 + 0).
Reading Span task (RSPAN)
(Daneman & Carpenter,
1980)
. This span is structurally identical to the previous one,
but is made up with different stimuli. In fact the tasks
consists in recalling a sequence of letters that appear on the
screen while judging if some phrases make logical sense or
not. As mentioned in the previous task we used the RSPAN
score.
Mr Cucumber (Case, 1985). This is the only
noncomputerized test. The task requires to recall the position of
stickers on a target image by pointing with the finger on a
figure without any sticker on. A point is given for each level
fully correctly recalled, one third of a point (0.33) is given
for each correct item beyond that level. The test must be
discontinued if the participant fail all the three item of the
same level. Dependent measure is the score obtained
(expected range 0-8).
Statistical Analysis
In order to validate our hypothesis we performed a latent
factor analysis with two different model: a one factor model
where Inhibition and WM are clustered in a unitary factor
and a two factor model in order to test the hypothesis of two
separate abilities strongly related. Descriptive statistics and
zero-order (Pearson) correlations among measures were
calculated. Outliers values at more than three standard
deviation of the mean were 27 and were excluded from the
analysis. In addition nine values were excluded from
memory scores because they did not respect the task
instruction to keep the percentage of symmetries/ phrases
correctly judged to at least the 75%. In total the values
excluded represent the 1.8% out of the full sample.
Two CFAs, based on covariance matrices, were conducted
using EQS 6.1 software
(Bentler, 2006)
.
The fit of each model to the data was evaluated by
examining multiple fit indices
(Schermelleh-Engel,
Moosbrugger, & Műller, 2003)
: the χ2 statistic, the root
mean square error of approximation (RMSEA), the
standardized root mean squared residual (SRMR), Bentler’s
comparative fit index (CFI), the non-normed fit index
(NNFI) and the Akaike Information Criterion (AIC). In
addition a measurement of invariance across gender and age
was performed, in order to verify the reliability of the model
and to test differences between the groups. In order to
perform this particular analysis through age, the sample was
first divided into two age groups: a first group from 14 to
16;11 years of age (n 120, 37 males- 83 females, mean age
15.7±11.8) and a second group from 17 to 19 (n 107, 42
males-65 females, mean age 18.3±9.3). Then we proceeded
by testing different levels of invariance, running separated
models
(see van de Schoot, Lugtig, & Hox, 2012)
in which
different parameters were progressively constrained; then,
as the models could be considered nested, we used the
chisquare test and other goodness-of-fit indices to compare the
models
(see also Lee et al., 2013)
.
3. Results
Descriptive statistics and correlations are summarized in
table 1.1 and 1.2.
Note: Stop SSD= Stop Signal Delay index; Symm_Span=
Visuo-Spatial Span Measure; RSPAN= Phonological-Span
Measure
All tasks seem to show a significant correlation pattern that
reflects an association among the tasks. This result confirm
that Inhibitory task and WM task are related both with each
other and also in terms of inter-Correlation between
different dimensions.
Confirmatory Factor Analysis (CFA)
To determine the EF structure, we tested two factor
models representing the two different theoretical hypothesis:
a one-factor model and a two factor model. The fit indices
for these models, summarized in Table 2, are good for the
two-factor model. Parameters for the one-factor model are
generally adequate or mediocre in the case of NNFI and
CFI. Based on the goodness of fit and AIC values, the
twofactor model appears to better fit the data than the more
parsimonious, single-factor model.
Note. RMSEA, Root mean square error of approximation;
SRMR, Standardised root mean squared residual; CFI,
Comparative Fit Index; NNFI, Non-normed fit index; AIC,
Akaike’s information criterion; Note: Stop SSD= Stop Signal
Delay; Symm_Span= Symmetry complex span; RSPAN= Reading
Span; EF=Executive function single-factor ; WM= working
memory.
The error-term squares are considered to be estimates of the
unexplained variance for each task. The factor loadings
were all significant (t values>2). We will only discuss the
two-factor model parameters, as the two-factor model
showed the best fit to the data. The estimate of the
correlation between the latent variables was quite large
(.68). The 90% confidence interval for the correlation was
[0.79 ,-0.53]. The proportion of variance in the individual
task scores explained by the latent variables varied across
tasks. The R2 values were .475 concerning the Go-NoGo
task, .277 concerning the visuo-spatial span measure, .275
concerning the balanced index for the Flanker task, .230
concerning the MR Cucumber span measure, .163 with
reference to the antisaccade and .127 relatively to the
Phonological span measure. Gender and age invariance
testing was conducted for the best-fitting two-factor model
only. Configural invariance was already assumed by
identifying a two-factor solution in females. Configural
invariance and metric plus covariance between the latent
factors models were invariant and showed a good fit, while
constraining the intercepts resulted in a deterioration of the
model’s fit: when the intercept of males are constrained to
be equal to those of females the chi-square significance
improve, suggesting a difference among the two groups. In
order to perform this analysis the sample was divided into
two age intervals: The first one from 14 to 16;11 years of
age while the second from 17 to 19 years of age. The
invariance configural model replicated on the 17-19 sample
group seem to show stronger reliability in terms of goodness
of fit. R2 show that also factorial weights in terms of
variance change, this suggest a different organization in
terms the influence that specific abilities may gain along the
years.
4.Discussion
Results show that all inhibitory task cluster in one factor,
moreover according to Pearson correlation all inhibitory
tasks result also to be related to each other (see table 1.1).
This find suggests a unitary nature of inhibition in
opposition to previous studies in adulthood
(e.g. Nigg,
2000; Friedman & Miyake, 2004)
and in childhood
(Gandolfi et al., 2014)
. In fact also in this stage of
development it seems that even if the task implied in this
study do elicit different sub-abilities they actually seem to
be connected to the same macro ability: Inhibition. Most of
the variability seem to rely on the Go-No Go, a task that
elicit a more behavioural and simple kind of inhibition, the
inhibition of an automatic response, an ability that is also
one of the first to register a good performance even in early
age
(see Best & Miller, 2010)
and one of the skills that is
more connected to the adaptive function of voluntary
control: being able to stop an overriding automatic response
in every-day life may be quite important both in terms of
social adaptability and self-preservation. The ability to
manage interferences is the second index that seem to
explain the second larger portion of variability of the
response data around the mean, this ability is another core
dimension of inhibition, because scaffold also the ability to
stay focused and to not get distracted by irrelevant
information. These results seem to reveal that in
adolescence, as in adulthood, these abilities are core
dimension in inhibitory control. On the other hand
antisaccade and stop signal delay indexes seem to explain
less portion of the global variability. This find may be due
to the fact that antisaccade task relies on more basic
processes that may be more difficult to control voluntarily .
The analysis of invariance through gender show that the
model fits both datasets, until strong invariance: this
finding suggests a difference among the two groups in terms
of quickness and accuracy: males may be more reactive than
females, and especially in terms of RT may differ from
females. On the other hand the analysis of invariance
through age group seems to show some differences : the
configural invariance replicated on the 17-19 sample group
seems to show stronger reliability in terms of goodness of
fit, this result may be due to the fact that we tried to test on
our sample the model tested in adulthood by
Miyake et al.
(2000)
and
Miyake and Friedman (2012)
, given that it
makes sense that data registered in younger adolescents
cluster in a model that is not as stable as the first one. R2
show that also factorial weights in terms of variance change
in the two models, this suggest a different organization in
terms the influence that specific abilities may gain along the
years.
5.Conclusions
Data confirm the separability of inhibition and WM, even
though these abilities result as well strongly related. This
find is consistent with results reported in previous research
in adulthood
(Miyake et al., 2000; Miyake & Friedman,
2012; Diamond, 2013)
. Moreover the model is invariant
across gender and age.
Data also seem to show a unitary dimension of inhibition in
adolescence, this find is not consistent with previous
research in childhood
(Gandolfi et al. 2014)
and in
adulthood
(Miyake et al., 2000; Miyake & Friedman, 2012;
Diamond; 2013)
. This find may be due to the specific task
implied in this study: most of the rely mainly on the ability
to inhibit responses (behavioural inhibition) and not on
other kind of inhibition
(see Nigg, 2000; Miyake et al.,
2000; 2012)
.