Professional Documents
Culture Documents
SelectiveStopping DS
SelectiveStopping DS
Manuscript ID Draft
Re
Manuscript Type: Research Article
developmentalscience@wiley.com
Page 1 of 31 Developmental Science
1
2
3 1 Research highlights
4
5
6 2 ● Although the development of global stopping has been investigated, little is
7 3 known about the development of selective stopping.
8
9
4 ● Here we investigate the qualitative and quantitative developmental changes in
10
11 5 stimulus-selective stopping in a large-scale and wide age-range cross-sectional
12
6 study.
13
14
15 7 ● The ability to stop selectively to some stimuli but not others instead of stopping
16 8 indiscriminately to all stimuli fully matures by early preadolescence.
17
18
19
9 ● The efficiency or speed of stopping (SSRT) continues to mature until young
20 10 adulthood both for global and selective implementations of stopping.
Re
21
22 11
23
vie
24
25 12
26
27 13
w
28
29
30
Co
31
32
33
34
py
35
36
37
38
On
39
40
41
42
ly
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
developmentalscience@wiley.com
Developmental Science Page 2 of 31
1
2
3 14 Abstract
4
5
6 15 Although progress has been made in elucidating the behavioural and neural
7 16 development of global stopping across the lifespan, little is known about the development
8
9 17 of selective stopping. This more complex form of inhibitory control is required in real-
10 18 world situations where ongoing responses must be inhibited to certain stimuli but not
11
12 19 others, and can be assessed in laboratory settings using a stimulus-selective stopping task.
13 20 Here we used this task to investigate the qualitative and quantitative developmental
14
15
21 changes in selective stopping in a large-scale cross-sectional study with three different
16 22 age groups (children, preadolescents and young adults). We found that the ability to stop
17
18
23 a response selectively to some stimuli (i.e., use a selective strategy) rather than non-
19 24 selectively to all presented stimuli (i.e., use a global, non-selective strategy) is fully
20
Re
21
25 mature by early preadolescence, and remains stable afterwards at least until young
22 26 adulthood. By contrast, the efficiency or speed of stopping (indexed by a shorter stop-
23
vie
24
27 signal reaction time or SSRT) continues to mature throughout adolescence until young
25 28 adulthood, both for global and selective implementations of stopping. We also provide
26
27
29 some preliminary findings regarding which other task variables beyond the strategy and
w
28 30 SSRT predicted age group status. Premature responding (an index of “waiting
29
30
31 impulsivity”) and post-ignore slowing (an index of conflict detection and proactive
Co
31 32 control) were among the most relevant predictors in discriminating between age groups.
32
33
33 Although present results need to be confirmed and extended in longitudinal studies, they
34 34 provide new insights into the development of a relevant form of inhibitory control.
py
35
36 35
37
38
On
38
43
44
45 39
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
developmentalscience@wiley.com
Page 3 of 31 Developmental Science
1
2
3 40 1. INTRODUCTION
4
5
6 41 Inhibition is a domain-general cognitive construct that encompasses the ability to
7 42 overcome internal or external drives and to control behaviour and lower-level mental
8
9 43 processes needed to remain goal-directed (Diamond, 2013). It can be further
10 44 subcategorised into cognitive inhibition (the stopping or overriding of a mental process
11
12 45 such as memories, thoughts or perceptions) and behavioural inhibition (the stopping or
13 46 overriding of a manifest behaviour) (Bari & Robbins, 2013; Hung, Gaillard, Yarmak, &
14
15
47 Arsalidou, 2018; MacLeod, 2007). Here we investigate the inhibition of inappropriate, no
16 48 longer required responses, a form of behavioural inhibition (Bari & Robbins, 2013;
17
18
49 Dalley & Robbins, 2017; MacKillop et al., 2016). This ability, called response inhibition,
19 50 is impaired in a wide range of neurodevelopmental and psychological disorders emerging
20
Re
21
51 in childhood and adolescence such as attention deficit/hyperactivity disorder (ADHD),
22 52 oppositional defiant and conduct disorders, eating disorders and substance abuse
23
vie
24
53 disorders (Bartholdy, Dalton, O’Daly, Campbell, & Schmidt, 2016; Groman, James, &
25 54 Jentsch, 2009; López-Martín, Albert, Fernández-Jaén, & Carretié, 2015; Schachar et al.,
26
27
55 2007). Moreover, inefficient response inhibition has been associated with poor academic
w
28 56 and occupational outcomes, as well as with a broad range of health and behavioural
29
30
57 problems (Diamond, 2013; Houben, Nederkoorn, & Jansen, 2014; Morgan, Farkas,
Co
35 60 stopping, which involves the rapid cancellation of all motor actions in response to a single
36 61 stimulus (Aron, 2011; Wessel & Aron, 2017; Wiecki & Frank, 2013). Research using
37
38 62 standard stop-signal tasks has been crucial to elucidate the neural and behavioural
On
39 63 mechanisms underlying this important executive function ability. At the neural level,
40
41 64 global stopping is thought to be mediated by the hyperdirect pathway that connects the
42
ly
65 inferior frontal gyrus and/or pre-supplementary motor area directly to the subthalamic
43
44 66 nucleus (Chen et al., 2020; Narayanan, Wessel, & Greenlee, 2020). This is the fastest way
45 67 to stop, but leads to global cancellation of thalamocortical motor programs. In everyday
46
47 68 life, however, situations where multiple response tendencies need to be globally stopped
48 69 are rare. Instead, real-world scenarios with multiple stimuli where some responses must
49
50 70 be suppressed while others must continue to be executed are much more frequent. Thus,
51 71 there has been an increasing interest in investigating the selective mechanisms of stopping
52
53 72 using more complex and ecologically valid versions of the stop-signal task (Aron, 2011).
54 73 Stopping selectively probably involves brain regions beyond those activated in global
55
56 74 stopping, including the striatum, the dorsolateral prefrontal cortex and the superior
57 75 parietal cortex (Aron, 2011; Majid, Cai, Corey-Bloom, & Aron, 2013; Sánchez-Carmona,
58
59 76 Albert, & Hinojosa, 2016; Sánchez-Carmona, Santaniello, Capilla, Hinojosa, & Albert,
60 77 2019). Indeed, it has been proposed that selective stopping may be implemented via the
3
developmentalscience@wiley.com
Developmental Science Page 4 of 31
1
2
3 78 indirect fronto-striatal pathway rather than the hyperdirect fronto-subthalamic nucleus
4
5 79 pathway (Aron, 2011; Majid et al., 2013; Wiecki & Frank, 2013).
6
7 80 Selective stopping refers both to the ability to interrupt certain responses but not
8
9 81 others (response-selective stopping), and to the ability to cancel the response to certain
10 82 stimuli but not others (stimulus-selective stopping) (Verbruggen & Logan, 2017). The
11
12 83 latter is the focus of the present study. It can be experimentally investigated using a
13 84 stimulus-selective stopping task, which is a modification of the traditional stop-signal
14
15
85 task used to examine global stopping. In the stop-signal task, participants are required to
16 86 respond as fast as possible to a frequent stimulus (“go”), but to cancel the response-in-
17
18
87 progress whenever another infrequent stimulus (“stop” signal) is presented with a variable
19 88 delay shortly after the go stimulus. Selectivity is introduced into the task through a third
20
Re
21
89 stimulus (the so-called “ignore” or “continue” signal), which is also presented just after
22 90 the go stimulus with the same frequency and delay as the stop signal. Participants are
23
vie
24
91 instructed to ignore this new stimulus, and thus to complete the previously initiated motor
25 92 response triggered by the go stimulus.
26
27 93 Most of the studies that have investigated selective forms of stopping using the
w
28
29 94 stimulus-selective stopping task assumed that all participants interrupted their responses
30 95 selectively to stop signals (for a review see Bissett & Logan, 2014). However, evidence
Co
31
32 96 from behavioural, electrophysiological and neuroimaging studies revealed that
33 97 individuals use different strategies to perform the task (Bissett & Logan, 2014; Sánchez-
34
py
35 98 Carmona et al., 2016; Sánchez-Carmona et al., 2019; Sebastian et al., 2017). Whereas
36 99 some implement the stopping process selectively to stop but not to ignore signals
37
38 100 (independent and dependent Discriminate then Stop strategies, iDtS and dDtS), others
On
39 101 stop indiscriminately to both signals and then resume their response when the signal
40
41 102 presented is an ignore (Stop then Discriminate strategy, StD). Within this framework,
42
ly
103 researchers have begun to determine what factors modulate the adoption of selective
43
44 104 versus non-selective strategies. Thus, task-related factors such as signal discrimination
45 105 difficulty and probability of occurrence of the signals, and individual-related factors such
46
47 106 as genetic variations in the dopaminergic system, have been shown to influence strategy
48 107 adoption in adults (Bissett & Logan, 2014; Rincón-Pérez et al., 2020; Sánchez-Carmona,
49
50 108 Rincón-Pérez, López-Martín, Albert, & Hinojosa, 2021; Verbruggen & Logan, 2015).
51
52 109 To our knowledge, only two preliminary studies have examined age-related
53
54 110 differences in the adoption of stopping strategies during a selective stopping task (Hsieh
55 111 & Lin, 2017; Rincón-Pérez et al., 2021). Whereas the former observed a similar pattern
56
57 112 of strategy use between younger adults (20-30 years) and older adults (61-76 years), the
58 113 latter found important changes in strategy adoption in middle childhood. Specifically, a
59
60 114 greater use of selective stopping strategies and less use of the global, non-selective
4
developmentalscience@wiley.com
Page 5 of 31 Developmental Science
1
2
3 115 strategy in children aged 10-11 years compared to children aged 6-7 years was found.
4
5 116 These results suggest that middle childhood may be a critical period of substantial
6 117 development in an individual’s ability to stop selectively. Still, further studies with larger
7
8 118 sample sizes and on a wide age-range from childhood to adulthood are needed to
9 119 substantiate and extend these preliminary findings. In this sense, an interesting question
10
11 120 that remains unexplored is whether the ability to stop selectively continues to develop
12 121 beyond middle childhood up until young adulthood.
13
14
15
122 Beyond identifying the strategy used to solve the task, the stimulus-selective stopping
16 123 task allows for the estimation of the latency of the stop process (stop signal reaction time,
17
18
124 SSRT). Longer SSRTs are indicative of poor, less efficient response inhibition and have
19 125 been found in children, adolescents and adults with a range of disorders characterized by
20
Re
21
126 impulsive symptoms, including ADHD and substance abuse disorders (Crosbie et al.,
22 127 2013; Rømer Thomsen, Blom Osterland, Hesse, & Feldstein Ewing, 2018; Smith,
23
vie
24
128 Mattick, Jamadar, & Iredale, 2014; van Hulst et al., 2018). The SSRT can be derived on
25 129 the basis of the horse race model (Logan, Cowan, & Davis, 1984), which posits that the
26
27
130 go process triggered by the onset of the go stimulus and the stop process triggered by the
w
28 131 onset of the stop signal compete regarding finishing time. When stopping finishes before
29
30
132 going, the already initiated response is successfully suppressed, but when going finishes
Co
31 133 before stopping, the response cannot be successfully interrupted. Importantly, this model
32
33
134 assumes stochastic (trial-by-trial SSRT variability is unrelated to trial-by-trial go reaction
34 135 time [RT] variability) and context independence (Go RT distribution is the same for stop
py
35
36
136 and go trials; for details, see Bissett, Jones, Poldrack & Logan, 2021; Verbruggen &
37 137 Logan, 2009).
38
On
39 138 It should be noted, however, that the independence assumption of the horse race
40
41 139 model is not met in a group of participants who use a particular selective stopping strategy
42
ly
140 (those adopting the so-called dependent DtS strategy, dDtS: Bissett & Logan, 2014; see
43
44 141 also Verbruggen & Logan, 2015). In these cases, the SSRT cannot be calculated through
45 142 the standard procedure of using the go RT distribution. By contrast, using the ignore RT
46
47 143 distribution has been recommended, taking into account that the validity of this procedure
48 144 needs to be confirmed (Bissett & Logan, 2014). Moreover, the SSRT for an individual
49
50 145 who uses the non-selective, global stopping (StD) strategy does not include the time
51 146 needed to discriminate between stop and ignore signals, because perceptual
52
53 147 discrimination occurs after the response has been interrupted. In contrast, the SSRT for
54 148 an individual who adopts a selective stopping strategy (whether or not it meets the
55
56 149 independence assumption) includes both the signal discrimination and stopping
57 150 processes. This remarks the importance of identifying and categorizing the strategy
58
59 151 adopted by each participant in developmental research on selective stopping, so that
60 152 results regarding the length of the SSRT are not confounded by differences between
5
developmentalscience@wiley.com
Developmental Science Page 6 of 31
1
2
3 153 participants in the dependency/independency of the go and stop process and in the chain
4
5 154 of processes involved in stopping during selective stop-signal tasks.
6
7 155 Developmental studies using traditional stop-signal tasks have found that SSRT
8
9 156 becomes faster (more efficient) with increasing age throughout childhood and at a lower
10 157 rate during adolescence (Tillman, Thorell, Brocki, & Bohlin, 2007; van de Laar, van den
11
12 158 Wildenberg, van Boxtel, & van der Molen, 2011; Williams, Ponesse, Schachar, Logan,
13 159 & Tannock, 1999). The fastest SSRT values have been observed in early adulthood,
14
15
160 suggesting that mechanisms supporting global stopping are not fully efficient until this
16 161 age period. These developmental changes in SSRT fit well with the results of most cross-
17
18
162 sectional and longitudinal studies investigating the maturation of simple or global
19 163 response inhibition, which suggest increases from childhood to young adulthood in the
20
Re
21
164 activation of key cortical regions of the global stopping network, including inferior frontal
22 165 gyrus and pre-SMA (Bunge, Dudukovic, Thomason, Vaidya, & Gabrieli, 2002; Cope et
23
vie
24
166 al., 2020; Durston et al., 2002; Rubia, Smith, Taylor, & Brammer, 2007). It should be
25 167 noted, however, that decreases and even nonlinear developmental patterns in inhibitory
26
27
168 control activation have also been found (Ordaz, Foran, Velanova, & Luna, 2013; Paulsen,
w
28 169 Hallquist, Geier, & Luna, 2015). Developmental research on selective stopping is much
29
30
170 more scarce and restricted to behavioural studies (Bedard et al., 2002; Kray, Kipp, &
Co
31 171 Karbach, 2009; van de Laar et al., 2011). These studies suggest a decrease in SSRT from
32
33
172 childhood to young adulthood, where selective stopping has been associated with longer
34 173 SSRTs and with a slower maturation than global stopping (van de Laar et al., 2011). These
py
35
36
174 conclusions, however, need to be substantiated further due to small sample sizes and
37 175 especially because it was assumed that all participants had performed the task using the
38
On
176 same strategy (i.e., stopping selectively employing the so-called iDtS).
39
40
41 177 The purpose of this cross-sectional study was to examine the qualitative and
42
ly
6
developmentalscience@wiley.com
Page 7 of 31 Developmental Science
1
2
3 191 SSRT) until young adulthood, both for global and selective stopping (Tillman et al., 2007;
4
5 192 van de Laar et al., 2011; Williams et al., 1999). We were interested in further testing
6 193 whether the magnitude of the reduction in SSRT differs between global and selective
7
8 194 implementations of stopping. Finally, we explored secondary task-related factors beyond
9 195 strategy and SSRT that could predict age group membership, including those associated
10
11 196 with attention-related processes and with other forms of inhibitory control such as
12 197 premature responding (“waiting impulsivity”) or proactive stopping (Aron, 2011; Voon
13
14 198 et al., 2014).
15
16 199
17
18 200 2. METHOD
19
20
Re
24
25
203 preadolescents aged 10-11 were recruited from regular local schools in Madrid (Spain),
26 204 along with 117 young adults aged 18-20 from a local university (Universidad Autónoma
27
w
28
205 de Madrid, UAM). Participants had normal or corrected-to-normal vision, and no history
29 206 of psychiatric, neurological or sensory impairment. None of them had repeated a grade,
30
nor did they need any curricular or special educational adaptations. Written informed
Co
207
31
32 208 consent to participate in the study was obtained from parents (with the child giving assent)
33
34
209 when the participant was underage, and from the participants themselves in the case of
py
35 210 young adults. The study was conducted according to the guidelines of the Declaration of
36
37
211 Helsinki, and was approved by the Research Ethics Committee of the UAM.
38
On
39 212 Forty-three participants were excluded from analyses due to one or several of the
40 213 following reasons: the strategy adopted to complete the task could not be identified (n=43;
41
42 214 see details below), the probability of responding on the stop trials (respond|signal) was
ly
43 215 higher than 0.75 or lower than 0.25 (n=5) and/or there was a high probability of omissions
44
45 216 (>30% of go and/or ignore trials; n=3). The final sample therefore consisted of 271
46 217 participants: 83 children aged 6-7 years (mean age±SD, 6.58±0.5), 86 preadolescents
47
48 218 aged 10-11 (10.13±0.36), and 102 young adults aged 18-20 (18.94±0.78). Gender
49 219 distribution did not differ between groups (chi-squared test; χ2=1.54, df=2, p=0.46).
50
51
52 220 2.2. Stimulus-selective stopping task
53
54 221 Participants completed a single run of a stimulus-selective stop-signal task composed
55
222 of three different stimuli: go, stop and ignore (see Figure 1). The go stimulus was
56
57 223 presented for 1500 milliseconds and consisted of a white arrow on a black background
58
224 pointing downwards. All task trials contained a go stimulus to which participants should
59
60 225 press the spacebar on a keyboard with the index finger of their dominant hand as fast as
7
developmentalscience@wiley.com
Developmental Science Page 8 of 31
1
2
3 226 possible. Trials where only the go stimulus appeared are called go trials, and represent
4
5 227 60% of the total trials. The go stimulus was followed by a red diamond (stop signal)
6 228 appearing around the go stimulus after a variable delay (stop signal delay, SSD) on
7
8 229 another 20% of the trials, called stop trials. In these, participants were asked to try to
9 230 cancel their motor response as soon as the stop signal appeared. In the remaining 20% of
10
11 231 the trials, called ignore or continue trials, the go stimulus was followed by a green square
12 232 (ignore signal) that likewise appeared around the go stimulus after a variable delay (ignore
13
14 233 signal delay, ISD). In these, participants were asked to press the spacebar even if they
15 234 saw the green square, and thus complete their already initiated motor response triggered
16
17 235 by the go stimulus. We used bright colours because salient and easily detectable signals,
18 236 such as the ones we employed here, minimize the influence of perceptual discrimination
19
20 237 processes on the SSRT and the probability that differences between groups can be
Re
21 238 attributed to them (Verbruggen et al., 2019). Salient signals also facilitate the adoption of
22
23 239 selective stopping strategies as opposed to a global non-selective strategy (Sánchez-
vie
24 240 Carmona et al., 2021) and might also minimize the probability of “trigger failures” on
25
26 241 stop and ignore trials (Verbruggen et al., 2019). Overall, the task consisted of 230 trials
27 242 (138 go, 46 stop and 46 ignore) that lasted approximately 8.6 minutes. The order of trials
w
28
29 243 was fully randomized.
30
Co
35 246 or 1000 ms with equal probability (the inter-trial interval, ITI). Then the go stimulus was
36 247 presented in all trials for 1500 milliseconds. Thus, the total trial duration was either 2000
37
38 248 or 2500 milliseconds. After the appearance of the go stimulus (and while it was still on-
On
39 249 screen), the stop signal was presented after the SSD on 20% of the trials. SSD was initially
40
41 250 set to 250 ms and then was adjusted dynamically from one stop trial to the next with a
42
ly
8
developmentalscience@wiley.com
Page 9 of 31 Developmental Science
1
2
3 263 (Verbruggen et al., 2019). Before the stimulus-selective stopping task, participants
4
5 264 performed a first practice block of 20 trials with only go stimuli (i.e., without stop or
6 265 ignore signals) to highlight the relevance of the go component of the stop-signal task
7
8 266 (Verbruggen et al., 2019), and then a second practice block of 30 trials with the same
9 267 characteristics described above to ensure understanding of the stimulus-selective stopping
10
11 268 task. It was designed and implemented in MATLAB using Psychtoolbox
12 269 (www.psychtoolbox.org) from the script called STOP-IT generated by Verbruggen,
13
14 270 Logan and Stevens (2008).
15
16 271 2.3. Task performance
17
18 272 2.3.1. Main task variables
19
20
Re
21 273 Stopping strategies. We were firstly interested in identifying the strategies used by
22 274 participants for solving the stimulus-selective stopping task. Each individual’s strategy
23
vie
24 275 was calculated by comparing their mean go RT with ignore RT and with failed stop RT
25 276 through two independent t tests (see Table 1), following the method proposed by Bissett
26
27 277 and Logan (2014). These comparisons result in three different possible strategies (two
w
28 278 selective and one global, non-selective): independent Discriminate then Stop (iDtS),
29
30 279 dependent Discriminate then Stop (dDtS) and Stop then Discriminate (StD). The already
Co
31 280 initiated motor response is selectively inhibited in both the iDtS and dDtS, whereas it is
32
33 281 globally inhibited in the StD. Bayes factor (BF) was employed to compare the evidence
34 for and against the null hypothesis without bias (Rouder, Speckman, Sun, Morey, &
py
282
35
36 283 Iverson, 2009). A BF of 1 indicates that there is no difference between RTs, and a BF≠1
37 284 means that there is a difference. We used Rouder’s Bayes factor calculator on the
38
On
1
2
3 300 signals in the dDtS that violates the context independence assumption, and here the SSRT
4
5 301 cannot be estimated reliably using the go RT distribution (Bissett et al., 2021; Bissett &
6 302 Logan, 2014). These authors also proposed that if responding slows equally on both stop
7
8 303 and ignore trials, the SSRT could be estimated through the integration method based on
9 304 the ignore RT distribution. This is what we did for this strategy (dDtS), but it is worth
10
11 305 noting that this solution is only valid as long as the assumption of equal slowing holds.
12 306 Therefore, the SSRTs computed using the ignore RT distribution should be interpreted
13
14 307 with caution.
15
16 308 2.3.2. Secondary task measures
17
18 309 Premature responses. Those responses emitted at any point in the ITI before the onset
19
20 310 of the go stimulus on a go trial. Importantly, we discounted those responses considered
Re
21 311 as atypically slow: if there was a response before the go stimulus but at the same time no
22
23 312 response was emitted in the previous trial, we considered this sequence as a slow response
vie
24 313 to the previous trial instead of a premature response to the current trial.
25
26
27 314 Post-signal behavioural adjustments. We compared the RT of go trials that occurred
w
28 315 either after a failed stop trial (post-stop error) or after a correct ignore trial (post-correct
29
30 316 ignore), with the RT of go trials that happened after a correct go trial (this “go-post-go”
Co
31 317 trial being the nearest one that precedes the error). This method prevents confounds
32
33 318 derived from participants’ changes in ability, motivation or response caution throughout
34 the task (Dutilh, van Ravenzwaaij, et al., 2012).
py
319
35
36
37
320 Intra-individual variability. A participant's variability in RT across trials within the
38 321 task. It was estimated by fitting the participants’ RTs to an ex-Gaussian distribution,
On
39
322 which is formed by the convolution of a normal and an exponential distribution with three
40
41 323 parameters: the mean and the standard deviation of the normal component (mu and sigma,
42
ly
324 respectively) and the mean and the standard deviation of the exponential component (both
43
44 325 described by tau). Since mu is merely a reflection of average speed, the intra-individual
45
326 variability is characterized particularly by sigma and tau. Sigma reflects the variability in
46
47 327 fast (normal) responses of the distribution, and tau reflects variability in the extreme slow
48
328 responses of the distribution (van Belle, van Raalten, et al., 2015). Of note, the ex-
49
50 329 Gaussian distribution was not fitted to the complete go RT distribution, because stop-go
51
330 and ignore-go trial sequences would introduce proactive/post-signal adjustments that
52
53 331 could bias response variability. Therefore, for this analysis we used only those go trials
54
332 that followed another go trial (go-post-go). This fitting process was analysed in the
55
56 333 DISTRIB toolbox for MATLAB (Lacouture & Cousineau, 2008).
57
58 334 Go and ignore omissions. A go omission is a go trial without a response, while an
59
60 335 ignore omission is an ignore trial without a response.
10
developmentalscience@wiley.com
Page 11 of 31 Developmental Science
1
2
3 336 Go RT slope. We computed the slope of the linear least squares fitting model with
4
5 337 respect to a subset of go RTs. Concretely, we selected only those go trials that followed
6 338 another go trial (go-post-go), thus trying to avoid the speed adjustments that probably
7
8 339 occur after stop and ignore signals. A slope value close to zero would suggest the
9 340 maintenance of speed throughout the task. However, the more different to zero, the more
10
11 341 it would suggest a speed increase (negative slope value) or a speed decrease (positive
12 342 slope value), indexing the participant’s ability to sustain attention throughout the task.
13
14
15
343 Proactive stopping. Following Verbruggen et al. (2019) we included a practice block
16 344 without stop and ignore signals, with two goals: i) to emphasize the importance of the go
17
18
345 task component to discourage waiting and ii) to obtain a measure of proactive stopping.
19 346 Since task instructions state that the stop signal may occur at any time, participants
20
Re
21
347 inevitably anticipate having to stop at each trial. Consequently, go trials, influenced by
22 348 the awareness of the presence of a stop signal, would show a proactive lengthening of
23
vie
24
349 RTs (Wessel, 2018). We can measure this proactive inhibition for each participant by
25 350 taking the difference between go trials included in the stimulus-selective stop-signal task
26
27
351 and go trials included in the practice block. Mean RT of go trials during the task was
w
31
32
33 354 Contingency tables were employed to examine potential differences in the use of
34 different strategies to perform the task among the three developmental age groups
py
355
35
36 356 (children, preadolescents and young adults). Since we expected certain strategies to be
37 357 chosen by fewer participants, the Freeman-Halton extension of Fisher’s exact test was
38
On
39 358 used instead of a chi-squared test (Freeman & Halton, 1951). Cramer’s V was reported as
40 359 a measure of effect size. Follow-up Z-tests for independent proportions with Bonferroni
41
42 360 correction for multiple comparisons were performed to detect significant differences
ly
43 361 among groups. SSRTs were then submitted to a 3x3 ANOVA with age group (children,
44
45 362 preadolescents, adults) and strategy (StD, iDtS, dDtS) as between-subjects factors in order
46 363 to assess differences between developmental age groups and a potential interaction
47
48 364 between strategy and age group. Of note, the SSRT was estimated as a function of the
49 365 strategy used by each participant, as previously described. Significant main effects and
50
51 366 interactions were further investigated using post-hoc t tests with Bonferroni correction
52 367 for multiple comparisons. Effect sizes were measured using partial eta-square (η2p) for F
53
54 368 values and Cohen’s d for t values.
55
56 369 Additionally, two multinomial regression analyses were performed to identify which
57
370 factors predicted age group membership. The first one included the two theoretical
58
59 371 predictors on which this study focused: the strategy used to perform the task and the
60
372 SSRT. The second one was an exploratory forward stepwise multinomial regression. This
11
developmentalscience@wiley.com
Developmental Science Page 12 of 31
1
2
3 373 model included the following potential predictors in addition to strategy and SSRT:
4
5 374 premature responding, post-stop error slowing, post-ignore success slowing, sigma, tau,
6 375 go omissions, ignore omissions, Go RT slope and proactive stopping. This data-driven
7
8 376 regression was intended to identify potential predictors of age group membership that
9 377 should be considered in future developmental studies on selective stopping. We used a
10
11 378 forward entry strategy using likelihood ratio test until no omitted variable contributed
12 379 significantly to predict age groups. We checked that the final model was the one with the
13
14 380 smallest AIC. There were no problems of multicollinearity (the lowest tolerance was .384
15 381 for go omissions predictor variable) nor overdispersed residuals were found. All reported
16
17 382 p-values are two-tailed. Analyses were conducted with SPSS 26 unless mentioned
18 383 otherwise.
19
20
Re
21
384 Prior to statistical testing, data were examined for outliers. For each task measure and
22 385 each age group, outliers were defined as those data points further than 1.5 times from the
23
vie
24
386 upper or the lower limit of the interquartile range. They were then replaced by the mean
25 387 ± 2 standard deviations of the corresponding group. The number of outliers only
26
27
388 represented 0.005% (13/2710) of the data set.
w
28
29 389
30
Co
31 390 3. RESULTS
32
33
34
391 We first identified the strategy used to perform the stimulus-selective stopping task
py
35 392 by comparing, for each participant individually, correct go, failed stop and correct ignore
36
37
393 RTs. To do this, we employed the method introduced by Bissett and Logan (2014), which
38 394 has been previously applied in the study of the neural, genetic and behavioural correlates
On
39
395 of selective stopping (Bissett & Logan, 2014; Rincón-Pérez et al., 2020; Sánchez-
40
41 396 Carmona et al., 2016; Sánchez-Carmona et al., 2019; Sebastian et al., 2017). Overall, we
42
ly
397 found that 52 participants used the global, non-selective stopping strategy (StD), whereas
43
44 398 219 adopted selective stopping strategies (199 iDtS and 20 dDtS). The strategy could not
45
399 be correctly classified in 43 participants (14 children, 14 preadolescents and 15 young
46
47 400 adults; 13.7% of the total sample). As described in the Method section, these participants
48
401 were excluded from further analysis. Means and standard deviations of go, failed stop and
49
50 402 ignore RTs, as well as of other variables of the task including SSD, SSRT and the
51
403 probability of responding on failed stop trials, for each group and strategy separately can
52
53 404 be found in Supplementary Material (Table S1), as recommended by consensus
54
405 guidelines on the use of stop-signal tasks (Verbruggen et al., 2019). Of note, the
55
56 406 probability of responding on stop trials [i.e., stopping failures, p(respond|signal)] did not
57
407 differ between age groups (F(2, 262)=0.58, p=0.56) or with regard to the age group x
58
59 408 strategy interaction (F(4, 262)=1.38, p=0.24). These results show that the adaptive-
60
409 algorithm was successful, and that any possible between-groups differences related to
12
developmentalscience@wiley.com
Page 13 of 31 Developmental Science
1
2
3 410 stopping failures, such as emotional frustration and error monitoring, were controlled (Li,
4
5 411 Huang, Constable, & Sinha, 2006; Sánchez-Carmona et al., 2016). By contrast, as
6 412 expected, the mean SSD differed between age groups (F(2, 262)=9.61, p<0.001), but not
7
8 413 regarding the interaction (F(4, 262)=2.18, p=0.07).
9
10 414 Contingency tables were then made to examine potential differences in the use of
11
12 415 strategies to perform the stimulus-selective stopping task among children, preadolescents
13 416 and young adults. Specifically, we examined whether the proportion of adoption of each
14
15
417 strategy differed among the three groups. A significant association between strategy and
16 418 age group was found (Fisher’s exact test=18.85, p=0.001, Cramer’s V=0.19; Figure 2 and
17
18
419 Supplementary Material, Table S2). Z-tests for independent proportions revealed that the
19 420 proportion of participants using the StD strategy was significantly greater in children
20
Re
21
421 (34.9%, n=29) than in preadolescents (11.6%, n=10) and young adults (12.7%, n=13). No
22 422 differences in its use were found between preadolescents and adults. By contrast, the
23
vie
24
423 proportion of participants using the iDtS strategy was significantly lower in children
25 424 (56.6%, n=47) than in preadolescents (81.4%, n=70) and young adults (80.4%, n=82).
26
27
425 Again, no differences were observed between preadolescents and young adults in the
w
28 426 adoption of this selective strategy. The three age groups did not differ significantly in the
29
30
427 use of the dDtS strategy (children: 8.4%, n=7, preadolescents: 7%, n=6; young adults:
Co
35
36 430 To assess differences between age groups and a potential interaction between strategy
37 431 and age group in SSRTs, we conducted a 3x3 ANOVA with age group (children,
38
On
39 432 preadolescents, adults) and strategy (StD, iDtS, dDtS) as between-subjects factors. The
40 433 main effects of both age group (F(2,262)=30.47, p<0.001, ƞ2p=0.19) and strategy (F(2,
41
42 434 262)=20.85, p<0.001, ƞ2p=0.14) were significant. Post-hoc t-tests with Bonferroni
ly
43 435 correction indicated greater SSRT for children compared with preadolescents (p<0.001;
44
45 436 Cohen’s d=0.63) and young adults (p<0.001; Cohen’s d=1.44), as well as preadolescents
46 437 versus young adults (p<0.001; Cohen’s d=0.89) as seen in Figure 2, irrespective of the
47
48 438 strategy adopted. With respect to the main effect of strategy, post-hoc t test with
49 439 Bonferroni correction revealed greater SSRT for dDtS compared with iDtS (p<0.001;
50
51 440 Cohen’s d=1.28) and StD (p<0.001; Cohen’s d=1.08) with no difference between iDtS
52 441 and StD (p=0.87, Cohen’s d=0.03), irrespective of age group. The interaction of age group
53
54 442 and strategy was not significant (F(4,262)=1.49, p=0.20): there was a reduction of the
55 443 SSRT with age in all strategies (lower SSRT values in the older age groups). However, it
56
57 444 should be noted that further post-hoc ANOVAs conducted for each strategy separately
58 445 revealed a larger effect size for SSRT differences between groups in the dDtS (difference
59
60 446 of 132.94 ms between young adults and children, F(2,17)=3.63, p<0.05, ƞ2p=0.29) and
13
developmentalscience@wiley.com
Developmental Science Page 14 of 31
1
2
3 447 the iDtS strategy (difference of 108.50 ms between young adults and children,
4
5 448 F(2,196)=53.71, p<0.001, ƞ2p=0.35), compared to the StD strategy (difference of 80.38
6 449 ms between young adults and children, F(2,49)=6.07, p<0.05, ƞ2p=0.19).
7
8
9 450 Finally, we ran multinomial regressions to study the effect of each task-related
10 451 predictor on developmental age groups while adjusting for the rest of predictors. We first
11
12 452 conducted a theoretically guided multinomial regression where the dependent variable
13 453 was age group with three levels (children, preadolescents and young adults). The
14
15
454 predictor variables were strategy with three levels (iDtS, dDtS, and StD; this one was the
16 455 reference category) and SSRT[1]. Overall, the model fits significantly better than the null
17
18
456 model (LRT(6)=116.51, p<0.001; AIC=492.67), Nagelkerke’s R-square=0.39. Table 2
19 457 shows the significant coefficients. A reduction in SSRT was associated with a higher
20
Re
21
458 likelihood of being in the young adult group relative to the children group (the reference
22 459 group). In addition, the adoption of iDtS strategy (relative to StD) increases the odds of
23
vie
24
460 being in the young adult relative to the children group by a factor of 32.22 adjusting for
25 461 the SSRT variable. The adoption of dDtS strategy (relative to StD) increases the odds of
26
27
462 being in the young adult group relative to the children group by a factor of 6.43. When
w
28 463 comparing the determinants between children and preadolescent groups, the SSRT
29
30
464 predictor was also significant. A reduction in SSRT was associated with a higher
Co
31 465 likelihood to be in the preadolescent group relative to the children group. Similarly, the
32
33
466 adoption of iDtS strategy (relative to StD) increases the odds of being in the preadolescent
34 467 group relative to the children group by a factor of 10.24, and the adoption of dDtS strategy
py
35
36
468 (relative to StD) increases the odds of being in the preadolescent group relative to the
37 469 children group (OR=5.22). Finally, when comparing the determinants between
38
On
470 preadolescent and young adult groups, only the SSRT predictor (not strategy) was
39
40 471 significant. A reduction in SSRT was also associated with a higher likelihood to be in the
41
472 young adult relative to the preadolescent group (the reference group). The model
42
ly
43 473 classifies 61.1% correct (vs. 33.6% hit rate, therefore improving the correct classification
44
474 of cases by 27.5%).
45
46
47 475 We then carried out an exploratory, data-driven multinomial logistic regression with
48
476 age group as the dependent variable. Both secondary task measures, SSRT and strategy
49
50 477 were included as predictors. Likelihood ratio test was the criterion used to include
51
478 variables step-by-step. The final model included six significant predictors. The first
52
53 479 predictor included in the stepwise model was premature responses (as it was the predictor
54
480 that most reduced the misfit with respect to the null model), followed by SSRT, the
55
56 481 adopted strategy, post-ignore slowing, ignore omissions and tau. In contrast, five
57
58
59 1 The interaction term between both predictors was included but a non-significant interaction effect was
60
found (LRT(4) = 7.628, p = 0.106). Therefore, only main effects were included in the model.
14
developmentalscience@wiley.com
Page 15 of 31 Developmental Science
1
2
3 482 predictors did not contribute to explaining age groups: post-stop error slowing, sigma, go
4
5 483 omission, Go RT slope and proactive stopping. The final model correctly predicts 78.5%
6 484 of the age group of the subjects (Nagelkerke’s R-square=0.75) and was the model with
7
8 485 the lowest AIC (311.084 vs 561.511 of the null model). Of note, an exploratory model
9 486 like this might show problems of overfitting, so a cross-validated sample is needed in
10
11 487 future studies. However, this is a preliminary step to examine which task-related factors
12 488 predict development age group status. Regression coefficients in the order in which the
13
14 489 variables were included can be found in the Supplementary Material, Table S3.
15
16 490
17
18
19
491 4. DISCUSSION
20
Re
24 494 participants) and wide age-range (from children aged 6-7 to young adults aged 18-20)
25 495 cross-sectional study. At the qualitative level, we observed age-related changes in the
26
27 496 proportion of the strategies used by participants to perform the stimulus-selective stop-
w
28 497 signal task. Specifically, we found an increase in the use of the independent selective
29
30 498 stopping strategy (iDtS) and a decrease in use of the global (non-selective) stopping
Co
31 499 strategy (StD) in preadolescents relative to younger children, with no further differences
32
33 500 between preadolescents and young adults. These findings suggest that by the end of
34
py
501 middle childhood (or early preadolescence), there has been a sufficient brain maturation
35
36 502 for the majority of participants to be able to adopt the independent selective stopping
37 503 strategy. Selective stopping is thought to be mediated by a network of brain areas that is
38
On
39 504 more complex than that used for global stopping, which would involve the participation
40 505 of the dorsolateral prefrontal cortex, superior parietal cortex and the striatum among
41
42 506 others (Aron, 2011; Bissett & Logan, 2014; Majid et al., 2013; Sánchez-Carmona et al.,
ly
43 507 2019). This highlights early preadolescence as the endpoint in the maturation of processes
44
45 508 that make it possible to solve the task by stopping selectively, as suggested by previous
46 509 studies (Rincón-Pérez et al., 2021) and in agreement with our hypotheses.
47
48
49 510 However, the question of when do children begin to adopt selective strategies to
50 511 achieve selective-stopping task demands needs to be addressed in future studies with
51
52 512 wider age ranges. For example, our data indicate that the selective iDtS strategy is the
53 513 most used in all age groups, even if there was a significant increase in its use in
54
55 514 preadolescents compared to children so that very few participants used the non-selective
56 515 StD strategy by early preadolescence. It thus seems clear that children at the beginning of
57
58 516 middle childhood are already able to stop selectively at least to an extent, even if the
59 517 ability is not completely mature. Therefore, the question of when selective stopping
60
15
developmentalscience@wiley.com
Developmental Science Page 16 of 31
1
2
3 518 begins to develop remains unanswered. Although there are no prior data on the early
4
5 519 development of selective stopping, inhibition in general seems to develop from early
6 520 childhood to young adulthood (Best & Miller, 2010; Cope et al., 2020). Some basic forms
7
8 521 of inhibitory control are present as early as the first year of life (Holmboe, Bonneville-
9 522 Roussy, Csibra, & Johnson, 2018), followed by a rapid development throughout the
10
11 523 preschool years (Garon, Bryson, & Smith, 2008; Wiebe, Sheffield, & Espy, 2012). In this
12 524 light, the earliest appearance of selective inhibitory abilities is yet to be identified.
13
14 525 Likewise, a deeper investigation regarding whether the use of these behavioural strategies
15 526 remains stable all across the lifespan (including aging) is also needed to complete the
16
17 527 picture. While there is evidence from cross-sectional studies pointing to a lack of
18 528 differences between younger and older adults (Hsieh & Lin, 2017), the full developmental
19
20 529 trajectory of selective stopping strategies should be confirmed by the findings of
Re
21 530 longitudinal studies. Furthermore, future research should also consider the modulatory
22
23 531 role of factors that are intrinsic to the task and that influence strategy adoption, such as
vie
24 532 difficulty in signal discrimination and variations in the frequency of stop and ignore
25
26 533 signals (Bissett & Logan, 2014; Sánchez-Carmona et al., 2021).
27
w
28 534 Of note, both the StD and the dDtS strategies were adopted by a reduced number of
29
30
535 participants in all age groups. This result is not surprising since signal discrimination was
Co
31 536 kept easy throughout the stimulus-selective stop-signal task, and prior reports have shown
32
33
537 that this scenario facilitates the adoption of the iDtS (Sánchez-Carmona et al., 2021). This
34 538 selective stopping strategy complies with the independence assumption of the horse race
py
35
36
539 model, which allows for a reliable estimation of the SSRT. Nonetheless, as we have
37 540 already noted there was a significant reduction in the use of the StD from the child to the
38
On
541 preadolescent group, without differences between preadolescents and adults. These
39
40 542 findings suggest that developmental changes in the use of this global strategy seem to
41
543 culminate in the preadolescent stage. It should also be noted that the dDtS strategy, while
42
ly
16
developmentalscience@wiley.com
Page 17 of 31 Developmental Science
1
2
3 556 Johnson, & Pascual-Leone, 2014; Humphrey & Dumontheil, 2016; Vara, Pang, Vidal,
4
5 557 Anagnostou, & Taylor, 2014). However, it should be noted that the magnitude of SSRT
6 558 reduction in young adults relative to children was greater in the selective strategies
7
8 559 compared to the global strategy. This finding suggests that changes in the speed or
9 560 efficiency of the stop process could be greater during these developmental stages for
10
11 561 selective than for global stopping. Therefore, we hypothesize that global stopping might
12 562 start developing earlier than selective stopping so that by middle childhood and
13
14 563 adolescence the gain in efficiency is less pronounced than for selective stopping. In any
15 564 case, even if the ability to stop selectively (i.e., choose a selective strategy) is mature at
16
17 565 the preadolescent period, the speed of stopping continues to mature throughout
18 566 adolescence until young adulthood both for global and selective stopping.
19
20
Re
21
567 Additionally, the main effect of strategy on the SSRT was significant. In this sense,
22 568 we found no significant differences in the length of the SSRT between the non-selective
23
vie
24
569 strategy StD and the selective strategy iDtS, irrespective of age group. This finding seems
25 570 at odds with the idea that the SSRT in the iDtS should be longer because it includes the
26
27
571 time taken to discriminate between stimuli. However, the SSRT in the dDtS strategy was
w
28 572 longer than for the other two strategies (with no difference between iDtS and StD), which
29
30
573 would suggest that the SSRT is elongated only when there is a dependence between
Co
31 574 responding and discriminating stop and ignore signals (Verbruggen & Logan, 2015).
32
33
575 Nonetheless, the estimation of the SSRT in the dDtS strategy (using the ignore RT
34 576 distribution) still lacks validation (Bissett & Logan, 2014), so these findings warrant
py
35
36
577 further confirmation from additional studies.
37
38 578 Results showing qualitative (strategy choice) and quantitative (SSRT) age-related
On
39 579 changes were further confirmed and extended by multinomial logistic regression
40
41 580 analyses. In the confirmatory analysis we found that both strategy and SSRT predicted
42
ly
17
developmentalscience@wiley.com
Developmental Science Page 18 of 31
1
2
3 593 membership: premature responses, post-ignore slowing, ignore omissions and tau. Some
4
5 594 are associated with inhibitory control processes (premature responses and post-ignore
6 595 slowing) and others with attentional processes (ignore omissions and tau). Of note, both
7
8 596 premature responses and post-ignore slowing were able to differentiate between the three
9 597 age groups, whereas ignore omissions discriminated between children and preadolescents
10
11 598 (and also between children and young adults), and tau only discriminated between
12 599 preadolescents and young adults. These results suggest that these secondary measures
13
14 600 derived from the stimulus-selective task can be considered as relevant predictors of age
15 601 group membership in future developmental studies.
16
17
18 602 The premature responses variable was one of the most relevant predictors in
19 603 discriminating between developmental age groups in the exploratory model. These
20
Re
21 604 responses signal a lack of inhibitory control since they are a measure of waiting
22 605 impulsivity, which is the inability to wait to emit responses until the stimulus appears.
23
vie
24 606 This waiting impulsivity is distinct but related to the stopping impulsivity that we have
25 607 already seen in the task through the SSRT analyses (Dalley & Robbins, 2017; Robbins &
26
27 608 Dalley, 2017; Voon, 2014; Voon et al., 2014). Developmental studies regarding
w
28 609 premature responses are scarce, but their results suggest that premature responding
29
30 610 correlates negatively with age in a sample of ADHD and control children aged 7-15
Co
31 611 (Rubia, Smith, & Taylor, 2007), while this correlation seems to disappear from age 16
32
33 612 onwards (Nord et al., 2019). In partial agreement with our results, these findings indicate
34 613 a decline in premature responding (and thus a reduction in waiting impulsivity) with age
py
35
36 614 that extends at least until well into adolescence. Our regression analyses show that
37 615 participants making less premature responses were more likely to be either in the young
38
On
39 616 adult group, or in the preadolescent compared to the child group, even though we cannot
40 617 pinpoint when during adolescence this aspect of inhibitory control may become stable, or
41
42 618 if this may happen even later. Nonetheless, premature responding seems to improve in
ly
43 619 middle childhood and possibly beyond preadolescence, which highlights the role of this
44
45 620 measure as an interesting marker of inhibitory control development in future studies.
46
47 621 Another important predictor that discriminated between all age groups was post-
48
49 622 ignore slowing. This is an adjustment of the reaction time in the form of slowing in the
50 623 trial after a successful ignore trial (which can be considered an instance of heightened
51
52 624 conflict brought about by the detection of the ignore signal). There is some debate
53 625 regarding why individuals slow down after an error or after conflict (Danielmeier &
54
55 626 Ullsperger, 2011; Dutilh, Vandekerckhove, et al., 2012; Wessel, 2018), but it seems clear
56 627 that the phenomenon involves cognitive control processes (Dutilh, Vandekerckhove, et
57
58 628 al., 2012). Within this framework, it could be argued that post-ignore slowing in the
59 629 stimulus-selective stopping task involves the activation of proactive inhibition processes
60
18
developmentalscience@wiley.com
Page 19 of 31 Developmental Science
1
2
3 630 aimed towards preventing potential errors and/or monitoring and resolving conflict.
4
5 631 Although prior research has only investigated slowing after errors (not after ignore trials),
6 632 developmental studies have mainly reported a decrease in slowing with age, even if there
7
8 633 are some mixed findings (Smulders, Soetens, & van der Molen, 2016). In line with these
9 634 findings, we observed post-ignore slowing improvements with age, which would suggest
10
11 635 a greater ability to process and solve the conflict generated by the ignore signal as
12 636 participants age. In any case, further studies are needed to disentangle the cognitive
13
14 637 processes involved in post-ignore slowing.
15
16 638 Finally, the other significant two predictors in the exploratory model were ignore
17
18 639 omissions and the tau parameter, even if they did not discriminate between all age groups.
19 640 It is worth noting that these two variables are both related to attentional processes needed
20
Re
21 641 to sustain performance throughout the task. Weaker attentional-related abilities result in
22 642 distractibility and attentional lapses, which lead to more omission errors and greater tau
23
vie
24 643 values (Lin et al., 2014; Vaurio, Simmonds, & Mostofsky, 2009). In agreement with the
25 644 view that middle childhood is critical for the improvement of attention-related functions
26
27 645 (Betts, McKay, Maruff, & Anderson, 2006; Klimkeit, Mattingley, Sheppard, Farrow, &
w
28 646 Bradshaw, 2004; Suades-González et al., 2017), we found that ignore omissions were
29
30 647 more likely to be committed by children. Still, an interesting question that arises from our
Co
31 648 data is why ignore omissions are a significant predictor of age group, but go omissions
32
33 649 are not, even if they both act as indexes of distractibility. A go omission implies ignoring
34
py
650 only the go stimulus, while an ignore omission involves disregarding both the ignore
35
36 651 signal and the go stimulus in the same trial. Thus, it could be argued that ignore omissions
37 652 are a sign of greater distractibility than go omissions. These results point to the fact that
38
On
39 653 ignore omissions, which cannot be calculated using other tasks, might be a more sensitive
40 654 measure to discriminate between age groups. Regarding intra-individual variability, we
41
42 655 only found significant differences for the tau parameter, which represents the
ly
43 656 exponentially distributed tail of the distribution of RT and has been related to attentional
44
45 657 lapses (Hwang Gu, Gau, Tzang, & Hsu, 2013; van Belle, van Raalten, et al., 2015). Again,
46 658 younger groups seem to display a greater intra-individual variability: tau discriminated
47
48 659 between preadolescents and young adults (and there was a near-significant trend for
49 660 adults relative to children). This agrees in part with previous literature that found
50
51 661 decreases in tau with age in samples spanning from childhood to young adulthood
52 662 (Thomson et al., 2020; van Belle, van Hulst, & Durston, 2015; van Belle, van Raalten, et
53
54 663 al., 2015).
55
56 664 In conclusion, by characterizing the qualitative and quantitative changes between the
57
58 665 three age groups performing a stimulus-selective stopping task, this study provides new
59 666 insights into the development of selective stopping. We found that the ability to stop a
60
19
developmentalscience@wiley.com
Developmental Science Page 20 of 31
1
2
3 667 response selectively reaches its final point by early preadolescence, and remains stable
4
5 668 afterwards at least until young adulthood. By contrast, the efficiency or speed of stopping
6 669 (SSRT) continues to mature throughout adolescence until young adulthood, both for
7
8 670 global and selective implementations of stopping. We also provide some preliminary
9 671 findings regarding other task-related variables related to cognitive control and attentional
10
11 672 processes involved in selective stopping that are of interest for future developmental
12 673 studies using a selective stopping task. Although current results need to be confirmed and
13
14 674 extended in longitudinal studies, they provide new insights into the development of a
15 675 relevant aspect of inhibitory control.
16
17
18 676 References
19
20 677 Aron, A. R. (2011). From reactive to proactive and selective control: developing a richer
Re
21 678 model for stopping inappropriate responses. Biological Psychiatry, 69(12), e55-
22 679 68. doi:10.1016/j.biopsych.2010.07.024
23
vie
24 680 Bari, A., & Robbins, T. W. (2013). Inhibition and impulsivity: behavioral and neural basis
25 681 of response control. Progress in Neurobiology, 108, 44-79.
26 682 doi:10.1016/j.pneurobio.2013.06.005
27
Bartholdy, S., Dalton, B., O’Daly, O. G., Campbell, I. C., & Schmidt, U. (2016). A
w
683
28
29 684 systematic review of the relationship between eating, weight and inhibitory
30 685 control using the stop signal task. Neuroscience & Biobehavioral Reviews, 64, 35-
Co
20
developmentalscience@wiley.com
Page 21 of 31 Developmental Science
1
2
3 709 Chen, W., de Hemptinne, C., Miller, A. M., Leibbrand, M., Little, S. J., Lim, D. A., . . .
4
710 Starr, P. A. (2020). Prefrontal-Subthalamic Hyperdirect Pathway Modulates
5
6
711 Movement Inhibition in Humans. Neuron, 106(4), 579-588.e573.
7 712 doi:https://doi.org/10.1016/j.neuron.2020.02.012
8 713 Cope, L. M., Hardee, J. E., Martz, M. E., Zucker, R. A., Nichols, T. E., & Heitzeg, M. M.
9 714 (2020). Developmental maturation of inhibitory control circuitry in a high-risk
10
715 sample: A longitudinal fMRI study. Developmental Cognitive Neuroscience, 43,
11
12
716 100781. doi:https://doi.org/10.1016/j.dcn.2020.100781
13 717 Crosbie, J., Arnold, P., Paterson, A., Swanson, J., Dupuis, A., Li, X., . . . Schachar, R. J.
14 718 (2013). Response inhibition and ADHD traits: correlates and heritability in a
15 719 community sample. Journal of Abnormal Child Psychology, 41(3), 497-507.
16
720 doi:10.1007/s10802-012-9693-9
17
18 721 Dalley, J. W., & Robbins, T. W. (2017). Fractionating impulsivity: neuropsychiatric
19 722 implications. Nature Reviews Neuroscience, 18(3), 158-171.
20 723 doi:10.1038/nrn.2017.8
Re
21
22
724 Danielmeier, C., & Ullsperger, M. (2011). Post-error adjustments. Frontiers in
23 725 Psychology, 2, 233. doi:10.3389/fpsyg.2011.00233
vie
24 726 Diamond, A. (2013). Executive functions. Annual Review of Psychology, 64, 135-168.
25
727 Durston, S., Thomas, K. M., Yang, Y., Uluğ, A. M., Zimmerman, R. D., & Casey, B. J.
26
27 728 (2002). A neural basis for the development of inhibitory control. Developmental
w
31
732 confound and a simple solution. Journal of Mathematical Psychology, 56(3), 208-
32
33 733 216. doi:https://doi.org/10.1016/j.jmp.2012.04.001
34 734 Dutilh, G., Vandekerckhove, J., Forstmann, B. U., Keuleers, E., Brysbaert, M., &
py
43
741 Freeman, G. H., & Halton, J. H. (1951). Note on an exact treatment of contingency,
44 742 goodness of fit and other problems of significance. Biometrika, 38(1-2), 141-149.
45 743 Garon, N., Bryson, S. E., & Smith, I. M. (2008). Executive function in preschoolers: a
46 744 review using an integrative framework. Psychological Bulletin, 134(1), 31-60.
47
745 doi:10.1037/0033-2909.134.1.31
48
49 746 Groman, S. M., James, A. S., & Jentsch, J. D. (2009). Poor response inhibition: at the
50 747 nexus between substance abuse and attention deficit/hyperactivity disorder.
51 748 Neuroscience & Biobehavioral Reviews, 33(5), 690-698.
52 749 doi:10.1016/j.neubiorev.2008.08.008
53
54 750 Holmboe, K., Bonneville-Roussy, A., Csibra, G., & Johnson, M. H. (2018). Longitudinal
55 751 development of attention and inhibitory control during the first year of life.
56 752 Developmental Science, 21(6), e12690. doi:https://doi.org/10.1111/desc.12690
57 753 Houben, K., Nederkoorn, C., & Jansen, A. (2014). Eating on impulse: The relation
58
59
754 between overweight and food-specific inhibitory control. Obesity, 22(5), E6-E8.
60 755 doi:https://doi.org/10.1002/oby.20670
21
developmentalscience@wiley.com
Developmental Science Page 22 of 31
1
2
3 756 Howard, S. J., Johnson, J., & Pascual-Leone, J. (2014). Clarifying inhibitory control:
4
757 Diversity and development of attentional inhibition. Cognitive Development, 31,
5
6
758 1-21. doi:https://doi.org/10.1016/j.cogdev.2014.03.001
7 759 Hsieh, S., & Lin, Y. C. (2017). Strategies for stimulus selective stopping in the elderly.
8 760 Acta Psychologica, 173, 122-131. doi:10.1016/j.actpsy.2016.12.011
9
10
761 Humphrey, G., & Dumontheil, I. (2016). Development of Risk-Taking, Perspective-
11 762 Taking, and Inhibitory Control During Adolescence. Developmental
12 763 Neuropsychology, 41(1-2), 59-76. doi:10.1080/87565641.2016.1161764
13 764 Hung, Y., Gaillard, S. L., Yarmak, P., & Arsalidou, M. (2018). Dissociations of cognitive
14 765 inhibition, response inhibition, and emotional interference: Voxelwise ALE meta-
15
16
766 analyses of fMRI studies. Human Brain Mapping, 39(10), 4065-4082.
17 767 doi:https://doi.org/10.1002/hbm.24232
18 768 Hwang Gu, S.-L., Gau, S. S.-F., Tzang, S.-W., & Hsu, W.-Y. (2013). The ex-Gaussian
19 769 distribution of reaction times in adolescents with attention-deficit/hyperactivity
20
Re
24 773 Examining the development of attention and executive functions in children with
25 774 a novel paradigm. Child Neuropsychology: a journal on normal and abnormal
26 775 development in childhood and adolescence, 10(3), 201-211.
27
w
28
776 doi:10.1080/09297040409609811
29 777 Kray, J., Kipp, K. H., & Karbach, J. (2009). The development of selective inhibitory
30 778 control: the influence of verbal labeling. Acta Psychologica, 130(1), 48-57.
Co
31 779 doi:10.1016/j.actpsy.2008.10.006
32
33 780 Lacouture, Y., & Cousineau, D. (2008). How to use MATLAB to fit the ex-Gaussian and
34 781 other probability functions to a distribution of response times. Tutorials in
py
22
developmentalscience@wiley.com
Page 23 of 31 Developmental Science
1
2
3 802 MacLeod, C. M. (2007). The concept of inhibition in cognition. In Inhibition in cognition.
4
803 (pp. 3-23). Washington, DC, US: American Psychological Association.
5
6 804 Majid, D. S. A., Cai, W., Corey-Bloom, J., & Aron, A. R. (2013). Proactive Selective
7 805 Response Suppression Is Implemented via the Basal Ganglia. The Journal of
8 806 Neuroscience, 33(33), 13259-13269. doi:10.1523/JNEUROSCI.5651-12.2013
9
10
807 Morgan, P. L., Farkas, G., Hillemeier, M. M., Pun, W. H., & Maczuga, S. (2019).
11 808 Kindergarten Children's Executive Functions Predict Their Second-Grade
12 809 Academic Achievement and Behavior. Child Development, 90(5), 1802-1816.
13 810 doi:https://doi.org/10.1111/cdev.13095
14 811 Narayanan, N. S., Wessel, J. R., & Greenlee, J. D. W. (2020). The Fastest Way to Stop:
15
16
812 Inhibitory Control and IFG-STN Hyperdirect Connectivity. Neuron, 106(4), 549-
17 813 551. doi:10.1016/j.neuron.2020.04.017
18 814 Nord, C. L., Kim, S.-G., Callesen, M. B., Kvamme, T. L., Jensen, M., Pedersen, M. U., .
19 815 . . Voon, V. (2019). The myeloarchitecture of impulsivity: premature responding
20
Re
28 822 Paulsen, D. J., Hallquist, M. N., Geier, C. F., & Luna, B. (2015). Effects of incentives,
29 823 age, and behavior on brain activation during inhibitory control: A longitudinal
30 824 fMRI study. Developmental Cognitive Neuroscience, 11, 105-115.
Co
31 825 doi:https://doi.org/10.1016/j.dcn.2014.09.003
32
33 826 Rincón-Pérez, I., Echeverry-Alzate, V., Sánchez-Carmona, A. J., Bühler, K. M.,
34 827 Hinojosa, J. A., López-Moreno, J. A., & Albert, J. (2020). The influence of
py
834 Robbins, T. W., & Dalley, J. W. (2017). Dissecting Impulsivity: Brain Mechanisms and
43
835 Neuropsychiatric Implications. In J. R. Stevens (Ed.), Impulsivity: How Time and
44
45 836 Risk Influence Decision Making (pp. 201-226). Cham: Springer International
46 837 Publishing.
47 838 Rømer Thomsen, K., Blom Osterland, T., Hesse, M., & Feldstein Ewing, S. W. (2018).
48 839 The intersection between response inhibition and substance use among
49
840 adolescents. Addictive Behaviors, 78, 228-230.
50
51 841 doi:https://doi.org/10.1016/j.addbeh.2017.11.043
52 842 Rouder, J. N., Speckman, P. L., Sun, D., Morey, R. D., & Iverson, G. (2009). Bayesian t
53 843 tests for accepting and rejecting the null hypothesis. Psychonomic Bulletin &
54 844 Review, 16(2), 225-237. doi:10.3758/pbr.16.2.225
55
56 845 Rubia, K., Smith, A., & Taylor, E. (2007). Performance of Children with Attention Deficit
57 846 Hyperactivity Disorder (ADHD) on a Test Battery of Impulsiveness. Child
58 847 Neuropsychology, 13(3), 276-304. doi:10.1080/09297040600770761
59
60
23
developmentalscience@wiley.com
Developmental Science Page 24 of 31
1
2
3 848 Rubia, K., Smith, A. B., Taylor, E., & Brammer, M. (2007). Linear age-correlated
4
849 functional development of right inferior fronto-striato-cerebellar networks during
5
6
850 response inhibition and anterior cingulate during error-related processes. Human
7 851 Brain Mapping, 28(11), 1163-1177. doi:https://doi.org/10.1002/hbm.20347
8 852 Sánchez-Carmona, A. J., Albert, J., & Hinojosa, J. A. (2016). Neural and behavioral
9 853 correlates of selective stopping: Evidence for a different strategy adoption.
10
854 Neuroimage, 139, 279-293. doi:10.1016/j.neuroimage.2016.06.043
11
12 855 Sánchez-Carmona, A. J., Rincón-Pérez, I., López-Martín, S., Albert, J., & Hinojosa, J. A.
13 856 (2021). The effects of discrimination on the adoption of different strategies in
14 857 selective stopping. Psychonomic Bulletin & Review, 28(1), 209-218.
15 858 doi:10.3758/s13423-020-01797-6
16
17 859 Sánchez-Carmona, A. J., Santaniello, G., Capilla, A., Hinojosa, J. A., & Albert, J. (2019).
18 860 Oscillatory brain mechanisms supporting response cancellation in selective
19 861 stopping strategies. Neuroimage, 197, 295-305.
20 862 doi:10.1016/j.neuroimage.2019.04.066
Re
21
22
863 Schachar, R., Logan, G. D., Robaey, P., Chen, S., Ickowicz, A., & Barr, C. (2007).
23 864 Restraint and cancellation: multiple inhibition deficits in attention deficit
vie
28
868 (2017). Neural Architecture of Selective Stopping Strategies: Distinct Brain
29 869 Activity Patterns Are Associated with Attentional Capture But Not with Outright
30 870 Stopping. The Journal of Neuroscience, 37(40), 9785-9794.
Co
31 871 doi:10.1523/jneurosci.1476-17.2017
32 872 Smith, J. L., Mattick, R. P., Jamadar, S. D., & Iredale, J. M. (2014). Deficits in
33
34
873 behavioural inhibition in substance abuse and addiction: A meta-analysis. Drug
py
24
developmentalscience@wiley.com
Page 25 of 31 Developmental Science
1
2
3 896 van Belle, J., van Hulst, B. M., & Durston, S. (2015). Developmental differences in intra-
4
897 individual variability in children with ADHD and ASD. Journal of Child
5
6
898 Psychology and Psychiatry, 56(12), 1316-1326. doi:10.1111/jcpp.12417
7 899 van Belle, J., van Raalten, T., Bos, D. J., Zandbelt, B. B., Oranje, B., & Durston, S. (2015).
8 900 Capturing the dynamics of response variability in the brain in ADHD.
9 901 Neuroimage: Clinical, 7, 132-141. doi:10.1016/j.nicl.2014.11.014
10
11 902 van de Laar, M. C., van den Wildenberg, W. P., van Boxtel, G. J., & van der Molen, M.
12 903 W. (2011). Lifespan changes in global and selective stopping and performance
13 904 adjustments. Frontiers in Psychology, 2, 357. doi:10.3389/fpsyg.2011.00357
14 905 van Hulst, B. M., de Zeeuw, P., Vlaskamp, C., Rijks, Y., Zandbelt, B. B., & Durston, S.
15
16
906 (2018). Children with ADHD symptoms show deficits in reactive but not
17 907 proactive inhibition, irrespective of their formal diagnosis. Psychological
18 908 Medicine, 48(15), 2515-2521. doi:10.1017/S0033291718000107
19 909 Vara, A. S., Pang, E. W., Vidal, J., Anagnostou, E., & Taylor, M. J. (2014). Neural
20
Re
24 913 Vaurio, R. G., Simmonds, D. J., & Mostofsky, S. H. (2009). Increased intra-individual
25 914 reaction time variability in attention-deficit/hyperactivity disorder across response
26 915 inhibition tasks with different cognitive demands. Neuropsychologia, 47(12),
27
w
28
916 2389-2396. doi:10.1016/j.neuropsychologia.2009.01.022
29 917 Verbruggen, F., Aron, A. R., Band, G. P., Beste, C., Bissett, P. G., Brockett, A. T., . . .
30 918 Boehler, C. N. (2019). A consensus guide to capturing the ability to inhibit actions
Co
39 925 and stop-change paradigms. Neuroscience & Biobehavioral Reviews, 33(5), 647-
40 926 661. doi:10.1016/j.neubiorev.2008.08.014
41 927 Verbruggen, F., & Logan, G. D. (2015). Evidence for capacity sharing when stopping.
42
ly
25
developmentalscience@wiley.com
Developmental Science Page 26 of 31
1
2
3 942 Wessel, J. R. (2018). An adaptive orienting theory of error processing. Psychophysiology,
4
943 55(3), e13041. doi:10.1111/psyp.13041
5
6 944 Wessel, J. R., & Aron, A. R. (2017). On the Globality of Motor Suppression: Unexpected
7 945 Events and Their Influence on Behavior and Cognition. Neuron, 93(2), 259-280.
8 946 doi:https://doi.org/10.1016/j.neuron.2016.12.013
9
10
947 Wiebe, S. A., Sheffield, T. D., & Espy, K. A. (2012). Separating the Fish From the Sharks:
11 948 A Longitudinal Study of Preschool Response Inhibition. Child Development,
12 949 83(4), 1245-1261. doi:https://doi.org/10.1111/j.1467-8624.2012.01765.x
13 950 Wiecki, T. V., & Frank, M. J. (2013). A computational model of inhibitory control in
14 951 frontal cortex and basal ganglia. Psychological Review, 120(2), 329-355.
15
16
952 doi:10.1037/a0031542
17 953 Williams, B. R., Ponesse, J. S., Schachar, R. J., Logan, G. D., & Tannock, R. (1999).
18 954 Development of inhibitory control across the life span. Developmental
19 955 Psychology, 35(1), 205-213. doi:10.1037/0012-1649.35.1.205
20
Re
21 956
22
23
vie
24
25
26
27
w
28
29
30
Co
31
32
33
34
py
35
36
37
38
On
39
40
41
42
ly
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
26
developmentalscience@wiley.com
Page 27 of 31 Developmental Science
1
2
3 957 Figure 1. Schematic illustration of the stimulus-selective stopping task used in this study.
4
5 958 ITI = inter-trial interval; SSD = stop-signal delay; ISD = ignore signal delay.
6
7 959 Figure 2. Differences in strategy use (left) and SSRT (right) by age group.
8
9 960 * = significant differences. Error bars: 95% Confidence Interval.
10
11 961
12
13 962
14
15
16
17
18
19
20
Re
21
22
23
vie
24
25
26
27
w
28
29
30
Co
31
32
33
34
py
35
36
37
38
On
39
40
41
42
ly
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
27
developmentalscience@wiley.com
Developmental Science Page 28 of 31
1
2
3
4 Table 1. Identification of strategies by reaction time comparisons using the method
5 proposed by Bissett and Logan (2014)
6
7
8 Selective Non-selective Uncategorized
9
10
11 Independent Dependent
12 Stop then Strategy not
Discriminate then Discriminate then
13 Discriminate (StD) identified
Stop (iDtS) Stop (dDtS)
14
15
16
17 Failed stop RT < Go Failed stop RT ≮ Go Failed stop RT < Go Failed stop RT ≮ Go
18 RT ≮ Ignore RT RT < Ignore RT RT < Ignore RT RT ≮ Ignore RT
19
20
Re
21 963
22
23
vie
964
24
25
26
27
w
28
29
30
Co
31
32
33
34
py
35
36
37
38
On
39
40
41
42
ly
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
28
developmentalscience@wiley.com
Page 29 of 31 Developmental Science
1
2
3 Table 2. Confirmatory multinomial regression model results (with strategy and SSRT
4
as predictors).
5
6
7 Determinants of young adults vs children CI 95%
8 Predictors Wald Lower Upper
9 Est. S.E. p-value OR
test bound bound
10
11 Intercept 6.68 1.03 42.33 <.001
12 iDtS vs StD 3.47 0.93 14.08 <.001 32.22 5.25 197.58
13 dDtS vs StD 1.86 0.47 15.40 <.001 6.43 2.54 16.28
14 SSRT -0.03 0.00 57.99 <.001 0.97 0.97 0.98
15
16 Determinants of preadolescents vs children CI 95%
17
Predictors Wald Lower Upper
18 Est. S.E. p-value OR
19
test bound bound
20 Intercept 2.35 0.86 7.47 .006
Re
24
25 Determinants of young adults vs preadolescents CI 95%
26
27
Predictors S.E. Wald p-value Lower Upper
Est. OR
w
Interval; SSRT = Stop-Signal Reaction Time; iDtS = independent Discriminate then Stop strategy; dDtS
35
= dependent Discriminate then Stop strategy, and StD = Stop then Discriminate strategy.
36
37 965
38
On
39
966
40
41 967
42 968
ly
43
44 969
45
46 970
47
48
49
50
51
52
53
54
55
56
57
58
59
60
29
developmentalscience@wiley.com
Developmental Science Page 30 of 31
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Re
19
20
21
vie
22
23
24 483x227mm (118 x 118 DPI)
25
w
26
27
28
Co
29
30
31
32
py
33
34
35
On
36
37
38
39
ly
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 developmentalscience@wiley.com
Page 31 of 31 Developmental Science
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Re
19
20
21 1069x410mm (38 x 38 DPI)
vie
22
23
24
25
w
26
27
28
Co
29
30
31
32
py
33
34
35
On
36
37
38
39
ly
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 developmentalscience@wiley.com