The Influence of Pubertal Maturation on Antisaccade Performance

Abstract

Adolescence is a period characterized by continued improvements in inhibitory control, and this persisting immaturity is believed to interact with affective/motivational behavior to generate the impulsive and risk-taking behavior evidenced at this time. Puberty is a central event of adolescence that has been shown to influence affective/motivational behavior. However, despite plausible mechanisms by puberty might influence inhibitory control, researchers have yet to test this possibility rigorously. Thus, we designed a study to examine the unique role of pubertal maturation, independent of age, in the development of inhibitory control. In order to minimize age-related variability while maximizing pubertal status variability, we recruited 78 participants (34 F) whose ages narrowly spanned the mean age of gonadarche for each sex (F: ages 11–13, M: ages 12–14). Two complementary measures were used to assess pubertal status: (1) circulating blood serum testosterone and estradiol levels reflecting internal manifestations of pubertal maturation, and (2) Tanner staging by a trained nurse reflecting pubertal maturation’s external manifestations. Inhibitory control was assessed using the antisaccade task, and findings were adjusted for the potential effect of age. Results revealed no association between testosterone levels and error rates or response latencies in either sex. In girls, estradiol levels were not associated with error rates, but were associated with faster response latencies. There was similarly no association between Tanner status and error rates, though females in more advanced pubertal stages showed faster response latencies. Power analyses indicate that findings of a lack of association did not reflect limited statistical power. Thus, in a study designed to isolate the effects of pubertal maturation independent of age, both external and internal indices of pubertal maturation converged to indicate that age-related improvements in cold antisaccade performance are independent of pubertal maturation.

Puberty and Inhibitory Control

Adolescence is a developmental period defined by the onset of puberty, and it is characterized by ongoing brain maturation and continued cognitive and affective development. Behaviorally, it is distinguished from childhood and adulthood by high rates of risk-taking behaviors such as unprotected sex, distracted driving, and impulsive, self-injurious behaviors (Dahl, 2004). Thus, despite being a period of peak physical health, a 200% increase in the incidence of mortality and morbidity occurs during adolescence (Resnick et al., 1997). Risk-taking behavior is theorized to reflect an interplay between affective/reward processing and inhibitory control, the ability to direct behavior through internally-represented goals in the face of competing stimuli (Steinberg, 2010). Puberty is thought to play a role in risk-taking behaviors in that it has been posited that the development of affective/motivational processes (and the brain networks that support them) are driven primarily by pubertal maturation whereas inhibitory control is not (Blakemore, Burnett, & Dahl, 2010; Forbes & Dahl, 2010; Steinberg, 2010). While there is evidence that pubertal status influences affective processes, studies have yet to test whether puberty may influence inhibitory control. Importantly, and contrary to prevailing theory, there is reason to suspect that inhibitory control may be affected by puberty: animal studies provide evidence for the presence of gonadal hormone receptors in the cortex (Kato, 1977; Merchenthaler, Lane, Numan, & Dellovade, 2004; Wang, Hara, Janssen, Rapp, & Morrison, 2010) and neuroimaging studies indicate that cortical thickness in areas of cortex that support inhibitory control is influenced by pubertal status (Bramen et al., 2012; Bramen et al., 2010; Goddings et al., 2014; Herting, Gautam, Spielberg, Dahl, & Sowell, 2015; Herting, Maxwell, Irvine, & Nagel, 2012; Neufang et al., 2009; Nguyen, McCracken, Ducharme, Botteron, et al., 2013; Nguyen, McCracken, Ducharme, Cropp, et al., 2013; Peper et al., 2009). Puberty encompasses the re-activation of multiple neurodendocrine systems that contribute to marked increases in circulating levels of testosterone, estradiol, and progesterone, and ultimately to the emergence of secondary sex characteristics. Hormone concentrations are in flux over the course of puberty, and their action on hormone receptors instigates gonadal maturation, the emergence of secondary sex characteristics, and changes in height and musculature. Pubertal maturation can contribute to behavioral and cognitive changes through multiple mechanisms, including a direct, internal effect via gonadal hormone binding to receptors in the brain (Sisk & Zehr, 2005) and an indirect, external effect whereby visible manifestations of secondary sex characteristics (e.g., breasts, facial hair) not only reflect circulating gonadal hormone levels, but also influence how others (e.g., peers, parents) may perceive and/or interact with an individual (Skoog & Stattin, 2014).

Whereas studies reveal that pubertal maturation – independent of age – is associated with the maturation of affective brain networks (Braams, van Duijvenvoorde, Peper, & Crone, 2015; Forbes et al., 2010; Moore et al., 2012; Op de Macks et al., 2011; Spielberg, Olino, Forbes, & Dahl, 2014) and contributes to increases in irritability, mood lability, and sensation-seeking (Buchanan, Eccles, & Becker, 1992; Sisk & Zehr, 2005; Steinberg et al., 2008), little is known about the role of puberty in the development of inhibitory control. There are, however, multiple lines of evidence suggesting that inhibitory control may be shaped by pubertal development. First, several studies, including studies conducted with the sample studied here, have demonstrated that the cortex within frontoparietal and striatal brain regions networks canonically involved in inhibitory control mature with advancing pubertal status (Bramen et al., 2012; Bramen et al., 2010; Goddings et al., 2014; Herting et al., 2015; Herting et al., 2012; Neufang et al., 2009; Nguyen, McCracken, Ducharme, Botteron, et al., 2013; Nguyen, McCracken, Ducharme, Cropp, et al., 2013; Peper et al., 2009). Second, developmental trajectories of cortical thickness in frontoparietal and striatal areas involved in cognitive control have been shown to be shaped by genetic variation in adrogen receptor expression (Raznahan et al., 2010), providing a plausible mechanism by which puberty can impact inhibitory control. Third, animal research indicates that gonadal hormone receptors populate these cortical areas known to support inhibitory control (Kato, 1977; Merchenthaler et al., 2004; Motta & Martini, 1984; Wang et al., 2010). And while there is extensive evidence for continued age-related improvements of inhibitory control in the age range corresponding to the pubertal period (Akshoomoff et al., 2014; Davidson, Amso, Anderson, & Diamond, 2006; Luciana & Nelson, 2002; Luna, 2009; Ridderinkhof & van der Molen, 1997; Williams, Ponesse, Schachar, Logan, & Tannock, 1999) and associated changes in brain function (Luna, Padmanabhan, & O’Hearn, 2010; Ordaz, Foran, Velanova, & Luna, 2013; Ridderinkhof, van den Wildenberg, Segalowitz, & Carter, 2004), it is unclear whether these changes reflect the influence of pubertal maturation itself or of other maturational processes.

The present study was designed specifically to investigate the role of pubertal maturation on inhibitory control in adolescence. To this end, we utilized: (a) a task sensitive to adolescent changes in inhibitory control; (b) an experimental design with the capacity to disentangle age and pubertal status; and (c) a sensitive and complementary set of pubertal measures. Firstly, the antisaccade (AS) task (Hallett, 1978) is a measure of inhibitory control that requires suppressing a reflexive visually elicited eye movement and using an internal representation to generate a voluntary saccadic response. Developmental improvements in AS performance have been well-characterized and are particularly sensitive to a protracted maturation during the teenage years (Fischer, Biscaldi, & Gezeck, 1997; Fukushima, Hatta, & Fukushima, 2000; Klein, 2001; Klein & Foerster, 2001; Luna, Garver, Urban, Lazar, & Sweeney, 2004; Munoz, Broughton, Goldring, & Armstrong, 1998; Nieuwenhuis, Ridderinkhof, van der Molen, & Kok, 1999; Ordaz, Davis, & Luna, 2010; Ordaz et al., 2013; Romine & Reynolds, 2005), including the specific age range we study here. In addition, brain networks supporting AS performance have been well-delineated through adult and developmental functional neuroimaging studies (Hwang, Velanova, & Luna, 2010; Ordaz et al., 2013; Velanova, Wheeler, & Luna, 2008, 2009) and through single-cell recordings of non-human primates (Munoz & Everling, 2004). Specifically, we know that AS performance is supported by specific frontal, parietal, and striatal brain regions across development (Luna et al., 2010).

Secondly, to disentangle the effects of pubertal maturation and age, participants were recruited to fall within a limited age range reflecting the peak age of gonadarche; this maximized variability associated with puberty and minimized age-related variability. Due to sex differences in pubertal timing, girls were recruited to be younger. Given sex differences in hormone levels, hormone receptor densities, and relations of pubertal maturation to brain and cognitive outcomes (Peper et al., 2009), we tested associations separately by sex.

Thirdly, because age is typically measured more precisely than puberty, care was taken to use sensitive indices of pubertal development. These included multiple measures of pubertal status that provide complimentary information about distinct yet overlapping aspects of gonadarcheal development (Dorn et al., 2003). These included nurse assessments of pubertal status rather than self-report measures and hormonal indices of pubertal status, which are not influenced by self-report biases or adiposity (Dorn, Dahl, Woodward, & Biro, 2006). We assessed testosterone levels in boys and girls because it has demonstrated effects on the brain in both sexes, and we assessed estradiol levels in girls. Testosterone and estradiol levels were assessed using blood spot assays, which are more accurate than salivary indices in both sexes (Shirtcliff, Granger, & Likos, 2002). Thus, we examined the relationship of circulating testosterone hormone levels and stage of pubertal development on AS performance.

Given the paucity of prior studies examining the role of pubertal status in inhibitory control, the following hypotheses were exploratory. Firstly, on the basis of research indicating that more advanced pubertal status (Tanner stage and gonadal hormone levels) is associated with more mature patterns of cortical development in specific frontoparietal and striatal regions that are well-established to support AS performance, we hypothesized that higher scores on all metrics of pubertal maturation would be associated with better AS performance, as measured by lower error rates and faster latencies to initiate correct responses. Secondly, given that testosterone levels are higher and more variable in males during puberty and the level of change in testosterone during puberty is much greater in males than females, we predicted larger effect sizes in males than in females. Thirdly, we predicted that higher estradiol levels would be associated with lower error rates and faster latencies in females.

Method

Participants

Adolescents were recruited from the community through advertisements, flyers, and demographically targeted phone lists. To minimize age-related variability and maximize puberty-related variability, participants were recruited to be within a narrow age range representing the mean age of gonadarche for each sex, which was 11–13 years for girls and 12–14 years for boys (Herman-Giddens et al., 1997; Karpati, Rubin, Kieszak, Marcus, & Troiano, 2002). Adolescents were included only if they were free of current and lifetime psychiatric disorders, had no history of head injury, serious medical illness, psychotropic medication use, and reported no alcohol or illicit drug use at the time of testing. Data from 78 individuals are included in the following analyses, as data from 20 individuals were lost due to mechanical and data collection errors. The sample described in this report have also been evaluated to assess relationships between pubertal maturation and brain activation (Forbes et al., 2010; Holm et al., 2009; Spielberg, Forbes, et al., 2014; Spielberg, Olino, et al., 2014) and between pubertal maturation and brain structure (Bramen et al., 2012; Bramen et al., 2010; Herting et al., 2015; Herting et al., 2012).

Study Procedure

Following the guidelines of the University of Pittsburgh Institutional Review Board, participants provided assent indicating that they understood the nature of the study and agreed to participate, while participants’ legal guardians provided written informed consent. Following the consents and assents, finger-stick procedures were used to obtain blood-spot samples for hormonal assays. Samples were obtained between 8:20 and 8:35 AM from all subjects to standardize in light of known circadian influences on hormone levels (Granger et al., 2003; Worthman & Stallings, 1997). Efforts were made to reduce hormonal variations among post-menarcheal girls by collecting data during the follicular phase. During the rest of the study visit, participants completed a physical examination to determine level of sexual maturity, followed by other cognitive and affective tasks in the laboratory and MR scanner environment. Participants were compensated for their participation.

Materials

Sexual Maturity

Adolescents underwent physical examination by a single nurse trained in the scientific study of puberty to determine stage of sexual maturation using the criteria specified by Marshall and Tanner (Marshall & Tanner, 1968). As breast and genital development in girls and boys, respectively, reflect changes in levels of gonadal steroids that are central to gonadarche and potentially influence brain development during puberty, participants were classified using the Tanner breast/genital scales. Indices are thought to reflect comparable aspects of development in boys and girls. While the literature states that the criteria for Tanner staging are reliable (Slora et al., 2009), evidence indicates that the steepest increment in gonadal hormone levels occurs between stages two and three (Knorr, Bidlingmaier, Butenandt, & Fendel, 1974). Thus, rather than treating Tanner staging values as an ordinal variable, our a priori approach was to classify participants as pre/early pubertal if they were Tanner stage one or two and as mid/late pubertal if they were Tanner stage three, four, or five on the breast/genital scale, consistent with our prior approach to examining affective aspects of pubertal development (Forbes, Williamson, Ryan, & Dahl, 2004).

Circulating Testosterone and Estradiol Levels

Testosterone and estradiol levels were assayed according to a standardized procedure (Worthman & Stallings, 1997) using blood-spot samples obtained via a minimally-invasive finger-stick procedure. Assays were a modification of a commercially available serum/plasma radioimmunoassay kit (T: Pantex, Santa Monica, CA, E₂: Diagnostic Systems Laboratories, Webster, TX). Blood spot measurements were converted to serum concentration estimates using published conversion equations (Shirtcliff, Reavis, Overman, & Granger, 2001) that were subsequently modified according to in-house validation studies to yield a single regression equation for boys and girls (for testosterone; estradiol was for collected for girls only) with a high blood spot-serum correlation (testosterone: r = 0.98, estradiol: r = 0.98). Serum testosterone and estradiol estimates yielded values within expected ranges for individuals within the age range studied. None of the participant estimates fell below the minimal detectable dose sensitivity criterion (testosterone: 14.2 ng/dL for boys and 14.0 ng/dL for girls, estradiol: 3.4 pg/mL), indicating that the assay was sensitive to gonadalhormone levels in prepubertal participants of both sexes. Inter-assay coefficients of variation were acceptable.

Antisaccade Task

The antisaccade (AS) task was administered using a PC running E-Prime software (version 1.0, Psychology Software Tools, Pittsburgh, PA, USA). Eye movement measurements were obtained using an Applied Science Laboratories (ASL, Bedford, MA) Model 504 table-mounted near-infrared eye tracker with a sampling rate of 60 Hz. Participants were seated and positioned 54 cm from a 15″ computer screen in a table-mounted chinrest with a Velcro head restraint to minimize head motion and to position the eyes at a fixed distance from the screen. Testing occurred in a darkened room for the purposes of eye tracking, and the testing session began with nine-point eye tracker calibration for each potential target location. An experimenter then explained the AS task to the participants, and this was immediately followed by practice trials to ensure that the participant understood the task directions. Participants’ performance was monitored in real time, and testing did not begin until they were able to perform three correct trials in a row. No participants required more than 10 trials before testing commenced. Real-time monitoring also permitted experimenters to immediately redirect participants if they moved their head or did not appear to be attentive to the task, which occurred minimally.

Participants were instructed to fixate on a central red fixation cross that was present for 500, 2000, 4000, or 6000 ms and signaled participants to prepare to suppress a saccade to a peripheral target that would appear in an unexpected location. The peripheral target, a yellow dot, appeared for 1000 ms at one of four locations: 4° and 8° of visual angle to the left or right of center fixation. Following this, the screen remained black for 200 ms before the instruction cue of the subsequent trial appeared. Both the red fixation cross and peripheral stimulus subtended approximately 1.5° vertical visual angle. A total of 16 AS trials, representing all combinations of the four instruction period lengths and four stimulus locations, were presented to ensure unpredictability of cue appearance. Each of the trial types were presented three times within a single task block, resulting in a total of 48 AS trials; trials were administered in a randomized order. Participants completed a total of 2 blocks, or 96 AS trials.

Eye movement recordings were analyzed offline using a combination of ILAB (Gitelman et al., 1999) and in-house programs written in MATLAB (MathWorks, Inc., Natick, MA). Saccades were identified using a velocity algorithm employing a 30°/sec criterion. A trained rater reviewed the graphical and numerical results generated by the computer algorithms to identify blink artifacts and occasional failures of the software to identify primary saccades. If necessary, modifications were made using the editing features available in ILAB. Each eye movement trial was scored by trained raters for accuracy and latency. Correct AS trials were defined as those in which the participant did not look at the peripheral target and instead generated a saccade towards the opposite visual field. Eye movements toward the peripheral target were defined as error trials. The majority of error trials were corrected errors (T1: mean 86%, SD 17%); that is, they were followed by a saccade to the correct location, indicating that participants were attending to the task but were unable to inhibit the automatic response towards the cue. Uncorrected error trials reflect a lack of attention to the task and were excluded on a trial-by-trial basis from further analyses. Trials that began with express saccades, characterized by an initial saccadic latency of less than 100 ms and generated towards the peripheral cue (Fischer & Ramsperger, 1984; Fischer & Weber, 1993), were dropped from analyses. In addition, trials with response latencies of over 1000 ms were dropped because they are thought to reflect inattention to the task, consistent with established protocol. For these analyses, we compiled two metrics of AS performance: the percentage of corrected error trials (inhibitory error rates) and the average latency to initiate correct antisaccades, with lower scores on each indicating better performance.

Our approach to the regression analysis was as follows. First, given the potential sex-specific relations between pubertal maturation and AS performance, we conducted regressions relating each measure of pubertal maturation (Tanner staging, circulating testosterone level, and circulating estradiol level) to AS performance within each sex. Second, to increase power, we calculated sex-specific z-scores for testosterone levels and included all participants in regressions relating testosterone level to AS performance. Third, to ensure that any potential results were not confounded by age, we re-ran all aforementioned regressions covarying for age. Regressions were conducted in SPSS version 23. Fourth, we report sensitivity power analyses in G*Power software version 3.0 (Faul, Erdfelder, Lang, & Buchner, 2007) to highlight that our study is adequately powered; details of the analysis are described in the Results.

Results

Participants included in the final sample included 34 girls and 44 boys. The sample reflected the racial demographics of Allegheny County, PA, with the following racial distribution: 78% White, non-Hispanic, 14% Black, non-Hispanic, and 8% multiracial.

Tanner Staging and Circulating Testosterone Levels

As per the study design, the sexes did not differ in the distribution of pre/early versus mid/late pubertal status (χ²(1) = 1.706, p = .192); 41% of girls were pre/early pubertal and 24% of boys were pre/early pubertal. Serum testosterone and estradiol levels were within published norms for each sex (Shirtcliff et al., 2001). Boys’ testosterone levels spanned a wider range (F(1,72) = 59.067, p = .000) and were significantly higher than girls (t(40.859) = −7.380; p = .000), as would be expected. Due to the differential (and expected) testosterone range in boys and girls, further analyses were initially conducted separately by sex and then sex-specific z-scores were calculated for regressions inclusive of both sexes (for the purposes of maximizing power). Regressions dummy coded for Tanner status indicated higher testosterone levels for individuals with more advanced Tanner status in both sexes (M: F(1,37) = 9.434, p = .004, R² = 0.203; F: F(1,29) = 8.268, p = .007, R² = 0.222), confirming a significant and expected relationship between hormonal and physical exam measures. Similarly, higher estradiol levels were associated with more advanced Tanner status in girls (F(1,29) = 11.309, p = .002, R² = .281.

Consistent with the study design whereby the sexes were matched on pubertal status rather than age, girls (mean age = 12.9 yrs, SD = 0.6) were significantly younger than boys (mean age = 11.8 yrs, SD = 0.6; t(76) = −8.189, p = 0.000). Within the restricted two-year age range studied, boys in the mid/late stages of puberty trended towards being significantly older than those in the pre/early stages (F(1,39) = 3.967, p = .053, R² = .092). Among boys in this restricted range, testosterone levels were significantly correlated with age (F(1,39) = 10.402, p = .003, R² = .211), despite the study design. Girls did not differ in age as a function of Tanner status (F(1,29) = 1.079, p = .308, R² = .036), nor did they evidence a significant association between age and testosterone (F(1,31) = .151, p = .700, R² = .005) or estradiol (F(1,31) = 2.098, p = .157, R² = .063).

Effects of Age and Sex on Antisaccade Performance

In light of evidence that adolescence is a period of continued change in AS performance (Fischer et al., 1997; Fukushima et al., 2000; Klein, 2001; Klein & Foerster, 2001; Luna et al., 2004; Munoz et al., 1998; Nieuwenhuis et al., 1999; Ordaz et al., 2010; Ordaz et al., 2013; Romine & Reynolds, 2005), we began by examining the extent to which age contributed to change in AS performance across the two-year age range studied within each sex in order to inform further analyses. Among boys, there was no significant effect of age on error rates (F(1,42) = 0.010, p = 0.822, R² = .001) or response latencies (F(1,42) = 1.679, p = .202, R² = .038). Among girls, error rates declined with age (F(1,32) = 5.328, p = .028, R² = .143). This significant result may reflect that the girls were younger than boys, in light of evidence that slopes of performance by age are steeper at the 11–13 age range than the 12–14 age range (Ordaz et al., 2010). There was no significant effect of age on response latency in girls (F(1,29) = 1.442, p = .240, R² = .047).

Not surprisingly, the boys, who were older, had lower error rates than girls (F(1,72) = 7.810, p = .007, R² = .098), which was no longer present when modelling with age as a covariate (t(75) = −.593, p = .555). There were no sex differences in response latencies (F(1,69) = .024, p = .8778, R² = .000).

Effect of Pubertal Status on Antisaccade Performance

Raw data are plotted in Figures 1 and 2. Regressions were run separately for each sex in light of the different metrics contributing to Tanner status and the different scaling of circulating testosterone level across sexes. Tanner status was dummy coded for the purposes of regressions using this variable as a predictor. For regressions with testosterone and estradiol as predictors, both linear and quadratic regressions were run given a lack of prior studies examining relationships among pubertal variables and AS performance. To enhance power, we also ran regressions combining boys and girls into a single regression; for the regression involving testosterone levels this required calculating sex-specific z-scores; we did not have estradiol levels for boys. All regressions were re-run controlling for age to confirm that these findings held, since age was associated with pubertal variables in boys and with error rates in girls. Results from all regressions are shown in Table 1.

Figure 1.

Scatterplots illustrating the relation between antisaccade error rates and metrics of pubertal maturation, segregated by sex.

Figure 2.

Scatterplots illustrating the relation between mean latency to initiate antisaccades on correct trials and metrics of pubertal maturation, segregated by sex.

Table 1.

Results from regressions for multiple metrics of pubertal maturation. Effect sizes for each regression are noted. Betas reported are for the linear pubertal-status term (Tanner staging or testosterone) in the linear regressions and for the quadratic pubertal-status term in the quadratic regression.

				Regression with pubertal status predictor				Regression with pubertal status predictor covarying for age (centered)
	R2	f2	E.s. powered to detect	Beta (un-standardized)	t	p	Beta (un-standardized)	t	p
Tanner staging - dichotomous variable, linear regression
Males
Error rates	0.012	0.201	medium	−0.042 (0.062)	−0.682	0.499	−0.059 (0.065)	−0.896	0.376
Latencies	0.001	0.201	medium	−3.087 (17.154)	−0.180	0.858	−10.616 (17.793)	−0.597	0.554
Females
Error rates	0.016	0.271	large	−0.065 (0.093)	−0.695	0.493	0.030 (0.090)	−0.329	0.745
Latencies	0.114	0.291	large	−40.634 (21.767)	−1.867	0.073+	−68.352 (29.265)	−2.336	0.027*
All participants
Error rates	0.026	0.112	medium	−0.077 (0.056)	−1.372	0.174	−0.033 (0.056)	−0.595	0.554
Latencies	0.037	0.115	medium	−21.597 (13.401)	−1.612	0.112	−27.382 (13.822)	−1.981	0.052+
Testosterone level - linear regression
Males (using z-transformed testosterone)
Error rates	0.011	0.201	medium	−0.019 (0.029)	−0.646	0.522	−0.032 (0.033)	−0.987	0.330
Latencies	0.004	0.201	medium	3.254 (8.760)	0.371	0.712	−0.683 (9.891)	−0.069	0.945
Females (using z-transformed testosterone)
Error rates	0.004	0.253	large	0.017 (0.046)	0.364	0.718	0.024 (0.043)	0.556	0.582
Latencies	0.022	0.281	large	−8.648 (10.964)	0.789	0.437	−9.884 (10.848)	10.848	0.370
All participants (using z-transformed testosterone)
Error rates	0.000	0.109	medium	−0.003 (0.027)	−0.109	0.913	0.013 (0.026)	0.518	0.606
Latencies	0.001	0.114	medium	−1.753 (6.794)	−0.258	0.797	−3.584 (6.957)	−0.515	0.608
Testosterone level - quadratic regression
Males (using z-transformed testosterone)
Error rates	0.086	0.255	large	−0.045 (0.026)	−1.774	0.084	−0.045 (0.026)	−1.733	0.092
Latencies	0.085	0.255	large	14.276 (7.746)	1.843	0.073+	14.571 (7.758)	1.878	0.068+
Females (using z-transformed testosterone)
Error rates	0.011	0.323	large	0.018 (0.041)	0.444	0.660	0.019 (0.038)	0.505	0.618
Latencies	0.036	0.360	large	−6.085 (9.498)	−0.641	0.527	−6.201 (9.363)	−0.662	0.514
All participants (using z-transformed testosterone)
Error rates	0.007	0.136	medium	−0.017 (0.024)	−0.695	0.490	−0.017 (0.023)	−0.746	0.458
Latencies	0.012	0.142	medium	5.309 (6.033)	0.880	0.382	5.352 (6.018)	0.889	0.377
Estradiol level – linear regression
Females
Error rates	0.005	0.253	large	−0.001 (0.003)	−0.382	0.705	0.000 (0.002)	0.189	0.852
Latencies	0.128	0.281	large	−1.195 (0.589)	−2.029	0.052+	−1.591 (0.581)	−2.738	0.011*
Estradiol level – quadratic regression
Females
Error rates	0.005	0.323	large	0.000 (0.000)	0.068	0.947	0.000 (0.000)	−0.964	0.343
Latencies	0.143	0.360	large	0.019 (0.027)	0.690	0.496	0.046 (0.026)	1.724	0.097+

Tanner status did not predict error rates in either sex nor latencies in boys, though there was a trend for girls in mid/late puberty to respond faster than those in the pre/early stages of puberty (B = −40.634, t = −1.867, p = .073); after controlling for age, these findings were significant (B = −68.352, t = −2.336, p = .027).

Testosterone level did not predict error rates nor latencies in either sex when modeled linearly, and these findings persisted after controlling for age. However, when the relationship between testosterone levels and response latencies in boys was modeled quadratically, a trend emerged for boys to show a decrease in response latencies with increasing testosterone levels (B = 14.276, t = 1.843, p = 0.073); these findings remained a trend after controlling for age (B = 14.571, t = 1.878, p = 0.068). Testosterone levels were not associated with error rates nor latencies when modeled quadratically in girls.

Estradiol level did not predict error rates in girls when modeled linearly or quadratically. Response latencies were associated with estradiol levels at a trend level when modeled linearly (B = −1.195, t = −2.029, p = .052), and this was significant after controlling for age (B = −1.591, t = −2.738, p = .011). When modeled quadratically, estradiol levels were associated with response latencies only after controlling for age (B = .046, t = 1.724, p = .097).

Sensitivity power analyses are reported for each regression. This indicates the size of an effect that the regression was powered to detect given the sample sizes used, ß = .80, and α= .05 (Faul et al., 2007). Resulting f² effect sizes are shown in Table 1, and an analysis with an f² between 0 and 0.085 indicates power to detect a small effect, 0.086–0.250 indicates a medium effect size, and 0.251–1 indicates a large effect size (Cohen, 1988). Values indicate this study was powered to detect at least a medium-sized effect for analyses that combined boys and girls.

Discussion

The aim of this study was to examine whether pubertal maturation contributes to well-established age-related improvements in antisaccade performance during the early adolescent period. Contrary to our hypotheses and despite sufficient statistical power to detect the effects of interest, both hormonal and physical indices converged to indicate that pubertal maturation is not associated with antisaccade performance in boys nor with error rates in girls. However, advanced physical and hormonal indices were associated with faster latencies to inhibit in girls after accounting for age. Furthermore, sensitivity analyses supported the veracity of the null findings, indicating that they were not merely driven by a lack of statistical power. Importantly, we utilized a design optimized to investigate pubertal development but minimize age effects. We further utilized gold-standard metrics most sensitive to pubertal maturation with high reliabilities to empirically assess both external and internal manifestations of pubertal status: physical nurse assessments and serum estimate gonadal hormone levels.

This is the first study to our knowledge that examines the effects of pubertal maturation on inhibitory control. While we know from an extensive literature that inhibitory control (Akshoomoff et al., 2014; Davidson et al., 2006; Luciana & Nelson, 2002; Luna, 2009; Ridderinkhof & van der Molen, 1997; Williams et al., 1999), and antisaccade performance in particular (Fischer et al., 1997; Fukushima et al., 2000; Klein, 2001; Klein & Foerster, 2001; Luna et al., 2004; Munoz et al., 1998; Nieuwenhuis et al., 1999; Ordaz et al., 2010; Ordaz et al., 2013; Romine & Reynolds, 2005), improves during the early adolescent period studied here (11 to 14 years), this study provides novel results indicating that pubertal maturation might not play a critical role in these improvements, except for speeding the female response. Importantly, we assessed both gonadal hormone levels and physical sexual maturity to investigate the multiple mechanisms by which antisaccade performance could plausibly be shaped by pubertal maturation: specifically, testosterone and estradiol, acting on brain networks implicated in antisaccade performance; and second, external manifestations of puberty that, in addition to serving a proxy for internal gonadal hormone levels, have social signaling effects.

Our results generally demonstrating the lack of association between gonadal hormone levels and antisaccade performance provide behavioral insight to an equivocal neuroimaging and animal literature that testosterone can influence function of cortical and subcortical networks implicated in antisaccade performance. Animal studies indicate the presence of androgen receptors in relevant cortical areas on which testosterone can act, though they are relatively sparse compared to affective/motivational areas of the brain (Kato, 1977; Motta & Martini, 1984). Further, developmental neuroimaging studies have shown that genotypic variability in androgen receptor, which relates to capacity for testosterone to exert effects on the brain, alters the rate of thinning in inhibitory control-relevant areas of cortex (Raznahan et al., 2010). Data from this sample and others suggest that individual differences in testosterone levels relates to inter-individual variability in cortical thickness and white matter integrity within inhibitory control networks, and testosterone has its greatest effects on brain development during the developmental period studied here (Bramen et al., 2012; Bramen et al., 2010; Goddings et al., 2014; Herting et al., 2015; Herting et al., 2012; Neufang et al., 2009; Nguyen, McCracken, Ducharme, Botteron, et al., 2013; Nguyen, McCracken, Ducharme, Cropp, et al., 2013; Peper et al., 2009).

Thus while plausible mechanisms exist by which circulating testosterone can impact brain networks, our results suggest that testosterone’s effects may not penetrate to the level of cognitive function, at least during early adolescence. Studies indicate that the organization and segregation of brain networks, including those that underlie cognitive control, are stable by childhood with only inter-network integration increasing (Marek, Hwang, Foran, Hallquist, & Luna, 2015), and these inter-network connections may be less impacted by testosterone. We do not know, however, if the effects of increased testosterone levels during puberty, which occurs in early adolescence, may be manifested later in adolescence, as perhaps still-maturing brain systems may need prolonged exposure to testosterone in order to influence cognitive performance. Indeed, there is evidence that it takes time for circulating testosterone levels during this pubertal/early adolescent period to influence behavioral outcomes (Tarter et al., 2007). Longitudinal research would be necessary to test a delayed onset hypothesis in which testosterone levels in early adolescence could influence antisaccade performance in late adolescence.

We also found that higher circulating estradiol levels were associated with faster latencies in girls. This could reflect the influence of white matter maturation as myelination speeds information processing. It is well-established that white matter matures earlier in girls as compared to boys (Asato, Terwilliger, Woo, & Luna, 2010; Giedd et al., 1999), presumably as a result of earlier pubertal onset in girls. Further, there is specific evidence that estradiol levels are associated with white matter volume (Herting et al., 2014) and white matter integrity (Herting et al., 2012) in human adolescents. In addition, rodent research employing experimental manipulations of ovarian steroids reveals that ovarian steroids contribute to changes in frontal cortex white matter volumes (Koss, Lloyd, Sadowski, Wise, & Juraska, 2015) and white matter integrity (Prayer et al., 1997).

Secondly, social signaling effects (both to oneself and others) could alter inhibitory control by changing an individual’s own self-perceptions, changing how others (parents, teachers, peers) interact with an individual, and/or provide opportunities that would not otherwise be available without a visible degree of physical maturity (e.g., dating, greater responsibilities, athletic involvement). Such sophisticated, challenging social contexts may provide more opportunities to engage inhibitory control, and facilitate its maturation. In our sample, Tanner status, which assesses an external manifestation of pubertal maturation that can serve as a social signal, was associated with antisaccade response latencies in girls but not boys. This female-specific effect is consistent with evidence that advancing pubertal development (as measured by external manifestations of pubertal development), especially relative to one’s peers, is associated with changes in social and aggressive behaviors only in girls; it has been suggested that the mechanism by which this operates is through social signaling effects (Skoog & Stattin, 2014). Take together with our findings, it is possible that girls may be more sensitive than boys to the effects of social signaling. Future research to explicitly test whether these additional opportunities afforded be external manifestations of pubertal maturation is necessary to draw such conclusions. Nonetheless, it is important to note that social signaling effects do not impact error rates in either sex, which reflect the capacity to generate an appropriate inhibitory response.

The finding that neither internal nor external manifestations of pubertal maturation are a major contributor to developmental improvements in antisaccade error rates raises the question: what generates these developmental improvements? Neuroimaging research using the antisaccade task suggests that maturation of brain connectivity, brain networks, and white matter maturation (Fjell et al., 2012; Hwang et al., 2010; Marek et al., 2015; Stevens, Kiehl, Pearlson, & Calhoun, 2007), combined with the capacity to sustain a task set (Velanova et al., 2009), and maturation of brain regions implicated in error monitoring (Ordaz et al., 2013; Rubia, Smith, Taylor, & Brammer, 2007; Velanova et al., 2008) facilitate developmental changes in antisaccade performance and in other inhibitory control tasks. Taken together with other studies relating pubertal maturation to brain development, these findings suggest that puberty has a limited role in facilitating the capacity to generate a goal-consistent inhibitory response. Instead, other processes such as biologically-programmed cellular events (e.g., neuronal signaling, synaptogenesis, dendritic pruning, myelination)(Luna, Marek, Larsen, Tervo-Clemmens, & Chahal, 2015) and frequent opportunities to successfully exert inhibitory control in service of one’s goals may instead lead to these brain changes, which in turn shape the maturation of inhibitory control. It also merits consideration that pubertal maturation may be associated with antisaccade performance in high-affect or motivationally salient contexts. That is, pubertal maturation may influence inhibitory control when it interacts with affective/reward systems.

A number of limitations of this study should be acknowledged. First, our study was limited to the antisaccade task, as we focused on the antisaccade task and did not collect other cognitive or temperamental measures of inhibitory control; thusour capacity to generalize conclusions beyond this task is limited. Future studies utilizing other metrics of inhibitory control (go/no-go task, stop-signal task, etc.) can reveal whether these findings are consistent across metrics and may therefore generalize to the broader construct of inhibitory control. Second, while our focus on a limited, early-pubertal age range allowed us to maximize pubertal-related variability relative to age-related variability, this prevented us from examining whether the effects of pubertal maturation are manifested in late puberty or may be delayed. Complementary studies examining the role of pubertal status on inhibitory control performance over a wide age range or longitudinally could address these questions. Third, we limited our investigation to the gonadal hormones testosterone and estradiol, but other hormones such as the adrenarcheal hormones DHEA and DHEAS were not investigated. As the changes in levels of these hormones occurs earlier, their effects may be more evident than those of gonadal hormones in the developmental range studied here.

Taken together, this study examining multiple mechanisms by which puberty could impact antisaccade performance indicates that pubertal maturation may contribute to the speed at which one can initiate an inhibitory response in girls, but underscores that non-pubertal factors are dominant in driving the maturation of the capacity to generate an inhibitory response. As the capacity to develop mature inhibitory control in adolescence has relevance for risk for developing psychopathology in adolescence, likelihood of engaging in risk-taking behaviors, and for the development of positive academic and social outcomes, it is essential to understand mechanisms supporting its development. Future studies are needed to further probe girls’ sensitivity to the effects of pubertal maturation of reaction times, to test whether puberty has a delayed impact on inhibitory control error rates or on any aspect of inhibitory control in boys, whether arousal can undermine inhibitory control, and whether experience and biologically programmed events lead to improvements in inhibitory control.

Research Highlights.

• While multiple lines of evidence suggesting that inhibitory control may be shaped by pubertal development, research has yet to examine whether pubertal maturation contributes to well-established developmental improvements in inhibitory control

• This study utilized a design optimized to investigate the role of pubertal maturation independent of age and collected gold-standard metrics (nurse assessed Tanner Stage and serum estimate testosterone levels) of pubertal development.

• Contrary to our hypotheses and despite sufficient statistical power to detect the effects of interest, both hormonal and physical indices converged to indicate that pubertal maturation is not associated with any metric of antisaccade performance in boys but is associated with latencies in girls.