1. Introduction

Sensation seeking is a personality trait associated with a propensity to seek out novel, highly stimulating experiences, and the willingness to take risks in pursuit of these experiences (M. Zuckerman & Kuhlman, 2000). Compared to low sensation seekers, individuals high in sensation seeking ascribe fewer consequences and less danger to their involvement in risky activities (Horvath & Zuckerman, 1993; M. Zuckerman & Kuhlman, 2000). In some instances, such as engagement in some acts of courage and bravery, evidence suggests that sensation-seeking may play an important role in promoting these functional behaviors (Neria, Solomon, Ginzburg, & Dekel, 2000). However, an extensive body of research has also documented links between sensation seeking and an array of deleterious outcomes such as severe substance use impairments (Ball, Carroll, & Rounsaville, 1994), comorbid psychopathology (Slutske et al., 2002), poorer treatment outcomes (Staiger, Kambouropoulos, & Dawe, 2007), and increased rates of physical injury and mortality (Cservenka, Herting, Seghete, Hudson, & Nagel, 2013). The high societal costs attributed to this facet of personality have led to its recognition as an important public health concern (Steinberg, 2008), and further underscores the need to improve our understanding of its development.

Sensation seeking has a genetic basis (Joseph, Liu, Jiang, Lynam, & Kelly, 2009; Terracciano et al., 2011), with heritability estimates generally ranging from 40% to 60% (Stoel, De Geus, & Boomsma, 2006). In addition, shared genetic factors underlie sensation seeking and externalizing psychopathology such as substance dependence and conduct disorder (Harden & Mann, 2015; Mann et al., 2016). Changes in the incentive processing system, particularly increased responsivity to reward, likely plays a key role in explaining individual differences in sensation seeking (Dalley et al., 2007; Harden & Mann, 2015; M. Zuckerman, 1984).

Research investigating neural sensitivity to rewards as a marker of sensation seeking has largely focused on key regions in the mesolimbic dopamine (DA) system, particularly the nucleus accumbens (NAcc) (Bornovalova et al., 2009; Ikemoto & Panksepp, 1999). Findings across basic and clinical science indicate that the DA system undergoes dramatic alterations beginning around the onset of puberty (Galvan, 2010; Wahlstrom, Collins, White, & Luciana, 2010). For example, animal research has shown that DA levels in the striatum increase during adolescence (Andersen, Dumont, & Teicher, 1997) and that the firing rate of DA neurons increase at the beginning in adolescence, peak during late adolescence, and decline into adulthood (McCutcheon & Marinelli, 2009). Similarly, research in humans suggests that compared to other age groups, adolescents appear to experience particularly exaggerated reactivity to rewards in the NAcc (Galvan, 2010; C. F. Geier, Terwilliger, Teslovich, Velanova, & Luna, 2010; Luciana, Wahlstrom, Porter, & Collins, 2012; Beatriz Luna, Paulsen, Padmanabhan, & Geier, 2013; Padmanabhan, Geier, Ordaz, Teslovich, & Luna, 2011; Paulsen, Hallquist, Geier, & Luna, 2015), a region that is a principal target of DA neurons (Haber & Knutson, 2010; Ikemoto & Panksepp, 1999).

Accumulating evidence suggests NAcc response to reward may be related to increased sensation seeking during late adolescence and adulthood. There is a particular interest in delineating this association in adolescence given that sensation seeking is often found to peak at this time (Steinberg, 2008). Initial studies report somewhat discordant findings. Using a monetary incentive delay task (MID) among a sample of 26 adolescents (ages 12–16), Bjork, Knutson, and Hommer (2008), found a positive relationship between NAcc activation to anticipated rewards and sensation seeking (p < .05). Similarly, using a rewarded delayed response task with children (ages 7–11) and adolescents (ages 13–17), Galvan, Hare, Voss, Glover, and Casey (2007) found an association between NAcc response and risk perception in adolescents but not in children. Finally, Cservenka et al. (2013) using a modified version of the Wheel of Fortune task found no differences in NAcc activity to reward feedback between High (n = 27) and Low (n = 27) sensation seeking adolescents (mean age = 13.94, SD = 1.05). In sum, although a direct relationship between NAcc reward response and sensation seeking has yet to be clearly established; if such a relationship does exist, it appears this association may vary by both task and age.

Emerging literature indicates that the role of the mesolimbic system may change through development as several lines of evidence suggest that there may be a peak in dopamine availability (Padmanabhan & Luna, 2014; Wahlstrom et al., 2010). fMRI studies have reported that activity in NAcc may be different in children than at older ages. Using a rewarded cognitive control task, Paulsen et al. (2015) found increased NAcc activation in children even in the absence of rewards was associated with better cognitive control, while greater reliance on NAcc later in development was associated with decreased control. Continued predominant reliance on NAcc at later ages may reflect persistent immaturities in transitioning to reliance on cognitive control regions. However, probing the predictive significance of neuroimaging data on later outcomes has been difficult. Typically neuroimaging analyses consider brain activation as the outcome of interest (i.e., dependent variable). Recently, more studies have begun to implement a brain-as-predictor approach, wherein brain activation is modeled as an independent variable that can be used to predict longitudinal outcomes. In this approach, regions of interest are defined a priori based on strong empirical and theoretical evidence. Next, neural activation in the a priori specified region or regions is measured, followed by the collection of longitudinal data on the outcome of interest (see Berkman & Falk, 2013 for a more detailed review). In this manner, an analytic model can be specified to probe the predictive validity of the hypothesized brain region to predict the longitudinal outcome data. This methodology provides a more direct manner for testing theoretical models, that can inform ecologically validity and long-term outcomes (Berkman & Falk, 2013; Gonzalez & Berman, 2010).

Current Study

The current study investigated activation in NAcc, given its well established role in reward processing, as a biomarker of sensation seeking across development. Reward sensitivity was assessed in 139 participants ranging in age from 10- to 25-years-old using an incentivized antisaccade task. Sensation seeking was assessed at the time of the scan and then at two follow-ups conducted one-and-a-half years apart. Reward sensitivity was specified as a predictor of trait sensation seeking using a latent variable modeling framework. Developmental effects were investigated by examining the moderating influence of participant’s age on the association between reward related brain activity and sensation seeking. Given previous evidence suggesting that the association between NAcc activation and markers of risk taking may differ with development (Galvan et al., 2007; Paulsen et al., 2015), we hypothesized that reward activation in the NAcc would similarly show a different association with sensation seeking across age.

2. Materials and Methods

2.1 Participants

Data was collected as part of an ongoing neuroimaging study focused on characterizing the neural basis of cognitive control across development. Participants included a community based sample of healthy volunteers from Pittsburgh and surrounding areas who underwent behavioral and neuroimaging testing. Volunteers were native English speakers screened by phone to ensure no neurological, psychiatric, or eye movement problems, a negative history of medications known to affect brain function or eye movements, and no first degree relatives with schizophrenia or bipolar disorder. In addition, all participants had far visual acuity of at least 20/40 (corrected or uncorrected) and were screened for non-removable metal on the body, claustrophobia, and weight under 300 pounds to ensure scanner eligibility.

This study initially included 171 participants, with the sample being further refined by only including cases which met the following inclusionary criteria: runs with fewer than 15% volumes having greater than .3 mm head motion between volumes; visits sharing 90% of whole-brain coverage with all other subjects; subject visits with greater than 50% accuracy in antisaccade performance per condition; visits with 20 or more correct antisaccade trials per condition; and subjects having available data on the sensation seeking measure. This resulted in a final sample of size of 139 participants (76 females) ranging in age from 10- to 25-years-old (M=16.43, SD=3.77). Participants were assessed for IQ using the Wechsler Abbreviated Scale of Intelligence (WASI; Wechsler, 1999), and demonstrated scores in the normal range (M=110.99, SD=11.64). There were no associations found between any study control variables (i.e., gender, IQ, ethnicity), however, participant’s gender was related to the study outcome (B = −.21, p < .05) and therefore controlled for in all study analyses. No differences were found between participants included in this study’s final sample when compared to those meeting exclusionary criteria in terms of IQ (t(169) = −1.13, p > .05), gender (χ²(1) = .973, p > .05), ethnicity (χ² = 1.67(1), p > .05), or scores on the sensation seeking measure (t(169) = −.244, p > .05). However, on average, excluded participants (M = 14.73, SD = 3.39) were younger than those included in the final study sample (M = 16.44, SD = 3.75; t = −2.60(169), p < .05). All experimental procedures in this study were approved by the Institutional Review Board at the University of Pittsburgh. Participants were paid US $75 for their participation in the study in addition to any performance-related earnings.

2.2 Procedures

Immediately prior to scanning, participants were asked to choose a reward to work towards based on their performance, which included a series of gift cards to retail locations or $25 cash. Participants were then instructed to rate how ‘valuable’ (7-point Likert scale) they considered their chosen reward and to indicate at least one item they might purchase with this compensation as a means to increase the salience of the reward. Prior to each trial, participants were informed whether they could win, lose, or neither win nor lose points on the upcoming trial by a set of colored bars. Potential reward trials were indicated by green filled bars above a central cross-hair and potential loss trials were indicated by red filled bars located below the central fixation. Neutral trials, where the subject’s performance would not affect the number of points they earned, were indicated by unfilled bars.

Subjects were informed that the number of filled bars would correlate with the number of points at stake for the upcoming trial and that the computer would automatically calculate these points, facilitating the subject to focus on the task. Following each trial, audio feedback was provided to indicate correct and incorrect responses. Subjects were remunerated based on the proportion of points earned out of a total of 280 using the following scale: 0–70 points (US $10), 71–140 (US $15), 141–210 (US $20), 211–280 (US $25.00 or gift card). This point-based approach allowed a separation between trial outcomes and dollar amounts, which was reasoned to more adequately adjust for potential differences in the subjective value of dollar amounts across age. This method has been previously used in a behavioral implementation of the current paradigm (Charles F Geier & Luna, 2012). Aspects of this dataset have been published elsewhere (see Paulsen et al., 2015).

2.3 Measures

Trait Sensation Seeking

The self-reported Sensation-Seeking Scale (SSS-V; M. Zuckerman, 1994, 2007), was used to assess trait-like levels of sensation seeking in the current study. The scale consists of 40 items that contain two different descriptive statements (e.g., I like to go off the high board vs. I do not like to go off the high board) and participants are asked to indicate which statement best describes them. This instrument was first administered to participants at the time of their fMRI scan and additionally during two follow-up evaluations spaced approximately 1.5 years apart (M = 1.66 years, SD = .30; M = 1.57 years, SD =. 31). The scale consists of a total sensation seeking score, as well as four 10-item subscales: 1) Thrill and Adventure-Seeking (TAS; “I often wish I could be a mountain climber”), 2) Experience-Seeking (ES; “I like to try new foods I have never tasted before”), 3) Disinhibition (DIS; “I rarely if ever do anything crazy”), 4) Boredom Susceptibility (BS; “I can’t stand watching a movie that I’ve seen before”). Across the three assessment waves, average total scores on this measure were 18.21 (SD = 6.61), 19.25 (SD = 7.11), and 18.76 (SD = 6.79) and reliability ranged from α = .70 −.73.

2.4 Design

The design for this incentivized antisaccade task has been described in detail elsewhere (Paulsen et al., 2015). Subjects were informed that they would see a cue indicating whether correct performance would result in a gain of points (Reward trials), incorrect performance would incur a loss of points (Loss trials), or neither correct nor incorrect performance would affect accrual of points (Neutral trials). Reward and Loss trials were worth +/− 5 points, respectively, which was indicated by the number of filled green or red bars appearing above or below the pre-target central fixation cue. Following each 1.5 second central fixation cue was a 1.5 second preparatory epoch, followed by a 1.5 second saccade event. The display of the saccade event contained a small yellow dot at one of six pseudo randomized peripheral locations; subjects were required to saccade away to the opposite location from the dot upon presentation. Immediately after the saccade, subjects heard a cash register sound for correct responses or a buzzer sound for incorrect responses. Intertrial intervals varied from 1.5–19.5 seconds following an exponential distribution. A total of 56 trials for each Reward, Neutral, and Loss condition were presented across 4 runs. An additional 72 partial trials with either a cue alone or cue and preparatory epoch without a saccade event were also presented to estimate better the hemodynamic response to each event type in other analyses (Ollinger, Corbetta, & Shulman, 2001; Ollinger, Shulman, & Corbetta, 2001). Similar to prior research using this task (e.g., Paulsen, Hallquist, Geier, & Luna, 2014), we collapsed across Cue, Delay, and Response epochs to gain more power in identifying our effects of interest.

2.5 Data Acquisition

Eye-tracking data were collected using a long-range optics eye-tracking system from Applied Science Laboratories (Model 504LRO; Bedford, MA). Eye-position was obtained via pupil-corneal reflection observed in the reflection of a head coil-mounted mirror with 0.5° of visual angle. Video monitoring was also used to ensure compliance. A 9-point calibration was performed prior to the experimental session and between runs when necessary. Stimuli were presented using E-prime software (Psychology Software Tools, Inc., Pittsburgh, PA) and projected onto a flat screen behind the scanner, visible to the subject through the coil-mounted mirror. To provide performance feedback, eye-data were scored on-line using EPrime software. Additionally, eye-data were scored off-line using an in-house automated scoring package developed in R to heuristically match manual scoring. This scoring package has been used in previous analyses (Paulsen et al., 2014).

Correct responses in the antisaccade task were defined as those in which the first eye movement during the saccade epoch with velocity greater than or equal to 30°/s (Gitelman, 2002) was made toward the mirror location of the peripheral cue and extended beyond a 2.5°/visual angle from central fixation. Incorrect responses occurred when the first saccade during the saccade epoch was directed toward the peripheral stimulus and exceeded the 2.5°/visual angle central fixation zone but were subsequently directed to the correct location, indicating that the instructions were being followed. Trials in which no eye movements were generated, or in which the tracker lost fixation, were excluded from analyses, resulting in 14% of trials being excluded.

2.6 MRI Data Acquisition

Imaging data were collected using a 3.0-T Siemens Trio scanner at the Magnetic Resonance Research Center, University of Pittsburgh, Pittsburgh, PA. A single-shot echo-planar imaging sequence sensitive to BOLD contrast (T2*) was performed. The acquisition parameters were TR = 1.5 s; TE = 25 ms; flip angle = 70°; in-plane resolution of 3.125 mm. Twenty-nine 4-mm thick axial slices with no gap were collected, aligned to the anterior and posterior commissure and covering the entire cortex and part of the cerebellum. A 3D volume magnetization prepared rapid acquisition gradient-echo (MP-RAGE) pulse sequence with 192 slices (1-mm slice thickness) was used to acquire structural images in the sagittal plane.

2.7 fMRI data preprocessing

Anatomical scans were registered to the MNI152 template (Fonov, Evans, McKinstry, Almli, & Collins, 2009) using both affine (FSL FLIRT) and nonlinear (FSL FNIRT) transformations. Functional images were preprocessed using tools from NiPy (Millman & Brett, 2007), AFNI (Cox, 1996), and the FMRIB software library (Smith et al., 2004). First, large transient spikes in voxel time series were interpolated downward using the AFNI 3dDespike program. Second, slice timing and motion correction were performed simultaneously using a four-dimensional registration algorithm implemented in NiPy (Roche, 2011). Non-brain voxels were removed from functional images by masking voxels with low intensity and by a brain extraction algorithm implemented in FSL BET. The alignment of subjects’ functional images to their anatomical scan was computed using the white matter segmentation of each image and a boundary-based registration algorithm (Greve & Fischl, 2009). Functional scans were then resampled into 3mm isocubic voxels and warped into MNI152 template space using the concatenation of the functional-structural and structural-MNI152 transforms. Images were spatially smoothed using a 5-mm full-width at half maximum kernel (FSL SUSAN). A .008Hz temporal high-pass filter was applied to remove slow-frequency signal changes.

2.8 First-Level Analyses

Fixed-effects analyses were run using FSL FEAT to generate parameter estimates (PE) at the visit level for each subject for Reward (Rew), Neutral (Neu), and Loss conditions, using the jittered intertrial interval as baseline. Nuisance regressors included the time-courses of two voxels from the right and left lateral ventricles to account for physiological noise, the six motion regressors used in motion correction, and the convolved regressors from error trials, trials that could not be rated due to missing eye-tracking data, or those without a saccade (i.e. partial trials). Thus, only correct trials were used in the analyses, with an event duration of 4.5 s to model the Cue, Preparatory, and Saccade epochs.

2.9 Group-Level Analysis

As the focus of the current study is on reward responsivity, only contrasts with the reward condition of the antisaccade task were examined (i.e., the loss condition was not included in any analysis). The primary contrast of interest was the reward > neutral antisaccade trial (i.e. reactivity to reward stimuli)1. Group analysis of this contrast was performed using a random effects analysis following the FSL FLAME mixed effects approach, which propagates estimation error from single-subject analyses into estimates of group effects (Woolrich, Behrens, Beckmann, Jenkinson, & Smith, 2004). Due to its involvement in various aspects of reward processing and sensation seeking behaviors, our region-of-interest (ROI) analyses were conducted on the bilateral NAcc using an anatomical mask from the Harvard-Oxford sub-cortical structural atlas provided in FSL. Mean BOLD parameter estimates were then extracted for each individual subject data (FSL featquery command) from the right and left NAcc. These values reflect individual differences in the magnitude of neural activation between reward and neutral trials within the NAcc.

In addition to ROI analyses, a voxel-wise exploratory whole-brain (WB) analysis was conducted for the Reward > Neutral contrast. To guard against false positives, a minimum cluster size of 52 contiguous voxels (voxelwise p < .001, cluster threshold of p < .05) was identified for the whole-brain analysis using AFNI’s AlphaSim. Effects reported beyond our a priori ROIs are included only for descriptive purposes.

2.10 Associations with Sensation Seeking

Analyses then examined whether the mean BOLD response to reward in the NAcc was related to trait-levels of sensation seeking, and whether this association varied by age. To assess stable levels of sensation seeking across time, a second-order latent trait model was specified using the 3 repeated assessments of sensation seeking2. In this model, subscales of the SSS-V (i.e., TAS, ES, DIS, BS) were used as indicators of a latent sensation seeking construct at each of the three assessments. These first-order latent sensation seeking factors were then used as indicators of a second-order factor representing time-stable (trait-level) individual differences in sensation seeking across all three time points. After the latent factor model was specified, the average BOLD response to reward, age and their interaction were included as predictors of the higher-order trait sensation seeking construct (for a visual depiction, see Figure 1). All analyses controlled for participant’s gender.

Figure 1 Neural Activation × Age Interaction Predicting Trait

Models were estimated using robust maximum likelihood estimation (MLR) in Mplus 7 (Muthén & Muthén, 1998–2012). Global fit indices used to examine model fit included the comparative fit index (CFI) and the root mean square error of approximation (RMSEA). Cutoff values of .90 or greater were used to indicate acceptable fit and .95 or greater to indicate good fit for CFI (Hu & Bentler, 1999; McDonald & Ho, 2002). RMSEA values between .05 and .10 were considered to represent an acceptable fit, while values less than .05 were considered to indicate good fit (Browne & Cudeck, 1993; McDonald & Ho, 2002).

In addition, longitudinal invariance testing was conducted to determine if the sensation seeking construct was being measured in a consistent manner across the study window. First, a baseline configural model was specified, allowing factor loadings and intercepts to vary across time. Subsequent models were specified by adding parameter constraints, until arriving at a final invariance model with factor loadings, intercepts, factor covariances, and item residual variances constrained to equality across each assessment wave. These models were compared via (Satorra-Bentler) chi-square difference testing, wherein a non-significant chi-square indicates that the more constrained model should be adopted. However, as the chi-square difference test is sensitive to sample size and violations of the normality assumption, we also assessed change in CFI and RMSEA fit indices across models. Changes in CFI equal to or less than .01 and changes in RMSEA of equal to or less than .015 have been proposed as demonstrating evidence of model invariance (Chen, 2007; Cheung & Rensvold, 2002).

3. Results

3.1 Behavioral

Study participants’ overall correct response rate (‘accuracy’) was high for the Reward (M = .92, SD = .08) and Neutral (M = .89, SD = .11) conditions, however, accuracy was significantly greater for Rewarded trials (t(135) = 4.09, p < .001). Similar to previously reported findings in research that has spanned a large developmental window (e.g., B. Luna, Garver, Urban, Lazar, & Sweeney, 2004; Paulsen et al., 2014), accuracy rates also improved with age and followed an inverse trajectory for both Reward (B = .45, t(1, 138) = 5.83, p < .001; R² = .20) and Neutral (B = .42, t(1, 138) = 5.39, p < .001; R² = .18) trials (see Figure 2). There was no evidence of an association between accuracy rates and sensation seeking for the reward (r = .12, p = .18) or neutral condition (r = .07, p = .40), or any evidence that age moderated this association.

Figure 2 Antisaccade Performance during Reward and Neutral Trials

Saccade latencies for correct responses were also examined to assess participants’ reaction time. Similar to the accuracy rate findings, a significant difference was found between Reward and Neutral conditions. Reaction times during Rewarded trials (M = 427.86, SD = 66.34) was significantly less than during Neutral trials (M = 436.76., SD = 70.41; t(135) = −4.21, p < .001). In addition, these results also indicated that an inverse trajectory provided an appropriate fit to the data, whereby reaction times decreased across age for the Reward (B = −.43, t(1, 138) = −5.44, p < .001; R² = .18) and Neutral (B = −.40, t(1, 138) = −5.02, p < .001; R² = .16) conditions (see Figure 2). Finally, as with accuracy rates, there was no evidence of any relationship between latency and sensation seeking in the reward (r = −.08, p = .38) or neutral conditions (r = −.05, p = .58).

3.2 Latent Trait Modeling of Sensation Seeking

Prior to investigating associations between neural activation and participant’s levels of sensation seeking, longitudinal invariance of the sensation seeking construct was examined. The baseline configural model was found to provide a good fit to the data (χ² = 45.64, df = 39, p = .215; CFI = .987, TLI = .979, RMSEA = .035). Similarly, the more constrained, final invariance model was also found to provide good model fit (χ² = 72.91, df = 61, p > .05; CFI = .977, TLI = .976, RMSEA = .037; also see Table S1). Comparing these two models revealed negligible differences among the absolute fit indices (i.e., CFI, TLI), while results of chi-square difference testing also indicated there was no significant difference between these models (χ² = 27.31, df = 22, p > .05). Therefore, these results were considered to support the longitudinal invariance of the sensation seeking construct in the current study.

The higher-order latent trait model of sensation seeking fit the data well (χ² (54) = 66.30, p = .121, RMSEA = .040, CFI = .977; see Figure 1). Analyses were also conducted to compare the fit of this latent trait model to that of a higher-order latent growth curve model. A comparison of these nested models provided evidence that the latent trait model fit the data better (BIC = 4615) compared to the growth curve model (BIC = 4924). Subsequent to these model testing steps, we investigated whether there was an association between brain activation to reward and trait-like levels of sensation seeking, as well as if an age effect moderated this relationship.

3.3 Functional Imaging Results

Mean ROI activation from the reward > neutral contrast was extracted for the right (MNI coordinates: 40, 72, 33) and left (MNI coordinates: 48, 67, 36) NAcc. There was no evidence of a linear or (inverse) quadratic association between participant’s age and reward activation in either the right (linear- β = .05, p = .53; inverse- β = −.02, p = .79) or left (linear- β = .09, p = .25; inverse- β = −.05, p = .54) NAcc regions. This is similar to our previous study using overlapping data (Paulsen et al., 2015), where we found that NAcc engagement during reward was robust across ages undermining the ability to uncover age effects.

In addition, results indicated there was a main effect of age on trait sensation seeking, but no main effect of NAcc neural activation predicting this outcome (see Table 1 and Table S2)3. However, further analyses revealed a significant age-by-activation interaction in the prediction of sensation seeking for the right NAcc (b = .14, SE = .09, β = .20, p < .05)4. An examination of Akaike (AIC) and Bayesian (BIC) Information Criterion also demonstrated improved model fit when this interaction term was included in the model (AIC = 4925; BIC = 4915) compared to a model in which it was not included (AIC = 4927; BIC =4918). At earlier ages, higher levels of reward-related activation in the right NAcc were associated with lower sensation seeking. However, at later periods of development an opposite pattern emerged. That is, reactivity to reward was associated with higher sensation seeking during later periods of development (see interaction plot in Figure 4). The overall R² for this model without the interaction term included was .16, while inclusion of the interaction term increased the value of R² to .21 (ΔR² = .05). This interaction was not found for the left NAcc (β = .15, p = .07). To place these findings into a more developmentally informative context, we further probed this interaction effect using the PROCESS macro (Hayes, 2015) to employ the Johnson-Neyman statistical approach (Johnson & Fay, 1950; Potthoff, 1964). This technique identifies ranges of a moderator variable (e.g., age) for which the influence of the predictor (e.g., NAcc activation to reward) on the outcome is statistically significant. This analysis indicated that prior to 12 years of age, increased NAcc reactivity was significantly associated with lower levels of sensation seeking, whereas at ages greater than 17, NAcc activation was significantly associated with increased sensation seeking (see Figure 4).

Figure 4 Regions of Significance PlotNote. Johnson-Neyman regions of significance plot. The figure shows the simple slope on the y-axis for the relationship between RNAcc activation and Trait Sensation Seeking, plotted against participant’s age on the x-axis. The dashed lines represent the upper and lower limit of the 95% confidence intervals surrounding the simple slopes (solid line) at each level of the moderator (age). The vertical solid lines represent the value of age at which the lower and upper limits of the confidence interval crosses the zero point and the relationship between RNAcc activation and Trait Sensation Seeking becomes significant. n’s = the number of participants within each of the age bins from the region of significance plot.

Table 1

Main Effects and Interactions of Neural Activity and Age Predicting Sensation Seeking

Open in a separate window Notes: Above analyses control for participant genderTAS = Thrill and Adventure-Seeking Subscale; ES = Experience-Seeking Subscale; DIS = Disinhibition Subscale; BD = Boredom Subscale

3.4 Exploratory Analyses

Exploratory whole-brain analyses also supported the NAcc findings from our ROI analysis. A significant cluster of neural activity (79 voxels) in the right NAcc and overlapping our ROI analysis exceeded whole-brain cluster correction for the Reward > Neutral condition (peak activation: 40, 66, 38, MNI coordinates; see Figure 3). As with the ROI results, an age-by-activation interaction (using the mean parameter estimate of this cluster) was found to be a significant predictor of trait levels of sensation seeking (b = .20, SE = .07, β = .25, p = .003). The whole-brain analysis also identified a small number of other regions outside of our hypothesized ROIs, that elicited significant activation in the Reward > Neutral condition. However, none of these regions demonstrated a significant main or interaction effect with participant’s age or level of sensation seeking. A complete table of results from the whole-brain analysis is available in the supplemental online materials (Table S3).Figure 3Whole Brain Activation in Reward > Neutral Contrast Notes. BOLD activation obtained from whole-brain analysis for clusters exhibiting significantly greater activity in the reward compared to neutral condition (minimum cluster size = 52 contiguous voxels; voxel-wise p < .001; cluster threshold of p < .05). Table S1 provides cluster size as well as coordinates and z-statistic for the peak voxel of each significant cluster

4. Discussion

This study characterized unique developmental effects of the relationship between sensitivity to reward and trait-levels of sensation seeking. We found that during earlier periods of development (i.e., late childhood to early adolescence), increased NAcc response to reward was linked to lower levels of sensation seeking at follow-up times. However, the period of late adolescence to early adulthood marked a transition point whereby increased reward activation was associated with higher levels of sensation seeking. This is the first empirical evidence, of which we are aware, to suggest that reward sensitivity may serve as an important biomarker that confers unique vulnerabilities for engaging in high-risk behaviors differentially across development.

The finding that NAcc reactivity exhibited differential associations with sensation seeking across development lends support to prior research suggesting an association between developmental changes in reward responsivity and cognitive control (Paulsen et al., 2014). Several studies now show that although overall NAcc response to reward cues may be similar across development, this responsivity is only associated with risk-taking behaviors in late adolescence and early adulthood, but not during childhood (Galvan et al., 2007; Luking, Luby, & Barch, 2014). This suggests that reward reactivity in the NAcc may reflect different aspects of development during earlier compared to later developmental stages. While research such as the dual systems model suggests that reward sensitivity follows a curvilinear pattern across adolescence, findings from this study indicate that important functional change may occur even if reward reactivity remains robust across development. It is also possible that changes in the relationship between NAcc reactivity and sensation seeking across development are due, in part, to changes in the functional role of DA during this time. Findings from a large body of literature suggest that there is substantial change in the DA system during adolescence, including shifts in firing rate patterns and overall levels of DA (e.g., Galvan, 2010; McCutcheon & Marinelli, 2009; Padmanabhan et al., 2011; Wahlstrom et al., 2010). It may be that these changes impact the general role of DA and/or influence the way in which neural regions within the DA system interact, which in turn affects behavioral responses.

It is also worth noting that the interaction effect found in this study appears to be driven largely by the disinhibition (DIS) and experience-seeking (ES) subscales of the sensation seeking measure. Compared to aspects of restlessness and socially acceptable means of arousal seeking (e.g., playing sports) assessed by the boredom (BS) and thrill and adventure seeking (TAS) subscales of the measure, items on the DIS and ES scales appear to tap into more atypical, high-risk behaviors such as defying authority and use of illicit substances (M. Zuckerman, 1994; M. Zuckerman & Como, 1983). However, future research that parses out these associations in a more fine-grained manner is needed before arriving at a clearer conclusion.

Together, these results provide evidence that the relationship between striatal indices of reward processing and sensation seeking increasingly couple at older adolescent ages when sensation seeking typically peaks. Earlier in development, indices of striatal sensitivity to reward processing may be a marker of developmental timing where more advanced maturation is coupled with less sensation seeking.

4.1 Limitations

This study was characterized by a number of strengths including the use of a relatively large sample, repeated assessments of study outcomes, and the incorporation of neuroimaging data into an SEM framework to assess prospective prediction. However, several limitations should also be noted. First, this study focused on a convenience sample of healthy participants and may not generalize to at-risk populations. Non-generalizable samples have the potential to distort aspects of brain-behavior relationships (Falk et al., 2013) and youth in the current study may not be representative of those most likely to engage in high risk behaviors. Some research suggests this is particularly important to keep in mind when interpreting findings related to clinically significant adolescent risky behavior that are based on brain responsivity of neurotypical subjects (for more detailed consideration of this topic, see Bjork & Pardini, 2015). In addition, the current investigation did not examine the relationship between sensation seeking and reward reactivity across distinct task specific phases of reward processing (e.g., anticipation and receipt of rewards). Although findings across several fMRI (for a review, see Plichta & Scheres, 2014), and PET studies (Leyton et al., 2002; Lind et al., 2005) have revealed no general differences in this pattern of associations (but also see Richards, Plate, & Ernst, 2013), future research specifically focused on disaggregating and testing these effects is needed.

The latent trait model of sensation seeking included data collected during the T1 assessment, at which time study fMRI data was also collected. Thus, the predictive SEM model was not fully prospective in its design and results should be interpreted with this in mind. Potential aspects of dynamic change that may stem from time-specific variations in the sensation seeking construct were not examined in this study. In addition, data from the current study was collected using an age-accelerated approach and developmental effects were assessed, in part, via the use of age data collected cross-sectionally. To help ensure that any mean shifts in sensation seeking were controlled for, sensation seeking scores from each assessment wave were regressed onto participant’s age at each phase (i.e., T1, T2, and T3). However, we note that future investigations should examine study findings in the context of a more traditionally implemented longitudinal design, consisting of a larger number of follow-up assessments. Such a design may facilitate easier implementation of methods to provide a more direct assessment of issues related to developmental change in sensation seeking, and for modeling the joint development of brain activation and sensation seeking (e.g., bivariate latent growth curves; growth mixture modeling). Finally, the current study focused on age as a moderating influence between reward response and sensation seeking. However, future research may also wish to consider other potentially viable models, such as neural response to reward mediating the path from age to sensation seeking, or other more complex relations (e.g., moderated mediation/mediated moderation models).

4.2 Conclusions

Although sensation seeking may serve an adaptive role in some contexts, it has also been linked to an individual’s propensity to engage in numerous high risk behaviors, many of which have injurious and costly outcomes. Understanding the mechanisms that underlie sensation seeking is essential for early identification and prevention purposes. Here we show that sensation seeking appears to demonstrate a unique association with reward-related activity in the nucleus accumbens across development. Developmental findings indicate that this response during late adolescence and early adulthood may provide important insight into sensation seeking and risk taking behaviors that persist into later development. Although these findings must be replicated in other samples, this suggests that reward sensitivity exhibited in the nucleus accumbens region may serve as a marker to the propensity for sensation seeking that may inform models of atypical development including substance use disorders and delinquent behaviors.

Link to Article

1. Introduction

Sensation seeking, a personality trait characterized by the pursuit of novel and stimulating experiences often involving risk (M. Zuckerman & Kuhlman, 2000), presents a complex interplay of potential benefits and detriments. While linked to acts of courage (Neria et al., 2000), high sensation seeking is also associated with negative outcomes like substance abuse and psychopathology (Ball et al., 1994; Slutske et al., 2002), making it a significant public health concern (Steinberg, 2008).

Extensive research suggests a genetic basis for sensation seeking, with heritability estimates ranging from 40% to 60% (Stoel et al., 2006) and shared genetic factors connecting it to externalizing disorders (Harden & Mann, 2015; Mann et al., 2016). Heightened reward sensitivity, particularly in the mesolimbic dopamine (DA) system, particularly the nucleus accumbens (NAcc), is thought to underlie this trait (Dalley et al., 2007; Harden & Mann, 2015; M. Zuckerman, 1984).

Adolescence represents a critical period for both DA system development and sensation seeking (Galvan, 2010; Steinberg, 2008; Wahlstrom et al., 2010). Studies have shown increased DA levels and neuronal firing rates in the adolescent striatum, peaking in late adolescence (Andersen et al., 1997; McCutcheon & Marinelli, 2009). Similarly, human research indicates heightened adolescent NAcc reactivity to rewards (Galvan, 2010; C. F. Geier et al., 2010; Luciana et al., 2012; Beatriz Luna et al., 2013; Padmanabhan et al., 2011; Paulsen et al., 2015).

While a direct link between NAcc reward response and sensation seeking remains elusive, initial studies suggest a possible age-dependent association (Bjork et al., 2008; Cservenka et al., 2013; Galvan et al., 2007). Given the dynamic role of the mesolimbic system throughout development (Padmanabhan & Luna, 2014; Wahlstrom et al., 2010), understanding this relationship is crucial.

This study adopts a novel "brain-as-predictor" approach (Berkman & Falk, 2013; Gonzalez & Berman, 2010) to investigate NAcc activation as a biomarker of sensation seeking across development. This method examines the predictive validity of brain activation on longitudinal outcomes, providing valuable insights into the ecological validity of theoretical models.

Current Study

We investigated NAcc activation, a key area for reward processing, as a predictor of sensation seeking across development in 139 participants aged 10-25 years. Reward sensitivity was assessed using an incentivized antisaccade task, while sensation seeking was measured at baseline and two follow-up assessments. We hypothesized that the association between NAcc reward activation and sensation seeking would be moderated by age, reflecting the developmental changes in reward processing.

2. Materials and Methods

2.1 Participants

Data was drawn from an ongoing neuroimaging study on cognitive control. The final sample comprised 139 healthy volunteers (76 females) aged 10-25 years (M=16.43, SD=3.77) from Pittsburgh and surrounding areas. Participants underwent rigorous screening, including neurological and psychiatric evaluations, and demonstrated normal IQ scores (M=110.99, SD=11.64). No significant differences in demographic or sensation seeking scores were found between included and excluded participants, except for age (excluded participants were younger). The study protocol was approved by the University of Pittsburgh IRB, and participants provided informed consent and were compensated for their time.

2.2 Procedures

Prior to scanning, participants selected a reward (gift card or cash) based on their performance and rated its value. The incentivized antisaccade task involved three conditions: Reward, Loss, and Neutral trials. Participants earned/lost points based on their performance in Reward/Loss trials, while Neutral trials had no point consequences. Earnings were converted to monetary rewards. This point-based approach, previously validated (Charles F Geier & Luna, 2012), allows for separating trial outcomes from dollar amounts, mitigating potential age-related differences in reward value perception.

2.3 Measures

Trait Sensation Seeking

The Sensation-Seeking Scale (SSS-V; M. Zuckerman, 1994, 2007) was administered at baseline and two follow-up assessments (approximately 1.5 years apart) to assess trait sensation seeking. The scale comprises a total score and four subscales: Thrill and Adventure-Seeking (TAS), Experience-Seeking (ES), Disinhibition (DIS), and Boredom Susceptibility (BS).

2.4 Design

The incentivized antisaccade task (Paulsen et al., 2015) involved presenting cues indicating potential gains (Reward), losses (Loss), or no change (Neutral) in points. Saccade events followed a central fixation cue, requiring participants to saccade away from a peripheral target. Feedback was provided after each trial. The study analyzed only correct trials, collapsing across Cue, Delay, and Response epochs for greater power.

2.5 Data Acquisition

Eye-tracking data were collected using an Applied Science Laboratories Model 504LRO system. Correct responses were defined as saccades directed away from the peripheral cue, exceeding a specific velocity and angular displacement. Incorrect responses and trials with missing data were excluded.

2.6 MRI Data Acquisition

Imaging data were acquired on a Siemens Trio 3.0T scanner. Functional images were acquired using a T2*-weighted echo-planar imaging sequence, while structural images were acquired using a 3D MP-RAGE sequence.

2.7 fMRI data preprocessing

Standard preprocessing steps were performed using tools from NiPy, AFNI, and FSL, including spike interpolation, slice timing and motion correction, brain extraction, spatial smoothing, and temporal filtering.

2.8 First-Level Analyses

Fixed-effects analyses in FSL FEAT generated parameter estimates (PE) for each subject and condition at each visit. Nuisance regressors included physiological noise, motion parameters, and error trials.

2.9 Group-Level Analysis

Group analysis focused on the reward > neutral contrast, using a random-effects approach (FSL FLAME) to account for estimation errors. ROI analyses focused on the bilateral NAcc, using an anatomical mask. Mean BOLD PEs were extracted for each subject. A voxel-wise whole-brain analysis was also conducted.

2.10 Associations with Sensation Seeking

A second-order latent trait model assessed stable sensation seeking levels across time using Mplus 7. Subscales of the SSS-V served as indicators of latent sensation seeking factors at each assessment. These factors then predicted a higher-order trait sensation seeking factor. The model included the average BOLD response to reward, age, and their interaction as predictors, controlling for gender. Model fit was assessed using CFI and RMSEA. Longitudinal invariance testing of the sensation seeking construct was performed.

3. Results

3.1 Behavioral

Participants demonstrated high accuracy rates in both Reward and Neutral conditions, with significantly higher accuracy in Reward trials. Accuracy rates and reaction times improved with age. No significant associations were found between accuracy/reaction time and sensation seeking.

3.2 Latent Trait Modeling of Sensation Seeking

The longitudinal invariance of the sensation seeking construct was confirmed. The higher-order latent trait model provided a good fit to the data, outperforming a latent growth curve model.

3.3 Functional Imaging Results

No linear or quadratic association was found between age and reward activation in the right or left NAcc. A significant age-by-activation interaction was found in predicting sensation seeking for the right NAcc, but not the left. Increased right NAcc activation was associated with lower sensation seeking at earlier ages, but higher sensation seeking at later ages.

3.4 Exploratory Analyses

Whole-brain analyses corroborated the ROI findings, revealing a significant cluster in the right NAcc for the Reward > Neutral contrast. The age-by-activation interaction using this cluster's mean PE significantly predicted trait sensation seeking. No other significant clusters were found.

4. Discussion

This study provides novel evidence for a developmentally specific relationship between NAcc reward sensitivity and sensation seeking. We found that heightened NAcc activation predicted lower sensation seeking in late childhood and early adolescence but predicted higher sensation seeking in late adolescence and early adulthood. This suggests a functional shift in the role of reward sensitivity, potentially reflecting the evolving interplay between the reward and cognitive control systems during development (Paulsen et al., 2014).

These findings align with previous research indicating that despite consistent NAcc reward responses across ages, the association with risk-taking behaviors emerges primarily in late adolescence and adulthood (Galvan et al., 2007; Luking et al., 2014). This suggests a changing functional significance of NAcc activity throughout development.

The observed shift may also be linked to developmental changes in the DA system (Galvan, 2010; McCutcheon & Marinelli, 2009; Padmanabhan et al., 2011; Wahlstrom et al., 2010). Alterations in DA levels and neuronal firing patterns during adolescence could influence the functional role of DA and inter-regional interactions within the DA system, impacting behavioral outcomes.

Interestingly, the interaction effect appears primarily driven by the Disinhibition and Experience-Seeking subscales of the sensation seeking measure, suggesting a specific link between NAcc reward reactivity and the propensity for high-risk behaviors.

4.1 Limitations

Despite the strengths of a large sample, repeated assessments, and prospective modeling, several limitations warrant attention. First, the convenience sample of healthy participants may limit the generalizability of findings to at-risk populations. Second, the study did not differentiate between task-specific phases of reward processing, which could provide further insights. Third, while the SEM model incorporated longitudinal data, the inclusion of T1 fMRI data with concurrently collected sensation seeking measures introduces limitations to the prospective nature of the analyses. Fourth, the cross-sectional age data and age-accelerated design, while controlled for mean-level shifts in sensation seeking, would benefit from further investigation using a traditional longitudinal design with multiple follow-up assessments. Lastly, exploring alternative models examining potential mediating or more complex relationships between age, reward response, and sensation seeking would be valuable.

4.2 Conclusions

Sensation seeking, while potentially adaptive, is strongly linked to risky behaviors. Our findings suggest that NAcc reward sensitivity demonstrates a developmentally specific association with this trait, transitioning from a protective factor in late childhood to a risk factor in late adolescence and adulthood. Further research is needed to confirm these findings. However, this suggests that NAcc reward reactivity could serve as a potential biomarker for identifying individuals at risk for high-sensation-seeking behaviors, informing early intervention and prevention strategies.

Link to Article

Sensation Seeking Across Development: Examining the Predictive Significance of Neural Reward Sensitivity

1. Introduction

Sensation seeking is a personality trait that makes people crave new and exciting experiences, even if it means taking risks (Zuckerman & Kuhlman, 2000). Compared to people who are not big sensation seekers, high sensation seekers tend to downplay the potential dangers of their risky actions (Horvath & Zuckerman, 1993; Zuckerman & Kuhlman, 2000). While this trait can sometimes be beneficial, leading to acts of courage (Neria et al., 2000), it's often linked to negative outcomes like substance abuse (Ball et al., 1994), mental health issues (Slutske et al., 2002), and even physical injuries (Cservenka et al., 2013). Because of these risks, understanding how sensation seeking develops is crucial (Steinberg, 2008).

We know that genetics play a role (heritability of 40-60%; Stoel et al., 2006), and that sensation seeking shares genetic roots with problems like addiction and conduct disorder (Harden & Mann, 2015; Mann et al., 2016). One key factor might be how our brains respond to rewards, specifically the role of the 'reward system' (Dalley et al., 2007; Harden & Mann, 2015; Zuckerman, 1984), particularly a brain area called the nucleus accumbens (NAcc) (Bornovalova et al., 2009; Ikemoto & Panksepp, 1999).

The reward system goes through big changes during puberty (Galvan, 2010; Wahlstrom et al., 2010). Adolescents seem to have especially strong reactions to rewards in the NAcc (Galvan, 2010; Geier et al., 2010; Luciana et al., 2012; Luna et al., 2013; Padmanabhan et al., 2011; Paulsen et al., 2015). This area is flooded with dopamine during rewarding experiences (Haber & Knutson, 2010; Ikemoto & Panksepp, 1999).

Some studies suggest a link between NAcc reward responses and sensation seeking in late adolescence and adulthood (Bjork et al., 2008; Cservenka et al., 2013; Galvan et al., 2007). However, it's unclear if this link exists earlier in adolescence, when sensation seeking often peaks (Steinberg, 2008).

Recent research indicates that the reward system might work differently at different ages. It seems there's a peak in dopamine activity during adolescence (Padmanabhan & Luna, 2014; Wahlstrom et al., 2010), and NAcc activity varies with age (Paulsen et al., 2015).

Traditionally, brain imaging studies look at how the brain responds to something. But now, researchers are using a "brain-as-predictor" approach, where brain activity is used to predict future behaviors. This method helps us test theories and see if brain activity can forecast real-world outcomes (Berkman & Falk, 2013; Gonzalez & Berman, 2010).

Current Study

We looked at NAcc activity in 139 people aged 10 to 25 while they did a task where they could win rewards. We measured their sensation seeking at three different times, roughly 1.5 years apart. Our goal was to see if NAcc reward responses could predict future sensation seeking, and if this link changed with age. We thought the link between NAcc activity and sensation seeking would be different at different ages, based on past research (Galvan et al., 2007; Paulsen et al., 2015).

2. Materials and Methods

2.1 Participants

Our participants were 139 healthy volunteers (76 female), aged 10-25 (average age 16.43 years), from Pittsburgh. They had normal IQs, no history of neurological or psychiatric problems, and no immediate family members with schizophrenia or bipolar disorder. They were paid for their time and could earn extra rewards during the task.

2.2 Procedures

Participants chose a reward (gift cards or cash) to work towards during a computerized task. Before each trial, they were told if they could win, lose, or neither win nor lose points based on their performance. Their earnings were based on their total points.

2.3 Measures

We measured sensation seeking using the Sensation-Seeking Scale (SSS-V; Zuckerman, 1994, 2007) at three different times. The SSS-V has 40 questions measuring overall sensation seeking and four subscales: Thrill and Adventure Seeking (TAS), Experience Seeking (ES), Disinhibition (DIS), and Boredom Susceptibility (BS).

2.4 Design

Participants performed an "antisaccade" task. This means they had to look in the opposite direction of a target that flashed on the screen. Some trials were rewarded, some resulted in a loss of points, and some were neutral. We recorded their brain activity during the task using fMRI.

2.5 Data Acquisition & Preprocessing

We used an eye tracker to monitor participants’ eye movements and an fMRI scanner to measure their brain activity. The fMRI data was then processed and analyzed to isolate brain activity related to reward.

2.6 First-Level and Group Level Analyses

We analyzed brain activity in the NAcc, a key region in the brain’s reward system. First, we looked at individual brain activity and then at group averages for the contrast between reward and neutral trials. We controlled for factors like age and gender.

2.7 Associations with Sensation Seeking

We used statistical models to examine the relationship between NAcc activation during reward and sensation seeking over time. These models considered age as a potential moderator of this relationship.

3. Results

3.1 Behavioral Results

Participants performed better (faster and more accurate) when there was a reward involved. Performance on the task also improved with age. There was no relationship between task performance and sensation seeking.

3.2 Sensation Seeking over Time

Our analyses showed that sensation seeking remained relatively stable over the three assessments.

3.3 Brain Imaging Results

We found that NAcc activation to reward didn't change much with age. However, the relationship between NAcc activation and sensation seeking did change with age.

Specifically, in late childhood and early adolescence (before age 12), more NAcc activity during reward was linked to lower sensation seeking later on. But in late adolescence and early adulthood (after age 17), the opposite was true: more NAcc activity predicted higher sensation seeking.

3.4 Exploratory Analyses

Our exploratory analyses, looking at the whole brain rather than just the NAcc, supported our main findings.

4. Discussion

This study shows that the brain's response to reward can predict future sensation seeking, but this link changes with age. The NAcc, a key part of the reward system, seems to play a different role in sensation seeking at different points in development.

Our results align with previous research suggesting that the reward system changes with age (Paulsen et al., 2014). While the NAcc responds to rewards throughout development, its link to risk-taking behaviors might emerge later (Galvan et al., 2007; Luking et al., 2014).

These findings suggest that the brain's reward system matures over time. In early adolescence, strong reward responses might actually lead to less risk-taking. But as the brain matures, strong reward responses might fuel sensation seeking, especially during late adolescence when sensation seeking naturally peaks.

4.1 Limitations

It’s important to note that our study has some limitations. First, we studied healthy volunteers, so our results might not apply to everyone, especially those already engaging in high-risk behaviors. Second, we didn't look at different stages of reward processing (like anticipation vs. receiving a reward), which could be important. Third, while we measured sensation seeking at multiple time points, a longer study with more frequent assessments would be ideal. Finally, we focused on age, but other factors might also be important.

4.2 Conclusions

Sensation seeking is complex. While it can be beneficial, it’s often linked to risky behavior. Our study provides valuable insights into how the brain's reward system might contribute to sensation seeking across development.

Our findings suggest that the link between NAcc reward responses and sensation seeking changes dramatically as we move from childhood into adulthood. This knowledge could be crucial for developing better ways to identify and help individuals at risk for the negative consequences of high sensation seeking.

Note: This rewrite aims to make the article accessible to undergraduate students while staying true to the original structure and content. Some technical terms have been simplified, and explanations have been added where appropriate.

Link to Article

1. Introduction

People high in sensation seeking are drawn to thrilling experiences and don't worry too much about the potential dangers (Zuckerman & Kuhlman, 2000). Sometimes, this can be a good thing, like when someone acts bravely in a dangerous situation (Neria et al., 2000). But, there's also a downside. Research shows that sensation seeking is linked to problems like substance abuse, mental health issues, and even physical injuries (Ball et al., 1994; Slutske et al., 2002; Staiger et al., 2007; Cservenka et al., 2013). Because of these risks, it's really important to understand what causes sensation seeking and how it develops.

Scientists believe that sensation seeking is partly influenced by our genes, with about 40% to 60% being heritable (Stoel et al., 2006). This means that some people are simply wired to be more sensation-seeking than others. There's also a connection between sensation seeking and how our brains respond to rewards. The area of the brain responsible for this is called the mesolimbic dopamine system, specifically a part called the nucleus accumbens (NAcc) (Bornovalova et al., 2009; Ikemoto & Panksepp, 1999).

Interestingly, the brain undergoes significant changes during puberty, particularly in the dopamine system, which is involved in reward processing (Galvan, 2010; Wahlstrom et al., 2010). Studies on animals and humans show that teenagers experience heightened responses to rewards in the NAcc compared to other age groups (Galvan, 2010; Geier et al., 2010; Luciana et al., 2012; Luna et al., 2013; Padmanabhan et al., 2011; Paulsen et al., 2015).

Previous studies have explored the link between NAcc activity and sensation seeking, but the findings are inconsistent. Some studies show a connection, while others don't (Bjork et al., 2008; Galvan et al., 2007; Cservenka et al., 2013). It seems that the relationship might depend on factors like age and the specific task used to measure brain activity.

New research suggests that the role of the mesolimbic system in reward processing may change as we grow (Padmanabhan & Luna, 2014; Wahlstrom et al., 2010). For example, children and adults may rely on different brain regions when making decisions in response to rewards (Paulsen et al., 2015).

To better understand these changes, scientists are using a "brain-as-predictor" approach. This involves measuring brain activity and then examining how well it predicts future behavior or outcomes (Berkman & Falk, 2013). This helps researchers understand how brain activity relates to real-life behaviors.

Current Study

This study aimed to investigate if NAcc activation, a brain area important for reward, could predict sensation seeking as people age. We studied 139 participants aged 10 to 25 using a task that involved rewards. We measured their sensation seeking levels at the beginning of the study and then again twice, about one and a half years apart. We then explored if brain activity during the reward task could predict their future sensation-seeking levels and if age played a role in this relationship. Given past research, we predicted that the connection between NAcc activation and sensation seeking would differ depending on age (Galvan et al., 2007; Paulsen et al., 2015).

2. Materials and Methods

2.1 Participants

Our participants were part of an ongoing study examining the brain basis of cognitive control. They were healthy volunteers from Pittsburgh, without any history of neurological or psychiatric disorders. We made sure they had good vision, no metal in their bodies (important for MRI scans!), and no family history of schizophrenia or bipolar disorder. Our final group consisted of 139 participants (76 females) aged 10 to 25 (average age = 16.43 years). We also measured their IQ, which was within the normal range. While we didn't find differences based on gender, ethnicity, or IQ, we did find that gender was related to sensation seeking, so we controlled for this in our analyses.

2.2 Procedures

Before the brain scan, participants chose a reward to work towards, like gift cards or cash. They then rated how much they valued their reward to make it more motivating. During the scan, participants completed a task where they could win, lose, or neither win nor lose points based on their performance. The points corresponded to different amounts of money or the chosen reward. This way, we could separate how good it felt to win from the actual dollar amount, which might matter less to younger people. This method has been used in similar studies (Geier & Luna, 2012). Parts of this data have been published elsewhere (Paulsen et al., 2015).

2.3 Measures

Trait Sensation Seeking

We used a questionnaire called the Sensation-Seeking Scale (SSS-V; Zuckerman, 1994, 2007) to measure how much participants enjoyed seeking new and intense experiences. The questionnaire had 40 items and measured different aspects of sensation seeking, like enjoying thrills, trying new things, disinhibition (acting without thinking), and being easily bored. Participants filled out this questionnaire three times: during the brain scan and at the two follow-ups.

2.4 Design

We used a task called an incentivized antisaccade task, where participants had to look away from a target that flashed on a screen (Paulsen et al., 2015). Before each trial, they were told whether they could win, lose, or neither win nor lose points based on their performance. The more points at stake, the more bars they saw on the screen. This helped them focus on the task without calculating points themselves. They heard a sound indicating whether they responded correctly or not.

2.5 Data Acquisition

We tracked participants' eye movements during the task to see where they were looking. This was done using a special camera system (Applied Science Laboratories, Model 504LRO). We calibrated the system before the task and between runs to ensure accuracy. We used E-Prime software (Psychology Software Tools, Inc.) to present the task and record responses. We also double-checked and scored the eye-tracking data afterwards using a special program developed by our team.

2.6 MRI Data Acquisition

We used a powerful MRI scanner (3.0-T Siemens Trio) at the University of Pittsburgh to take images of participants' brains. We used a technique called functional MRI (fMRI) to measure brain activity during the task. fMRI works by detecting changes in blood flow in the brain, which is linked to neural activity. We also took high-resolution images of their brain structure for anatomical reference.

2.7 fMRI data cleaning

Before analyzing the fMRI data, we had to "clean" it. This involved removing noise and artifacts caused by head movements, breathing, and other factors. We used special software programs (NiPy, AFNI, FSL) to align the brain images, reduce noise, and prepare the data for analysis.

2.8 First-Level Analyses

For each participant, we examined their brain activity during the reward trials compared to the neutral trials. This gave us an idea of how their brains responded to the possibility of winning points.

2.9 Group-Level Analysis

Next, we combined the data from all participants to look for brain regions that showed increased activity during the reward trials compared to neutral trials. Our primary focus was on the NAcc because of its known role in reward processing and sensation-seeking behaviors. We used special software (FSL) to compare brain activity between the two conditions and identify any significant differences.

2.10 Associations with Sensation Seeking

Finally, we examined if the level of brain activity in the NAcc during reward trials was related to participants' scores on the sensation seeking questionnaire. We also wanted to see if this relationship changed with age. To do this, we used a statistical technique called latent variable modeling, which allowed us to look at how brain activity and age together predicted sensation seeking over time.

3. Results

3.1 Behavioral

As expected, participants performed better (higher accuracy) and faster (shorter reaction times) when they could earn rewards compared to when they couldn't. Interestingly, performance also improved with age, meaning older participants were both more accurate and faster at the task (see Figure 2). This is consistent with previous research showing that cognitive abilities, like attention and control, improve as we get older. We did not find a relationship between sensation seeking and task performance.

3.2 Latent Trait Modeling of Sensation Seeking

Before examining the brain imaging results, we wanted to make sure that our measure of sensation seeking was stable across time. In other words, we wanted to be sure that a person's level of sensation seeking remained relatively consistent across the three times we measured it. Our analyses confirmed that the sensation seeking questionnaire was reliable and measured a stable trait over time.

3.3 Functional Imaging Results

Our primary interest was in examining activity in the NAcc, the brain region associated with reward processing. We didn't find a direct link between age and NAcc activity during reward trials. This means that, on average, NAcc activity in response to rewards was similar across different ages.

However, we did find a fascinating interaction between age and NAcc activity in predicting sensation seeking. Specifically, for younger participants (around ages 10-12), greater NAcc activation during reward trials was associated with lower sensation seeking scores later on. But, for older participants (around ages 17-25), higher NAcc activity was linked to higher sensation seeking scores (see Figure 4). This suggests that the relationship between NAcc activity and sensation seeking changes as we transition from childhood to adulthood.

3.4 Exploratory Analyses

In addition to our primary analysis focusing on the NAcc, we also examined activity across the entire brain. This analysis confirmed that the NAcc was indeed a key region involved in reward processing, but it also highlighted other brain areas that might be important. However, these other regions didn't show the same age-related interaction with sensation seeking that we observed in the NAcc.

4. Discussion

Our study provides evidence that the relationship between brain activity during reward processing and sensation seeking changes with age. This might explain why some teenagers are more drawn to risky behaviors than others. It seems that for younger individuals, increased NAcc activity might actually be linked to less sensation seeking, possibly reflecting a more cautious approach. However, as we mature, the same level of NAcc activity might become associated with a greater tendency towards sensation seeking.

These findings support previous research suggesting that the brain's reward system undergoes significant changes during adolescence, which might contribute to increased risk-taking behaviors (Paulsen et al., 2014). The exact reasons for this shift are still unclear, but it likely involves complex interactions between brain development, hormones, and environmental factors.

4.1 Limitations

While our study provides valuable insights into the relationship between brain activity and sensation seeking, it's important to acknowledge some limitations. First, our participants were healthy volunteers, so our findings might not apply to individuals with mental health or substance abuse problems. Second, we only measured sensation seeking at three time points, so we couldn't capture every fluctuation in this trait. Future studies with more frequent assessments would provide a more complete picture. Finally, we only focused on age as a factor influencing sensation seeking, but other factors like peer influence and personality traits might also be important.

4.2 Conclusions

Despite these limitations, our study provides important evidence for a developmental shift in the relationship between brain reward activity and sensation seeking. These findings could have important implications for understanding why some individuals are more prone to risky behaviors during adolescence. By identifying these neural markers early on, we might be able to develop interventions to promote healthier decision-making and reduce risk-taking in young people.

Link to Article

Do Our Brains Tell Us to Look for Thrills?

Some people love trying new and exciting things, even if it means taking risks. This is called sensation seeking. People high in sensation seeking may not always think about the dangers of their actions (Horvath & Zuckerman, 1993; M. Zuckerman & Kuhlman, 2000). Sometimes this can be good, like when someone is brave in a tough situation (Neria, Solomon, Ginzburg, & Dekel, 2000). But sensation seeking is also linked to problems like drug use and getting hurt (Ball, Carroll, & Rounsaville, 1994; Cservenka et al., 2013). So, it's important to understand why some people love thrills more than others.

Sensation seeking is partly determined by our genes, kind of like our eye color (Joseph, Liu, Jiang, Lynam, & Kelly, 2009; Terracciano et al., 2011). Our brains also play a big role. A part of the brain called the nucleus accumbens (NAcc) lights up when we get rewards, like winning a game or getting a good grade (Bornovalova et al., 2009; Ikemoto & Panksepp, 1999). Scientists think the NAcc might work differently in people who are high in sensation seeking, making them more excited by rewards (Dalley et al., 2007; Harden & Mann, 2015; M. Zuckerman, 1984).

This study looked at how the NAcc and age affect sensation seeking. We asked 139 people aged 10 to 25 to do a computer task where they could win points. We also measured their sensation seeking using a questionnaire. We found that the way the NAcc responded to rewards was different for younger and older people. In younger kids (around 10-12 years old), those whose NAcc lit up more for rewards were actually less likely to be sensation seekers later on. But in older teens and young adults (around 17-25 years old), a more active NAcc was linked to higher sensation seeking.

What does this mean?

This is one of the first studies to show that how our brains respond to rewards might be linked to how much we like thrills as we grow up. Our brains are still developing during our teenage years, and this might change how the NAcc affects our behaviors.

It's important to remember that this study doesn't mean that everyone with a certain brain pattern will love risks, or that there's a "thrill-seeking" part of the brain. But it does suggest that understanding how our brains respond to rewards could help us understand and maybe even prevent risky behaviors in the future.

Link to Article