Participants (N = 43; mean age, 23.0 years; range, 18–35 years; 20 male, 23 female) were right-handed and free from any history of psychiatric or neurological disorders. Each participant provided written informed consent after additional screening for physical or medical conditions that would affect their eligibility for fMRI. The study protocol was approved in accordance with guidelines instituted by the Washington University Human Research Protection Office, and data were collected by the senior author (KJ) at Washington University in St. Louis. Participants were compensated for their participation ($10 per h for the behavioral session, $25 per h for the fMRI session). Of the 45 participants recruited into the study, two were eliminated due to the small number of choices (<10) for the delayed option in the fMRI session.

We analyzed a data set collected in an fMRI experiment of an intertemporal choice task involving real liquid rewards where human participants directly experienced choice, delay, and consumption of the rewards (Jimura et al., 2013; Figure 1). In this experiment, participants performed the intertemporal decision-making task (Figure 1) in two separate (behavioral and fMRI) sessions.

The analyses of the choice and delay periods were published previously (Jimura et al., 2013), and were not analyzed in the current study. The current study analyzed the data while participants consumed the liquid rewards only, which were not analyzed in the previous study (Jimura et al., 2013).

The behavioral session aimed to measure individuals’ delay discounting of real liquid rewards (Jimura et al., 2009, 2011, 2013). Prior to the behavioral session, participants were asked to choose one favorite drink that would serve as the reward from a list consisting of apple, orange, grape, grapefruit, and cranberry juices, lemonade, and water. No participants requested to change the reward drink in the fMRI session.

At the beginning of each trial, two alternatives were presented on the left and right sides of the screen, respectively: one involved a larger reward (20 or 40 squirts) available after a delay (10, 30, or 60 s), while the other consisted of a variable smaller amount available immediately (Figure 1A). Participants were instructed to press one of two corresponding response buttons to indicate their preference.

During the delay, a fixation cross was presented on the center of the screen. At the time of reward delivery, participants saw a visual message indicating the reward was ready. Importantly, participants were able to control the rate of liquid flow. Reward delivery continued as long as the button was held down; if the button was released, delivery paused and then resumed when the button was pressed again. During reward delivery, the amount remaining (in squirts) was displayed below a red horizontal bar whose length corresponded to the number of squirts still available. After the participant finished drinking, a fixation cross was presented.

To estimate individuals’ delay-discounting rates, the current study used three delay conditions (10, 30, 60 s) for the larger amount (40 squirts), and two delay conditions (10, 30 s) for the smaller amount (20 squirts; Jimura et al., 2009). On the first trial of each delay condition, the choice was between a larger delayed amount and an immediate reward that was half of the delayed amount. For each delay condition, the amount of the immediate reward after the first trial was adjusted based on the participant’s preceding choice. If the participant had chosen the smaller, immediate reward on the preceding trial, then the amount of the immediate reward was decreased by half (i.e., 10 and five squirts for the 40- and 20-squirt conditions, respectively); if the participant had chosen the larger, delayed reward on the preceding trial, then the amount of the immediate reward was increased by half (Jimura et al., 2009, 2011). The adjustment amount was five squirts in the third trial in the 40-squirt condition. The subjective value of the delayed reward was estimated to be 1 ml (i.e., 2.5 squirts) more or less than the amount of the immediate reward available in the last trial (third and second trials in the 40- and 20-squirt conditions, respectively), depending on whether the delayed or immediate reward had been chosen on that trial.

After the behavioral session, the participants practiced drinking liquid rewards in a supine position with a mock scanner setup. When drinking liquid rewards, they were encouraged to use jaw movements and mouth muscles for swallowing, but not to move their heads.

During fMRI scanning, participants performed an intertemporal decision-making task that was similar to that of the behavioral session. The primary difference was that the choice options for each trial were prespecified (rather than adjusted across the session), but set in an individualized manner based on a discounting profile estimated from the behavioral session. Three conditions (60 s/40 squirts, 30 s/40 squirts, 30 s/20 squirts) were used to measure brain activity during the delay period. The value of the immediate reward was systematically manipulated so that across trials, its value was smaller than the subjective value of the delayed reward, estimated for each participant based on their choice profile in the behavioral session. This manipulation of the immediate reward amount biased decisions toward delayed options, as the reward value was always smaller than the subjective value of the delayed reward, providing more opportunity to measure brain activity during consumption of delayed rewards (Jimura et al., 2013). When drinking the liquid rewards, the participants were instructed to use jaw movements and mouth muscles without moving their heads.

Both anatomical and functional images were available from each participant. High-resolution anatomical images were acquired using an MP-RAGE T1-weighted sequence [repetition time (TR), 9.7 s; echo time (TE), 4.0 ms; flip angle (FA), 10°; slice thickness, 1 mm; in-plane resolution, 1 × 1 mm²]. Functional [blood oxygen level-dependent (BOLD)] images were acquired using a gradient echo-planar imaging sequence (TR, 2.0 s; TE, 27 ms; FA, 90°; slice thickness, 4 mm; in-plane resolution, 4 × 4 mm²; 34 slices) in parallel to the anterior–posterior commissure line, allowing complete brain coverage at a high signal-to-noise ratio. Each functional run involved 512 volume acquisitions.

For each participant, the degree of behavioral impulsivity was quantified by calculating the area under the discounting curve (AuC; Myerson et al., 2001; Sellitto et al., 2010; Jimura et al., 2011, 2013; Tanaka et al., 2020). The AuC represents the area under the observed subjective values at a given delay; more specifically, the AuC was calculated as the sum of the trapezoidal areas under the indifference points normalized by the amount and delay (Myerson et al., 2001). Both subjective value and delay were normalized for the purposes of calculating the AuC, which, as a result, ranged between 0.0 (maximally steep discounting) and 1.0 (no discounting). It has been argued that the AuC is the best measure of delay discounting for use in individual difference analyses, because it is theoretically neutral (i.e., assumption-free) and also psychometrically reliable (Myerson et al., 2001).

Each participant was classified into one of three groups, namely steep (STP), shallow (SHL), and intermediate (INT) discounting, based on their AuC values. The groups were identical to those analyzed in the previous study (Jimura et al., 2013).

Participants performed 19.5 ± 2.6 (mean ± SD) trials, and choose the delayed alternative in 83.0 ± 12.6% of the trials. The number of trials where the delayed alternative was chosen did not correlate with behavioral impulsivity (AuC) measured in the behavioral session (see “Materials and Methods” section) [r = 0.15, t₍₄₁₎ = 0.97, P = 0.33]. The mean drinking duration was 7.5 ± 2.1 s.

Participants were classified into three groups based on AuC values reflecting the delay-discounting pattern (see “Materials and Methods”section), as in the previous study (Jimura et al., 2013: steep discounters (N = 15; high impulsivity), shallow discounters (N = 15; low impulsivity) and an intermediate group (N = 13).

The liquid was delivered from outside the scanner room through a plastic tube, which enabled participants to consume the liquid during fMRI administration (see “Materials and Methods” section). However, it is well known that head movements during fMRI lead to significant artifacts and signal losses in images. To evaluate whether our data were contaminated by the motion-derived artifacts and signal losses, we first assessed head movements and MRI images while participants were drinking liquid rewards.

As shown in Figure 2A, head movements were greater during liquid consumptions than during button presses made in a money discounting distractor task performed in the same scanning sessions (Figure 1B; Jimura et al., 2013, 2018). However, the absolute magnitude of head motion during consumption was small, with maximum translations of (0.06, 0.11, 0.24 mm) ± (0.02, 0.03, 0.09 mm; mean ± SD), and maximum rotations of (0.27, 0.09, 0.06) ± (0.09, 0.03, 0.02) degrees, along the x-, y-, and z-axes, respectively. Moreover, the maximum head movement parameters in the six axes (see “Materials and Methods” section) did not correlate with the AuC (|r|s < 0.28; |t|s < 1.88; Ps > 0.07, uncorrected).

Nonetheless, it is known that jaw movements can yield significant instability in echoplanar images, as reported in nonhuman primate scanning (Keliris et al., 2007). However, the instability seemed absent in the current study, as we observed comparable image quality during liquid consumption and the fixation periods (Figure 2B).

Given these results, we felt confident in assuming that movement-derived contamination during liquid consumption was less obvious in the current study than the previous study (Keliris et al., 2007). Our recent study also showed similar results (Tanaka et al., 2020).

The current results highlight functional segregation and integration in the prefrontal cortex and the ventral striatum during the consumption of delayed liquid rewards. Figure 7 summarizes our results and illustrates activity magnitudes and signal flows between the aPFC, aVS, and pVS for three levels of impulsivity.

In the aPFC, activation is greater in impulsive individuals, resulting in enhanced signaling toward the aVS and an elevation of its activity. On the other hand, the pVS transmits reward-related signals to the aVS, and then aVS activity is further amplified in impulsive individuals. Interestingly, the signals from both the aPFC and pVS appear to be associated with both the magnitude of aVS activation and the level of impulsivity.

These mechanisms can be interpreted as showing that the reward-related signals from the aPFC and pVS are aggregated into the aVS, which is critical for the degree of behavioral impulsivity. These prefrontal-striatum mechanisms are compatible with those suggested by prior studies analyzing connectivity between aPFC and VS (Diekhof and Gruber, 2010; Jimura et al., 2013; Tanaka et al., 2020).

A diffusion tensor imaging study of the human VS showed that both the anterior and posterior parts of the VS are connected to the anterior ventral part of the PFC and the orbitofrontal cortex, including the aPFC in the current study. Interestingly, the anterior part is also connected to more dorsal parts of the PFC (Tziortzi et al., 2014). The dorsal parts of the PFC are involved in working memory and choice difficulty in intertemporal choice, which is enhanced in self-controlled individuals (Jimura et al., 2018). The anatomical connection between the anterior VS and dorsal PFC may regulate the aPFC-aVS-pVS mechanisms during consumption of liquid rewards, as illustrated by the reduced activity in the intermediate discounting group (Figure 4).

The current study focused on the consumption phase of intertemporal choice, whereas previous studies analyzed delay and choice phases (Jimura et al., 2013; Tanaka et al., 2020). These collective results are derived from a single dataset, allowing us to speculate regarding possible functional mechanisms involved in the distinct phases of intertemporal choice behavior: choice, delay, and consumption.

In all three phases, impulsivity in decision-making was associated with increased VS activation, consistent with previous reports (Tanaka et al., 2004; Hariri et al., 2006; Kable and Glimcher, 2007; Ballard and Knutson, 2009; Pine et al., 2009). On the other hand, aPFC activation differed among the phases. During the choice and early delay periods, self-controlled (less impulsive) individuals exhibited increased aPFC activity (Jimura et al., 2013; Tanaka et al., 2020). In contrast, in the current study, highly impulsive individuals exhibited increased aPFC activity during the consumption period. A possible unified explanation of these findings is that the aPFC is associated with currently available utility, i.e., future anticipation during delay and reward consumption during drinking. The former may be more valuated in self-controlled individuals to maximize future reward attainment, whereas the latter may be evaluated in impulsive individuals when consuming a reward after a delay.

Greater aPFC activation during a choice period in less impulsive individuals may reflect resistance to impulsive choice, which reduces VS activity (Diekhof and Gruber, 2010). During the delay period, aPFC regions may encode an anticipatory utility signal associated with delayed rewards, which is the extra utility derived from the pleasure of waiting for a reward delivered in the future (Loewenstein, 1987; Berns et al., 2006, 2007; Peters and Buechel, 2010; Jimura et al., 2013; Tanaka et al., 2020).

On the other hand, as mentioned above, highly impulsive individuals showed greater aPFC activation during the consumption period, suggesting marked current utility when individuals consume liquid rewards. Because highly impulsive individuals do not prefer to wait to obtain a larger reward, the utility of a liquid reward would become greater upon completion of the delay, eliciting greater aPFC activation when consuming delayed rewards.

Another possibility is that the aPFC encodes current pleasure related to reward attainment, as suggested by aPFC regions showing anticipatory utility effect that is thought to reflect pleasure of waiting (Loewenstein, 1987; Jimura et al., 2013; Tanaka et al., 2020). Thus, the receipt of a delayed reward may provide greater pleasure for highly impulsive individuals. Alternatively, while consuming the reward, highly impulsive individuals may retrieve episodic information about past experiences of rewards, resulting in greater aPFC activation. However, additional evidence is needed to directly support the role of the aPFC in relation to pleasure.