Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Aug 11.
Published in final edited form as: Psychol Music. 2017 Aug 22;46(5):734–748. doi: 10.1177/0305735617723721

Better late than never (or early): Music training in late childhood is associated with enhanced decision-making

Kirsten E Smayda 1, Darrell A Worthy 2, Bharath Chandrasekaran 1,3,4,5
PMCID: PMC8356733  NIHMSID: NIHMS997019  PMID: 34385757

Abstract

Decision-making is critical to everyday life. Here we ask: to what extent does music training benefit decision-making? Supported by strong associations between music training and enhanced cross-domain skills, we hypothesize that musicians may show decision-making advantages relative to non-musicians. Prior work has also argued for a “critical period” for cross-domain plasticity such that beginning music training early enhances sensorimotor brain regions that mature early in life. Given that brain regions supporting decision-making begin maturing late in childhood, we hypothesized that an advantage in decision-making may only be present in musicians who began music training later in childhood. To test this hypothesis, young adults who began music training before and after 8 years of age (early-trained musicians, ET; late-trained musicians, LT, respectively) and non-musicians (NM) performed a decision-making task. We found a decision-making advantage in LT relative to ET and NM. To better understand the mechanism of the LT advantage, we conducted computational modeling on participant responses and found that LT were less biased by recent outcomes and incorporated longer strings of outcomes when deciding among the choice options. These results tentatively suggest that music training may confer decision-making enhancements, and carry strong implications for the utility of music training in childhood.

Keywords: cognition, computational modeling, critical period, decision-making, music training

Decision-making is a critical skill required for everyday functioning. We use it to make choices, guide our actions, and complete tasks throughout the day. We can choose options that are riskier or safer and those decisions may lead to positive or negative outcomes immediately or after some delay. Prior work has shown that decision-making relies heavily on the prefrontal cortex across species (Barraclough, Conroy, & Lee, 2004; Bechara, Damasio, Tranel, & Anderson, 1998; Broche-Pérez, Herrera Jiménez, & Omar-Martínez, 2016; Hare, O’Doherty, Camerer, Schultz, & Rangel, 2008; Li, Lu, D’Argembeau, Ng, & Bechara, 2009). More specifically, Hare and colleagues found that activity in the medial orbitofrontal cortex is associated with the predicted amount of reward each option will confer (i.e. goal values), while the central orbitofrontal region is associated with measuring the net reward of choosing an option (i.e., decision values; Hare et al., 2008). Goals and decision values are critical components of the decision-making process that guide the participant to actions that result in the largest net benefit. Additionally, animal studies suggest converging evidence implicating the orbitofrontal region in updating expected outcomes during a decision-making task (Sul, Kim, Huh, Lee, & Jung, 2010). A recent study also implicates the medial prefrontal cortex in tracking differences between a received reward and the expected reward on a trial-by-trial basis (Samanez-Larkin, Worthy, Mata, McClure, & Knutson, 2014). Lastly, prior studies have found that damage to the ventromedial prefrontal cortex (Bechara, Damasio, Damasio, & Anderson, 1994), dorsolateral and dorsomedial prefrontal cortex (Bechara et al., 1998; Manes et al., 2002) results in impaired decision-making ability. Together, these results suggest that while different subregions of the prefrontal cortex may be responsible for specific reward-based processes during a decision-making task, the prefrontal cortex as a whole is a critical brain region for successful decision-making.

The prefrontal cortex demonstrates protracted development through childhood (Gogtay et al., 2004), suggesting that as children age, and their prefrontal cortices develop, they should show enhanced decision-making abilities. Supporting this notion, prior work suggests that in a classic decision-making task, the Iowa Gambling Task (IGT), in which the participant must choose between decks that lead to net gains or net losses, the tendency to avoid disadvantageous deck choices increases from childhood through adolescence to adulthood (Hooper, Luciana, Conklin, & Yarger, 2004), and overall performance also increases throughout childhood (Crone & van der Molen, 2004). The IGT has also been used as a measure of incentive-based decision-making across a variety of populations including healthy adolescents (Hooper et al., 2004), healthy older adults (Wood, Busemeyer, Koling, Cox, & Davis, 2005), and clinical populations such as patients with schizophrenia (Shurman, Horan, & Nuechterlein, 2005), bipolar disorder (Ono et al., 2015), and damage to the ventromedial prefrontal cortex (Bechara et al., 1994).

Music training across the lifespan confers advantages in domains well beyond music (Herholz & Zatorre, 2012; Kraus & Chandrasekaran, 2010) such as speech perception (Kraus & Chandrasekaran, 2010; Parbery-Clark, Anderson, Hittner, & Kraus, 2012; Parbery-Clark, Skoe, Lam, & Kraus, 2009) and cognitive abilities (Costa-Giomi, 1999; Moreno et al., 2011; Schellenberg, 2005) and frontally-mediated cognitive abilities such as working memory (Bergman Nutley, Darki, & Klingberg, 2014; George & Coch, 2011; Kraus, Strait, & Parbery-Clark, 2012), processing speed (Bugos, 2010; Bugos & Mostafa, 2011), and cognitive control (Bialystok & DePape, 2009; Pallesen et al., 2010). Music training beginning before the age of seven relative to later in life, can enhance sensorimotor abilities that persist beyond childhood (Steele, Bailey, Zatorre, & Penhune, 2013). These findings support the concept of a “critical period,” which suggests that introducing a novel skill that utilizes brain regions that are undergoing significant development will confer long-lasting cross-domain benefits (Hensch, 2005; Steele et al., 2013). Given that regions implicated in decision-making, namely the prefrontal cortex, do not develop until late childhood, we posit that beginning to play music in late childhood may provide an enduring enhancement in decision-making relative to beginning to play music earlier in childhood or never at all.

The goal of the current study is to examine the extent to which music training impacts decision-making, as measured by the Iowa Gambling Task. We hypothesize that musicians will show a performance advantage in decision-making relative to non-musicians, supported by the literature showing associations between music training and enhanced cognitive ability. We also hypothesize that if there is an advantage, it may only be present in musicians who began music training later in childhood, given that the brain regions supporting decision-making begin maturing later in childhood (Gogtay et al., 2004).

We examined decision-making performance using behavioral and computational modeling approaches in three groups: Adults who began playing music at or before the age of 8, classified as “early-trained” musicians (ET), adults who began playing music after the age of 8 classified as a “late-trained” (LT) musicians. Adults who never played an instrument were considered non-musicians (NM). We predicted that late-trained musicians would perform better than both early-trained musicians and non-musicians. Although accuracy data is informative for showing group differences, we fit computational models to response patterns to identify the mechanism behind any performance differences between groups. Lastly, we included data from a “many labs” collaboration as a Super Control group in our behavioral analyses. The purpose of including the Super Control group is to compare any group effect to a large, established well-studied sample that is representative of healthy young adult performers on the Iowa Gambling Task (Steingroever et al., 2015).

Method

Participants

A total of 69 participants (aged 18–35 years) were recruited from The University of Texas at Austin community and were part of a larger multi-day study, which was occurring at the same time as testing for the present study. Written consent was obtained from all participants and the Institutional Review Board of The University of Texas at Austin approved the experimental procedure. Participants were classified as non-musicians (NM), early-trained musicians (ET), or late-trained musicians (LT). Operationally, non-musicians had less than 2 years of music training and were not currently playing any instrument. ET had at least 8 years of music training beginning at or before age 8, while LT had at least 8 years of music training beginning after the age of 8. Several intervention studies examining neural indices of cross-domain enhacements from music training have used 8 years as the starting point of music training (Chobert, Francois, Velay, & Besson, 2014; Moreno et al., 2009). These studies consistently find evidence of music-training induced plasticity even past 8 years of age. We used 8 years as the cut-off to be consistent with prior literature examining early versus late training as well as accommodating for the work on brain plasticity induced by music training past the “so-called” sensorimotor critical period (Bailey & Penhune, 2012; Steele et al., 2013). Both ET and LT actively practiced instruments at the time of testing and did not significantly differ in the number of hours they currently played, t(44) = 0.44, p = .66, the number of instruments they played, t(44) = 0.85, p = .40, or their age, t(44) = 1.11, p = .27. Participant demographic details can be found in Table 1. Reported participant ages for the Super Control group from the “many labs” collaboration were not complete, but from the five studies that published age information, the mean age of participants is 25.6 years (datasets from Horstmann, 2012; Kjome et al., 2010; Premkumar et al., 2008; Steingroever, Šmíra, Lee, & Pachur, n.d.; Wood et al., 2005), and for the two studies which did not report ages, the samples included only undergraduates (Maia & McClelland, 2004; Worthy, Pang, & Byrne, 2013). No music experience metric was reported for any of the studies included in the “many labs” collaboration. Several participants across both early-trained and late-trained groups in the present study played multiple instruments, as represented in the instrument count in Table 2.

Table 1.

Demographic information of participants.

n Age
Mean (SD)		 Years playing music
Mean (SD)		 Age began playing
Mean (SD)		 Hours currently play
Mean (SD)
Non-musicians 23 23.91 (4.47) 0.00 (0.00) n/a 0.00 (0.00)
Early-trained musicians 23 23.70 (5.34) 16.91 (4.82) 5.61 (1.85) 7.42 (5.82)
Late-trained musicians 23 25.17 (3.50) 13.65 (3.72) 11.43 (2.11) 6.46 (8.78)
Open in a new tab

Table 2.

Instrument counts across musician groups.

Strings Brass Woodwind Keys Percussion Voice Other
Early-trained musicians 11 1 6 18 7 10 3
Late-trained musicians 17 2 4 12 3 7 2
Open in a new tab

Procedure

The Iowa Gambling Task was run on PCs using Matlab with Psychtoolbox (version 2.5). Participants were seated at a computer and instructed that on each trial, they were to choose a card from one of four decks (A, B, C, D) by pressing “Z,” “W,” “P” or “/” on the keyboard. They were told that after each card choice, they would be shown the number of points gained or lost, and that their goal was to maximize the number of points gained. The participants were unaware that two of the decks (C and D) led to a net gain of points (and are therefore considered “good decks”) while the other two led to a net loss of points (A and B; considered “bad decks”). In addition, the decks varied by magnitude and frequency of points won or lost such that choosing decks B or D resulted in large but infrequent losses, while choosing decks A or C led to more frequent and smaller-magnitude losses. The task consisted of five 20-trial blocks and was identical to the task in Worthy, Pang, et al. (2013).

Neuropsychological testing measure descriptions

Participants completed several neuropsychological tests. The Stroop Test was administered to assess selective attention and inhibitory control ability (Stroop, 1935). A “relative interference score” was calculated as the number of words read in the “incongruent” condition normalized by the average of the number of words read in the “color” and “word” conditions. Operation Span was used to assess complex working memory ability. At a computer, participants were asked to remember a string of letters while completing a secondary math problem between each letter presented in the sequence. After each sequence of letters was presented, participants were asked to recall the sequence of letters in the order they were presented. Additionally, participants were told to maintain an accuracy of 85% or higher on the math problems (Unsworth, Heitz, Schrock, & Engle, 2005; see also Xie, Maddox, Knopik, McGeary, & Chandrasekaran, 2015). Lastly, the WAIS-III Digit Span was administered to assess verbal working memory. For this task, participants were required to repeat strings of numbers read aloud by the researcher; first exactly how the researcher read them, and then backwards (Wechsler, 1997).

Results

Iowa Gambling Task behavioral results

Overall performance on the IGT was our primary measure and was calculated as the proportion of good minus bad deck selections for each subject in five 20-trial blocks (C + D – A – B; Worthy, Pang, et al., 2013). A mixed ANOVA with group as a between-subjects variable (non-musician, early-trained, late-trained) and block as a within-subjects variable (1 through 5), revealed significant main effects of group, F(2, 66) = 5.10, p < .01, partial η2 = 0.13, block, F(4, 264) = 8.19, p < .001, partial η2 = 0.11, and an interaction between group and block, F(8, 264) = 2.56, p < .05, partial η2 = 0.07. To decompose the significant interaction, we ran a series of one-way ANOVAs at each block comparing the effect of group on overall performance. We found a significant effect of group in blocks 4 and 5. Pairwise comparisons of overall performance between groups in blocks 4 and 5 suggest an LT advantage relative to NM in block 4, t(44) = 3.29, p < .005, and ET and NM in block 5, t(44) = 4.61, p < .001; t(44) = 2.08, p < .05, respectively. These results are displayed in Figure 1(a).

Figure 1.

Figure 1.

Open in a new tab

Overall performance on the IGT. Error bars represent standard error of the mean (a). Overall performance of musicians on the IGT plotted with the age they began playing music (b).

We also compared LT to a 441-participant “Super Control” group of young adults who performed the 100-trial IGT (shown in gray in Figure 1(a); Steingroever et al., 2015). To determine if the late-trained musicians performed significantly different from the Super Control group, we ran a mixed model ANOVA with group as a between-subjects variable (Super Control group, LT musician) and block as a within-subjects variable (1 through 5). We found significant main effects of group, F(1, 462) = 6.96 p < .009, partial η2 = 0.01, and block, F(4, 1848) = 83.67, p < .0001, partial η2 = 0.15, and interaction between group and block, F(4, 1848) = 2.39, p = .05, partial η2 = 0.005. These results suggest that late-trained musicians perform significantly better than the Super Control group.

Within just the early- and late-trained musician groups, a linear regression of age of onset of music training as a predictor of overall performance suggests that the later an individual begins playing music in childhood, the greater the benefit to their decision-making skills as young adults, F(1, 44) = 8.96, p < .01, R2 = 0.17, b = 0.03; displayed in Figure 1(b). Interestingly, although the two musician groups differed significantly in their years of playing music, LT: M = 13.65, ET: M = 16.91; t(44) = 2.57, p < .05, years of playing music was not a significant predictor of performance, F(1, 44) = 1.90, p = .17, R2 = 0.04; b = −0.01.

Deck choice results

We also compared differences in deck choices between our four participant groups (late-trained musicians, early-trained musicians, non-musicians, and the Super Control group) in two steps. First, we compared deck choices between non-musicians, late-trained musicians, and early-trained musicians; and second, we compared deck choices between the late-trained musicians and the Super Control group. For our first analysis, we performed a two-way ANOVA on proportion of deck chosen with group (non-musicians, late-trained musicians, early-trained musicians) as a between-subjects variable and deck (A, B, C, D) as a within-subjects variable. We did not find a significant main effect of group, F(2, 66) = 1.24, p = .300, but we did find a significant effect of deck, F(3, 198) = 11.14, p < .001, partial η2 = 0.14. There was no significant interaction of group and deck, F(6, 198) = 2.14, p = 0.051. Results are displayed in Figure 2 and suggest that Deck D was chosen most often (M = 0.32), then Decks B (M = 0.26) and C (M = 0.24), and Deck A was chosen the least often (M = 0.19). Post-hoc pairwise comparisons (with Holm correction) of selections in each deck suggest a group difference in only Deck B selection. Late-trained musicians selected Deck B significantly less (M = 0.21) than both early-trained musicians, M = 0.28; t(44) = 2.97, p < .05, and non-musicians, M = 0.28; t(44) = 2.38, p < .05.

Figure 2.

Figure 2.

Open in a new tab

Proportion of decks chosen by non-musicians, early-trained musicians, late-trained musicians, and the Super Control group. Error bars represent standard error of the mean.

To compare deck choices between late-trained musicians and the Super Control group, we ran a two-way ANOVA on proportion of deck chosen with group and deck (A, B, C, D) and found significant main effects of deck, F(3, 1386) = 98.43, p < .001, partial η2= 0.18, and group, F(1, 462) = 14.91, p < .0002, partial η2 = 0.03, and a significant interaction of deck and group, F(3, 1386) = 4.36, p < .006, partial η2 = 0.5. Post-hoc analyses at each deck suggest late-trained musicians and the Super Control group differ in the proportion they chose a deck in only Deck B, F(1, 462) = 10.95, p < .002, partial η2 = 0.02. Results are displayed in Figure 2.

Neuropsychological testing results

To determine if there are any other differences between our participant groups that may mediate the late-musician advantage in the Iowa Gambling Task, we ran a series of planned pairwise t tests with Holm correction between the neuropsychological test scores of our three groups. Our results suggest there are no differences between groups with respect to Stroop Interference score [ET and LT: t(44) = 0.50, p = 1.00; ET and NM: t(44) = 0.65, p = 1.00; LT and NM: t(44) = 0.25, p = 1.00], Digit Span Total score [ET and LT: t(44) = 0.50, p = .93; ET and NM: t(44) = 0.65, p = .91; LT and NM: t(44) = 1.08, p = 0.87], and Operation Span score [ET and LT: t(38) = 0.88, p = 1.00; ET and NM: t(41) = 0.88, p = 1.00; LT and NM: t(39) = 0.71, p = 1.00].

Computational modeling results

To understand the mechanism involved in the LT performance advantage, we fit a series of computational models to the data (Worthy, Pang, et al., 2013). The full details of the models are provided in the online Supplementary Material. We fit a total of four reinforcement learning (RL) models: The Prospect Valence Learning Delta and Decay models (PVL2-Delta & PVL2-Decay), Valence-Plus-Perseveration (VPP) model, and the Expectancy-Frequency-Perseveration (EFP) model. These RL models all assume that participants track and update expected values for each action and then make decisions based on a relative comparison of the value of each deck. They differ in the ways that they account for specific processes like recency, loss aversion, frequency of gains versus losses, and perseveration. We also fit a win-stay-lose-shift model (WSLS; Worthy, Hawthorne, & Otto, 2013). This model fundamentally differs from the RL models because it makes no assumption about the expected value of alternatives, but instead assumes that participants stay with the same option or switch to a different option depending upon whether the outcome on the previous trial was a net gain or loss. Critically, in the IGT, the WSLS model assumes that participants will switch randomly between the three alternative decks on switch trials, rather than decks with higher expected values. Another critical assumption of the WSLS model is that it relies only on the outcome of the previous trial, and is thus heavily recency-biased. We also fit a Baseline or null model. Readers who are interested in a comprehensive account of each model beyond the basic mechanisms are directed to the online Supplementary Material.

We next used BIC or Schwarz weights to compare the relative fit of each model (Wagenmakers & Farrell, 2004). BIC weights are calculated from BIC to obtain a continuous measure of goodness-of-fit (see online Supplementary Material, section 1.2). These weights can be simply interpreted as the probability that the model is the best model given the data set and the set of candidate models (Wagenmakers & Farrell, 2004). We computed the BIC weights for each model for each participant. The average weights for participants in each group are shown in Table 3. Average weights for the WSLS model were nearly twice as high for participants in the NM and ET groups (both M = 0.50) compared to participants in the LT group (M = 0.27). This suggests that participants in the LT group relied on a simple WSLS strategy much less than participants in the other groups and that they may have incorporated a more complex RL strategy by comparing the expected values of each action.

Table 3.

BIC (Schwarz) weights for each model.

Model PVL2-Delta PVL2-Decay VPP EFP WSLS Baseline
Non-musician 0.06 (0.20) 0.37 (0.41) 0.00 (0.00) 0.05 (0.20) 0.50 (0.42) 0.02 (0.06)
Early-trained musician 0.06 (0.16) 0.22 (0.32) 0.10 (0.29) 0.09 (0.20) 0.50 (0.42) 0.03 (0.07)
Late-trained musician 0.18 (0.26) 0.22 (0.31) 0.14 (0.34) 0.15 (0.26) 0.27 (0.38) 0.04 (0.17)
Open in a new tab

Note. Standard deviations are listed in parentheses.

We also examined the best-fitting parameter estimates for each model across groups (Table 4) to compare specific decision-making sub-processes assumed by the models. For instance, the best-fitting recency or learning rate parameter (Φ) was lower for participants in the LT group than for NM and ET participants for both the PVL2-Delta and EFP models. Because the recency parameter represents how strongly someone weights recent trial outcomes in updating expected values, a lower recency parameter suggests that LT may have valued options based on a longer series of recent outcomes while NM and ET may have based decisions on only the most recent outcomes. Greater reliance on recent actions has been associated with WSLS strategy use in prior work as well (Otto, Taylor, & Markman, 2011; Worthy, Otto, & Maddox, 2012). Enhanced memory of more distant outcomes could account for LT participants’ ability to avoid selecting Deck B, which yields very large but infrequent losses, less than NM and ET participants as well as the Super Control group.

Table 4.

Average best-fitting parameter estimates for each model and group.

Non-musician Early-trained musician Late-trained musician
IGT
PVL2-Delta
α 0.40 (0.34) 0.42 (0.32) 0.53 (0.35)
λ 3.00 (2.15) 2.16 (2.16) 3.09 (2.01)
Φ 0.56 (0.43)* 0.48 (0.43)† 0.21 (0.32)*†
c 0.24 (0.27) 0.21 (0.32) 0.32 (0.35)
p 25.10 (37.5)*† −7.00 (36.0)‡ −8.70 (38.40)*
PVL2-Decay
α 0.30 (0.35) 0.41 (0.28) 0.30 (0.32)
λ 1.34 (1.73)‡ 2.93 (2.18)†‡ 1.69 (2.00)†
Φ 0.43 (0.25) 0.55 (0.34) 0.58 (0.32)
c 0.56 (0.80) 0.23 (0.40) 0.46 (0.50)
VPP
α 0.45 (0.29) 0.40 (0.37) 0.42 (0.35)
λ 2.40 (2.43) 1.95 (2.46) 2.68 (2.05)
Φ 0.30 (0.39) 0.32 (0.33) 0.18 (0.27)
C 1.84 (0.98) 2.08 (0.90) 2.51 (1.06)
w 0.53 (0.42) 0.63 (0.39) 0.69 (0.36)
k 0.42 (0.35) 0.52 (0.38) 0.43 (0.34)
ε pos 0.30 (0.53)‡ −0.12 (0.58)‡ 0.06 (0.57)
ε neg 0.10 (0.83)‡ −0.41 (0.62)†‡ 0.07 (0.78)†
EFP
Λ 1.36 (2.04)* 1.71 (2.19) 2.88 (2.30)*
Φ 0.52 (0.40)* 0.57 (0.44)† 0.28 (0.33)*†
c 1.02 (1.13) 1.12 (1.02) 0.72 (0.52)
p 2.7 (6.40) 2.90 (15.9) −2.30 (22.6)
ω 0.48 (0.36) 0.48 (0.38) 0.56 (0.41)
WSLS
P(stay | win) 0.55 (0.20)‡ 0.33 (0.24)‡ 0.51 (0.32)
P(shift | loss) 0.59 (0.34)‡ 0.80 (0.22)†‡ 0.57 (0.34)†
Open in a new tab

Note. Standard deviations are listed in parentheses. All tests were two-tailed.

*

Significant difference between Non-musician and Late groups at p < .05,

†

Significant difference between Early and Late groups at p < .05,

‡

Significant difference between Non-musician and Early groups at p < .05.

Discussion

We examined the extent to which music training confers long-lasting decision-making benefits. Our results suggest that music training impacts decisional processes differently depending on the age at which the musician began his or her training. In particular, we showed that late-trained musicians (who began after age 8) show a performance advantage in the Iowa Gambling Task, a well-studied decision-making task, relative to early-trained musicians (who began playing at or before age 8) and non-musicians. The performance advantage is observed in the last two blocks of the task when the gains and losses associated with each deck have already been learned. Critically, “age of onset” but not “years playing” was a significant predictor of performance in the Iowa Gambling Task in musician groups. One possible explanation for these results is that music training beginning late in childhood capitalizes on the period of significant maturation in the prefrontal cortex, a region of the brain implicated in optimal performance on the IGT (Bechara et al., 1994; Hare et al., 2008; Samanez-Larkin et al., 2014). Per a critical period theory, taxing the neural regions implicated in decision-making during a period of significant growth can lead to long-lasting functional benefits in decision-making ability (Hensch, 2005). Because the prefrontral cortex does not begin a stage of rapid development until late childhood, introducing a novel skill that relies on the prefrontal cortex during this stage of rapid development can have a profound and positive impact on other functions that the area subserves.

Importantly, our results suggest that the age at which novel skill acquisition begins, and not necessarily the ages during which the skills are developed beyond initial training, may be an important metric for characterizing long-lasting impact on the brain regions the skill utilizes. This is because the process of learning a novel procedural skill can be decomposed into at least two stages (Karni et al., 1998; Miyachi, Hikosaka, & Lu, 2002; Yin et al., 2009). The first stage is characterized by rapid skill attainment at the onset of the learning process. Imaging research suggests that the initial fast learning of a procedural skill in the brain can characterized by first a reduction in activation in the primary motor cortex, followed by an increase in the area of activation in the primary motor cortex (Karni et al., 1998). Subcortically, initial fast learning is associated with increased activation in the association regions of the basal ganglia (Miyachi et al., 2002). The second stage is considered to be slower with neural changes occurring within the context of the template set up during the initial, fast stage of learning. The early learning changes to the motor cortex during procedural skill learning are thought to “set up” the motor cortex for a task-specific process and serve as a template for ongoing modification of the slower learning process (Karni et al., 1998; Miyachi et al., 2002).

The literature on the neural mechanisms of learning a procedural skill supports the results of the current study. As would be expected from the literature, we found that the best predictor of performance on the Iowa Gambling Task was when our participants began playing music, which corresponds to the initial (or fast learning) stage of procedural skill acquisition. Although the early-trained musicians played their instruments during late childhood, they were by that point beyond the initial stage of skill acquisition, and therefore may have received less cognitive benefit from taxing the decision-making regions of the brain with music-learning.

The findings of the current study are corroborated by a large literature supporting the notion that beginning music training at an age at when the underlying brain regions are undergoing significant development may grant long-term performance enhancements in other skills those brain regions serve. For instance, research suggests that beginning music training at an early age (before age 7 years) can produce significant structural and behavioral enhancements in perceptual-motor tasks and the associated regions because major neural development in perceptual-motor regions occur before the age of seven (Bailey & Penhune, 2012; Bailey, Zatorre, & Penhune, 2014; Steele et al., 2013).

Our computational modeling results suggest that late-trained musicians more successfully compared reward expectations of each option in order to perform well, the process of which has been associated with the medial prefrontal cortex (Samanez-Larkin et al., 2014). In addition, NM and ET participants were more likely to rely on the most recent outcomes and on recency-based WSLS strategies more than LT participants who may have avoided the disadvantageous Deck B more successfully by better retention of the infrequent, but very large, losses (−1,250 points) it provided. It is possible that late music training enhances individuals’ ability to incorporate a longer series of events when engaging in complex cognitive processes like decision-making. Interestingly, the results of our linear regression also support this notion, as years of music-playing, was not a significant predictor of performance in the Iowa Gambling Task, but age of onset of playing was.

Our study design is similar to previous research that has examined the advantage of early music training, relative to late music training on brain plasticity (Bailey & Penhune, 2012; Steele et al., 2013). As in previous studies, a limitation is that our data are correlational in nature. We compared non-musicians and two types of musicians in a classic decision-making task, but this method does not allow us to draw a causal inference regarding the effects of age of onset of music-playing and decision-making later in life. Although additional analyses have ruled out a number of obvious alternative explanations, there may be other factors (e.g. personality; Corrigall, Schellenberg, & Misura, 2013) that influence when a person begins playing music that may mediate the intruiging relationship we find in the present study. Future work should seek to further refine this relationship using a longitudinal methodology to more directly test the effect of beginning music training earlier or later in life on decision-making ability.

The results of the current study have real-world implications: in addition to the rich sensorimotor benefits of early music playing may have on the developing brain, music training may also confer long-lasting benefits in complex cognitive functions like decision-making. To take a real-world example, the music classes offered during many children’s elementary and high-school education in America (which correspond to the age of music-training onset in the present “late-trained musician” group) may result in improved decision-making ability as an adult. Understanding tools that can serve individuals who show profound decision-making deficits as young as 8 years old could have a dramatic effect on the rate of risky behaviors later in life. Specifically, our results lead to an exciting and testable hypothesis: learning to play music during elementary and high school may lead to dramatic enhancements on late-maturing brain regions that mediate decision-making abilities.

The results of the current study also suggest that music training should be added as a measure of individual differences for future studies in decision-making. The Super Control group data which came from a “many labs” collaboration is composed of young adults with a mean age of 25.6 years. Music experience was not reported for any of the studies included in the Super Control group; therefore, we cannot exclude the possibility that music training was a factor in any results presented. Our results suggest that music-training metrics may be easy and worthwhile measurements to obtain in future decision-making studies. Therefore, we recommend that future decision-making studies include measurements of music training to help elucidate any effects that might be otherwise obscured in a musically heterogeneous sample.

We also cannot completely rule out alternative explanations for our results. For instance, it can be argued that very few aspects about a child’s environment throughout childhood are under the child’s control. This means that the age at which a child begins playing music, or any other engaging activity, is most likely determined by the parents. It is possible that children who were introduced to music at an early age may have different motivations for pursuing music training than the children who began playing music later in childhood. It is reasonable to assume that parents may have had a larger influence in the decision to play music at an earlier age than later, suggesting that early-trained musicians would have a different motivation to play relative to the late-trained musician. Research also suggests that several factors such as self-regulation, deliberate practice strategies, and self-perception of competence all significantly influence musical achievement. These factors will likely differ in their relationship to musical achievement depending on what age the musician begins practicing (Bonneville-Roussy & Bouffard, 2015). For instance, a younger musician may practice a certain number of hours per week because of parental involvement, which shapes their practice schedule and influences how they self-regulate their practicing later in life. Although the current study was not equipped to measure the existence of such motivational differences, we cannot rule out that a difference in the primary contributor of motivation (i.e., parent or self) may have resulted in the current results.

A limitation of the current study is that our results are based on a group of young adults who had access to music education during childhood and may not generalize to the rest of the population. Factors such as socioeconomic status and race are important factors that can affect access to music education (Palmer, 2011). Because the participants in this study were among a cohort who had access to music education in schools or as an extracurricular activity and were later recruited in a university setting, our results are limited to such a cohort. Future studies should endeavor to include participants from diverse socioeconomic statuses, and races in order to better understand the ways in which music training may affect the developing brain.

“Future directions should also include a test for the existence of an upper age limit for beginning music training by which receiving music training after such an age would no longer provide the cognitive benefits explored in this paper. In our current sample, late-trained musicians began playing music between the ages of 9 and 16 years old.” Because the neural regions supporting decision-making continue to develop into adulthood, it is plausible that beginning music training at any point between late childhood and adulthood might confer decision-making benefits (the correlational results displayed in Figure 1(b) suggests this may be the case). However, a study similar in recruitment methodology to Savion-Lemieux, Bailey, and Penhune (2009) including several age groups binned within a late-trained musician group, and an early-trained control group would be better suited to test such a phenomenon. Lastly, we encourage researchers to replicate the current study and publish both significant and non-significant findings in order to better understand the impact of music training in late childhood on decision-making abilities.

In conclusion, the goal of the present study was to test the extent to which the music training effects decision-making. Using the Iowa Gambling Task, we found that musicians who began training at age 8 or later showed improved decision-making ability relative to musicians who began training before the age of 8 or non-musicians. Using computational modeling, we found that the late-trained musician performance advantage was a result of using longer strings of trial outcomes to inform their current trial’s strategy. Our results have important implications for science-based approaches to incorporating music education in school systems, cognitive development, and the development of ecologically valid tools to enhance individual deficits in decision-making.

Supplementary Material

SupplementalMaterial

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors would like to acknowledge NIA grant 5F31AG052308-02 to KS.

References

  1. Bailey JA, & Penhune VB (2012). A sensitive period for musical training: Contributions of age of onset and cognitive abilities. Annals of the New York Academy of Sciences, 1252(1), 163–170. doi: 10.1111/j.1749-6632.2011.06434.x [DOI] [PubMed] [Google Scholar]
  2. Bailey JA, Zatorre RJ, & Penhune VB (2014). Early musical training is linked to gray matter structure in the ventral premotor cortex and auditory–motor rhythm synchronization performance. Journal of Cognitive Neuroscience, 26(4), 755–767. doi: 10.1162/jocn_a_00527 [DOI] [PubMed] [Google Scholar]
  3. Barraclough DJ, Conroy ML, & Lee D (2004). Prefrontal cortex and decision making in a mixed-strategy game. Nature Neuroscience, 7(4), 404–410. doi: 10.1038/nn1209 [DOI] [PubMed] [Google Scholar]
  4. Bechara A, Damasio A, Damasio H, & Anderson S (1994). Insensitivity to future consequences following damage to human prefrontal cortex. Cognition, 50, 7–15. [DOI] [PubMed] [Google Scholar]
  5. Bechara A, Damasio H, Tranel D, & Anderson SW (1998). Dissociation of working memory from decision making within the human prefrontal cortex. Journal of Neuroscience, 18(1), 428–437. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bergman Nutley S, Darki F, & Klingberg T (2014). Music practice is associated with development of working memory during childhood and adolescence. Frontiers in Human Neuroscience, 7. doi: 10.3389/fnhum.2013.00926 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bialystok E, & DePape A-M (2009). Musical expertise, bilingualism, and executive functioning. Journal of Experimental Psychology: Human Perception and Performance, 35(2), 565–574. doi: 10.1037/a0012735 [DOI] [PubMed] [Google Scholar]
  8. Bonneville-Roussy A, & Bouffard T (2015). When quantity is not enough: Disentangling the roles of practice time, self-regulation and deliberate practice in musical achievement. Psychology of Music, 43(5), 686–704. [Google Scholar]
  9. Broche-Pérez Y, Herrera Jiménez LF, & Omar-Martínez E (2016). Neural substrates of decision-making. Neurologia, 31(5), 319–325. [DOI] [PubMed] [Google Scholar]
  10. Bugos J (2010). The benefits of music instruction on processing speed, verbal fluency, and cognitive control in aging. Music Education Research International, 4, 1–9. [Google Scholar]
  11. Bugos J, & Mostafa W (2011). Musical training enhances information processing speed. Bulletin of the Council for Research in Music Education, 187, 7–18. [Google Scholar]
  12. Chobert J, Francois C, Velay J-L, & Besson M (2014). Twelve months of active musical training in 8- to 10-year-old children enhances the preattentive processing of syllabic duration and voice onset time. Cerebral Cortex, 24(4), 956–967. doi: 10.1093/cercor/bhs377 [DOI] [PubMed] [Google Scholar]
  13. Corrigall KA, Schellenberg EG, & Misura NM (2013). Music training, cognition, and personality. Frontiers in Psychology, 4, 222. doi: 10.3389/fpsyg.2013.00222 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Costa-Giomi E (1999). The effects of three years of piano instruction on children’s cognitive development. Journal of Research in Music Education, 47(3), 198–212. doi: 10.2307/3345779 [DOI] [Google Scholar]
  15. Crone E, & van der Molen M (2004). Developmental changes in real life decision making: Performance on a gambling task previously shown to depend on the ventromedial prefrontal cortex. Developmental Neuropsychology, 25(3), 251–279. [DOI] [PubMed] [Google Scholar]
  16. George EM, & Coch D (2011). Music training and working memory: An ERP study. Neuropsychologia, 49(5), 1083–1094. doi: 10.1016/j.neuropsychologia.2011.02.001 [DOI] [PubMed] [Google Scholar]
  17. Gogtay N, Giedd JN, Lusk L, Hayashi KM, Greenstein D, Vaituzis AC, … Thompson PM (2004). Dynamic mapping of human cortical development during childhood through early adulthood. Proceedings of the National Academy of Sciences of the United States of America, 101(21), 8174–8179. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Hare TA, O’Doherty J, Camerer CF, Schultz W, & Rangel A (2008). Dissociating the role of the orbitofrontal cortex and the striatum in the computation of goal values and prediction errors. Journal of Neuroscience, 28(22), 5623–5630. doi: 10.1523/JNEUROSCI.1309-08.2008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Hensch TK (2005). Critical period plasticity in local cortical circuits. Nature Reviews Neuroscience, 6(11), 877–888. doi: 10.1038/nrn1787 [DOI] [PubMed] [Google Scholar]
  20. Herholz SC, & Zatorre RJ (2012). Musical training as a framework for brain plasticity: Behavior, function, and structure. Neuron, 76(3), 486–502. doi: 10.1016/j.neuron.2012.10.011 [DOI] [PubMed] [Google Scholar]
  21. Hooper CJ, Luciana M, Conklin HM, & Yarger RS (2004). Adolescents’ performance on the Iowa Gambling Task: Implications for the development of decision making and ventromedial prefrontal cortex. Developmental Psychology, 40(6), 1148–1158. doi: 10.1037/0012-1649.40.6.1148 [DOI] [PubMed] [Google Scholar]
  22. Horstmann A (2012). Iowa Gambling Task: There is more to consider than long-term outcome. Using a linear equation model to disentangle the impact of outcome and frequency of gains and losses. Frontiers in Neuroscience, 6, 61. doi: 10.3389/fnins.2012.00061 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Karni A, Meyer G, Rey-Hipolito C, Jezzard P, Adams MM, Turner R, & Ungerleider LG (1998). The acquisition of skilled motor performance: Fast and slow experience-driven changes in primary motor cortex. Proceedings of the National Academy of Sciences, 95(3), 861–868. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Kjome KL, Lane SD, Schmitz JM, Green C, Ma L, Prasla I, … Moeller FG (2010). Relationship between impulsivity and decision making in cocaine dependence. Psychiatry Research, 178(2), 299– 304. doi: 10.1016/j.psychres.2009.11.024 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Kraus N, & Chandrasekaran B (2010). Music training for the development of auditory skills. Nature Reviews Neuroscience, 11(8), 599–605. [DOI] [PubMed] [Google Scholar]
  26. Kraus N, Strait DL, & Parbery-Clark A (2012). Cognitive factors shape brain networks for auditory skills: Spotlight on auditory working memory. Annals of the New York Academy of Sciences, 1252(1), 100–107. doi: 10.1111/j.1749-6632.2012.06463.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Li X, Lu Z-L, D’Argembeau A, Ng M, & Bechara A (2009). The Iowa Gambling Task in fMRI images. Human Brain Mapping, 410–423. doi: 10.1002/hbm.20875 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Maia TV, & McClelland JL (2004). A reexamination of the evidence for the somatic marker hypothesis: What participants really know in the Iowa gambling task. Proceedings of the National Academy of Sciences of the United States of America, 101(45), 16075–16080. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Manes F, Sahakian B, Clark L, Rogers R, Antoun N, Aitken M, & Robbins T (2002). Decision-making processes following damage to the prefrontal cortex. Brain, 125, 624–639. [DOI] [PubMed] [Google Scholar]
  30. Miyachi S, Hikosaka O, & Lu X (2002). Differential activation of monkey striatal neurons in the early and late stages of procedural learning. Experimental Brain Research, 146(1), 122–126. doi: 10.1007/s00221-002-1213-7 [DOI] [PubMed] [Google Scholar]
  31. Moreno S, Bialystok E, Barac R, Schellenberg EG, Cepeda NJ, & Chau T (2011). Short-term music training enhances verbal intelligence and executive function. Psychological Science, 22(11), 1425–1433. doi: 10.1177/0956797611416999 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Moreno S, Marques C, Santos A, Santos M, Castro SL, & Besson M (2009). Musical training influences linguistic abilities in 8-year-old children: More evidence for brain plasticity. Cerebral Cortex, 19(3), 712–723. doi: 10.1093/cercor/bhn120 [DOI] [PubMed] [Google Scholar]
  33. Ono Y, Kikuchi M, Hirosawa T, Hino S, Nagasawa T, Hashimoto T, … Minabe Y (2015). Reduced prefrontal activation during performance of the Iowa Gambling Task in patients with bipolar disorder. Psychiatry Research: Neuroimaging, 233(1), 1–8. doi: 10.1016/j.pscychresns.2015.04.003 [DOI] [PubMed] [Google Scholar]
  34. Otto AR, Taylor EG, & Markman AB (2011). There are at least two kinds of probability matching: Evidence from a secondary task. Cognition, 118(2), 274–279. doi: 10.1016/j.cognition.2010.11.009 [DOI] [PubMed] [Google Scholar]
  35. Palmer CM (2011). Challenges of access to post-secondary music education programs for people of color. Visions of Research in Music Education, 18. Retrieved from http://www-usr.rider.edu/~vrme/v18n1/visions/Palmer-Access%20to%20Music%20Ed%20for%20People%20of%20Color.pdf [Google Scholar]
  36. Pallesen KJ, Brattico E, Bailey CJ, Korvenoja A, Koivisto J, Gjedde A, … Carlson S (2010). Cognitive control in auditory working memory is enhanced in musicians. PLoS ONE, 5(6). doi: 10.1371/journal.pone.0011120 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Parbery-Clark A, Anderson S, Hittner E, & Kraus N (2012). Musical experience strengthens the neural representation of sounds important for communication in middle-aged adults. Frontiers in Aging Neuroscience, 4, 30. doi: 10.3389/fnagi.2012.00030 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Parbery-Clark A, Skoe E, Lam C, & Kraus N (2009). Musician enhancement for speech-in-noise. Ear and Hearing, 30(6), 653–661. [DOI] [PubMed] [Google Scholar]
  39. Premkumar P, Fannon D, Kuipers E, Simmons A, Frangou S, & Kumari V (2008). Emotional decision-making and its dissociable components in schizophrenia and schizoaffective disorder: A behavioural and MRI investigation. Neuropsychologia, 46(7), 2002–2012. doi: 10.1016/j.neuropsychologia.2008.01.022 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Samanez-Larkin GR, Worthy DA, Mata R, McClure SM, & Knutson B (2014). Adult age differences in frontostriatal representation of prediction error but not reward outcome. Cognitive, Affective, & Behavioral Neuroscience, 14(2), 672–682. doi: 10.3758/s13415-014-0297-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Savion-Lemieux T, Bailey JA, & Penhune VB (2009). Developmental contributions to motor sequence learning. Experimental Brain Research, 195(2), 293–306. doi: 10.1007/s00221-009-1786-5 [DOI] [PubMed] [Google Scholar]
  42. Schellenberg EG (2005). Music and cognitive abilities. Current Directions in Psychological Science, 14(6), 317–320. doi: 10.1111/j.0963-7214.2005.00389.x [DOI] [Google Scholar]
  43. Shurman B, Horan WP, & Nuechterlein KH (2005). Schizophrenia patients demonstrate a distinctive pattern of decision-making impairment on the Iowa Gambling Task. Schizophrenia Research, 72(2–3), 215–224. doi: 10.1016/j.schres.2004.03.020 [DOI] [PubMed] [Google Scholar]
  44. Steele CJ, Bailey JA, Zatorre RJ, & Penhune VB (2013). Early musical training and white-matter plasticity in the corpus callosum: Evidence for a sensitive period. Journal of Neuroscience, 33(3), 1282–1290. doi: 10.1523/JNEUROSCI.3578-12.2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Steingroever H, Fridberg DJ, Horstmann A, Kjome KL, Kumari V, Lane SD, … Wagenmakers E-J (2015). Data from 617 healthy participants performing the Iowa Gambling Task: A “many labs” collaboration. Journal of Open Psychology Data, 3, e5. doi: 10.5334/jopd.ak [DOI] [Google Scholar]
  46. Steingroever H, Šmíra M, Lee MD, & Pachur T (n.d.). Do intuitive and deliberate decision makers perform differently on the Iowa gambling task? Retrieved from Steingroever et al. (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Stroop JR (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18(6), 643–662. [Google Scholar]
  48. Sul JH, Kim H, Huh N, Lee D, & Jung MW (2010). Distinct roles of rodent orbitofrontal and medial prefrontal cortex in decision making. Neuron, 66(3), 449–460. doi: 10.1016/j.neuron.2010.03.033 [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Unsworth N, Heitz RP, Schrock JC, & Engle RW (2005). An automated version of the operation span task. Behavior Research Methods, 37(3), 498–505. [DOI] [PubMed] [Google Scholar]
  50. Wagenmakers EJ, & Farrell S (2004). AIC model selection using Akaike weights. Psychonomic Bulletin & Review, 11(1), 192–196. doi: 10.3758/BF03206482 [DOI] [PubMed] [Google Scholar]
  51. Wechsler D (1997). Wechsler Adult Intelligence Scale – Third edition. San Antonio, TX: Harcourt Brace & Company. [Google Scholar]
  52. Wood S, Busemeyer J, Koling A, Cox CR, & Davis H (2005). Older adults as adaptive decision makers: Evidence from the Iowa Gambling Task. Psychology and Aging, 20(2), 220–225. doi: 10.1037/0882-7974.20.2.220 [DOI] [PubMed] [Google Scholar]
  53. Worthy DA, Hawthorne MJ, & Otto AR (2013). Heterogeneity of strategy use in the Iowa gambling task: A comparison of win-stay/lose-shift and reinforcement learning models. Psychonomic Bulletin & Review, 20(2), 364–371. doi: 10.3758/s13423-012-0324-9 [DOI] [PubMed] [Google Scholar]
  54. Worthy DA, Otto AR, & Maddox WT (2012). Working-memory load and temporal myopia in dynamic decision making. Journal of Experimental Psychology: Learning, Memory, and Cognition, 38(6), 1640–1658. doi: 10.1037/a0028146 [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Worthy DA, Pang B, & Byrne KA (2013). Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task. Frontiers in Psychology, 4, 640. doi: 10.3389/fpsyg.2013.00640 [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Xie Z, Maddox WT, Knopik VS, McGeary JE, & Chandrasekaran B (2015). Dopamine receptor D4 (DRD4) gene modulates the influence of informational masking on speech recognition. Neuropsychologia, 67, 121–131. doi: 10.1016/j.neuropsychologia.2014.12.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Yin HH, Mulcare SP, Hilário MRF, Clouse E, Holloway T, Davis MI, … Costa RM (2009). Dynamic reorganization of striatal circuits during the acquisition and consolidation of a skill. Nature Neuroscience, 12(3), 333–341. doi: 10.1038/nn.2261 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

SupplementalMaterial
NIHMS997019-supplement-SupplementalMaterial.pdf (140.3KB, pdf)

RESOURCES