A Twin Study of Musical Pitch and Rhythm ... - Semantic Scholar

7 downloads 0 Views 469KB Size Report
Springer Science+Business Media New York 2015. Abstract Both genetic and environmental factors are known to play a role in our ability to perceive music, but.
Behav Genet DOI 10.1007/s10519-015-9774-y

ORIGINAL RESEARCH

The Nature and Nurture of Melody: A Twin Study of Musical Pitch and Rhythm Perception Erik Seesja¨rvi1 • Teppo Sa¨rka¨mo¨1 Isabelle Peretz3,4 • Jaakko Kaprio2



Eero Vuoksimaa2 • Mari Tervaniemi1



Received: 17 June 2015 / Accepted: 23 November 2015 Ó Springer Science+Business Media New York 2015

Abstract Both genetic and environmental factors are known to play a role in our ability to perceive music, but the degree to which they influence different aspects of music cognition is still unclear. We investigated the relative contribution of genetic and environmental effects on melody perception in 384 young adult twins [69 full monozygotic (MZ) twin pairs, 44 full dizygotic (DZ) twin pairs, 70 MZ twins without a co-twin, and 88 DZ twins without a co-twin]. The participants performed three online music tests requiring the detection of pitch changes in a two-melody comparison task (Scale) and key and rhythm incongruities in single-melody perception tasks (Out-ofkey, Off-beat). The results showed predominantly additive genetic effects in the Scale task (58 %, 95 % CI 42–70 %), shared environmental effects in the Out-of-key task (61 %, 49–70 %), and non-shared environmental effects in the Off-beat task (82 %, 61–100 %). This highly different Edited by Yoon-Mi Hur

Electronic supplementary material The online version of this article (doi:10.1007/s10519-015-9774-y) contains supplementary material, which is available to authorized users. & Teppo Sa¨rka¨mo¨ [email protected] 1

Cognitive Brain Research Unit (CBRU), Institute of Behavioural Sciences, University of Helsinki, Siltavuorenpenger 1B, P.O. Box 9, 00014 Helsinki, Finland

2

Department of Public Health, University of Helsinki, Helsinki, Finland

3

International Laboratory for Brain, Music, and Sound Research (BRAMS) and Centre for Research on Brain, Language and Music (CRBLM), Montreal, Canada

4

Department of Psychology, Universite´ de Montre´al, Montreal, Canada

pattern of effects suggests that the contribution of genetic and environmental factors on music perception depends on the degree to which it calls for acquired knowledge of musical tonal and metric structures. Keywords Music perception  Pitch  Rhythm  Twins  Genetics  Environment

Introduction Musical abilities develop under a complex set of genetic predispositions and environmental input (Trainor and Hannon 2012). In its basic form, music is comprised of series of sound events that change either in the spectral (pitch) or temporal (rhythm) domain over time (Krumhansl 2000). Psychologically, the perception of a musical melody involves two parallel cognitive processes: discriminating the pitch, duration, and time interval of consecutive tones and perceiving whether the tones follow the tonal structure or musical scale (key) of the melody. These processes seem to develop at different ages, suggesting a different degree of biological innateness and musical enculturation. Even small infants are able to detect changes in pitch (Chang and Trehub 1977; Trehub et al. 1999) and rhythm and meter (Hannon and Johnson 2005; Winkler et al. 2009) in melodies and also move spontaneously to the rhythm of music (Zentner and Eerola 2010). Infants are, however, equally sensitive to both out-of-key and within-key changes in melodies, whereas children and adults are better at detecting out-of-key changes (Corrigall and Trainor 2010; Trainor and Trehub 1992), indicating that tonality perception develops later. Overall, auditory-cognitive networks become increasingly specialized throughout childhood for encoding the musical structure of one’s own

123

Behav Genet

culture (Trainor and Corrigall 2010; Trainor and Hannon 2012), leading to implicit knowledge of musical syntax of that culture in adulthood (Bigand and Poulin-Charronnat 2006; Brattico et al. 2006; Tillmann et al. 2000). The ability to perceive music or discriminate between musical stimuli varies considerably between individuals, and the key question is how much of this variability is caused by genetic and environmental factors. Genetic family studies indicate that pitch-related musical deficits (congenital amusia) and abilities (absolute pitch) aggregate in families and are inheritable (Baharloo et al. 2000; Peretz et al. 2007) and that musical aptitude and creativity are linked to specific genes responsible for auditory and neurocognitive development (Gingras et al. 2015; Oikkonen et al. 2015; Ukkola et al. 2009). While providing evidence for the genetic basis of musicality, these studies address special populations (amusic and musically trained persons) and therefore do not directly answer the question about the role of genetic and environmental factors on music perception in the general population. To date, only two research groups have addressed this topic in populationbased twin studies. Drayna and collaborators studied pitch perception in twins using an updated version of the Distorted Tunes Test (DTT, Kalmus and Fry 1980) where subjects are presented popular melodies and the task is to judge whether each melody is correct (original melody of the song) or incorrect (contains altered-pitch tones). They found a strong additive genetic effect (0.71) and no significant environmental effects (Drayna et al. 2001). Recently, Ulle´n and collaborators studied music perception and musical training in a large twin sample using the Swedish Music Discrimination Test (SMDT) where pairs of tones or tone sequences are compared for differences in pitch, melody, and rhythm (Ulle´n et al. 2014). Across tasks, they found moderate-tostrong additive genetic effects (0.30–0.61) with little shared environmental contribution (Mosing et al. 2014a). Interestingly, there was also a clear genetic effect on the amount of music practice, which accounted for the association between practice and performance on the SMDT. In summary, current evidence suggests that there is a strong genetic component to the development of musicrelated auditory discrimination skills, whereas the contribution of shared environmental factors seems to be negligible. However, due to the nature of the stimuli and tasks used in previous studies, it remains unclear whether this holds true also for other higher-level aspects of music cognition, such as tonality processing. The stimuli used by Ulle´n and others were relatively simple acoustically (sinusoidal tones or isochronous piano tone sequences), which limits their generalizability as indices of more complex music cognition. The studies also employed tasks that have a strong memory component, making

123

performance in them partly dependent on either long-term memory (previous exposure to and familiarity with the popular melodies; Drayna et al. 2001) or working memory (maintaining a memory representation of the preceding tone/melody and making a comparison against it; Mosing et al. 2014a). Therefore, the derived heritability estimates may reflect not only musical ability, but also the moderateto-high heritability of general cognitive capacity and memory (Deary et al. 2009; Lee et al. 2010; Volk et al. 2006). Indeed, Mosing et al. (2014b) found that there were phenotypic correlations between intelligence and SMDT performance and that the covariation between these variables could be explained by shared genetic influences. The goal of the present population-based twin study was to gain more insight into the relative contribution of genetic and environmental factors on different aspects of music cognition using the naturalistic, unfamiliar melodies from the Montreal Battery of Evaluation of Amusia (MBEA; Peretz et al. 2003, 2008). We included three tasks measuring the discrimination pitch changes in pairs of piano melodies (Scale test; Peretz et al. 2003) and the perception of key and rhythm violations in single melodies (Out-of-key and Off-beat tests; Peretz et al. 2008). Importantly, the Scale and Out-of-key tests both measure the detection of musical key violations in melodies but differ partly in their perceptual and cognitive demands: while performance on both tests benefits from implicit tonal knowledge, the Scale test additionally entails a memory-based comparison process and is therefore more similar to the tasks used previously (Drayna et al. 2001; Mosing et al. 2014a). In contrast, the Out-of-key and Offbeat tests have the same overall task demands, but differ in the measured musical domain (pitch vs. rhythm). We aimed to (1) verify the previously reported genetic effects on musical pitch discrimination using more naturalistic and unfamiliar melodic stimuli and to (2) explore the impact of genetic and environmental factors on musical tasks more sensitive to implicit perception and knowledge of musical tonal and metric structures.

Methods Subjects The sample consisted of individuals from a populationbased longitudinal FinnTwin16 study, which includes altogether 2733 Finnish twin pairs born in 1975–1979 (Kaprio et al. 2002; Kaprio 2013). In the present study, invitations to participate were mailed between September 2012 and February 2013 to altogether 507 Finnish-speaking twin pairs [253 monozygotic (MZ) pairs, 254 dizygotic (DZ) pairs] who were randomly selected among those who

Behav Genet

had answered the wave five of the FinnTwin16 Internet Survey in 2010–2012. The sample represented all educational levels and all regions of Finland. Zygosity was determined using a validated questionnaire method (Sarna et al. 1978) with an additional questionnaire for younger twins (Goldsmith 1991). Furthermore, zygosity has been confirmed with DNA tests for those twins in whom the questionnaires were not able to provide definite classification and for those who have participated in earlier laboratory studies of FinnTwin16. In the present sample, the zygosity of 41 % of the twin pairs has been confirmed with DNA tests showing 95 % agreement with the questionnaires. The Ethical Committee of the Department of Public Health of the University of Helsinki approved the present study protocol, and informed consent was obtained from all participants. Of the 1014 persons invited to participate, 386 (38.1 %) completed our online music perception test. There were no statistically significant differences between responders and non-responders regarding age, sex or zygosity. There were statistically significant differences in response rates between different education levels (F(3.98, 2015.61) = 2.50, p = 0.041) with the lowest rate (19 %) among participants with only primary education and the highest rate (43 %) among participants with a Master’s degree or higher. Data from two individuals whose responses indicated that they had not actually performed the task (performance was well below chance level or the task was completed in less time than stimulus presentation takes) were excluded from the analyses. Thus, the final sample comprised 384 subjects including 69 full MZ twin pairs, 44 full DZ twin pairs, 70 MZ twins without a co-twin, and 88 DZ twins without a co-twin. Assessment of Music Perception Music perception was assessed with the Scale test from the Montreal Battery of Evaluation of Amusia (MBEA; Peretz et al. 2003) and the Out-of-key and Off-beat tests from newer online version of the MBEA [Online Test of Amusia (OTA); Peretz et al. 2008]. Although the MBEA and OTA are primarily designed for studying congenital or acquired amusia, their use in population-based studies of music perception is warranted by the tests’ 1) natural and ecologically valid musical (melodic) stimulus material, 2) robust psychometric properties (i.e., reliability, validity, and sensitivity), 3) rapid and easy applicability (performed online, duration ca. 20 min), and 4) clear focus on skills of the ordinary music listener (in contrast to musical aptitude or talent). As stimuli, all three tests use 30 novel melodies composed specifically for the MBEA by Ire`ne Delie`ge in order to be unfamiliar to subjects. The melodies are in major key, 9.6 tones long on average, and played by computer at a tempo of 120 beats/min.

In the Scale test, two example trials followed by 30 experimental trials are presented. Each trial consists of two piano melodies separated by an interval of 2 s (Fig. 1, examples A and B). In 15 trials, the second melody is identical to the first and in 15 trials it has been altered by changing the pitch of one tone to be out-of-scale while retaining the original melodic contour. The task is to judge whether the melodies are identical or not (same-different classification). In the Off-beat and the Out-of-key tests, two example trials and 24 experimental trials are presented. In each trial, a single melody is presented. In 12 trials, the melody is modified so that one tone is altered either in terms of pitch (Out-of-key) or time (Off-beat). In the Offbeat test, the alteration is an additional silence of 5/7 of the beat duration (357 ms) prior to the onset of the tone, thereby locally disrupting the meter of the melody or introducing an off-beat tone, without changing anything else (Fig. 1, examples C and D). In the Out-of-key test, the pitch of the altered tone is shifted by one or half a semitone, thereby creating a tone that is out-of-key, a ‘‘foreign’’ or ‘‘wrong’’ pitch in the musical context, without changing anything else (Fig. 1, examples C and E). The melodies are presented with 10 different timbres (e.g., piano, saxophone, guitar). The task is to decide after each trial whether or not the melody contained a musically anomalous or incongruent tone (yes/no classification). The online presentation of the test was implemented using standard Web browser technologies (i.e., HTML, PHP, and Flash), with the stimuli presented in MP3 format and the responses automatically recorded and tabulated using Microsoft Excel. The participants were asked to perform the test alone in a quiet room, use headphones if possible, and adjust the volume of the audio system to a comfortable level using three sample tones before starting the test. The three tests were presented in fixed order (Scale, Off-beat, Out-of-key) and within all tests, the trials were presented in fixed order that had been randomized beforehand. A total score (a sum of the correct answers in each subtest) was calculated as well. In addition to the three tests, the participants answered four self-evaluation questions about their capacity to perceive music (see Table 1 and Peretz et al. 2008). The questions and task instructions were presented in Finnish. Previously, the Finnish version of MBEA/OTA has been tested with 61 healthy subjects and the distributions of the scores were found to be similar to the original test (Hausen et al. 2013). Data Analysis The estimation of genetic and environmental effects was performed by comparing groups of identical (monozygotic, MZ) and non-identical (dizygotic, DZ) twins. Since MZ twins have an identical genome whereas DZ twins share on

123

Behav Genet Fig. 1 Examples of melodies used in Scale test (a, b), Offbeat test (c, e) and Out-of-key test (d, e). Violation of scale (a) and incongruencies in rhythm (d) and key (e) are circled

Table 1 Demographical characteristics of the participants All

MZ

DZ

MZ and DZ comparison (test value, p value)

Male

109

50

59

F(1, 270) = 2.64, p = 0.105

Female Age

275

158

117

35.00 (1.50)

34.96 (1.52)

35.04 (1.48)

F(1, 270) = 0.16, p = 0.685

Primary level

10

3

7

F(3.94, 1062.64) = 0.87, p = 0.477

Secondary level

93

51

42

Lowest level tertiary

47

24

23

Bachelor level

110

66

44

Master level or higher

124

64

60

4.52 (0.67)

4.55 (0.64)

4.47 (0.70)

F(1, 270) = 1.17, p = 0.280

Yes

53

24

29

F(1, 270) = 1.58, p = 0.210

No

331

184

147

183

157

25

19

Gender

Mean (SD) Educational level

Do you recognise a familiar melody? Mean (SD)a Do you lack sense of music?

Do you notice if someone is singing off-key? Yes

340

No 44 Do you notice if someone plays a wrong note? Yes

290

154

136

No

94

54

40

MZ monozygotic participants, DZ dizygotic participants a

Measured with a 5-point Likert scale (never, rarely, sometimes, often, very often)

123

F(1, 270) = 0.12, p = 0.734

F(1, 270) = 0.40, p = 0.527

Behav Genet

average 50 % of their segregating genes and assuming that there are no systematic differences in developmental environment between MZ and DZ twins (equal environmental similarity assumption), the main difference between these groups is the different degree of genetic relatedness. Using structural equation modelling, it is possible to parcel out the relative contributions of additive genetic (A) effects (summed effects of all the alleles that influence the given phenotype), dominant genetic (D) effects (effects due to one allele having a stronger influence on the given phenotype than another allele in the same locus), shared environmental (C) effects (environmental factors that make individuals similar), and non-shared environmental (E) effects (environmental factors that make individuals different, including also the measurement error of the model) on a given trait. The quantitative genetic modelling was done with Mx software version 1.68e (Neale 2004), while initial descriptive data analyses were performed with IBM SPSS Statistics version 22.0 and Stata 11 software. The data was first screened for any missing entries, outliers and persons who might have answered by guessing, and consequently, the data of two subjects were removed (see above). The three tests and total score were all modelled individually with a univariate model using the maximum likelihood method (Neale and Cardon 1992). There were no differences in means or variances between randomly ordered first and second twins, or between zygosity groups. As the data includes only twins reared together, it is not possible to model dominant genetic effects and shared environmental effects simultaneously, which leads to a selection between an ACE and an ADE model (Posthuma et al. 2003), referred to as full models. Based on within-pair twin correlations, an ADE model was used for the Off-beat test while an ACE model was used for the other tests and the total score. The full ACE (or ADE) model did not result in a significant change in fit relative to saturated model in total score or in any of the tests (p [ 0.05). Next, simpler models (AE, CE and E) were estimated and compared to the corresponding full model in order to find the most parsimonious model that still provides a good fit to the data. For these comparisons, the test statistics (D-2LL) follow a mixture of different v2 distributions with various degrees of freedom (Dominicus et al. 2006). The corrected probability values were calculated using a method described by Dominicus et al. (2006). Models were also compared using Bayesian Information Criterion (BIC), where smaller value indicates preferred model. All the models included gender and age as covariates. Furthermore, models with educational level as an additional covariate were also estimated. These results were very similar to those presented here and therefore are not shown.

Results To rule out any possible artificial genetic effects caused by gender, age, educational level or self-evaluated musical background, the zygosity groups were compared on these background variables while taking the twin-clustered structure of the data into account. There were no significant differences between MZ and DZ twins on any of these variables (Table 1). Looking at the main effects of the background variables on test performance, we observed a gender effect in the Off-beat subtest where men scored somewhat higher than women [men: mean 83.52 (SD 8.15), women: mean 80.94 (SD 8.63); F(1, 270) = 7.11, p = 0.008]. There were no other significant effects. The distribution statistics of scores in the three tests and the total score are shown in Table 2. In the Scale test, the Off-beat test, and the total score, the scores followed a normal distribution; in the Out-of-key test, the distribution was slightly negatively skewed (see Fig. S1 in Supplemental Material). However, this effect was minimal and similar for MZ and DZ twins. Regarding the overall relationship between the three tasks across subjects, there was a moderate correlation between the Scale and Out-of-key tests (r = 0.49, p \ 0.001), whereas the correlation of the Off-beat was non-significant to Scale and only modest to Out-of-key (r = 0.14, p = 0.007). Correlations between the four self-evaluation questions about music perception ability and the test scores are presented in Table 3. All self-evaluation questions correlated significantly with each other and with the Scale, Out-ofkey, and total scores. The Off-beat score correlated only with the first self-evaluation question (Do you recognise a familiar melody?) and with the Out-of-key and total scores. The within-pair twin correlations are presented in Table 4. In the Scale test, the correlations were fairly high and statistically significant in both MZ and DZ twin pairs. In the Off-beat test, the correlation was moderate and statistically significant in the MZ pairs but non-significant in the DZ pairs. In the Out-of-key test and the total score, the correlations were high and statistically significant in both zygosity groups. Comparison statistics between full and subsequent models are shown in Table 5 and the estimates for variance explained by different latent components in Table 6. Estimates are shown only for full models and those models whose fit to the data was not significantly worse than that of the corresponding full model. In the Scale test, the AE model was the most parsimonious model that fit the data well. In both the ACE and AE models, the estimates of additive genetic effects were strong (0.50 and 0.58, respectively), while the estimate of shared environmental effects in the ACE model was nonsignificant.

123

Behav Genet Table 2 Distribution of scores in the MBEA tests and total score (N = 384)

Min

Max

PS

Mean

SD

Scale test

16 (53.3 %)

30 (100 %)

2 (2.9 %)

24.9 (83.2 %)

3.0 (9.9 %)

Out-of-key test

11 (45.8 %)

24 (100 %)

43 (11.2 %)

20.4 (85.1 %)

2.9 (12.2 %)

Off-beat test

12 (50 %)

24 (100 %)

1 (0.3 %)

19.6 (81.7 %)

2.1 (8.6 %)

Total score

40 (51.1 %)

77 (98.6 %)

0 (0 %)

64.7 (83.3 %)

5.7 (7.3 %)

Min minimum score (% of full score), Max maximum score (% of full score), PS participants with perfect score (% of participants), Mean average score (% of full score), SD standard deviation (% of full score)

Table 3 Correlations between four self-evaluation questions and test scores Do you recognise a familiar melody?

Do you lack sense of music?

Do you notice if someone is singing off-key?

Do you notice if someone plays a wrong note?

Scale test

Out-ofkey test

Offbeat test

Total score

Do you recognise a familiar melody?

1

-0.33**

0.34**

0.30**

0.27**

0.28**

0.22**

0.37**

Do you lack sense of music?

-0.33**

1

-0.38**

-0.48**

-0.32**

-0.43**

-0.07

-0.41**

Do you notice if someone is singing off-key?

0.34**

-0.38**

1

0.74**

0.27**

0.34**

0.09

0.35**

Do you notice if someone plays a wrong note?

0.30**

-0.48**

0.74**

1

0.40**

0.46**

0.02

0.44**

Scale test Out-of-key test

0.27** 0.28**

-0.32** -0.43**

0.27** 0.34**

0.40** 0.46**

1 0.49**

0.49** 1

0.06 0.14**

0.75** 0.83**

Off-beat test

0.22**

-0.07

0.09

0.02

0.06

0.14**

1

0.50**

Total score

0.37**

-0.41**

0.35**

0.44**

0.75**

0.83**

0.50**

1

** Correlation is significant at the 0.01 level (2-tailed)

Table 4 Within-pair twin correlations in the MBEA tests and total score

Scale test

MZ pairs (N = 69)

DZ pairs (N = 44) 0.38 (0.09–0.61)

r (95 % CI)

0.58 (0.40–0.72)

p

\0.001

0.010

Out-of-key test

r (95 % CI)

0.63 (0.46–0.75)

0.67 (0.47–0.81)

p

\0.001

\0.001

Off-beat test

r (95 % CI)

0.31 (0.08–0.51)

-0.20 (-0.47–0.10)

p

0.010

0.202

r (95 % CI)

0.66 (0.50–0.78)

0.65 (0.44–0.79)

p

\0.001

\0.001

Total score

These are correlations prior to adjustment for sex and age effects

In the Off-beat test, the E model, where all the variation is explained by non-shared environmental effects, was the best fitting model (the fit of the AE and E models was not significantly worse than the fit of ADE model, but the more parsimonious E yielded better BIC). In the ADE model, the estimates of dominant and additive genetic effects were non-significant.

123

In the Out-of-key test, the CE model was the most parsimonious model that fit the data well. In the ACE model, the estimate of additive genetic effects was zero. In both ACE and CE models, the estimates of shared environmental effects were strong (0.59 and 0.61, respectively). In the total score, the results were similar to the Out-ofkey test. The CE model was the most parsimonious model

Behav Genet Table 5 Model fitting results for the MBEA tests and total score

Model Scale test

Out-of-key test

Off-beat test

Total score

-2LL

df

D - 2LL

Ddf

P value

BIC

0.391

344.94

ACE

2813.01

378

AE

2813.09

379

0.08

1

347.70

CE

2816.69

379

3.68

1

0.040

346.74

E

2846.36

380

33.35

2

\0.001

358.78

ACE

2946.35

378

AE

2953.82

379

7.47

1

0.003

CE

2946.35

379

0

1

0.500

411.58

E

3004.57

380

58.22

2

\0.001

437.88

414.37 415.31

ADE

2726.71

378

AE

2728.25

379

1.54

1

0.107

304.55 302.53

E

2730.88

380

4.17

2

0.056

301.04

ACE

2548.41

378

AE

2556.46

379

8.05

1

0.003

216.63

CE

2548.41

379

0

1

0.500

212.61

E

2611.17

380

62.76

2

\0.001

241.18

215.40

Best fit models for each test and total score are shown in bold -2LL minus 2 log likelihood, df degrees of freedom, D - 2LL difference in the -2LL values against the full model, Ddf difference in degrees of freedom against the full model, BIC Bayesian Information Criterion

Table 6 Estimates for variance explained by different latent components in different models

Model

A

C

D

E

Scale test

ACE

0.50 (0.00–0.69)

0.07 (0.00–0.50)



0.43 (0.31–0.60)

Out-of-key test

AE ACE

0.58 (0.42–0.70) 0.03 (0.00–0.46)

– 0.59 (0.19–0.70)

– –

0.42 (0.30–0.58) 0.38 (0.27–0.51)

CE



0.61 (0.49–0.70)



0.39 (0.30–0.51)

ADE

0.00 (0.00–0.34)



0.24 (0.00–0.44)

0.76 (0.56–0.99)

AE

0.18 (0.00–0.39)





0.82 (0.61–1.00)

E







1.00 (1.00–1.00)

ACE

0.01 (0.00–0.44)

0.63 (0.23–0.73)



0.36 (0.25–0.47)

CE



0.64 (0.53–0.73)



0.36 (0.27–0.47)

Off-beat test

Total score

Estimates are shown only for full models and those models whose fit to the data was not significantly worse than that of the corresponding full model A additive genetic effects, D dominant genetic effects, C shared environmental effects, E non-shared environmental effects

that fit the data well. In the ACE model, the estimate of additive genetic effect was non-significant. In the ACE and CE models, the estimates of shared environmental effects were strong (0.63 and 0.64, respectively). In order to control for the potential effects of individuals’ decision making processes and response bias, we also applied the signal detection theory and recalculated the models using d0 values [d0 = z(hit rate) - z(false alarms)] for each subtest. This analysis yielded similar results as the original scores: the Scale subtest was best explained by an AE model (A = 0.61, E = 0.39), the Off-beat subtest by an AE model (A = 0.34, E = 0.66), the Ouf-of-key subtest by a CE model (C = 0.50, E = 0.50) and the total score by a CE model (C = 0.54, E = 0.46).

Discussion Our main finding was the highly contrasting pattern of genetic and environmental effects observed in the three tasks measuring different aspects of music cognition. In full models, the Scale test showed a clear additive genetic effect (0.50) and no shared environmental effect (0.07) whereas the opposite pattern was observed in the Out-ofkey test with strong shared environmental effect (0.59) and no additive genetic effect (0.03). In contrast to both of these, the variance in the Off-beat was mainly accounted for by non-shared environmental effects (0.76) with genetic effects being small and non-significant (0.00–0.24).

123

Behav Genet

The additive genetic effect in the Scale test is comparable and in line with the previous population-based twin studies which reported notable genetic effects for the DDT (Drayna et al. 2001) and for the SMDT Melody test (Mosing et al. 2014a). In this regard, our results replicate and extend previous findings that variation in musical pitch discrimination is predominantly genetic in nature and that this is seen also with naturalistic and unfamiliar melodic stimuli. The novel result of the present study was the high environmental effects and non-existent or small genetic effects observed in the Out-of-key and Off-beat tests, which are in contrast with the results of previous studies and the Scale test. In general, shared environmental effects of the magnitude observed in the Out-of-key test are rare in adult twin studies (Boomsma et al. 2002; Polderman et al. 2015; Vinkhuyzen et al. 2009), although not unheard-of (Coon and Carey 1989). The result is unlikely due to selection bias or methodological artefacts because the other tests showed distinct differences between MZ and DZ pairs, and the distribution of scores did not differ by zygosity or between full and no co-twin pairs. A plausible explanation for the contrasting results between the Out-of-key test and the Scale test (and the DDT and SMDT) is the general structure and cognitive demands of the tasks. The Scale test and the SMDT Melody test are both discrimination tasks where pairs of melodies are compared for pitch changes. Performance relies both on the ability to perceive whether the second melody contains a scale-violating note and the memorybased comparison of the two melodies (keeping the first melody in mind while listening to the second and comparing each pitch of the second melody to the memory representation of the first melody). By contrast, in the Outof-key test and DDT, only one melody is presented at a time and the ‘‘correctness’’ of the pitches in each melody has to be judged using prior musical knowledge. However, the DDT uses popular melodies, which makes performance partly dependent on previous exposure to the original songs (familiarity), whereas in the Out-of-key test the melodies are unfamiliar and therefore performance relies exclusively on knowledge (i.e. internal model) of musical keys, which is not inborn but acquired through experience (Trainor and Corrigall 2010) and largely implicit and automatic in adults (Bigand and Poulin-Charronnat 2006; Brattico et al. 2006; Tillmann et al. 2000). Thus, while performance on all these tasks is influenced by tonality perception, the Scale test, SMDT, and DDT also require additional working or longterm memory processing, making them cognitively more demanding and thereby viable to be driven more by genetic factors. In line with this, also Mosing et al. (2014b) found support for genetic pleiotropy between intelligence and SMDT performance.

123

Regarding rhythm perception, Mosing et al. (2014a) reported a moderate additive genetic effect in the SMDT Rhythm task, which involves discriminating two short rhythmical sequences for a change in time interval. In contrast, the Off-beat test yielded notably smaller withinpair twin correlations than the other tests, and the models indicated high non-shared environmental effects and small and non-significant genetic effects. Since the task format in the Off-beat and Out-of-key tests is the same, the disparate results compared to Mosing et al. (2014a) may be attributable to different task demands, as discussed above. The dissimilar pattern of genetic/environmental effects in our three tests may be related to differences in the cognitive and neural processing of tonal/melodic and rhythmic information and their differential sensitivity to environmental influence. Interestingly, the Off-beat test did not correlate with the Scale test and only modestly with the Out-of-key test (for similar results, see Hausen et al. 2013). Moreover, we found previously that music education is a significant predictor of Out-of-key performance, whereas Off-beat performance is more linked to the self-perceived importance of music and musical mood regulation (Hausen et al. 2013; Saarikallio et al. submitted). Previous twin studies which have probed the level of musical training or ability using self-report questionnaires have reported higher shared environmental effects than genetic effects, especially in women (Coon and Carey 1989; Vinkhuyzen et al. 2009). According to recent meta-analyses of fMRI and PET studies on music (Janata 2015; Samson et al. 2011), melody- and tonality-related processing engages primarily superior temporal areas (STG/STS), with the addition of prefrontal areas in tasks involving an active memory component, whereas the processing of rhythm recruits additional motor regions, including the striatum, cerebellum, and motor cortical areas. Rhythm is known to be a key component in music-induced positive affect and reward and the associated dopaminergic striatal functions (Trost et al. 2014; Witek et al. 2014), which according to recent twin studies are best predicted by non-shared environment effects rather that genetic or shared environmental effects (Menne-Lothmann et al. 2012; Stokes et al. 2013). Thus, it is possible that rhythm perception and its link to emotions and everyday musical engagement is explained mostly by the non-shared environment, but since this component also includes the measurement error of the model, the strength of the effect is uncertain. The present study had some limitations. First, the response rate to the online test was relatively low (38 %), although comparable to the Mosing et al. study (33 %). As a result, the sample size of the present study was relatively small. Although this produces overall power limitations, it does not have significant bearing on the observed main

Behav Genet

results, namely the differential pattern of genetic and environmental effects on the three subtests. Second, the sample comprised more females than males (72 vs. 28 %). Although the gender distribution did not differ significantly between MZ and DZ twins and gender was included in the models as a covariate, the results may nevertheless be more applicable to females than to the general population as whole. We only observed a gender effect in the Off-beat subtest in which men scored somewhat higher than women. Due to small sample size, especially in men, unfortunately we were underpowered to perform sex-limitation models in order to explore if the relative proportion of genetic and environmental effects differ between men and women. Third, the musical background of the participants was not controlled, making it possible that the sample could be more musically oriented than the population in general. Fourth, the participants did the test online and not in laboratory settings. However, since the situation was the same for both MZ and DZ twins, it is unlikely to cause the different pattern of results in the three tasks. Finally, 61 % of the participants had a Bachelor’s or higher degree, suggesting that higher education levels were somewhat over-represented. However, the genetic modelling was also performed with education level as a third covariate, and these results were almost identical to those presented here. In conclusion, the current study provides novel insight into the ongoing debate about the role of genes and environment in the development of musical skills by showing for the first time that different aspects of music cognition are driven by genetic and environmental factors to a different degree. Together with previous findings (Drayna et al. 2001; Mosing et al. 2014a), our results suggest that pitch discrimination ability is primarily determined by genetic factors whereas the implicit perception of musical tonal and metric structures is mediated more by shared and non-shared environmental factors. Thus, music perception or musical skills should not be regarded as a single, unitary domain driven mainly by genetic influence, but a complex, multifaceted phenomenon affected by both genetic and environmental factors. Acknowledgments We would like to thank Anja Ha¨ppo¨la¨ and Kauko Heikkila¨ for their invaluable help in data collection and Viljami Salmela for assistance on statistical analyses. Data collection of FinnTwin16 has been supported by the National Institute of Alcohol Abuse and Alcoholism (AA-12502, AA-00145, AA-09203) and by the Academy of Finland (100499, 205585, 118555, 141054, 264146). Additional support for the present work was provided by the Paolo Foundation, Finland, and the Academy of Finland (AoF) Post-Doctoral (257077, 257075) and Academy Professor (265240, 263278) programs.

Compliance with Ethical Standards Conflict of Interest Erik Seesja¨rvi, Teppo Sa¨rka¨mo¨, Eero Vuoksimaa, Mari Tervaniemi, Isabelle Peretz, and Jaakko Kaprio declare that they have no conflict of interest. Human and Animal Rights and Informed Consent All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed consent was obtained from all individual participants included in the study.

References Baharloo S, Service SK, Risch N, Gitschier J, Freimer NB (2000) Familial aggregation of absolute pitch. Am J Hum Genet 67(3):755–758 Bigand E, Poulin-Charronnat B (2006) Are we ‘‘experienced listeners’’? A review of the musical capacities that do not depend on formal musical training. Cognition 100(1):100–130 Boomsma DI, Busjahn A, Peltonen L (2002) Classical twin studies and beyond. Nat Rev Genet 3(11):872–882 Brattico E, Tervaniemi M, Na¨a¨ta¨nen R, Peretz I (2006) Musical scale properties are automatically processed in the human auditory cortex. Brain Res 1117(1):162–174 Chang H, Trehub SE (1977) Auditory processing of relational information by young infants. J Exp Child Psychol 24(2):324–331 Coon H, Carey G (1989) Genetic and environmental determinants of musical ability in twins. Behav Genet 19(2):183–193 Corrigall KA, Trainor LJ (2010) Musical enculturation in preschool children: acquisition of key and harmonic knowledge. Music Percept 28(2):195–200 Deary IJ, Johnson W, Houlihan LM (2009) Genetic foundations of human intelligence. Hum Genet 126(1):215–232 Dominicus A, Skrondal A, Gjessing HK, Pedersen NL, Palmgren J (2006) Likelihood ratio tests in behavioral genetics: problems and solutions. Behav Genet 36(2):331–340 Drayna D, Manichaikul A, de Lange M, Snieder H, Spector T (2001) Genetic correlates of musical pitch recognition in humans. Science 291(5510):1969–1972 Gingras B, Honing H, Peretz I, Trainor LJ, Fisher SE (2015) Defining the biological bases of individual differences in musicality. Philos Trans R Soc Lond B Biol Sci 370(1664):20140092 Goldsmith HH (1991) A zygosity questionnaire for young twins: a research note. Behav Genet 21(3):257–269 Hannon EE, Johnson SP (2005) Infants use meter to categorize rhythms and melodies: implications for musical structure learning. Cogn Psychol 50(4):354–377 Hausen M, Torppa R, Salmela VR, Vainio M, Sa¨rka¨mo¨ T (2013) Music and speech prosody: a common rhythm. Front Psychol 4(566):1–16 Janata P (2015) Neural basis of music perception. Handb Clin Neurol 129:187–205 Kalmus H, Fry DB (1980) On tune deafness (dysmelodia): frequency, development, genetics and musical background. Ann Hum Genet 43(4):369–382

123

Behav Genet Kaprio J (2013) The Finnish Twin Cohort Study: an update. Twin Res Hum Genet 16(1):157–162 Kaprio J, Pulkkinen L, Rose RJ (2002) Genetic and environmental factors in health-related behaviors: studies on Finnish twins and twin families. Twin Res 5(5):366–371 Krumhansl CL (2000) Rhythm and pitch in music cognition. Psychol Bull 126(1):159–179 Lee T, Henry JD, Trollor JN, Sachdev PS (2010) Genetic influences on cognitive functions in the elderly: a selective review of twin studies. Brain Res Rev 64(1):1–13 Menne-Lothmann C, Jacobs N, Derom C, Thiery E, van Os J, Wichers M (2012) Genetic and environmental causes of individual differences in daily life positive affect and reward experience and its overlap with stress-sensitivity. Behav Genet 42(5):778–786 Mosing MA, Madison G, Pedersen NL, Kuja-Halkola R, Ullen F (2014a) Practice does not make perfect: no causal effect of music practice on music ability. Psychol Sci 25(9):1795–1803 Mosing MA, Pedersen NL, Madison G, Ulle´n F (2014b) Genetic pleiotropy explains associations between musical auditory discrimination and intelligence. PLoS One 9(11):e113874 Neale M (2004) Mx software. Virginia Commonwealth University, Richmond Neale M, Cardon L (1992) Methodology for genetic studies of twins and families. Kluwer Academic Publishers, Dordrecht Oikkonen J, Huang Y, Onkamo P, Ukkola-Vuoti L, Raijas P, Karma K et al (2015) A genome-wide linkage and association study of musical aptitude identifies loci containing genes related to inner ear development and neurocognitive functions. Mol Psychiatry 20(2):275–282 Peretz I, Champod AS, Hyde K (2003) Varieties of musical disorders. Ann N Y Acad Sci 999(1):58–75 Peretz I, Cummings S, Dube´ M (2007) The genetics of congenital amusia (tone deafness): a family-aggregation study. Am J Hum Genet 81(3):582–588 Peretz I, Gosselin N, Tillmann B, Cuddy LL, Gagnon B, Trimmer CG et al (2008) On-line identification of congenital amusia. Music Percept 25(4):331–343 Polderman TJ, Benyamin B, de Leeuw CA, Sullivan PF, van Bochoven A, Visscher PM et al (2015) Meta-analysis of the heritability of human traits based on fifty years of twin studies. Nat Genet 47(7):702–709 Posthuma D, Beem AL, De Geus EJC, van Baal CM, von Hjelmborg JB, Iachine I et al (2003) Theory and practice in quantitative genetics. Twin Res 6(5):361–376 Saarikallio S, Hausen M, Sa¨rka¨mo¨ T, Tervaniemi M (submitted) If you love music you can hear the beat: daily music engagement relates to temporal abilities of music perception Samson F, Zeffiro TA, Toussaint A, Belin P (2011) Stimulus complexity and categorical effects in human auditory cortex: an

123

activation likelihood estimation meta-analysis. Front Psychol 1:241 Sarna S, Kaprio J, Sistonen P, Koskenvuo M (1978) Diagnosis of twin zygosity by mailed questionnaire. Hum Hered 28(4):241–254 Stokes PR, Shotbolt P, Mehta MA, Turkheimer E, Benecke A, Copeland C et al (2013) Nature or nurture? Determining the heritability of human striatal dopamine function: an [18F]DOPA PET study. Neuropsychopharmacology 38(3):485–491 Tillmann B, Bharucha JJ, Bigand E (2000) Implicit learning of tonality: a self-organizing approach. Psychol Rev 107(4):885– 913 Trainor LJ, Corrigall KA (2010) Music acquisition and effects of musical experience. In: Fay RR, Popper AN (eds) Springer handbook of auditory research: music perception. Springer, New York, pp 89–127 Trainor LJ, Hannon EE (2012) Musical development. In: Deutsch D (ed) The psychology of music, 3rd edn. Academic Press, London, pp 423–497 Trainor LJ, Trehub SE (1992) A comparison of infants’ and adults’ sensitivity to western musical structure. J Exp Psychol Hum Percept Perform 18(2):394–402 Trehub SE, Schellenberg EG, Kamenetsky SB (1999) Infants’ and adults’ perception of scale structure. J Exp Psychol Hum Percept Perform 25(4):965–975 Trost W, Fru¨hholz S, Scho¨n D, Labbe´ C, Pichon S, Grandjean D et al (2014) Getting the beat: entrainment of brain activity by musical rhythm and pleasantness. Neuroimage 103:55–64 Ukkola LT, Onkamo P, Raijas P, Karma K, Ja¨rvela¨ I (2009) Musical aptitude is associated with AVPR1A-haplotypes. PLoS One 4(5):e5534 Ulle´n F, Mosing MA, Holm L, Eriksson H, Madison G (2014) Psychometric properties and heritability of a new online test for musicality, the Swedish Musical Discrimination Test. Pers Indiv Differ 63:87–93 Vinkhuyzen AAE, Van der Sluis S, Posthuma D, Boomsma DI (2009) The heritability of aptitude and exceptional talent across different domains in adolescents and young adults. Behav Genet 39(4):380–392 Volk HE, McDermott KB, Roediger HL 3rd, Todd RD (2006) Genetic influences on free and cued recall in long-term memory tasks. Twin Res Hum Genet 9(5):623–631 Winkler I, Haden GP, Ladinig O, Sziller I, Honing H (2009) Newborn infants detect the beat in music. Proc Natl Acad Sci USA 106(7):2468–2471 Witek MA, Clarke EF, Wallentin M, Kringelbach ML, Vuust P (2014) Syncopation, body-movement and pleasure in groove music. PLoS One 9(4):e94446 Zentner M, Eerola T (2010) Rhythmic engagement with music in infancy. Proc Natl Acad Sci USA 107(13):5768–5773