The Development of Affective Responses to Modality and Melodic Contour

Music Perception Spring 1995, Vol. 12, No. 3, 279-290 1995 by the regents of the university of California The Development of Affective Responses to Modality and Melodic Contour GINA M. GERARDI Oberlin College and Conservatory of Music LOUANN GERKEN Department of Psychology, SUNY at Buffalo Although it is well established that melodic contour (ascending vs. descending) and modality (major vs. minor) evoke consistent emotional responses in adult listeners, the mechanisms underlying musical affect are unknown. One possibility is that the mechanisms are based on innate perceptual abilities (e.g., Helmholtz, 1885/1954). Another possibility is that the ability to associate various aspects of music with emotion is learned through exposure to one's musical culture (e.g., Serafine, 1988). The current research examines the affective responses to major and minor ascending and descending melodies by 5-year-olds, 8-yearolds, and college students. Affective responses to modality did not appear until age eight and affective responses to contour appeared only in the college students. These results are consistent with previous developmental perception experiments on contour and modality (Imberty, 1969; Krumhansl & Keil, 1982; Morrongiello & Roes, 1990) and extend the understanding of the relation of perception, cognition, and culture in determining musical affect. can have profound effects on people of all cultures, and these effects are frequently expressed by emotional states. One aspect of music that has been associated with emotion for more than 300 years is melodic contour. In the chorale prelude of J. S. Bach, "Through Adam's fall all is spoiled," all four voices use descending patterns that aurally signify the fall from grace to sin (Grout &c Palisca, 1988). Another aspect of music that has historically been associated with emotion is modality. Manipulation of modality extends back to the modal system of the ancient Requests for reprints may be sent to Gina Gerardi, Department of Psychology, University of Oregon, Eugene, OR 97403. e-mail: gerardi@oregon.uoregon.edu. 279

280 Gina M. Gerardi & LouAnn Gerken Greeks. From their system, only the ionian and aeolian modes, known today as major and minor, respectively, are commonly used in Western music. According to the stereotype of today, major and minor modes convey positive and negative affective connotations, respectively. Although there is wide agreement about the affective quality of various aspects of music, little is understood about how emotional associations to musical variables develop. To date, only a few studies have examined children's emotional responses to music. Two studies in which children were presented with excerpts from classical music demonstrated that beginning at age five, children are able to assign "happy" or "sad" labels in a manner very similar to adults (Cunningham &c Sterling, 1988; Terwogt & Grinsven, 1988). Although these studies are informative, they do not indicate which aspects of music influence affective judgments. In general, there are strong correlations among mode, tempo, timbre, and other musical variables (Cunningham &c Sterling, 1988; Hevner, 1935; Scherer &C Oshinsky, 1977). Therefore, determining if contour or modality in particular are salient is impossible without more controlled stimuli. Several other developmental studies have examined infants' and children's sensitivity to musical contour, an aspect of music that has been historically associated with affect. Thorpe (1986) used the reinforced head-turn procedure to train 7- to 10-month-old infants to respond to contour violations in melodies. She found that infants are able to detect contour violations as small as a half-step. Examining older listeners, Morrongiello and Roes (1990) taught 5- and 9-year-olds to point to a line drawing that best matched the contour of a melody. They found that both age groups were able to learn the task relatively quickly (in approximately 9-11 trials) with no significant developmental differences. Although these studies do not deal specifically with the emotional effects of musical contour, they indicate that contour is perceived very early. Thus they open the door to the possibility that the emotional effects of contour are present very early as well. We will examine this possibility in conjunction with another emotionally relevant factor in music, modality. A number of infant studies have examined sensitivity to diatonic scale structure - the pattern of whole-steps and half-steps that determines if a piece is in a major or minor mode (Trainor & Trehub, 1992; Trehub, 1987; Trehub &c Trainor, 1990). For example, Cohen, Thorpe, and Trehub (1987) tested 7- to 11 -month-old infants' ability to discriminate contour changes between standard and test melodies. They found that infants were better able to detect changes when the standard was diatonic than when it was nondiatonic (also see Trehub, Thorpe, & Trainor, 1990). Other studies of children's ability to go beyond discrimination to more complex musical judgments suggest that a firm grasp of diatonic scale structure develops

Modality and Melodic Contour 281 between 5 and 9 years of age (Krumhansl &c Keil, 1982; Morrongiello & Roes, 1990). Relatively few studies, however, have investigated children's sensitivity to modality itself - a specific pattern within a diatonic scale. Imberty (1969, cited in Hargreaves, 1986) found that 8-year-olds, but not 7-yearolds, were able to detect when a melody had been changed from the major to the minor mode. Only recently have affective responses to modality per se been examined. Kastner and Crowder (1990) investigated children's affective responses to major and minor versions of simple melodies presented both with and without accompaniment. Their goal was to contrast physiological and cultural accounts of musical affect: Based on Helmholtz's physiological theory, children should show adultlike affective responses earlier for accompanied melodies, because consonance and dissonance arise specifically from the triads (chords) that occur in accompanied melodies. Conversely, based on a cultural exposure account, unaccompanied melodies should elicit adultlike affective responses earlier, because most children's songs are presented unaccompanied (Cohen et al., 1987; Simonton, 1984). Thirty-eight children between the ages of 3 and 12 years were introduced to four faces, denoting happy, neutral, angry, and sad emotions. They were then exposed to 12 melodies and asked to point to the face that best expressed the emotional quality of the music. Kastner and Crowder report that even their youngest subjects differentially associated major melodies with happy or neutral faces and minor melodies with sad or angry faces, suggesting that whatever the basis of musical affect is, it appears quite early. One surprising result from this study was an interaction between mode and accompaniment, such that minor accompanied and major unaccompanied melodies were most likely to be associated with angry/sad versus happy/neutral emotions, respectively. Kastner and Crowder interpret children's negative responses to minor accompanied melodies as support for Helmholtz's physiological theory of music and emotion. They suggest that children's more positive response to unaccompanied than accompanied major melodies is due to the familiarity of unaccompanied major melodies (Zajonc, 1980). Kastner and Crowder's finding that their youngest children associated accompanied minor melodies with negative emotions is apparently at odds with findings by Cunningham and Sterling (1988), who found that 4-yearolds were not able to associate minor melodies with negative emotions at a rate greater than chance, even though their stimuli were unaltered excerpts from symphonic works and therefore contained multiple cues for affect. However, it is possible that, as Kastner and Crowder suggest, the nonverbal, nonanalytical task demands of their study allowed even very young children to demonstrate an effect of modality. Kastner and Crowder's hypothesis that young children find familiar

282 Gina M. Gerardi ôc LouAnn Gerken musical attributes to be more positive than unfamiliar attributes is apparently at odds with a study by Hargreaves and Castell (1986), who reported that 4- and 5-year-old children did not discriminate in their preference ratings between familiar and unfamiliar melodies, whereas older children did. It is particularly interesting to note that most of Hargreaves and Castell's familiar melodies were common nursery rhymes and were therefore probably composed in major. Preschool children rated these melodies to be as appealing as examples that were "far statistical approximations" to music (also see Davies, 1969). This study suggests that familiarity does not necessarily lead to higher liking ratings among preschool children. Consistent with Hargreaves and Castell's findings, it is possible that Kastner and Crowder's youngest subjects did not treat major melodies as unequivocally positive, but rather as neutral. In pilot work, Kastner and Crowder found that children would sometimes describe major melodies as "OK" or "plain" and minor melodies as "angry," hence the inclusion of neutral and angry faces in addition to happy and sad ones. The fact that they combined happy and neutral responses raises the possibility that major was perceived predominantly as neutral, not necessarily as intrinsically positive. This is not the case for adult listeners, who describe major melodies as emotionally positive (e.g., Hevner, 1935). Therefore, in order to determine the age at which children can demonstrate adultlike affective responses, it might be better to exclude the neutral option. In the research reported here, we examined children's affective responses to the modality of unaccompanied melodies, employing a method similar to that of Kastner and Crowder, but requiring either a clearly positive or negative response. In addition, we included two opposing musical contours to investigate the affective relevance of contour and to determine the relative ages at which contour and modality have their effects. Methods SUBJECTS The final sample of 68 subjects consisted of 20 five-year-olds (mean, 4.99 years; range, 4.42-5.66 years), 23 eight-year-olds (mean, 8.22 years; range, 7.58-9.08 years), and 25 college students (mean, 19.17 years; range, 18-22 years). (Two 5-year-olds who had been tested were excluded from the final sample because they did not attend to the task or because of a lack of differentiation in responses to stimuli.) Five-year-olds were enrolled in nursery schools in the area of Oberlin College, 8-year-olds were students at a local public elementary school, and college students were enrolled in an introductory psychology course at Oberlin College. The college students received course credit for their participation and were aware that the study would involve music and affect in advance of testing. Approximately equal numbers of males and females of each age were recruited.

Modality and Melodic Contour 283 STIMULI Four unaccompanied melodies were chosen from Music for Sight Singing by Richard Ottman (1956), a standard conservatory text used to develop the ability to read and sing unfamiliar melodies accurately. Three of these melodies were from Part I: #208, #209, and #221. The fourth melody was from Part III: #49. These particular melodies were chosen because they were composed in either major or minor and displayed a predominantly ascending or descending melodic line (average percentage of melodic intervals in a single direction, 67%). In addition, the four possible combinations of duple/triple meter and simple/compound beat division were represented: one melody each was composed in 4, 4, g, and g time. We chose melodies exhibiting several different patterns of meter and beat division only to maintain the external validity of the stimuli; these dimensions were not considered variables for investigation. Three melodies used sequences that exemplified the predominant direction of the melody at two levels, both within the pattern and between adjacent patterns. The remaining melody displayed a pattern that repeated itself in slightly altered form. All melodies were eight measures in length, contained between 13 and 26 notes (mean, 19 notes) and began and ended on pitches of the tonic triad. Each melody underwent transformations of both modality and contour in addition to key transposition. If the melody was originally ascending, it was written out backwards to become descending. If the melody was originally composed in major, it was altered to reflect minor. In line with compositional conventions, the third scale degree was lowered one half-step in both ascending and descending versions of the minor melodies, but scale degrees six and seven were lowered only in the descending versions. Thus, each melody generated a total of four stimuli (ascending-major, descending-major, ascending-minor, descending-minor) and served as its own control (see Figure 1). Each set of four related melodies was transposed such that either C, D, F, or G served as tonic in order to keep all melodies within the range B, a minor second below middle C, to F, an eleventh above. Tessitura (note span) of melodies averaged an octave (range, perfect fifth to perfect twelfth). By using several different keys, we were also able to thwart any tendency for listeners to develop an increasingly strong sense of tonic as melodies were presented. Because the tonic contributes significantly to one's sense of key, and therefore to the modality as well, the use of only one key might make subjects more sensitive to the modality of melodies that occurred late in the experiment. Melodies were randomized such that one of each set occurred within a block of four adjacent stimuli. The score of the stimuli in their final form is available on request from the first author. The computer program Finale was used with an Apple Macintosh II si microcomputer Fig. 1. Four melody types were generated from each melody. Ascending-major, descendingmajor, ascending-minor, and descending-minor versions of one of the four original melodies are presented (top to bottom).

284 Gina M. Gerardi & LouAnn Gerken to transfer the melodies onto computer disc. Because three of the four melodies began with a partial measure (pickup), and the computer relates what is written without any expression, it was necessary to add slight emphasis to strategic pitches in order for the music to convey a sense of meter. One accent (velocity, 30) was programmed into each melody on the downbeat of the first full measure. On beat two of this measure, for the g melodies, or on the downbeat of the second full measure for the 4, 4, and g melodies, an additional accent (velocity, 40) was added. Once programmed, the melodies were played by a Yamaha DXS synthesizer. The timbre chosen resembled a marimba. All melodies were played back at 96 beats/min for the main beat unit (quarter or dotted quarter note). Melodies averaged 9.25 s in length (range, 7-10 s) and were separated by pauses of approximately 6 s. Total duration of melodies and rests equaled 4 min 18 s. A Sony stereo cassette-corder (WM-D3) recorded the melodies directly from the synthesizer onto a Maxell Type I normal tape. During testing, melodies were played on either a Panasonic slimline or Califone 3430 tape player at a comfortable audio level. Two schematic drawings of faces, one obviously happy and the other sad, were taken from Kastner and Crowder (1990). The faces were approximately 2x3 inches and were juxtaposed on two 8x5 sheets of drawing paper. PROCEDURE All preschool and elementary school children were tested individually during school hours. Each child was brought to a small room and seated at a table either opposite or next to the experimenter. Elementary school children were informed that the experimenter hoped to learn how children heard the different feelings in music. Preschool children were asked if they would like to hear some songs and play a game. All children were asked if they noticed that not all music sounded the same, and all of them indicated that they had. They were told that they would hear 16 very short melodies, or tunes. Each melody would be followed by a little pause in which they should point to the face that matched the melody the best. Children were then encouraged to describe the two faces, and all were able to correctly identify the faces as happy or sad. The experimenter further explained that she would record their decisions. No decision was recorded while the music was playing, even when a child pointed well before the example was finished; the experimenter looked up at the child only in between examples. The placement of the sad face on the right or left was alternated between subjects. On the relatively few occasions that a child failed to point to one of the faces at the end of the melody, the experimenter encouraged the child to respond by saying "Can you choose one?" All of the subjects whose data are reported selected a face for each melody and did so within the 6-s interval that separated the melodies. One child in the preschool group generally failed to attend to the task and consequently failed to select a face after many of the melodies. This child's data were excluded from the analyses. College students were tested as a group. Each subject received a packet of four quartersheets of paper. The first page was numbered 1-4 and subjects were asked to continue numbering, four to a page. They were then told that they would hear sixteen melodies, each followed by a pause of approximately 5 s. Their task was to indicate whether a happy or sad label was most appropriate for each melody. All instructions were both spoken and written on the blackboard. Results For each melody, a happy response was coded as one and a sad response as zero. As noted above, there were no response failures, therefore the happy and sad responses are complements of each other. The sixteen melo-

Modality and Melodic Contour 285 Fig. 2. Mean scores for melody types (ascending-major, descending-major, ascendingminor, or descending-minor) by age. Responses to each of the four melodies within a category were coded as one if "happy" or zero if "sad." dies were organized into four categories: ascending-major, descendingmajor, ascending-minor, and descending-minor, and each subject received one score in each category. The scores in each category ranged from zero if all four melodies in that category were labeled sad, to four if all four melodies were labeled happy (see Figure 2). Scores were subjected to 3 Age (5 years, 8 years, adult) x 2 Contour (ascending, descending) x 2 Modality (major, minor) analysis of variance (ANOVA). Age was a between-subjects factor, and Contour and Modality were within-subject factors. There was a significant main effect of Age [5 years, 1.83; 8 years, 2.14; college 2.25; F(2,65) = 10.34, p <.001]. Pairwise comparisons (Newman- Keuls, p <.05) demonstrated that the 5-year-olds made fewer happy responses than either of the other two groups. There was a strong main effect of Modality, such that major melodies were perceived as more positive than minor melodies [major, 2.56; minor, 1.62; F(l,65) = 35.43, p <.0001].

286 Gina M. Gerardi & LouAnn Gerken There was a trend for ascending melodies to be perceived more positively than descending melodies [ascending, 2.21; descending, 1.96; F(l,65) = 3.46, p <.07]. Age interacted significantly with both Modality [F(2,65) = 16.46, p <.0001] and Contour [F(2,65) = 4.02, p <.025]. To determine the nature of these interactions, 2 Modality x 2 Contour ANOVAs were performed for each age group separately (see Figure 2). A Wilcoxon Matched Pairs Signed Ranks test was also used to analyze the data within each age group by items. For 5-year-olds, the by-subjects analysis revealed no main effects of either Modality (major, 1.75; minor, 1.90; F < 1) or Contour (ascending, 1.83; descending, 1.83; F < 1), nor was there a significant interaction (F < 1). In the analysis by items, major was rated more positively than minor in only two of the eight pairs (Wilcoxon rank =11, n.s.), and ascending melodies were rated higher than descending melodies in four of the eight pairs (Wilcoxon rank = 14.5, n.s.). In the analysis by subjects, 8-year-olds rated major melodies as significantly more positive than minor melodies [major, 2.57; minor, 1.72; F(l,22) = 7.79, p <.02]. They showed no main effect of Contour (ascending, 2.13; descending, 2.15; F < 1); however, they did show a marginal Modality x Contour interaction [JF(1,22) = 3.44, p <.08]. Interestingly, they perceived major-ascending melodies as most positive and minor-ascending melodies as most negative. This pattern was borne out in the analysis by items: major was rated more positively than minor in seven of the eight pairs (Wilcoxon rank = 3, p <.05, two-tailed), but ascending was rated more positively than descending in only four of the eight pairs (Wilcoxon rank = 14, n.s.). In the analysis by subjects, college students rated major melodies as significantly more positive than minor melodies [major, 3.20; minor, 1.30; F(l,24) = 77.36, p <.0001]. They also rated ascending melodies as significantly more positive than descending melodies [ascending, 2.60; descending, 1.90; F(l,24) = 15.47, p <.001]. The Modality x Contour interaction was not significant (F < 1). This pattern was confirmed in the analysis by items: major melodies were rated more positively in all eight pairs (Wilcoxon rank = 0, p =.01, two-tailed), and in only one pair was the descending member rated more positively than the ascending member (Wilcoxon rank = 3.5, p <.05, two-tailed). Discussion With regard to modality, our 5 -year-old subjects did not show a significant major versus minor difference, whereas the 8-year-olds and adults did. This developmental pattern is consistent with research conducted by Imberty (1969), which suggests that children are unable to associate emo-

Modality and Melodic Contour 287 tion with major and minor melodies until their early school years. With regard to contour, our data show that predominantly rising versus falling melodies are not given an emotional connotation until sometime after age eight. Thus, it appears that the emotional connotations given to both mode and contour must develop with cognitive maturity and/or exposure to the musical culture. Although our data are consistent with those of many previous studies, there are some apparent differences. The most obvious difference is the one between our 5 -year-olds, who did not affectively distinguish major versus minor melodies, and the preschool children studied by Kastner and Crowder (1990), who reportedly did. This difference cannot be explained by difficulties posed by verbal and/or highly analytic task demands, because we required our subjects to make nonverbal, nonanalytic responses of the type elicited by Kastner and Crowder. We attribute the apparent difference between the studies to two factors. One is that the present study used only unaccompanied melodies, whereas Kastner and Crowder used both accompanied and unaccompanied melodies. The other factor is that the present study required subjects to choose between happy and sad faces, whereas Kastner and Crowder included neutral and angry faces as well. If we interpret Kastner and Crowder's findings in light of the methodological differences between the current study and theirs, the possibility emerges that their youngest subjects, like our 5-year-olds, failed to distinguish between major and minor unaccompanied melodies, but rather gave largely neutral responses to both. The fact that the 5-year-olds in the current study chose the happy and sad faces almost equally often suggests that they would have frequently chosen a neutral face, had one been an option. Recall that Kastner and Crowder counted neutral responses to major melodies as correct (i.e., "happy"). This fact, coupled with their subjects' negative responses to minor accompanied melodies, could make it appear that the youngest children distinguished between major and minor melodies. This interpretation of Kastner and Crowder's data is consistent with the significant interaction they found between mode and accompaniment. If such an interpretation is correct, then the apparent differences between the two studies disappear, thereby supporting the contention that the ability to assign emotional connotations to unaccompanied melodies must develop over time. The remaining challenge to this view is children's negative responses to Kastner and Crowder's accompanied minor melodies. Kastner and Crowder propose that this aspect of their data supports Helmholtz's (1885/1954) notion that dissonance arises between minor chords and the overtone series, leading to negative affect. However, the finding is at odds with other research on children's emotional responses to accompanied music (Cunningham& Sterling, 1988; Imberty, 1969). Clearly this is an issue that requires further research.

288 Gina M. Gerardi &C Lou Ann Gerken Let us now consider an explanation for the present findings. With regard to contour, it is possible that the ability to associate particular contours with emotion is one that is learned with exposure to cultural norms. With regard to modality, two developmental changes are detectable in the current data. Five-year-olds' responses to modality are largely undifferentiated and cluster near chance (2 out of 4). By age eight, children begin to experience more positive affect toward major unaccompanied melodies, although their responses to minor unaccompanied melodies still hover in the neutral range. This shift in responses to major melodies gives rise to the more positive scores seen in 8-year-olds than in 5-year-olds. Interestingly, Cunningham and Sterling (1988) also found that the ability to associate major melodies with happy emotions appeared at a younger age than the ability to associate minor with sad emotions. We think that the 8-yearolds' positive responses to major melodies may stem from the fact that most songs for children are in a major mode (Cohen et al., 1987; Simonton, 1984). Thus, children can be presumed to have greater experience with major melodies. The concept of exposure effect, posited by Zajonc (1980) and discussed by Kastner and Crowder (1990), suggests how familiarity may be sufficient to induce preference. But, as is apparent in the study by Hargreaves and Castell (1986), it is only when children have attained sufficient cognitive maturity to demonstrate recognition of familiar melodies that the familiarity preference noted by Zajonc might be seen. Just as Hargreaves and CastelPs 4- and 5-year-old subjects did not show a preference for melodies that they had previously heard, we suggest that the 5- year-olds in the present study did not find major melodies more familiar and hence more positive. In contrast, we suggest that our 8 -year-olds were positively influenced by the relative familiarity of major melodies. By adulthood, responses to major melodies become even more positive, and responses to minor melodies are no longer neutral, but move well into the negative range. The greater differentiation of modality seen in adults suggests an additional developmental step. Perhaps by adulthood, listeners are able to recognize both major and minor modes and make culturally appropriate connotations. This contrasts with the 8-year-olds, whom we believe did not necessarily identify minor melodies as a separate type, but merely treated them as unfamiliar. The developmental step between 8- year-olds and adults might reflect the ability to organize music into patterns that can be remembered and compared. In particular, we propose that a developing sensitivity to tonal scale structure facilitates (or even underlies) modal differentiation. Interestingly, previous studies of the ability to organize tones into a tonal hierarchy suggest that this ability does not arise before age eight (Krumhansl &c Keil, 1982; Serafine, 1988). This is even the case when children have received explicit training in the Western idiom. Thus Serafine found that Suzuki musical training did not im-

Modality and Melodic Contour 289 prove children's judgments of tonal closure until age eight, at which time they performed at adult levels. In contrast, musically untrained 11 -yearolds continued to perform at chance levels for unaccompanied melodies, although they were almost perfect when accompaniment provided information about tonality. Similarly, Morrongiello and Roes (1990) found that musically trained 9-year-olds were better able to choose a line drawing that best matched a musical contour when the melody was tonal as opposed to atonal. Musically untrained 9-year-olds and trained and untrained 5 -year-olds did not perform significantly differently for tonal and atonal melodies. The current findings on the development of adultlike emotional interpretations of musical contour and modality are especially interesting in light of the infant studies discussed earlier. With regard to contour, several studies have demonstrated that the ability to differentiate even highly similar contours is present almost from birth (Thorpe, 1986; Trehub & Trainor, 1990; Trainor & Trehub, 1992). Seen in this context, our research suggests that although the ability to perceive contour may appear quite early, the ability to assign contour an emotional connotation takes time to develop. In a similar vein, studies by Trehub and her colleagues on infants' ability to detect changes in melodies constructed of diatonic versus nondiatonic scales might be taken to indicate that sensitivity to Western tonal structure is present very early in life (Cohen et al., 1987; Trainor & Trehub, 1990). However, it is possible that these studies tap basic perceptual or cognitive properties of Western scales, which reflect simple frequency ratios for important intervals (Trainor&c Trehub, 1990). Therefore, although sensitivity to some properties of scales may be present quite early, the ability to infer the harmonic structure, and thereby the modality, of music is one that requires sufficient musical exposure and cognitive maturity (Krumhansl & Keil, 1982; Serafine, 1988). In summary, the current study, seen within the context of previous research, suggests a complex and interesting interplay of perception, cognition, and culture in the development of musical affect. The nature of this interplay must now be explored in future developmental and crosscultural research.1 References Cohen, A. J., Thorpe, L. A., & Trehub, S. E. Infants' perception of musical relations in short transposed tone sequences. Canadian Journal of Psychology, 1987, 41, 33-47. 1. We thank Tom Bever, William Friedman, Peter Jusczyk, David Smith, and two anonymous reviewers for helpful comments and suggestions. We also thank the children and teachers who made the research possible. The research was supported by NSF grant #DBS9 1020952 to LAG.

290 Gina M. Gerardi & LouAnn Gerken Cunningham, J. G., & Sterling, R. S. Developmental change in the understanding of affective meaning in music. Motivation and Emotion, 1988, 12, 399-413. Davies, J. B. An analysis of factors involved in musical ability, and the derivation of tests of musical aptitude. Unpublished doctoral dissertation. University of Durham, 1969. Grout, D., & Palisca, C. A history of Western music, 4th ed. New York: W. W. Norton and Co., 1988. Hargreaves, D. J. The developmental psychology of music. London: Cambridge University Press, 1986. Hargreaves, D. J., & Castell, K. C. Development of liking for familiar and unfamiliar melodies. Paper presented at the Eleventh International Seminar of the International Society for Music Education, Frankfurt, 1986. von Helmholtz, H. L. F. The sensations of tone, 4th ed. (A. J. Ellis, Trans.). New York: Dover, 1954. (Original work published 1885) Hevner, K. The affective character of the major and minor modes in music. American Journal of Psychology, 1935,47, 103-108. Imberty, M. Vaquisition des structures tonales chez l'infant. Paris: Klinksiek, 1969. Kastner, M. P., & Crowder, R. G. Perception of the major/minor distinction: IV. Emotional connotations in young children. Music Perception, 1990 8, 189-202. Krumhansl, C. L., ÔC Keil, F. C. Acquisition of the hierarchy of tonal function in music. Memory and Cognition, 1982, 10, 243-251. Morrongiello, B. A., & Roes, C. L. Developmental changes in children's perception of musical sequences: effects of musical training. Developmental Psychology, 1990, 5, 814-820. Ottman, R. Music for sight singing. Englewood Cliffs, NJ: Prentice Hall, Inc., 1956. Scherer, K. R., ÔC Oshinsky, J. S. Cue utilization in emotional attribution from auditory stimuli. Motivation and Emotion, 1977, 1, 331-346. Serafine, M. L. Music as cognition. New York: Columbia University Press, 1988. Simonton, D. K. Melodic structure and note transition probabilities: a content analysis of 15,618 classical themes. Psychology of Music, 1984, 12, 3-6. Terwogt, M. M., & Grinsven, F. Recognition of emotions in music by children and adults. Perception and Motor Skills, 1988, 67, 697-698. Thorpe, L. A. Infants categorize rising and falling pitch. Paper presented at the meeting of the International Conference on Infant Studies, Los Angeles, April 1986. Trainor, L. J., & Trehub, S. E. A comparison of infants' and adults' sensitivity to Western musical structure. Journal of Experimental Psychology: Human Perception & Performance, 1992, 18, 394-402. Trehub, S. E. Infants' perception of musical patterns. Perception & Psychophysics, 1987, 41, 635-641. Trehub, S. E., & Trainor, L. J. Rules for listening in infancy. In J. T. Enns (Ed.), The development of attention: Research and theory. North-Holland: Elsevier, 1990. Trehub, S. E., Thorpe, L. A. & Trainor, L. J. Infants' perception of good and bad melodies. Psychomusicology, 1990, 9, 5-15. Zajonc, R. B. Feeling and thinking: preferences need no inferences. American Psychologist, 1980,35,151-175.