Proceedings paper



Steven M. Demorest, School of Music, University of Washington


Numerous studies have established the importance of phrase cues in melodic perception and recognition. We know that listeners use phrase cues to help make sense of musical structure (Gregory, 1978; Lerdahl & Jackendoff, 1983; Sloboda & Gregory, 1980), to learn new songs (Sloboda, 1977, Sloboda & Parker, 1985), and to recognize previously learned songs (Tan, Aiello & Bever, 1981; Chiappe & Schmuckler, 1997). We know less about when such strategies develop. Since not all of the world's music has the same phrase structure, we can assume that a certain amount of our response to musical phrase information is learned at some point, but when?

A recent study of children's melodic memory found that children as young as 7 years old use phrase cues as a grouping strategy in their memory for melodies even when they are taught the song as an entire unit (Demorest, 1999). Children from grades k-1 and grade 4 were asked to reconstruct two previously-learned songs by putting four melody blocks in the right order. In one condition the melody blocks matched the phrases of the melody, in the other, the blocks were of similar length, but divided against the phrase break. All of the children reconstructed a regular and a irregular set of blocks. The hypothesis was that if phrases cues were important in children's melodic representations then it should be harder to reconstruct melody blocks that did not match their internal representations. There was a significant difference in the number of times subjects had to listen to the blocks and move the blocks with the irregular condition requiring more listenings and more moves to complete the task. This was true for both the younger and the older children, though older children required fewer operations overall to complete the task in either condition.

In this earlier study the children were taught two common (though unfamiliar) children's melodies in English to maximize the ecological validity of the song-learning task. A number of studies have found evidence for the integration of text and melody in the musical memory of adults (Crowder, Serafine, & Repp, 1990; Serafine, Crowder & Repp, 1984; Serafine, Davidson, Crowder & Repp,1986) and children (Chen-Hafteck, 1999; Feierabend, Saunders, Holahan, & Getnick,1998; Morrongiello & Roes, 1990). The relationship between text and melody seems to be complex, and may be affected by subject experience, the nature of the memory task, and the subjects' culture. For children, text seems to be a crucial dimension in melodic recognition, and plays a significant role in melodic memory (Feierabend, Saunders, Holahan, & Getnick,1998; Morrongiello & Roes, 1990). It is possible that the children in my earlier study responded more to text phrase cues than musical phrase cues to help them memorize the songs, and then used that information when reconstructing the melody. This study investigates the role of musical phrase cues in children's melodic memory in the absence of meaningful text phrase cues.


The participants were 39 children aged 6-11 years from the Northwestern United States. They were taken from kindergarten (n=9), grade 3 (n=15), and grade 5 (n=15) of a local elementary school where they received general music instruction twice a week for 30 minutes. All of the children were taught two standard children's songs by their music teacher using the whole song or immersion method of rote teaching (i.e. not phrase by phrase) as a regular part of their music class. The two songs were Appalachian folk melodies, but both songs were set to an unfamiliar language (Maori) to remove text cues. This was considered the best solution to the issue of text cues since learning songs without text (e.g. on "la") would be quite unnatural for children this age, whereas learning songs in another language is not uncommon.

Phrase memory was tested using a reconstruction paradigm in which subjects are asked to demonstrate their memory for a melody by reconstructing a familiar four phrase melody on a computer from four pieces or "blocks" that were placed out of order (Figure 1). Children clicked on a block to hear the melody fragment it contained and then placed the blocks in a left-to-right sequence to reconstruct the song. The reconstruction approach is more involved than simple recognition, but does not rely on children's performance skills in testing their melodic recall.

Each subject performed the reconstruction task on the two melodies they had been taught in music class. In the regular phrase condition, the four melody blocks represented the four phrases of the melody while in the irregular phrase condition, the blocks were broken either before or after the natural phrase break. Figure 2 shows the regular and irregular divisions for Melody 1. The number of notes per block is very similar, but the difference is in the placement of the division. Thus if the memory task were simply a matter of remembering notes, there should be no difference in task difficulty between the two conditions. If however, children rely on structural information such as phrases as a memory aid, then the irregular condition would not match their internal melodic representation, and should be more difficult to reconstruct. All children reconstructed one melody in each condition. As with the earlier study, the hypothesis was that if phrase groupings were important in melodic memory, it should be more difficult to reconstruct the irregular phrase groupings than the regular. This difficulty would be reflected statistically as a within-subject difference in the total number of times subjects had to check the blocks, or block sequences, in solving the puzzle. In addition, the study looked at between-subject differences in performance by age and participation in private music lessons.

Figure 1. The reconstruction task using Impromptu (Bamberger & Hernandez, 1992-2000).


Figure 2. One of the test melodies in regular and irregular phrase groupings shown with the Maori text.


The primary question of the study was the influence of melodic grouping on children's reconstruction performance in the absence of text cues. Repeated measures analysis of variance revealed a significant difference in performance between the regular and irregular phrase conditions. Subjects had to listen an average of 2.67 more times to reconstruct the irregular melody blocks than the regular melody blocks [F (1, 33) = 12.77, p=.001]. This was true regardless of the age of the subject. Figure 3 shows the mean scores by age group for the regular and irregular phrase conditions. There were no significant between subject differences due to age [F (2, 33) = 0.69, p=.507] or musical training [F (1, 33) = 0.03, p=.864], but there was a significant age by training interaction [F (2, 33) = 4.74, p<.05]. This interaction can be seen in the graph in Figure 4 which demonstrates that while students with private training in grades three and five needed fewer hearings overall to complete their reconstructions, kindergartners with private training performed worse than their untrained counterparts.

Figure 3. Mean number of hearings required by each age group in the two phrasing conditions.


Figure 4. Mean number of hearings required for subjects with and without private training.


Reconstructing melodies whose segments do not correspond to melodic phrase breaks is a more difficult task for children even without the benefit of meaningful text cues to aid in segmenting the melody. This suggests that children as young as 6 years old are recognizing and employing purely musical phrase cues in song acquisition and memorization. In fact, when the performance of these students was compared with those of the earlier study, they actually required fewer moves overall to reconstruct the same melody in either condition. This difference may be due to a number of factors including the quality of music instruction at the different schools and other aspects of musical background, but it does suggest that the lack of text didn't impair student's performance. Indeed some research has suggested that simultaneous presentation of text and music can actually hamper song-learning and song performance (Goetze, 1986; Levinowitz, 1989; Welch, Sergeant, & White, 1995/1996). Future research should directly compare student's reconstruction performance with and without text cues in song-learning.

The lack of an age-related difference in overall performance contradicts the findings of the earlier study (Demorest, 1999) and is not consistent with overall developmental improvements memory. Perhaps the lack of a meaningful text reduced the memory advantage for older students in the study, forcing them to rely on their musical memory alone. It would be interesting to see if age-related differences in melodic memory are dependent on text conditions in future studies.

The interaction between age and private study is not too surprising given the relatively short time that the kindergartners had been studying privately. It is unlikely that any instructional benefits would be present after such a short time, and it is unusual for most children to be receiving lessons at that young age regardless of their ability. The improvement in performance for older children may indicate the benefits of private training, or simply reflect the natural tendency of students with musical interests or aptitude to seek extra instruction.

There was one other issue which should be considered in interpreting the data. As with the earlier study, some students could not successfully complete the task. In this study 50% of the kindergartners tested could not reconstruct both melodies successfully. There was however, no pattern regarding whether they failed on the regular or the irregular phrase condition and some failures were do to stopping voluntarily, difficulties using the computer and attention span. As other studies have found, young children often have difficulty with melodic memory tasks (Feierabend, Saunders, Holahan, & Getnick,1998). It was interesting to note that the average age of the subjects who did not successfully complete both conditions was 2 months younger than those who did and included all of the subjects under age six (3).

The failure rate of the youngest subjects might also be reflective of a greater memory challenge involved in learning a song without the benefit of meaningful text cues. This would be consistent with the results of a study of preschooler's melodic recognition performance, that found that learning songs with text yielded a better performance on a subsequent recognition task (Feierabend, Saunders, Holahan, & Getnick,1998). Perhaps students who couldn't use musical phrase cues to help them memorize the song were simply unable to reconstruct any melody, regardless of phrase condition, though they could sing the songs from memory as a group in class.

The reconstruction paradigm was created to attempt to address some of the shortcomings of recognition and recall methodologies (Chen-Hafteck, 1999; Sloboda & Parker 1985). In the future, students' performance on reconstruction tasks should be compared directly to their performance on these other memory measures to determine the relative difficulty and concurrent validity of the various tasks for measuring children's melodic memory. The reconstruction paradigm may be a more challenging memory task for young children by virtue of the need to manipulate and sequence musical segments or the presentation of all songs without text.

The songs students learned for this study had a text, but the Maori text offered no linguistic cues to aid in phrase identification. While some studies have demonstrated that text can be an important aid in song learning for young children (Feierabend, Saunders, Holahan, & Getnick,1998; Morrongiello & Roes, 1990), children may also be able to respond to purely musical cues when segmenting melodies in long-term memory. Continued research on the interactions of musical cues and text cues in children's melodic memory has the potential to reveal much about the nature of musical memory and its development. It may also lead to better teaching strategies for developing children's musical thinking.


Bamberger, J. S., & Hernandez, A. (1992-2000). Impromptu. [Software, Version 1.0.2] New York: Oxford University Press.

Chen-Hafteck, L. (1999). Discussing text-melody relationship in children's song-learning and singing: A Cantonese-speaking perspective. Psychology of Music, 27, 55-70.

Chiappe, P., & Schmuckler, M.A. (1997). Phrasing influences the recognition of melodies. Psychonomic Bulletin and Review, 4, 254-259.

Crowder, R. G., Serafine, M. L., & Repp, B. (1990). Physical interaction and association by contiguity in memory for the words and melodies of songs. Memory and Cognition,18, 469-476.

Demorest, S. M. (1999). The role of phrase groupings in children's memory for melodies: A preliminary study. In Suk Won Yi (Ed.) Music, Mind, and Science (Seoul: Seoul National University Press) p. 410-419.

Feierabend,J. M., Saunders, T.C., Holahan, J.M., & Getnick, P.E. (1998). Song recognition among preschool age children: An investigation of words and music. Journal of Research in Music Education, 46, 351-359.

Goetze, M. (1986). Factors affecting accuracy in children's singing. Dissertation Abstracts International, 46, 2955A.

Gregory, A. H. (1978). Perception of clicks in music. Perception and Psychophysics, 24, 171-174.

Lerdahl, F., & Jackendoff, R. (1983). A Generative Theory of Tonal Music. Cambridge, MA: MIT Press.

Levinowitz, L. (1989). An investigation of preschool children's comparative capability to sing songs with and without words. Bulletin of the Council for Research in Music Education, 100, 14-19.

Morrongiello, B. A., & Roes, C. L. (1990). Children's memory for new songs: Integration or independent storage of words and tunes? Journal of Experimental Child Psychology, 50, 25-38.

Serafine, M.L., Crowder, R.G., & Repp, B.H. (1984). Integration of melody and text in memory for songs. Cognition, 16, 285-303.

Serafine, M.L., Davidson, J., Crowder, R.G., & Repp, B.H. (1986). On the nature of melody-text integration in memory for songs. Journal of Memory and Language, 25, 123-135.

Sloboda, J. A. (1977). Phrase units as determinants of visual processing in music reading. British Journal of Psychology, 68, 117-124.

Sloboda, J. A., & Gregory, A. H. (1980). The psychological reality of musical segments. Canadian Journal of Psychology, 34, 274-280.

Sloboda, J. A., & Parker, D. H. (1985). Immediate recall of melodies. In P. Howell, I. Cross & R. West (Eds.), Musical Structure and Cognition (pp. 143-167). London: Academic Press.

Tan, N., Aiello, R., & Bever, T. G. (1981). Harmonic structure as a determinant of melodic organization. Memory and Cognition, 9, 533-539.

Welch, G. F., Sergeant, D. ., & White, P. J. (1995-1996). The singing competencies of five-year-old developing singers. Bulletin of the Council for Research in Music Education, 127, 155-162.


  Back to index