Revisiting singing proficiency in three-year-olds

What do we know about the singing ability of three-year-old toddlers?

According to the contour scheme theory (Davidson, 1985; McKernon, 1979), toddlers’ small vocal range impedes their ability to sing. With a pitch range spanning only a few semitones, toddlers necessarily compress the intervals of the target song to fit their available vocal scheme or range (Davidson, 1985; Dowling, 1984; McKernon, 1979). Children’s vocal pitch range is thought to expand gradually with development, which allows for increased singing accuracy (Davidson, 1985; Dowling, 1984; Flowers and Dunne-Sousa, 1990; McKernon, 1979; Rutkowski, 2013). It is notable, however, that insecure adult singers also compress intervals when singing (Pfordresher & Brown, 2007; Pfordresher, Brown, Meier, Belyk, & Liotti, 2010), raising the possibility that compression of intervals in song singing can reflect something other than immature vocal structures.

More compelling evidence against developmental limitations in vocal range as the major factors in young children’s inaccurate singing is the fact that three-year-olds’ vocal range in singing can be increased considerably by guided practice within a relatively short time (Jersild & Bienstock, 1931). Expansion of singing range is also possible with two-year-olds (De Vries, 2005). Moreover, infants and toddlers use a large vocal range during free play (Fox, 1990; Mang, 2001; Reigado, Rocha, & Rodrigues, 2011) and a larger range in pitch-matching exercises than they do when they sing songs (Flowers & Dunne-Sousa, 1990; Welch, Rush, & Howard, 1991). If young children can produce a large range of pitches during play, it is not obvious what prevents them from using their full vocal range when singing.

Just as early language development depends critically on the amount and nature of language input, especially in one-on-one contexts (Huttenlocher, Haight, Bryk, Seltzer, & Lyons, 1991; Ramírez-Esparza, García-Sierra, & Kuhl, 2017), the amount and nature of musical input must contribute to the large variation in toddlers’ singing proficiency (Kelley & Sutton-Smith, 1987; Moog, 1976). In fact, there are substantial variations in singing proficiency in school-age children (Flowers & Dunne-Sousa, 1990; Leighton & Lamont, 2006) and in adults (Pfordresher & Brown, 2007).

Viewing toddlers as capable singers

Reports of toddlers as capable singers have not received much attention from scholars and textbook writers in the field of music education, perhaps because of their focus on the type of singing acquisition that begins at around five years of age with the onset of formal schooling (Rutkowski, 2013; Welch, 2006). Nevertheless, there are noteworthy reports of singing skills in early childhood.

Moog (1976) described recordings and observations of the singing of 500 German children and the comments of their parents. He reported that “by the age of two, every child of normal development can sing” (p. 42). He was not referring to complete performances of songs but rather describes the predominant singing behaviour of this period, which he called “babbling songs.” He noticed, however, that some toddlers were capable of more advanced singing behaviours: stating that “one-third of children aged one to two begin singing songs which bear some resemblance to what has been sung to them” (p. 42). He found that most three-year-olds were capable of “imitative singing” or singing standard songs and song phrases heard in their environment. He concluded that “about half of the three-year-olds could sing the words, rhythm and pitch of a whole song more or less correctly” (p. 43). Moog’s view of three-year-olds as capable singers is markedly different from the conventional view that the singing of three-year-olds is largely inaccurate and limited in pitch range (Flowers and Dunne-Sousa, 1990). In cultures where singing is a common everyday practice for all (e.g. Zimbabwe), most three-year-olds are capable singers (Kreutzer, 2001). Even Western toddlers’ (20 to 40 months of age) average pitch range was found to be roughly nine semitones (range: 7–12 semitones) when they spontaneously sang a song that was highly familiar to them—“Twinkle, Twinkle, Little Star”—in the comfort of their home environment (Gudmundsdottir & Trehub, 2018).

Longitudinal studies of singing in preschool children are consistent with the view that children have different developmental trajectories (Kelley & Sutton-Smith, 1987; Moog, 1976; Whiteman, 2001). For example, some toddlers begin imitating songs at around two years of age, using simple syllables like loo or la instead of words (De Vries, 2005; Honig, 1995), or otherwise exhibit greater interest in melodies than in words (Stadler Elmer, 2012).

Scholars and educators of young children must consider the multifaceted nature of song singing. Songs have complex structures that integrate melodic pitches and rhythms. Most songs also have words, often but not always with meaningful content. The words of a song can be considered as providing percussive texture. Words are sounds, first and foremost, and are most likely treated as such by very young singers (Forrester & Borthwick-Hunter, 2015; Gudmundsdottir & Trehub, 2018). Some songs also feature bodily gestures that may be as important to young children as the syllabic sounds, rhythm, pitches, or melodic contour. At early stages in the song-learning process, such gestures may be the most important feature of a song for young toddlers (Forrester, 2010).

Learning to perform a song also involves learning the phrasing and expressive aspects (Stadler Elmer, 2012). For toddlers, singing a particular song may be linked to specific feelings or moods. A child who is learning the art of singing a song may not understand how crucial a stable tonal centre is to “culturally acceptable” performances or that intervals should be performed exactly as in the model song in repeated performances. Remarkably, some children seem to understand the importance of interval accuracy and consistency at an early age (Honig, 1995; Kreutzer, 2001; McGraw, 2017; Moog, 1976; Stadler Elmer, 2012). It is clear that frequent one-on-one practice with an adult facilitates the development of singing accuracy (De Vries, 2005). Even in group lessons, considerable progress in singing is possible with very young children (Jersild and Bienstock, 1931).

Testing three-year-olds

One factor contributing to the issue of limited data on singing in three-year-olds is their shyness about singing, especially in the presence of unfamiliar adults. Researchers, even parents, have reported that toddlers tend to stop singing when their singing becomes the focus of attention (Barrett, 2011; Dean, 2015). A number of toddlers and preschoolers regard their singing as a private venture, reacting negatively to others, even family members, who join in their singing (Dean, 2015; Forrester, 2010; De Vries, 2005). As a result, three-year-olds may not reveal their singing skills in a standardized test situation or in any evaluation conducted by an adult stranger (Crais, 1995, 2011). It is of considerable interest and importance to develop age-appropriate procedures for eliciting singing responses from this age group.

The present study of three-year-old children was guided by three research questions: 1) What is the most advanced singing ability observed in typically developing three-year-olds? 2) How common is their ability to produce contour-preserving renditions of standard songs? c) How commonly do they use a singing range that exceeds a few semitones?

Method

Standard-phrase (The ‘Ba-bou’ song)

The ‘Ba-bou’ song is a well-known song that is popular among three-year-olds in the local area (see Figure 1). The song has a refrain with the repeating non-words “ba-bou” and “tra-la-la.” This catchy phrase is often sung repeatedly by three-year-olds during free play. This refrain was selected as the Standard-phrase because of the likelihood that three-year-olds would be interested in singing it by themselves. At the same time, the phrase is challenging because of its six changes in melodic contour. In the testing session, the research assistant sang the song with appropriate hand gestures together with each child and then encouraged them to sing the refrain by themselves at least once.

OPEN IN VIEWER

Figure 1.

The Standard-phrase in the protocol: a refrain from a song highly familiar to the children in the study.

Results

Participation varied among tasks because children could choose to play one game and skip another. The Self-selected song had the highest participation at 87%; singing single pitches elicited 81% participation on average. Voice lift and the standard song phrase elicited responses from 82% of participants.

Voice lift

The children produced a large range of high siren pitches that were ascending glissandos. The highest pitches ranged from D4 to C6, with the pitch C5 as the most frequent high pitch (Figure 2). The majority (72%) of the children who participated in this item lifted their voice to C5 or above (Figure 2).

OPEN IN VIEWER

Figure 2.

The highest pitches produced by the children when imitating a siren. Some 72% of the children landed on C5 or higher in vocal ascending glissandos.

Matching single pitches

Pitch-matching accuracy differed across pitches. The most frequently matched pitch was C4 (middle C) with a 57% success rate, and the least frequently matched pitch was C5 with a 12% rate. A3, F4, and A4 resulted in pitch-matching success of 38%, 25%, and 32%, respectively (Figure 3).

OPEN IN VIEWER

Figure 3.

The pitch-matching task included five pitches. Middle C (C4) was the pitch most successfully matched by the three-year-olds.

Standard song phrase

Children’s melodic phrase productions were classified as interval preserving (IP), contour preserving (CP), or contour violating (CV). A majority of the children (61%) reproduced the phrase in an interval- or contour-preserving manner, the remaining children made some errors in melodic contour (see Figure 4). Of the subset of children who preserved the melodic contour, a third of the children (21% of the whole sample) also preserved the intervals.

OPEN IN VIEWER

Figure 4.

Preservation of melodic contour (CP) was found in the majority of the children’s renditions of the Standard-phrase and their Self-selected songs. Interval preservation (IP) occurred most often in Self-selected song. Contour violating (CV) renditions were more common in the Standard-phrase than in Self-selected song.

Self-selected song

The Self-selected songs were in all cases standard children’s songs from the repertoire of the preschools. If a child could not think of a song to sing, the researcher would suggest a couple of songs from the preschool repertoire. None of the children invented their own song during this item. Children’s renditions of Self-selected songs were analysed according to the following scheme:

1 Correct intervals with key stability.

2 Correct intervals but changing tonal centre.

3 Some intervals preserved but not consistently.

4 Contour preserved but not intervals.

5 Mostly chanting with little or no hint of melody.

Two expert judges (early childhood music education specialists with formal vocal training) applied this scheme to the children’s performances (see Figure 5), with high inter-rater reliability (α = .96). Only 5 children (15%) failed to produce contour-preserving renditions of their Self-selected songs. The remaining children produced contour-preserving performances (85%), some of them even singing the correct intervals with a stable tonal centre (24%).

OPEN IN VIEWER

Figure 5.

Singing accuracy in Self-selected song.

Pitch range of Self-selected song

The average pitch range for the Self-selected song was 8.29 semitones (range = 4–12 semitones, see Figure 6) with nine semitones as the most frequent singing range (n = 14, see Figure 6). The lowest pitch was F3 and the highest was C#5. Figure 7 displays the song range for each child’s Self-selected song (n = 34).

OPEN IN VIEWER

Figure 6.

Singing ranges used in Self-selected song. Fourteen children used the range of nine semitones when singing a Self-selected song.

OPEN IN VIEWER

Figure 7.

Singing ranges per individual child singing Self-selected song. The Y-axis displays the pitch names. The range sizes are in an ascending order from left to right. The horizontal line marks the middle C (C4).

Discussion

The aim of the present study was to provide evidence of the singing abilities of typically developing three-year-olds by means of engaging, age-appropriate tasks. An overall response rate exceeding 80% on most tasks attests to the generalizability of the findings for this population. Test participation is not always reported in previous singing research. However, participation in singing tasks in widely cited studies has been as low as 53% with five-year-olds (Welch, Sergeant, & White, 1996). The high level of participation by the toddlers in the present study confirms that the given tasks were realistic for this age group and attests to the appropriateness of the protocol applied.

Most children in this sample demonstrated the ability to lift their voices in upper registers and singing standard songs in a manner that preserved the melodic contour. In fact, most of the children singing a Self-selected song preserved the melodic contour and nearly a third of them further demonstrated the ability to preserve the intervals of the target song. These findings resonate with reports by Moog (1976) and Kreutzer (2001) suggesting that most three-year-olds are capable of singing songs in a culturally correct manner. The present findings do not corroborate results indicating that three-year-olds rarely demonstrate ability to correctly replicate songs from the surrounding culture (Davidson, 1985; Davidson, McKernon, & Gardner 1981; Flowers & Dunne-Sousa, 1990; McKernon, 1979).

Even though children at this age frequently invent their own songs, especially during free play (e.g. Björkvold, 1992; Sundin, 1998; Whiteman, 2001), they have gained considerable knowledge of the standard song repertoire in their culture and demonstrate that they have mastered a standard song when they get the opportunity. Indeed, when children are observed by adults as “inventors” of songs, the children themselves may not be aware that they are inventing songs as from the point of view of adults. The act of inventing songs is merely a playful mode of being for three-year-olds. Therefore, it does not come naturally to start inventing a song when an adult asks them to sing a song outside of the context of spontaneous play. The children in this study seemed to have a clear concept of a standard song because none of them invented their own song when asked to sing a song of their choice.

Interestingly, the children performed a Self-selected song with greater accuracy than the phrase from a standard song they were all asked to sing. This was evident, even though the Self-selected song was the last item in the protocol, which could have resulted in poorer performances due to fatigue. The superior performance of Self-selected song might be explained by the effect of familiarity and practice, by way that they probably had sung the Self-selected song more often than the Standard-phrase. The children may have been more confident and comfortable singing a song that they chose themselves rather singing a phrase on demand, even in this friendly and playful setting. Nevertheless, this standard phrase is quite commonly sung with and by children of this age in Iceland. The difficulty of the Standard-phrase (nine-semitone range with six changes in melodic contour) does not surpass the most common Self-selected songs such as the “Colour song” (the same melody as “Twinkle, Twinkle, Little Star” with a nine-semitone range and seven changes in melodic contour). Therefore, it seems unlikely that the difficulty of the standard phrase greatly surpassed the difficulty of the Self-selected songs. The measured difference between the Self-selected song and standard phrase demonstrates how much the context and content of test items can affect performance accuracy. However, further studies on the effect of practice and song complexity on toddler’s singing are recommended.

The children matched the pitch C4, or middle C more often than the other pitches that were either lower or higher. A follow-up study should investigate whether the same success rate of pitch matching can be reached with pitches closer to C4. A closer look at the children’s responses suggests that non-matches are sometimes wrong by an octave, either above or below the target. This was most notable when the children were asked to sing A4, as six of them sung the pitch A3 an octave below. Interestingly, when asked to sing the pitch A3, there were only 12 children (38%) who sang that pitch. Other pitch matching errors in the single pitch matching test that could be mentioned were occasional misses by less than a whole-step (200 cents) from the target pitch (but more than a semitone). In conclusion, the task of matching single pitches using their voices may be confusing to some three-year-olds and they may not fully understand that the intent of imitation is to match a pitch rather than, for example, a timbre. Alas, it is not safe to draw the conclusion that absence of pitch matching in a given task means inability to match pitch. In fact, the different results for individual pitches suggest that the children in this study were somewhat selective of which pitches they did successfully match, perhaps based on the comfort range of their voice. Ostensibly, some three-year-olds may interpret the game of pitch matching according to their own liking and may be unaware of the aim of the task and the importance of matching the modelled pitches.

The present study demonstrates that proficient singing skills are not as rare in a population of normally developing three-year-olds as might be expected according to theories on singing acquisition in toddlerhood (Davidson, 1985; Davidson, McKernon, & Gardner, 1981; Dowling, 1984). With only a few exceptions, the three-year-olds in this study were able to sing contour preserving renditions of regular children’s songs using vocal ranges that approximated the target songs which is in line with reports on toddlers as capable singers (Kreutzer, 2001; McGraw, 2017; Moog, 1976; Stadler Elmer, 2012). The average vocal range used to sing a Self-selected song was over eight semitones and reaching up to 12 semitones, which surpasses the expected singing range of three-year-olds by a great margin. In fact, studies have suggested that the singing range of three-year-olds is usually no larger than seven semitones (Flowers & Dunne-Sousa, 1990) and five-year-olds might use only three to four semitones when singing songs with words (Rutkowski & Miller, 2003; Welch, Rush, & Howard, 1991). The findings of the present study suggest that assumptions may have been made previously about the singing ability and singing skills of toddlers based on too little evidence. Also, the limited data collected with the population of three-year-olds in previous research may not demonstrate optimal abilities of children because of the limitations of methods employed. Further singing data need to be collected using testing methods appropriate for the delicate population of toddlers.