Task-Based Variability in Children's Singing Accuracy

Abstract

The purpose of this study was to explore task-based variability in children's singing accuracy performance. The research questions were: Does children's singing accuracy vary based on the nature of the singing assessment employed? Is there a hierarchy of difficulty and discrimination ability among singing assessment tasks? What is the interrelationship among different tasks and how few tasks might be employed in a comprehensive measure of accurate singing? A 2 X 4 factorial design was used to examine the performance of 4th grade children (n = 120) in both solo and doubled response conditions. Every child sang four task types: solo pitch, interval, pattern, and the song Jingle Bells. To account for the effect of tonal memory, test items in all tasks were presented in four total pitches by an adult female vocal model. Each task type contained five items, and the fifth item replicated the first to provide a measure of stability. Scoring was done by the researcher and one other judge with high reliability. Pitch matching was scored dichotomously and song singing was scored using an eight-point scale. Data were transformed to a 0-1 scale to express difficulty and discrimination indices. The results indicated that there was significant task-based variability in children's singing accuracy. Difficulty levels varied by task type, with patterns and songs indicating lower performance than single pitches and intervals. Performance was significantly higher for all tasks in the doubled condition than in the solo condition, and a significant interaction indicated task-based performance varied by response mode. Students who indicated a history of private lessons (n = 54) evidenced significantly higher performance than those without. An exploratory factor analysis demonstrated that all tasks load onto one factor. Internal reliability was satisfactory, and the results suggest that a minimum of three items can be included in each task in future research for a reliability coefficient of .75, or a minimum of four items for a coefficient greater than .80. These singing tasks were significantly inter-correlated, and the easiest item was also lowest in range. Vocal scooping and its implications for singing accuracy assessment were discussed. The three main findings were that doubled singing was more accurate than solo singing, summative assessment should include as many task types as is feasible, and within- and between-student performance should be compared using the same task type. Future research should explore variables affecting differing variability between lower- and higher-performing singers. Additionally, there is a relationship between history of private lessons and singing accuracy, and motivation and general musical experience should be explored as mediating or moderating variables so that teachers can best encourage student development. Last, it remains possible that doubled singing primes students to perform better in solo singing. Future research should examine under which conditions this may be true so that teachers can better target remediation for the majority of students who sing better when doubled by another voice.