Four phases were involced in the VPMD. The first phase focused on singing as 'chant-like'. Welch explained that 'the diviation between singing and speech are not clear, particularly for young children (Welch, 1998). If children persist the vocal behaviour into the later childhood, they are more often labelled as 'out-of-tune' (Welch, op.cit.).

(from Welch, 1979) Bentley's research (1968) reported that the proportion of monotones decreased with increasing age. from aged 7 (27 per cent.boys, 11 per cent girls.) to aged 12-14 years (7-8 per cent. boys. 2 per cent. girls.). This is a large-scale investigation in poor pitch singing.

(from Welch, 1979) Cleall (1970) tested 1,216 subjects, aged from 4 to 66-year-old. He reported that the pitch 'A4' was too high for 56 per cent of infant boys and 44 per cent of infant girls. (pitch range)

the 'real' monotones' : those who sing out of tune at all pitches.

(from Welch, 1979) Plumridge (1972) who tested 504 subjects in six schools ranging from 5 to 7 years of age. The tune Three Blind Mice was explored in several keys. She found only one child who exhibited two semitones to be classified as 'non-singer'. 13 other children with a limited singing range of three semitones are included in the data. - the relationship between comfortable singing range and pitch accuracy, if children have smaller comfortable singing range, it might be harder for them to sing a song which include a big range of pitches.

(from Welch, 1979) (Leont'et, 1969) the frequency of singing activity.

(from Welch, 1979) Wolner and Pyle (1933) singing in tune enabled children to distinguish changes in pitch, and respond accurately in a pitch test.

(Mang, 2006) Mang summarised that echoing individual pitches or short melodic patterns provided by an auditory stumulus, and subjects usually were asked to vocally reprodece immediately. Sining a complete song was 'a more complicated process', as children need to manage all elements, such as the lyrics and rhythms. Thus Mang though that compared with replicating pitches or short melodic patterns, singing a complete song and pitch matching could show more singing skills and competency. However, singing patterns does not related to singing songs . In a longitudinal research from Welch et al., (1996; 1997; 1998), they reported that children's singing accuracy of pattern was significantly higher than that of songs. When chosing materials, Mang prefered to use a complete song, as she though singing songs was 'one of the central activities of the elementary school music curriculum, and that song performance can provide a much more musical expressive experience for children' (op cit. p 163). The author agreed with Mang, as singing a song could be efficicy to build a more completed competency of singing accuracy.

Furthermore, different analysises were required between singing single pitches or patterns and singing a complete song. Fundamental frequency could used to measure single pitches or patterns (e.g. Trollinger, 2013), but a more general ratings, such as Voval Pitch Matching Development by Welch (1985), was required to measure singing a complete song. The results of two methods could be significantly different. Boys could sing more accurate than girls in patterns (Welch et al, 1996, 1997, 1998; Trollinger, 2013). However, gilrs sang more accurate than boys in songs in the study of Welch et al.

Welch's VPMD ratings was measured singing accuracy of song melodies (Welch, 1998). It developed from Welch (1986) which summarised from a varities of ealier literatures on children's singing accuracy.

Phase 2 stressed that children could sing more pitches, and the melodic outline begins to follow the general contours. Some mistakes were taken as self-invented and 'borrow' elements from children's music culture. Keys did not related to the rating.

Phase 3 could be rated if the main melodic shape are accurate, but some tonality change may occur. Phase 4 was no significant melodic or pitch error.

Instead of looking keys, lyrics, the VPMD focused on accuracies of pitch and rhythm .

The potensial problem of the VPMD rating is that it might be difficult to decide between Phase 2 and Phase 3. However, repeating recordings and a big number of participants could minimize the problem.

(Mang, 2006) investigated 120 aged 7 and 9 children's singing competency by age, sex and language. 60 children were Cantonese monolinguals and other were English bilinguals. A criterion song 'Happy Birthday' were used, and sang by English for both groups, and recordings were judged by two expers. Mang used VPMD and SVDM to evalua children' singing accuracy and vocal register, respectively. Highly aggreement between two judgrs about two scales (Pearsonn correlation: .821). Children sang the song in the comfortable singing range. Mang reported that gender and language were highly related to two scores. In terms of scores of VPMD, girls (mean: 3.56) sang slightly more accurate than that of boys (mean: 3.25), and Cantonese monolinguals (mean 3.6) sang more accurate than English bilingual (mean 3.22). Actuattly, aged 9 children (mean: 3.45) sang slightly more accuracy than that of aged 7 (mean: 3.365), but the difference was not significant. Mang explained that Cantonese -speaking children acquired singing voice earlier than that of English bilingual children. The familiar song might cause no difference on singing accuracy by age. Mang pointed that the bais was the study was that all children were from middle to lower middle class, and it could not represent the whole population.