Being a fluent reader implies reading precisely, with a reading rate, phrasing and expressivity adapted to the text (Kuhn et al, 2010). However many pupils confuse reading fluently and reading fast, mainly because of the use of the number of word correctly read in a minute to assess fluency (Dowd and Baretlett, 2019). In this project, we are interested in acting on this representation by iinvolving them in expressivity and phrasing assessment by introducing peer and auto assessment. In this preliminary study, we wanted to assess the reliability of peer and auto-assessment compared to teachers and expert assessment of reading fluency using an scale adapted from the multidimensionnal fluency scale (MFDS) (Rasinski, 2004) for the use of children. 70 pupils and 3 teachers from grade 4 and 5 participated in this preliminary study. The pupils were recorded while reading a text with an instruction of expressivity. Then after a 30 min training with the adapted MFDS, participants were asked to asess 10 recorded readings of grade 4 children using the scale. Finally, the pupils were individually asked to assess their own reading by listening to the recording. We controlled for their reading rate and reading prosody abilities, as well as their competence belief in reading. The recordings were also assessed by the 3 teachers and 4 experts (researcher and student in reading science). We compared the inter-rater agreement and the results between each groups of rater on both peer and auto-assessment. The results showed a medium reliability of peer assessment compared to teachers and experts. Linear regression shown on impact on reading prosody, reading rate and self competence on the assessment, both in auto and peer-assessment. In futur, it could be interesting to train pupils to peer and auto-assessment on a regular basis and follow the impact on their reading abilities and self competence.