What do consistency estimates tell us about reliability in holistic scoring ?

Abstract EN

Essay writing assessment is a largely used test in many examination types.

Preferred fortheir “validity and authenticity” (Hamp-Lyons, 2003:163), direct writing tests are prevailing in entrance, placement examinations, as well as in continuous assessment.

However, when compared to indirect writing testing, their level of reliability is questioned and their scoring procedures incriminated.

True scoring does not exist; errors stem from various sources: theraters, their training, the task (Huot, 1990); rendering essay marking doubtful, and raters’scoring inconsistent.

This study reports on a large scale, high-stake writing proficiency test taken by 441 students.

The essays were holistically scored on a 7-point scale by 16 raters.

ThePearson correlation coefficient was used for assessing the degree of consistency between raters.

The coefficientwas calculated for each pair of judges in the 25 groups of students.Results show positive correlation, but consistency in relationship has revealed some degree of variability between the paired samples.

The range of correlations fell between .16 and .91.

with the majority between .50 and .74.

These findings raise issues about the factors that threaten consistency of scoring in writing tests.