Gender differences for constructed-response mathematics items

The present study explored gender differences in answers to constructed-res ponse mathematics items. Features relevant and irrelevant to the scoring ru bric but possibly related to gender differences were identified by a conten t expert after a review of the literature. Raters were trained to score the identified features for approximately 500 papers evenly divided across two grade levels and between genders. The papers rated for these features had holistic scores assigned by local teachers using a state-provided rubric. T he relationship between these holistic scores, studied features, objective scores, and gender differences was explored through standardized mean diffe rences, correlations, and covariance analyses. The results indicate that ru bric-relevant variables were highly predictive of holistic scores and accou nted for some of the gender differences. especially at seventh grade. Howev er, the possibly rubric-irrelevant feature of number of words written was s tatistically significant even when the rubric-relevant features were contro lled.