Development of a Shortened form of the Fennema-Sherman Mathematics Attitudes Scales

Abstract

Data from 196 Irish school children were analyzed using principal components analysis with a varimax rotation to determine whether the underlying factor structure of the Fennema-Sherman Mathematics Attitudes Scales (FSMAS) fit the dimensions suggested by the position of the 108 items on nine subscales. Results indicated a factor structure virtually identical to a previous study that used a different sample, with the items being reduced to six separate components rather than nine as suggested by the scales' developers. Based on this factor structure, the authors attempted the development of a shortened form of the FSMAS. Internal consistency estimates of the reliability of scores on the whole scale and each subscale for both the original and the short form were favorable, with alpha coefficients ranging from .79 to .96.

Shortened Versions of Fennema-Sherman Mathematic Attitude Scales Employing Trace Information

“…To explore the characteristics of the current sample, we used the same factor analysis method employed by Mulhern and Rae (1998) to compare their results with those of the present study. Both studies identified six factors, and the percentages of variance accounted were 45.7% in Mulhern and Rae and 41.5% for the present study.…”

“…Because the BI methods presented below use the trace information criterion directly, they do not depend on any secondary results as is the case when employing Mulhern and Rae's (1998) factor analysis method. Details on this point are presented in the discussion section.…”

“…However, a similar study conducted by Mulhern and Rae (1998), identified only six factors suggesting that the FSMAS might not be measuring the domain-specific scales exactly as specified by Fennema and Sherman (1976a, b). Mulhern and Rae (1998) went on to develop a shortened version of the FSMAS (FSMAS-SF) that contained slightly less than half the items.…”

“…Fatigue alone would call into question the validity of responses to items near the end of the instrument. Development of Mulhern and Rae's (1998) FSMAS short form (FSMAS-SF) involved two stages. In the first stage, factor analysis was used to establish the internal structure of the whole scale.…”

“…Originally developed for applied science and technology applications, the BI-method selects items using the trace information criterion and can be applied to either the correlation or covariance matrix. Therefore, the purpose of our study was to apply Bhargava and Ishizuka's (1981) BI-method to develop short forms of the FSMAS that would show acceptable reliability and good predictive validity and then to compare the BI-results against Mulhern and Rae's (1998) factor analysis results. While the BI-method implements item selection using the familiar stepwise and forward selection procedures, we were also interested in extending it to employ an all-possible subset selection procedure, dubbed the All-pos method, and compare these results, too.…”