Although test anxiety (TA) has been shown to be prevalent among Australian university students, the 17-item German Test Anxiety Inventory (TAI-G) has only recently been shown to be a valid measure for use with Australian university students. However, contention exists within the literature as to whether the Lack of Confidence subscale is better conceptualized as a correlate of TA as opposed to a constituent factor because it has been found to have the weakest subscale intercorrelation. It has been suggested that this may be due to the positive item wording of this subscale, which is in contrast to wording of the other subscales. To test this, the Lack of Confidence subscale items were worded negatively for this study and previously established 30- item, 20-item, and 17-item models of the TAI-G were tested using confirmatory factor analysis (CFA) with 473 Australian university students. In line with expectations, the 17-item TAI-G provided the best fit to the data. Furthermore, in contrast to previous findings, there was no notable difference between models conceptualizing the reworded Lack of Confidence subscale as a correlate of TA and models incorporating it as a factor of TA. Moreover, the reworded Lack of Confidence subscale was found to have the strongest subscale intercorrelations. It was concluded that the 17-item TAI-G with the negatively worded Lack of Confidence subscale provided the most valid measure of TA, with semantically opposed items being detrimental to validity. Implications of these findings are discussed.