Speaker
Description
The incorporation of reverse-worded items in the development of psychometric instruments has been a longstanding practice. Indeed, reverse-worded items are routinely utilized to mitigate response style biases, including acquiescence bias, extreme response bias, and unthinking responding. The present study aimed to contribute novel insights regarding the wording effect. Specifically, we employed the Item Response Theory (IRT) approach to investigate this phenomenon, focusing on the Balanced Emotional Empathy Scale (BEES). A total of 312 participants (Mage = 21.35, sd = 8.25) completed the BEES. To explore the characteristics of the BEES items, we separately applied the IRT Graded Response Model to both non-reverse and reverse items. Additionally, we assessed age-related invariance. Our findings supported the unidimensionality of both item types when analyzed independently. Notably, the non-reverse items exhibited higher discrimination and information compared to the reverse items, while maintaining internal coherence. The BEES demonstrated satisfactory precision at the lower and middle-lower levels of the trait (Cronbach’s α = 0.85). However, age-related invariance was only evident within the medium to low trait levels. In conclusion, our findings caution against uncritical adoption of reverse-worded items as a standard practice, given their inherent conceptual challenges. Although we do not outright exclude the use of reverse items, we emphasize the importance of employing cost-effective tools and vigilantly contemplating any methodological artifacts that could impact research outcomes. Our empirical findings support a new parsimonious version of the BEES, comprising solely fifteen positively worded items.
