The Effects on Response Behavior of Different Response Scales on Semantic-Differential Items
(A2020-59009)

AUTHORS

Lars Bergkvist, Zayed University

KEYWORDS

semantic-differential; measurement; validity

ABSTRACT

Marketing researchers rely extensively on semantic differential (SD) items for self-report measures of common constructs. The response scales for SD items in marketing vary with respect to whether they have qualifying adverbs (e.g., “extremely” or “very”) before the endpoint adjectives or not, and whether the response categories are labeled or not. This study investigates the effects of four different SD response formats on response behavior. The results show that an unlabeled response scale without endpoint qualifiers yield significantly different results with respect to extreme response style (ERS), use of the mid-point category, and in mean index scores than response scales with full labels or the endpoint qualifiers “extremely” and “very.” Moreover, the results show that the precision of a fully labeled response scale is higher than for an unlabeled response scale. The recommendation is that marketing researchers use fully labeled response scales for SD items in future research.