Researchers often assume a strong, linear relationship between regular- and reverse-keyed items, with responses on regular-keyed items (e.g., agree) perfectly mirroring those on reverse-keyed items (e.g., disagree). The current research challenges this received view and propounds a possible nonlinear relationship, partly due to the logical tendency of midlevel respondents to disagree with both types of items. In four examples (reported human height, job satisfaction, positive-negative affect, and self-esteem; total N = 50,544), a nonlinear model consistently explained additional item variance beyond a linear model. We further demonstrate that this relationship is moderated by item characteristics such as item extremity (job satisfaction) and item softening (self-esteem). Suboptimal modeling of the relationship may result in the apparent bidmensionality of a construct that characterizes regular- and reverse-keyed items as separate factors. User-friendly syntax for the examination of nonlinearity is provided to enhance the accessibility of the procedure.

研究人员通常假定常规键项和反向键项之间存在很强的线性关系，对常规键项（例如，同意）的响应完美地反映了对反向键项（例如，不同意）的响应。当前的研究挑战了这种公认的观点并提出了一种可能的非线性关系，部分原因是中层受访者不同意这两种类型的项目的逻辑倾向。在四个例子中（报告的人类身高、工作满意度、积极-消极影响和自尊；总N= 50,544），非线性模型始终如一地解释了线性模型之外的额外项目方差。我们进一步证明这种关系受项目特征的调节，例如项目极限（工作满意度）和项目软化（自尊）。关系的次优建模可能导致结构的明显双维性，该结构将常规和反向键控项目描述为单独的因素。提供了用于检查非线性的用户友好语法，以增强程序的可访问性。