lnRT

lnRT is the natural logarithm of reaction time (RT), a common transformation used in cognitive psychology and related fields. RT data are typically positively skewed, with a long tail of slower responses. Applying the natural log to RT helps stabilize variance and bring the distribution closer to normal, which improves the fit and interpretability of many parametric statistical analyses, including linear and mixed-effects models.

Calculation and practice: RT is measured in units such as milliseconds or seconds. The standard lnRT is obtained by applying the natural logarithm to the raw RT values, i.e., ln(RT). Because RT cannot be negative and can be very small, some researchers add a small constant before transformation (for example, ln(RT + c)) to avoid undefined values or extreme instability when RTs are near zero. The choice of constant should be reported, as it affects interpretation.

Interpretation: When using lnRT as the outcome in a model, coefficients correspond to multiplicative changes in the raw RT. A one-unit change in a predictor that yields a coefficient beta implies that RT is multiplied by exp(beta). To interpret results on the original scale, researchers often back-transform model estimates or transform predicted lnRT values back to RT via exponentiation, with percent changes approximated by (exp(beta) − 1) × 100%.

Considerations: While lnRT can improve model assumptions, interpretation becomes less direct than for untransformed RT. Back-transformation is required to convey findings in practical units. Other approaches for RT data include modeling with ex-Gaussian or diffusion frameworks, which address skewness and latent processing dynamics in different ways.