Abstract
The influences of the measurement scale on interpretations of age-related slowing are discussed. Although the customary practice is to describe performances as response latencies (sec/response), a plausible alternative is to describe performances as response speeds (responses/sec). Different scales may lead to contradictory conclusions because nonlinear transformations of time (such as speed) reduce or remove age x complexity interactions. Reconciliation is difficult because choice of measure appears more dependent on theoretical than psychometric considerations. Certain assumptions of cognitive theory require that untransformed response latencies serve as the dependent measure, whereas those of behavioral-learning approaches suggest scales that give increasing weight to latency reductions.