THE REAL TEST BIAS

Abstract

ABSTRACT Efforts to hold schools accountable produce pressures to improve performance on the tests (usually multiple‐choice) used in school evaluation. Because of their greater economy, such tests have virtually driven out tests requiring constructed responses. If tests do not represent the broad spectrum of skills that ought to be taught, they produce bias against teaching what they do not measure. Studies show that format makes little difference for basic skills tests but does make a difference for complex tasks involving problem solving. Bias produced by conventional test formats appears to favor teaching the abilities associated with crystallized intelligence and to discourage teaching that might develop fluid intelligence. The greater cost of less efficient tests might be justified by using them in instruction to provide practice and feedback.