RETRIEVAL-PROCESSES UNDERLYING CONFIDENCE IN COMPREHENSION JUDGMENTS

People's ratings of confidence in ability to answer questions about previously studied expository passages are virtually uncorrelated with test performance. This correlation would be attenuated if confidence ratings were influenced by momentary accessibility (e.g., latency to initial access of probed information) and if test performance were influenced by other aspects of retrieval (e.g., total recall over an interval). To see if momentary accessibility influences confidence ratings, Experiment 1 manipulated accessibility by using self- versus other-generated title probes. Both confidence and accessibility measures varied with generation condition. Experiment 2 provided three findings. First, it assessed whether accessibility is the only common factor underlying confidence and domain familiarity ratings by observing whether confidence and familiarity covary with accessibility partialed out. Marginal covariation implied either some common factor other than accessibility, imperfect measurement of accessibility, or both. Second, negative correlations of rating latencies with rating magnitudes implied that access latency, momentary access rate, or both, affect the ratings. Third, calibration of comprehension was poor, but nonzero.