LONG-TERM REPRESENTATIONS IN ABSOLUTE IDENTIFICATION
This paper reports empirical data collected to investigate response biases and practice effects in unidimensional absolute identification. Response biases are manifested by an average error (R(N)âˆ’S(N)) that is different from zero when considering all stimuli. Such biases result in shifts of all data points up or down on impulse graphics for sequential dependencies. Individual participant data analysis showed systematic response biases not observed when considering group data. This result provides an additional modeling constraint that might be incompatible with some current models of absolute identification. The data also show that performance in unidimensional absolute identification improves with practice. The results reported here contradict the generally admit postulate that unidimensional absolute identification do no yield stable long-term representations of the stimulus-set.