Efficiently measuring recognition performance with sparse data

We examine methods for measuring performance in signal-detection-like tasks when each participant provides only a few observations. Monte Carlo simulations demonstrate that standard statistical techniques applied to ad’ analysis can lead to large numbers of Type I errors (incorrectly rejecting a hypothesis of no difference). Various statistical methods were compared in terms of their Type I and Type II error (incorrectly accepting a hypothesis of no difference) rates. Our conclusions are the same whether these two types of errors are weighted equally or Type I errors are weighted more heavily. The most promising method is to combine an aggregated’ measure with a percentile bootstrap confidence interval, a computerintensive nonparametric method of statistical inference. Researchers who prefer statistical techniques more commonly used in psychology, such as a repeated measurest test, should useγ (Goodman & Kruskal, 1954), since it performs slightly better than or nearly as well asd’. In general, when repeated measurest tests are used,γ is more conservative thand’: It makes more Type II errors, but its Type I error rate tends to be much closer to that of the traditional .05 α level. It is somewhat surprising thatγ performs as well as it does, given that the simulations that generated the hypothetical data conformed completely to thed’ model. Analyses in which H—FA was used had the highest Type I error rates. Detailed simulation results can be downloaded fromwww.psychonomic.org/archive/Schooler-BRM-2004.zip.     

Primes and flankers: Source confusions and discounting

Visual identification of briefly presented target words is affected by the presence of nondiagnostic prime words that immediately precede the target, flanker words simultaneously presented adjacent to the target, and visual masks that immediately follow the target in the same location. Priming is duration dependent: In a forced choice target identification task, brief primes produce a strong preference to choose the primed alternative, whereas long primes have the opposite effect. The ROUSE model (Huber, Shiffrin, Lyle, & Ruys, Psychological Review 108:149?C182, 2001) predicts this interaction by positing that prime features are confused with target features and that evidence regarding the prime features is discounted less for short primes and more for long primes, when both are compared with the optimal level. In the present study, we augmented the typical short-term priming experiment by adding flankers that appeared simultaneously with the target and remained for a short or long duration. In the experiment, we replicated previous priming effects and produced novel effects of flanker duration. ROUSE accounted for both the priming and flanker findings with the previously posited processes, but with different quantitative parameters for flankers: Relative to optimal levels of discounting, all flanker features were underdiscounted, but longer flankers were discounted more than short flankers.     

Priming in a free association task as a function of association directionality

Two experiments investigated priming in free association, a conceptual implicit memory task. The stimuli consisted of bidirectionally associated word pairs (e.g., BEACH-SAND) and unidirectionally associated word pairs that have no association from the target response back to the stimulus cue (e.g., BONE-DOG). In the study phase, target words (e.g., SAND, DOG) were presented in an incidental learning task. In the test phase, participants generated an associate to the stimulus cues (e.g., BEACH, BONE). In both experiments, priming was obtained for targets (e.g., SAND) that had an association back to the cue, but not for targets (e.g., DOG) for which such a backward association was absent. These results are problematic for theoretical accounts that attribute priming in free association to the strengthening of target responses. It is argued that priming in free association depends on the strengthening of cue-target associations.     

An expectancy model for memory search

A model for memory search is proposed in which S first forms an “expectancy” regarding the item to be tested on the next trial, then carries out a memory search. It is proposed that an expected item is encoded faster (or perhaps responded to more quickly), but the memory scanning process for expected and nonexpected items is otherwise identical. Assuming a serial exhaustive scanning process, we were able to fit much of the data in the literature. In addition, we tested the model by having S give his expectancy aloud before each trial. The data showed about a 100-msec advantage for expected items that did not interact with memory load. The model fit this data reasonably well.     

Attending to forty-nine spatial positions at once.

Simultaneous attention to 49 spatial positions resulted in the processing of threshold information from one of those positions essentially identical to the processing when the subject knew in advance that that position would be tested. This result held true when the task consisted of detection of the presence of a briefly presented dot. The same result held true for 9 spatial positions when the task consisted of report of the briefly presented letter in the target position.     

List-strength effect: I. Data and discussion

Extra items added to a list cause memory for the other items to decrease (the list-length effect). In one of the present studies we show that strengthening (or weakening) some items on a list harms (helps) free recall of the remaining list items. This is termed the list-strength effect. However, in seven recognition studies the list-strength effect was either absent or negative. This held whether strengthening was accomplished by extra study time or extra repetitions. The seven studies used various means to control rehearsal strategies, thereby providing evidence against the possibility that the findings were due to redistribution of rehearsal or effort from stronger to weaker items within a list. Current models appear unable to predict these results. We suggest that different retrieval operations underlie recall and recognition, as in the SAM model of Gillund and Shiffrin (1984), which can be made to fit the results with certain relatively minor modifications.     

List-strength effect: II. Theoretical mechanisms

Ratcliff, Clark, and Shiffrin (1990) examined the list-strength effect: the effect of strengthening (or weakening) some list items upon memory for other list items. The list-strength effect was missing or negative in recognition, missing or positive in cued recall, and large and positive in free recall. We show that a large number of current models fail to predict these findings. A variant of the SAM model of Gillund and Shiffrin (1984), involving a differentiation hypothesis, can handle the data. A variant of MINERVA 2 (Hintzman, 1986, 1988) comes close but has some problems. Successful variants of a variety of composite and network models were not found (e.g., Ackley, Hinton, & Sejnowski, 1985; Anderson, 1972, 1973; Metcalfe Eich, 1982; Murdock, 1982; Pike, 1984). The results suggest constraints on the future development of such models.