Misinterpretation of conditional statements in Wason's selection task.

Errors may be made on Wason's selection task because either (a) the rule to be tested is misunderstood, or (b) reasoning from that rule is inaccurate, or both. We report two experiments using the experimental paradigm introduced by Gebauer and Laming in which subjects are given six problems in succession. We use the subset of cards selected by each subject as (a) an indication of how the rule is understood and, when that selection is consistent throughout all six problems (so that we can infer a consistent understanding of the rule), as (b) a basis for evaluating the accuracy of the subject's reasoning according to three independent criteria. Experiment 1 adds an exactly parallel contextual version of the task to permit comparison between performances (by the same subjects) on the two versions. Experiment 2 repeats Exp. 1, but with negatives inserted in the conditional rule. Most subjects make a consistent selection of cards throughout all six problems, but typically appear to misunderstand the rule. This is so in both abstract and contextual tasks and replicates the finding by Gebauer and Laming. Most misunderstandings consisted of either (a) reading the simple conditional rule as a bi-conditional or (b) substituting "top/underneath" for "one side/other side". In Exp. 1 subjects seldom misevaluated the rule they appeared to be testing, but such "errors" of evaluation were common in Exp. 2. Negatives confuse the subjects and should not be used in any conditional application that matters. In Exp. 2 (but not 1) there was a significant correlation between interpretations of the two tasks. We provide an explanation of "matching bias" (it results from the confluence of the two common misunderstandings above) and comment on "mental models" which are, at present, unable to accommodate the variety of results we present here. We also relate our experimental paradigm to the conditional inference task and to truth tables.

Improving the detection of cancer in the screening of mammograms.

OBJECTIVE: Examination of figures from screening at several different centres, and from the incidence of interval cancers in those programmes, suggests that a single radiologist reading by him/herself would miss 16 to 31% of cancers detectable in the mammograms. This study investigates how the accuracy of breast cancer screening could be improved. METHOD: Draw a comparison between screening for breast cancer and other inspection tasks that have been studied in the experimental psychological laboratory to suggest two low cost fixes that might improve the accuracy of breast cancer screening. RECOMMENDATIONS: (1) Radiologists should not work for more than half an hour at a time; (2) two radiologists double reading should inspect the same series of mammograms in inverse order; (3) there should be a clerical check to see whether the efficiency of detection decreases with time on task; and (4) a more sophisticated engineering of the screening procedure might be accomplished by assembling a library of mammograms for which the correct diagnosis is known. If these library mammograms were mixed randomly with those to be inspected in some suitable ratio, it would then be possible to provide the radiologist with immediate feedback on the accuracy of (some of) his/her diagnoses. From experimental studies of inspection tasks this should improve diagnostic accuracy. These recommendations, of course, require evaluation in field trials.

Screening cervical smears.

A plausible explanation is offered why an experienced pathologist engaged in the screening of cervical smears should have issued a large number of false-negative diagnoses. The explanation centres on the interactions which occur between successive judgements when a sequence of similar stimuli are inspected. Briefly, those interactions have the character of an assimilation of each judgement to its predecessor, and the resultant process is capable of 'running away'. A procedure is suggested, involving the provision of immediate knowledge of results for a small proportion of the smears inspected, which would prevent such systematic misdiagnoses from recurring in the future.

The human element in medical screening.

An explanation is offered for the reasons why an experienced pathologist engaged in the screening of cervical smears should have issued a large number of false negative diagnoses. The explanation centres on the interactions which occur between successive judgments when a long sequence of similar stimuli are inspected. Briefly, those interactions have the character of an assimilation of each judgment to its predecessor, and the resultant process is capable of "running away". A procedure is suggested, in which the provision of immediate knowledge of results for a small proportion of the smears inspected would prevent such systematic misdiagnoses from recurring in the future. The implications of this natural experiment for medical screening in general are spelt out.

Springer's lines and Hermann's grid.

Hermann's grid is a complete matrix of black squares; Springer's figure has alternate squares whited out. Both figures generate illusory elements, bright and dark lines in the white spaces, with different elements seen from different distances. The visibility of these different elements is shown to be related to the detectability of sinusoidal gratings of wavenumbers corresponding to certain critical components of the Fourier decomposition of the figure. Evidence from the detection of gratings shows that near- and sub- (but not supra-) threshold contrasts are subject to a fourth power compression which shows, in the present figures, as a defect in the percept, equivalent to the deletion of (a part of) the corresponding Fourier components. There are some residual problems in understanding the perception of Hermann's grid.

F. Hegelmaier: on memory for the length of a line.

A very early student project undertaken by Friedrich Hegelmaier (1833-1906), published in German in 1852, is republished in English translation. Slight though the experimental work is, it nevertheless occupies a unique place in the history of experimental psychology. It is the source whence Fechner had the method of constant stimuli, a method that continued in use as the preferred psychophysical method, substantially in the form described here, for more than a century. The experiment is arguably the first experiment in the modern sense of a systematic preplanned body of observations and has the glaring faults that one would expect in a very first experiment. Finally, Hegelmaier suggests the use of two simultaneous tasks as a means to investigate human performance, a full hundred years before that idea was realized in practice. If only he had continued in experimental psychology!