Answer sheet of volunteer 6 (group A), containing the sequence of plates with the Landolt Cs and the closed circles(plate 1 is blank) and the corresponding responses. Filled symbols indicate the answer "closed."
Graph of the noises and utterances during the examination of volunteer 12 (group A). The high-amplitude complexes represent turning over the plates; the subsequent low-amplitude complexes represent the verbal responses.
Graf MH, Roesen J. Ocular MalingeringA Surprising Visual Acuity Test. Arch Ophthalmol. 2002;120(6):756–760. doi:10.1001/archopht.120.6.756
To describe a visual acuity test for use in identifying psychogenic visual impairment and malingering.
The test contained 32 white plates with a black Landolt C printed in the center. The sequence of the 4 alternative directions of the C was not predictable. After plate 21, 4 circles were interspersed among the remaining plates. The test is administered at a distance from which the subject is presumed to be able to recognize the optotypes. He or she is requested to identify the direction of the C within 2 seconds, and the responses are noted. Two elements were evaluated: (1) The number of correct answers was compared with the corresponding value of the distribution function of the binomial formula that represents the probability of reaching this rate of correct answers by pure chance. (2) The response to the first circle (appearing after plate 21) was compared with the responses to the previous 19 Cs. The test was administered to 20 volunteer pseudomalingerers and 15 patients believed to be true malingerers.
Malingering was detected in 14 (74%) of 19 volunteers included in the evaluation and in 12 (80%) of 15 patients by too many or too few correct answers to the stimuli or by their response to one of the circles.
This test is useful in proving malingering. It may also provide evidence of a minimum visual recognition acuity within the gross dimension of the actual acuity.
DECREASED VISUAL acuity is one of the most common nonorganic complaints encountered in the practice of ophthalmology.1- 3 It may be psychogenic or the result of malingering. Most of the tests used in assessing this complaint can only roughly approximate the true visual recognition acuity.4- 6 Objective methods using eye-movement recordings7- 9 or visual evoked cortical potentials10 also only estimate acuity, as their accuracy is limited by patient cooperation and by the fact that response to the detection stimuli used is not necessarily equivalent to the recognition of a corresponding optotype. Many subjective tests are based on feints, deluding the subject by a change in the test distance and the detail size of the optotype. We present another visual acuity test method that is effective in proving presumptive malingering.
A statistical evaluation of answers to a large number of optotype presentations is a useful means to establish a test model,11 as a very high rate or a very low rate of correct responses proves the recognition of an optotype. For example, psychogenic superposition is proved if none of 32 Snellen Es are identified correctly, because the probability of this result by pure chance is low. (The probability of an incorrect answer is 3 out of 4 for each stimulus. The probability of 32 consecutive incorrect answers is(3/4)32 ≈.0001). The probability of a maximum number of correct responses (k) out of 32 optotypes with 4 possible alternative answers can be determined by the distribution function of the binomial formula: If k = 1, P ≈ .001; if k = 2, P ≈ .07; and if k = 3, P ≈ .03. If a subject yields an abnormally low rate of correct answers, malingering or a psychogenic disturbance is likely. Such a melding of psychology and statistics can be informative in these cases.
Miller5 described a young man, claiming to be completely blind in 1 eye, who was shown a series of photographs from a popular men's magazine that elicited no response. The young man then asked if he could see the set of photographs one more time because he thought he"just might have seen something out of the eye," suggesting malingering. Another of Miller's patients, who had been shown several phrases suggesting actions that one would think were anatomically impossible, suddenly began to laugh at a stimulus and admitted that he had been "got." Miller considered this test somewhat too confrontational to recommend. Nevertheless, evoking a telltale reaction to a surprising stimulus is a seminal idea. Optotypes are comparatively monotonous stimuli; therefore, we hypothesized that a surprise element might be an effective addition to a visual acuity test. In our model, Landolt Cs of an easily recognizable size were presented to the subject, and his or her response was noted. After a learning phase, a circle of the same size was shown. We compiled a test book and administered it to a group of volunteer pseudomalingerers (group A) and a group of malingerers (group B) who visited our department. The purpose of the study was to assess the subjects' reaction to the test, its possible applications, and the range of revelatory responses to the stimuli.
A ring binder containing 37 white plates (20 × 30 cm) composed the test book. Plate 1 was blank. Thirty-two plates each had a black Landolt C printed in the center. The gaps in the C were at the 12-o'clock, 3-o'clock, 6-o'clock, and 9-o'clock positions. Each direction occurred 8 times in an unpredictable sequence. (Copies of the test plates are available from the authors.) The diameter of the C was 14.5 mm, corresponding to a visual acuity of 0.5 if recognized at 5 m. Four plates (22, 27, 31, and 35) showed a closed circle of the same size and proportion as the C. Using the Landolt C avoids overestimating the recognition acuity, which can occur when the Snellen E, a periodic stimulus, is used.
The ring binder contents were presented at a distance from which the subject should easily recognize the Landolt C. At distances shorter than a half meter, persons 40 years and older were tested with adequate near vision correction. If the subject stated that he or she was not able to recognize the optotype at any distance, eg, 0.05 m, the test was discontinued. The distance was increased in decadal logarithmic steps. The feigned visual acuity was proportional to the farthest distance from which at least 3 of 4 optotypes were identified correctly. Because of the size of the optotypes, the acuity corresponded to one tenth of that distance in meters. Then it was explained to the subject that now a so-called threshold determination of visual acuity would be performed and the optotypes would be presented at a greater distance, so that it would be difficult or impossible to identify the gap in the C. Using a forced-choice strategy, the subject had to try to identify or otherwise guess the direction of the C. The maximum time for reflection was set at 2 seconds with regard to the large number of questions prescribed in the test protocol. The examiner explained this procedure to be legally bound, although it might seem rather strange to a layperson. A distance was chosen that corresponded to a visual acuity at least 3 log units (50%) less than the true visual acuity(group A or the presumed true visual acuity (group B). While the examiner(M.H.G.) noisily turned over plates 1 through 37, the observer (J.R.) noted the subject's responses and reactions on an answer sheet (Figure 1). In group A, utterances during the examination were recorded on a commercial automodulating digital system with a tie clip microphone. This sound documentation was not used in group B.
Informed consent was given by all subjects. In group A, 20 adult volunteers(9 male and 11 female medical students and hospital staff; age range, 21-53 years) with a visual acuity of at least 1.0 were tested to verify the applicability of the model and to observe possible telltale responses. Some of the subjects were acquainted with the statistical method of visual acuity assessment. They were asked to feign as credibly as possible a given reduction of binocular visual acuity, with a reward promised for success. Group B consisted of 15 patients who were suspect of malingering or functional visual loss (Table 1), who undertook the test as part of an expert examination, predominantly striving for a medical certificate of blindness. The major motive for malingering was presumed to be financial. Malingering was proved by other methods (Table 1) or by statistical evaluation of the results of the test itself. The test was performed monocularly or binocularly, depending on the individual. The test can be administered by one examiner, but the presence of a second observer is recommended.
The responses to the stimuli were analyzed with regard to 2 criteria:(1) The direction of the C as identified by the subject was compared with the actual direction. If 17 or more answers were correct, the subject passed the tested acuity level. Fewer than 3 correct answers or more than 16 opposite answers were judged to be evidence of intentionally false responses, because the probability of attaining these rates by chance is low.11,13(2) The reaction to plate 22 (the first circle) was compared with the reactions to plates 3 through 21. Plate 2 (the first C) was excluded, because subjects tended to be argumentative at the beginning of the test. Among group A, the answers were noted, and the delay between turning over the plate and the volunteer's answer was measured using a graphically visualized recording of the soundtrack(Figure 2) (Cool Edit 2000, version 1.1; Syntrillium Software, Phoenix, Ariz). The latency of response to the circle was considered suspicious if it was longer than any previous latency. Similarly, among group B, the verbal responses were recorded and the delays were evaluated, but no sound recording was done.
One male nurse was excluded, as he was unable to feign visual acuity. The remaining 19 volunteer pseudomalingerers feigned visual acuities ranging from 0.01 to 0.32 (median, 0.08). The test level exceeded the feigned acuity by 3 to 16 lines (median, 6 lines). Fourteen of the 19 volunteers were detected as malingerers. The test analysis factors were applied: (1) Nine volunteers failed the test because of too many or too few correct answers. Two volunteers gave more than 16 and 2 gave fewer than 3 correct responses. Four volunteers with 15 or 16 correct answers each had results that were inconsistent with the much lower feigned acuity, compared with the normal psychometric function. One volunteer answered with the opposite direction of the C on 16 plates.(2) Ten volunteers showed a marked reaction to the circle, either by laughing or giving an answer ("nothing" or "what's that?") that revealed they could differentiate the C from the circle. One young woman, for example, promptly answered "closed" when shown plate 22. She then repeated the same response to some of the following stimuli, trying to cover her mistake. Another volunteer raised his eyebrows and, before answering "top," hesitated for 4.0 seconds, significantly longer compared with the previous 19 plates, for which the latency ranged from 1.3 to 2.8 seconds. A suspicious latency, exceeding the maximum latency of the previous 19 plates, occurred with 6 volunteers.
The results of group B, consisting of 15 patients, are summarized in Table 1. The test detected 12 malingerers. Three patients were uncooperative. The following test evaluation findings were observed: (1) The cooperation of 6 patients (2, 5, 6, 10, 12, and 13) significantly improved during the test because of repeated forced-choice questioning and positive feedback. In 5 patients (2, 5, 6, 12, and 13), after first stating that they did not recognize the optotype at all, the percentage of correct answers exceeded 60%. Four patients (9, 10, 11, and 14) had suspiciously few(<3) correct responses. Patient 1 gave only 3 correct answers, a rate that would also be unlikely by pure chance. A suspiciously high number of responses in the opposite direction of the C occurred in patients 3 and 14. (The probability of 16 accidental responses in the opposite direction is .002.) (2) Six patients(2, 4, 6, 9, 10, and 12) showed a reaction to the first closed circle (eg, answering "closed"). One woman hesitated for 7 seconds when plate 22 was shown and another older woman took 25 seconds, while their previous latencies were much shorter. The time lags between the stimulus and the response were estimated by the observer by counting silently.
This convenient test identified a significant percentage of malingerers in both groups. A fundamentally different response between the 2 groups was laughter as a reaction to the gapless C, which occurred only among the pseudomalingerers. In true malingering, hesitation or discussion was the dominant reaction. Two limitations of the test include the necessity of obtaining informed consent to electronic recording of the answers, and the use of response latency as an objective finding. Compared with the responses to the previous 19 plates, there was about a 5% probability that the longest pause would occur at plate 22. This calculation is appropriate if the differences between the latencies are small. In the case of a much longer hesitation over plate 22 than over the previous plates, it is difficult to quantify a statistical value in nongaussian distribution of the latencies. Therefore, the verbal response to plate 22 will usually be the dominant factor used in the evaluation. For example, if at plate 22 the subject starts to comment but the 19 previous plates had evoked answers without any discussion, this behavior is more obvious than a P value of .05 would express. If the subject showed a similar response to one of the previous plates (except plate 2), however, repeated discussion would not necessarily be suspicious. Such statistical considerations may be important concerning forensic issues. A one-word answer such as "closed" is unequivocal, which would probably be the most common reaction (Table 1).
The other strategy used, namely, to provide the chance either to achieve a sufficient rate of correct answers or to lie to a statistically evident degree, detected 14 (74%) of the 19 pseudomalingerers in this series. Although many in group A were acquainted with the forced-choice procedure and knew that the rate of correct responses should be about 25%, the malingering in 9 people became statistically apparent. One volunteer who surpassed a 60% rate of correct answers in his attempt to avoid an abnormally steep slope of the psychometric function exceeded his tested visual acuity. Another volunteer attained a 55% rate at a level of 3 log units above his stated acuity. In 4 volunteers, their results of 15 or 16 correct answers each contrasted with their feigned acuity 4 to 16 log units below the tested level. One nurse subconsciously chose the opposite direction of the C in 16 of 32 answers. Among 6 (40%) of 15 patients in group B, cooperation significantly improved during the test, demonstrating the importance of a forced-choice strategy in obtaining results. In uncooperative subjects, repeated motivational and positive feedback are required. If a patient does not recognize numbers or letters, despite normal Snellen-E or Landolt-C acuity, the differential diagnosis of alexia should be considered.
Two subjects from group A and 4 from group B gave fewer than 3 correct answers and thus yielded evidence of malingering, because the chance of achieving such a low rate of correct answers by mere accident is slim. In the introduction, some values of the distribution function of the binomial formula were given, representing the statistical probability of achieving a certain number of correct answers out of 32 answers by pure chance (eg, with the eyes closed) if the chance of each single correct answer is 1 out of 4. A rate of 3 correct responses out of 32 would be unlikely (2.5%). If only 2 answers are correct, intentionally false statements may be suspected, because the probability of such a result occurring by pure chance is less than 1%.11,12 The absence of any correct response to 32 consecutive questions is presumed to be blatant malingering. In these cases, one could continue the test on another level, but sooner or later, the subject would detect the method. When a sufficiently high or an extremely low rate of correct responses is expected after the first 20 plates, the circles should not be presented but kept in reserve for future testing. Generally, performance of the test at least 3 lines below the presumed acuity is recommended, because the intention would become obvious if performed too near the actual visual threshold. When recognition of the optotype becomes difficult, the subject may notice that guessing can cause a correct answer. When the test is not applicable because the subject refuses to given any answer or monotonously gives the same anwer to every stimulus (as in patients 7, 8, and 15), other objective methods of evaluation are required.
These results demonstrate that statistical analysis of test responses is a useful addition to other critical and objective methods for diagnosing functional visual impairment and malingering.3- 5,7 Compared with other methods using resolution or detection stimuli, this convenient test offers the advantage of proving a discrete minimum visual recognition acuity within the gross dimension of the actual acuity.
Submitted for publication November 30, 2001; final revision received February 25, 2002; accepted February 28, 2002.
Corresponding author and reprints: Michael H. Graf, MD, Department of Ophthalmology, Strabismology, and Neuroophthalmology, University of Giessen, Friedrichstrasse 18, D-35385 Giessen, Germany (e-mail: firstname.lastname@example.org).