An Eye Tracking Study on Feigned Schizophrenia

The literature on symptom validity assessment has grown significantly in the past 20 years (see statements from the National Association of Neuropsychology (NAN) in 2005 and the American Academy of Clinical Neuropsychology (AACN) in 2007). NAN has indicated that the assessment of symptom validity is an essential part of a neuropsychological examination (Bush et al., 2005). The AACN specified that objective measures of symptoms exaggeration should be used during cognitive testing, and that omitting such tests necessitates solid justification (Heilbronner et al., 2009).

Assessing the validity of test results is important because failure to identify invalid results can lead to incorrect diagnoses and treatment in clinical settings, inaccurate conclusions in research settings, and wrongful outcomes in forensic settings, all of which may waste time and resources. Moreover, when failing to assess the validity of test performance happens in a forensic context, it may be dangerous for the client or others. Failure to spot an invalid test may have harmful consequences for further medical and legal decisions, including decisions about treatment, such as medication administration, participation in certain activities, and decisions about competence to stand trial or conditional release (Segal & Buress, 2006).

Two forms of validity assessment are performance validity and symptom validity (Larrabee, 2012). Performance validity refers to the validity of the actual ability to perform a task, assessed either by standalone tests or by atypical performance on neuropsychological tests. Performance Validity Tests (PVTs) assess the credibility of performance on cognitive tasks. Symptom validity instead refers to the validity of symptomatic complaints on self-report measures. Symptom Validity Tests (SVTs) are self-reported validity indicators of psychiatric manifestations, and assess patients’ distorted, unusual, or inconsistent endorsement of symptoms. While symptom validity and performance validity appear to be independent, they are not mutually exclusive (McCaffrey et al., 2003; Van Dyke et al., 2013), and including both SVTs and PVTs in the same assessment presumably improves detection of invalid results (Boone, 2013; Fox & Vincent, 2020; Giromini et al., 2020). In fact, Gottfried and Glassmire (2016) examined the relationship between psychiatric and cognitive feigning strategies using both a SVT (i.e., the Structured Interview of Reported Symptoms, SIRS; Rogers et al., 1992) and a PVT (i.e., the Test of Memory Malingering, TOMM; Tombaugh, 1996) among 150 forensic psychiatric inpatients adjudicated incompetent to stand trial. Their results show that different strategies are used when psychiatric and cognitive symptoms are feigned. Therefore, the authors suggested that clinicians should screen for specific types of feigning and evaluate both feigning strategies in forensic assessment. SVTs and PVTs are often used for the evaluation of an individual patient to determine if the clinical presentation of that patient is valid, and forensically, to address the issue of malingering – which is defined as the “intentional production of false or grossly exaggerated physical or psychological symptoms, motivated by external incentives” (American Psychiatric Association, 2013, p. 726). The primary reasons why people engage in feigning include to obtain financial compensation, drugs, or medical care; to avoid work or military service, and to evade criminal prosecution or reduce criminal culpability (Rogers & Bender, 2018). While several studies have reported on the utility of using PVTs in neuropsychological evaluations (Bush et al., 2005; Heilbronner et al., 2009), less research has assessed the utility of including performance tasks in forensic psychiatric evaluations, although tools such as the TOMM, the Inventory of Problems-29 (IOP-29; Viglione et al., 2017; Viglione & Giromini, 2020), and the recently introduced memory module of IOP-29 (i.e., the IOP-M; Giromini et al., 2020) have been studied with forensic populations (Delain et al., 2003; Gierok et al., 2005; Weinborn et al., 2010; Fazio et al., 2015; Fazio et al., 2016; Viglione et al., 2017; Roma et al., 2019) and/or simulation studies focused on feigned schizophrenia (e.g., Banovic et al., 2021; Šömen et al., 2021).

Malingered psychosis is a particularly relevant issue in this context. Pierre et al. (2004) observed that almost 30% of inmates seen in psychiatric services of a large American prison reported having malingered psychotic symptoms in order to obtain psychotropic drugs. Furthermore, appearing to suffer from psychosis can be especially appealing to defendants charged with serious crimes, as evidence of mental illness can result in mitigated sentencing, including avoiding capital punishment (Resnick, 1999). In line with the hypothesis that PVTs may be useful to assess the credibility of presented psychiatric complaints, Green et al. (2012) reported that performance measures are able to increase detection of suspected feigning of psychotic symptoms over SVTs alone. Accordingly, Green et al. (2012) suggested that PVTs should become a standard of practice in assessment situations where there are prominent secondary gain issues. Along similar lines, Fox and Vincent (2020) stated that PVTs should always be used also in the assessment of the credibility of PTSD related complaints. Similar considerations were recently reported by Giromini et al. (2020).

Eye Tracking as Performance Validity Measure

Eye tracking is a promising approach to enhance assessment of performance validity (Barry & Ettenhofer, 2016). Research has established that ocular behaviors are reliable biomarkers of both conscious and unconscious cognitive processes, and they have distinguished deceptive responding from honest responding in experimental paradigms. Eye-tracking research has documented that oculomotor behaviors provide a quantitative measure of real-time attention (Duchowski, 2002). Also, it is well-established that eye tracking can provide valuable insights into cognitive processes in both healthy adults and clinical populations in a covert manner (Domagalik et al., 2012). Indeed, eye tracking technology has been effectively applied to deception detection (Proudfoot et al., 2016; Van Hooft & Born, 2012). Previous research has addressed eye behavior in the context of lying and deceiving (DePaulo et al., 2003), and investigated the effect that faking and response latencies have on eye behavior. For example, Van Hooft and Born (2012) explored whether eye tracking technology may yield information that can be used for identifying faking behavior. Their findings demonstrated that eye behavior can increase our understanding of the response processes when faking (vs answering honestly), and thus, it may be potentially useful in detecting faking behavior, as in their study it improved detecting rates over and beyond response latency measures. Many other studies focused on how deceivers exhibit different eye-gaze fixation patterns (Schwedes & Wentura, 2012) and pupil dilation variations (Bradley et al., 2008; Cook et al., 2012; Webb et al., 2010) relative to truth-tellers. Also, Tomer et al. (2020) integrated eye tracking technology with the Word Memory Test (WMT; Green, 2003) in order to test its usefulness for detecting feigned cognitive impairment. This was found to measure cognitive processes and showed initial promises for the detection of feigned cognitive impairment. Similarly, Kanser et al. (2020) examined the incremental utility of eye tracking on a clinical PVT in distinguishing adults diagnosed with TBI from adults coached to feign cognitive impairment. Using eye tracking, the authors found that participants feigning TBI showed multiple signs of greater cognitive effort than both TBI patients and healthy comparisons, and they concluded that eye tracking may be an important complement to traditional accuracy scores on PVTs.

Extensive research has highlighted the differences in performance on eye movement tasks in neurological and psychiatric disorders such as Parkinson’s disease (MacAskill et al., 2012), Alzheimer’s disease (Crawford et al., 2013), and schizophrenia (Broerse et al., 2001). Inspection of eye movements might thus be particularly useful in the assessment of bona fide versus feigned schizophrenia. A wide body of research has demonstrated that patients affected by schizophrenia differ from non-clinical controls on smooth pursuit as well as anti-saccade eye tracking tasks (Hutton & Ettinger, 2006)Footnote 1. More specifically, on the smooth pursuit task, participants with schizophrenia often show lower velocity gain (i.e., the ratio of peak velocity to target velocity) on the task and have increased difficulty accurately following the target compared to healthy controls. O’Driscoll and Callahan (2008) reported robust meta-analytic findings on the differences between the pursuit quality of patients with schizophrenia and that of non-psychiatric controls. Effect sizes for specific measures (e.g., maintenance gain) of pursuit were large (d = .87). For the anti-saccade task, participants with schizophrenia often display a significantly higher number of inhibitory errors on the task compared to controls and some studies have found increased latencies (Hutton & Ettinger, 2006). Using a very large sample from the Consortium on the Genetics of Schizophrenia (COGS), Radant and colleagues (2010) demonstrated that the anti-saccade performance of schizophrenia participants was impaired compared to both a Community Comparison Sample (CCS) and unaffected first-degree relatives (respectively, d = 1.18 for the patients with schizophrenia – CCS difference, and d = 1.10 for the patients with schizophrenia – relatives difference). However, to date, no published study has investigated whether or not it is possible for healthy volunteers to feign the oculomotor impairments of people with schizophrenia.

The Current Study

The current study sought to investigate whether experimental participants instructed to feign engaged in smooth pursuit and anti-saccade tasks would be able to reproduce oculomotor patterns typical of patients with schizophrenia when instructed to do so. More broadly, we aimed to investigate the extent to which eye tracking technology could improve our understanding of the examinee’s oculomotor performance when feigning schizophrenia. We designed an experimental feigning study, in which three eye tracking tasks were administered to 83 individuals. Half of the participants (n = 43) were asked to try to feign schizophrenia, the other half (n = 40) were asked to complete all tasks following standard instructions.

All participants engaged in a smooth pursuit task, a pro-saccade task (that served as a baseline measure), and an anti-saccade task using the standard procedures in our lab (Crawford et al., 2019; Wilcockson et al., 2019). Control participants were asked to do their best to perform as instructed on all three tasks. Participants instructed to feign were informed that patients with schizophrenia typically find it difficult to complete these tasks and thus they were asked to do their best to reproduce impaired eye movement patterns shown by individuals affected by schizophrenia.

Due to the automatic nature of eye movements on these tasks (Over et al., 2006), we predicted that the abnormal characteristics displayed by participants with schizophrenia would be difficult for the “Feigners” to replicate accurately. More specifically, we predicted there might be performance differences between Controls and “Feigners”, however these would not be comparable to performance characteristics of real participants with schizophrenia, as reported in the literature. If participants instructed to feign behave in a similar manner to “real” patients affected by schizophrenia then we would expect them to show same pattern of findings: relatively preserved eye-movements in the pro-saccade task, higher RMSE and lower maintenance gain in smooth pursuit; longer anti-saccade mean latencies and a lower proportion of correct anti-saccades. We hypothesized that participants instructed to feign would deviate from this pattern. If this hypothesis was supported, then we would have behavioral evidence that participants instructed to feign might be unable simulate the eye-movement effects that have been reported in people with schizophrenia.

It should be noted that this research project also included the administration of a SVT, i.e., the IOP-29 (Viglione et al., 2017; Viglione & Giromini, 2020), to all participants. However, the current article only focuses on eye tracking variables so that any information concerning that portion of the project will be reported elsewhere. As for the smooth pursuit, we took into consideration two of the most widely used parameters in the literature and we hypothesized that participants instructed to feign would show a lower maintenance gain and a higher RMSE, compared to control participants. As for the anti-saccade task, we expected lower mean latency and a lower proportion of correct trials in the “feigners” group, in comparison to our honest-responding group. We did not hypothesize any difference in the pro-saccade task, as it represented our baseline measure.

Method

Participants

The sample was comprised of 83 native English-speaking participants (64 women) ranging in age from 18 to 57 years old (M = 23.35; SD = 6.84). All participants were recruited in northern England via a university’s online recruitment portal and snowball sampling on that university’s campus. Participants in the two groups did not differ in age, t(57) = 1.26, p = 0.20, and gender, χ2 (1, N = 83) = .007, p = .93.

Participation eligibility criteria required that participants (a) were not currently affected by any neurological or psychological disease, (b) had no history of psychiatric disorders, (c) were not taking any psychoactive or psychotropic drugs, and (d) did not have any specific eye disorder such as retinal degeneration, diplopia, macular edema, strabismus, optic neuropathy, palsy of cranial nerves III, IV, or VI. Before beginning the experimental procedure, participants were also asked if they had disease that might affect the visual system (e.g., myasthenia gravis, sarcoidosis, demyelinating disease). Moreover, at the beginning of the experiment – and before each task – the calibration and validation procedures were carried out (which, in most cases, cannot be completed if the participant has problems with the visual system). Individuals with myopia or astigmatism were allowed to participate because effects of these conditions can be controlled with the eye tracker device.

In the next sections, should readers encounter jargon with which they are unfamiliar, we provide a brief jargon-free summary (see Appendix A) at the end of the article, along with appendices on the vignettes used in the experiment (see Appendix B and C).

Materials and Measures

Eye Tracking Measurement

Eye movements were recorded using the EyeLink Desktop 1000 with a sampling rate of 500 Hz. A chinrest was used to minimise head movements. The experiment was conducted in a monocular mode. The camera-eye was located 40 cm from the participant’s eye. Prior to the start of each task, the eye tracker was calibrated and validated using a 9-point calibration. The Eyelink software reported suitable gaze accuracy, within 0.25°-0.50° of visual angle. The stimuli were created and controlled via Experiment Builder Software version 1.10.1630. The settings used are those recommended for a proper use of the device.

Smooth Pursuit Task

The smooth pursuit task is a simple eye movement task in which participants follow a target with their eyes as it moves smoothly backward and forward on the horizontal plane. Important measurements include velocity gain and the number of saccades (see Appendix A) that occur during pursuit. Prior to the start of the trial, participants were presented with a fixation point on the left-hand side of the screen. Participants were then instructed to follow the red target as smoothly as they could with their eyes. The target moved horizontally from left to right two times.

Pro-saccade Task

We included the pro-saccade task as a baseline task to ensure there were no anomalies concerning performance of the two groups at this level of control. Therefore, no participant was asked to feign schizophrenia in the pro-saccade task. Participants completed 32 trials including four practice trials and were required to fixate on a white target in the centre of the screen. Then, a green dot appeared to the left or right of the central fixation at 4° for 2000 milliseconds (ms), and, this time, participants were asked to shift their visual attention to the location of the green stimulus. The display parameters and timings for the pro-saccade task were the same as for the anti-saccade task. Participants in pro-saccade tasks typically show very fast saccadic reaction times (< 200 ms) and made virtually no direction errors (Pratt & Trottier, 2005).

Anti-saccade Task

The anti-saccade task is an important variant of the pro-saccade task. Subjects are required to suppress a reflexive saccade toward the sudden appearance of visual stimulus (pro-saccade) and generate a saccade with the same amplitude as the target eccentricity but in the opposite direction (anti-saccade). The task is aimed at investigating the voluntary and flexible control of movement. Two processes are needed for the anti-saccade task: goal activation and inhibitory control (i.e., suppression of the automatic pro-saccade).

For this task, participants completed 32 trials, including four practice trials. Participants were first presented with a white central fixation target for 750-1000 ms. Time intervals between target presentations were randomized in order to avoid anticipatory responses. The fixation point was then removed and followed by a green target presented randomly to the left or right side of the central fixation at 4° for 2000 ms. A blank interval screen was displayed for 200 ms between the initial appearance of the green target and the extinguishment of the white fixation target producing a 2000 ms gap in presentations. Participants were instructed to shift their visual attention in the opposite direction of the green target stimulus. For example, if the stimulus was presented to the left of the white fixation, a successful trial would have required the participant to have looked towards the right. Failure to inhibit a reflexive saccade is considered an error (Levy et al. 2004). A blank interval screen was displayed for 3500 ms between trials.

Procedure

The study was approved by the Lancaster University Faculty of Science Ethics committee. Prospective participants were first met in a quiet room to ensure study eligibility and to provide written consent. The Feigning group was asked to try to feign schizophrenia during two out of three eye-movement tasks (i.e., smooth pursuit and anti-saccade tasks). Participants in this group received a vignette (see Appendix B) explaining schizophrenia symptoms and asking them to imagine themselves in a situation in which they would wish to feign schizophrenia. Then, they were asked to perform in a manner that would convince the examiner that they had schizophrenia (that is, without over-exaggerating or appearing as a feigner). Participants were informed that if they could produce test results that look like those of a true patient affected by schizophrenia, they would be entered in a lottery to win £25.

The Control group was asked to read and memorize a vignette (see Appendix C) about somebody else feigning schizophrenia. The vignette was presented to them as a memory test, to ensure they were committed to genuine participation. Next, they were instructed to complete all eye-movement tasks honestly, following standard instructions. As well as the “feigners”, the Control group was informed that they would be entered into a lottery to win £25 for completing all tasks.

The three eye movement tasks were presented in the same order, (i.e., smooth pursuit, pro-saccade, and anti-saccade). Each participant was paid £5 upon completion of the experimental procedure. Participants also completed SVT, the Inventory of Problems – 29 (IOP-29; Viglione et al., 2017; Viglione & Giromini, 2020), before these tasks. However, the results of this task are outside the scope of our hypotheses and are not reported here (please, contact the corresponding author for further information).

Data Analysis

The eye tracking data from the smooth pursuit task was extracted using R Studio [version 3.5.2 (2018-12-20)] before it was analysed using SPSS. R was used to calculate the mean velocity gains, the Root Mean Square Error (RMSE)Footnote 2 and the saccade lengthsFootnote 3.

The R script identified samples that were fixations (e.g., not saccades, but likely pursuit, see Appendix A). We then computed the average velocity of the target and eye movements for those samples and divided target velocity by eye movement velocity to get the velocity gain. This calculation excluded samples where the target velocity was less than two degrees; this is due to them being indicative of delayed and slowed saccades. Next, we computed the distance between the target and eye in degrees based on the visual angle for each sample and calculated the root of the square of the average of those distances (RMSE). If the eye velocity matched the target velocity exactly, the smooth pursuit would have a velocity gain value of 1, and a RMSE value of 0. Four participants’ data were removed from the smooth pursuit analysis due to poor calibration and incomplete eye tracking data. Poor calibration eye tracking data was classified in cases where the eye tracker repeatedly or for a long period lost the pupil during recording and therefore lead to gaps and saccade jumps during the smooth pursuit task due to the eye tracker failing to accurately record the participants saccades. Three participants were removed from the feigning group and one from the control group.

For the pro-saccade and anti-saccade tasks, first, the eye tracking data was examined. Anticipatory saccades with latencies less than 80 ms initiated prior to the presentation of the stimuli were excluded from analyses. Delayed saccades made after 500 ms were excluded due to inattention. Saccades with an amplitude less than 2° were excluded due to the indication of an anticipatory saccade. After filtering the data from the 28 trials, the average number of trials included for the pro-saccade and anti-saccade task was 25.12 (“Feigners”: 23.09, Controls: 27.40) and 25.01 (“Feigners”: 24.02, Controls: 26.08), retrospectively.

We then conducted a series of t-tests to compare results based on participant group (“Feigners” vs. Controls). The dependent variables investigated were mean latency, latency standard deviation, amplitude, peak velocity, number of successful trials, and error percentage (for a definition of these parameters, see Appendix A). For the anti-saccade task, corrected error percentage and corrected error latencies were also investigated. A corrected error was determined when a participant made an error by looking towards the target, but then corrected this error by shifting visual gaze to the correct location.

Results

For the smooth pursuit task, Table 1 displays the means and SDs for VGain and RMSE for Controls and “Feigners”. Results showed that Controls displayed higher VGain than “Feigners”, t(59.4) = 3.3, p = .002, d = 0.71. No significant differences were found between groups on the RMSE variable, t(41) = 1.5, p = .14, d = 0.32.

Table 1 Smooth Pursuit Performance for Control and Feigning Groups
Full size table

Table 2 shows the pro- and anti-saccade average latency times and proportion of correct trials observed for Controls and “Feigners”. All differences were statistically significant with relatively large effect sizes (.70 ≤ d ≤ 1.14). In line with our instructions, participants instructed to feign committed more errors than Controls on the anti-saccade task (d = 1.14). However, unexpectedly and in contrast with our instructions, participants instructed to feign also committed more errors than Controls on the pro-saccade task (d = .78) (Figure 1). Lastly, participants instructed to feign also took longer than Controls performing both the pro-saccade (d = .70) and anti-saccade (d = .72) tasks.

Table 2 Pro- and Anti-Saccade Performance for Control and Feigning Groups.
Full size table
Fig. 1
figure 1

Representation of Proportion of Correct Trials on Pro-Saccade Tasks by Group

Full size image

Additional Analyses

Because participants instructed to feign significantly differed from Controls on the VGain of the smooth pursuit task and on both anti-saccade tasks related variables (i.e., Proportion of Correct and Mean Latency), we then compared these findings against previously published data observed in patients with schizophrenia. This was done to further investigate the extent to which our participants instructed to feign could accurately reproduce the oculomotor performance typically observed in individuals genuinely affected by schizophrenia.

With regard to the smooth pursuit tasks, O’Driscoll and Callahan’s (2008) meta-analysis found that patients with schizophrenia differed from non-psychiatric controls on VGain with a mean Cohen’s d of .87 (95% CI = .74 to .99). The effect size between Controls and “Feigners” found in the current study was d = .71, outside of O’Driscoll and Callahan’s (2008) estimated 95% CI range. Participants instructed to feign in our study reduced their VGain less than patients with schizophrenia typically do.

For the anti-saccade tasks, we compared our findings to those published in Radant et al.’s (2010) study on 219 patients affected by schizophrenia. The proportion of correct trials by our “Feigners” (M = .53, SD = .28) did not significantly differ from those of their patients with schizophrenia (M = .60, SD = .26), t(260) = 1.59, p = .11, d = .27. Results from Radant et al. (2010) on latency used 60 trials, whereas our study used 26 trials, so the results from the two data sets should be compared with caution. Nevertheless, in Radant et al. (2010), patients with schizophrenia showed a notably longer latency to correct (M = 425 ms; SD = 99) compared to healthy controls (M = 392 ms; SD = 70), with a considerable effect (d = .41). In our study, participants instructed to feign also showed a notably longer latency to correct, compared to healthy Controls, but the effect size is greater (d = .70). That is, our participants instructed to feign took relatively more time than bona fide schizophrenia patients to complete the anti-saccade task. The difference in the effect size found in our study versus in Radant et al. (2010) research, however, is only marginally significant, z = 1.74, p = .08.

Discussion

Before the onset of specific PVT studies in cognitive assessments and forensic evaluations, test performance validity was always a concern and, in recent years, it has become a critical issue in both neurocognitive and neurobehavioral assessments. Up until a couple of decades ago, the validity of a given performance test was usually determined by clinical judgment and on observed test behaviors. In an attempt to bring greater objectivity in reporting test performance validity in cognitive assessments, specific performance validity measures began to be routinely applied during neuropsychological assessment batteries (Carone & Bush, 2012).

In particular, eye-tracking technology could be not only a feasible, but also beneficial, addition to PVTs. Tomer and colleagues’ (2020) found that measuring eye movements can successfully differentiate simulators from honest controls. These findings laid the groundwork for the use of eye-tracking devices as a novel technology able to help in deception detection. Their results pointed to strategies used by simulators and cognitive processes that underlie the feigning of cognitive impairments. More recently, Kanser et al. (2020) demonstrated that measuring eye behavior may enhance assessment of performance validity. Relying on previous literature on eye movements as reliable biomarkers of unconscious cognitive processes (Barry & Ettenhofer, 2016; Cook et al., 2012; Griffin & Oppenheimer, 2006; Kanser et al., 2017; Vrij et al., 2011), the authors examined the incremental utility of eye behavior on PVTs in discriminating between people who suffered Traumatic Brain Injury (TBI) and people coached to feign cognitive impairment. Their findings, consistent with previous literature, support the idea that oculomotor behavior improves deception detection, distinguishing deceptive from honest responding in experimental paradigms.

In the current study, we aimed to investigate whether experimental participants instructed to feign engaged in smooth pursuit and anti-saccade tasks would be able to reproduce oculomotor patterns typical of patients with schizophrenia, if instructed to do so. We compared the performance of our feigning group with archival data from O’Driscoll and Callahan meta-analysis (2008) and from Radant et al.’s study (2010). The former was chosen because it represents the most recent meta-analysis on pursuit type of movements, the latter because, to our knowledge, it is the study with the largest schizophrenic cohort.

O’Driscoll and Callahan (2008) used meta-analytic procedures to quantify patient-control differences in eye tracking. Specifically, they focused on pursuit type of movements. Calkins et al. (2008) reported that smooth pursuit dysfunction is considered one of the most promising candidate endophenotype of schizophrenia (Erlenmeyer-Kimling & Cornblatt, 1987; Holzman, 1987; Iacono, 1983, 1988, 1998; Iacono & Grove, 1993; Lee & Williams, 2000; Siever & Coursey, 1985; Siever et al., 1982; Venables, 1991). In particular, RMSE appears to be a meaningful measure of eye tracking performance that seems to be unable to be voluntarily replicated by participants instructed to feign. With regard to anti-saccade movements, studies (Ettinger et al., 2006; Stefanis et al., 2007) show that they satisfy most criteria put forward for endophenotypes by Gottesman and Gould (2003). Several studies have found that patients affected by schizophrenia perform significantly worse on the anti-saccade task than non-clinical controls, and this result has been replicated in existing published studies (e.g., Fukushima et al., 1988; Crawford et al., 1995; Crawford et al., 1998; Hutton et al., 1998).

Taken together, the results of our study demonstrate that although “Feigners” differed from Controls on both the smooth pursuit and pro- and anti-saccade tasks, they could not accurately reproduce eye movements typically observed in bona fide psychiatric patients affected by schizophrenia. In the smooth pursuit tasks, there were no differences in RMSE between “Feigners” and Controls, and participants instructed to feign reduced their VGain less than patients with schizophrenia typically do. In the anti-saccade tasks, our participants who were instructed to feign made more errors than Controls and their performance indeed resembled, to some extent, that of patients with schizophrenia. Participants instructed to feign, however, overall took more time than bona fide patients with schizophrenia to complete the anti-saccade tasks, giving a strong cue to their feigning status. This finding might be explained by the increased cognitive load experienced by the Feigning group as opposed to the Control group, as the amount of time spent completing a task tends to increase with increased cognitive demand (Baker et al., 1992). Alternatively, it is possible our participants instructed to feign might have believed that patients affected by schizophrenia would perform the task slowly, and thus, they may have intentionally performed the task slowly. Regardless of the case, the findings suggests that reaction times have the potential to contribute to differentiating valid versus invalid schizophrenia presentations. Perhaps more importantly, contrary to the instructions, our participants instructed to feign also committed more errors than Controls on the pro-saccade task, which patients with schizophrenia do not normally do. In fact, while almost all Controls made no errors, some of our participants instructed to feign made a notable number of errors also on these tasks, even though they were instructed to not feign any eye-tracking deficit. Interestingly, participants instructed to feign also took longer than Controls performing the pro-saccade task. These results might be due to a generalization effect. Despite not being informed on pro-saccade performance of patients with schizophrenia, participants might still assume that patients are impaired on this task. It may also be the case that deliberate feigning on one task interferes with participants’ ability to complete other tasks correctly. Future research could try to determine the causes for this reduced performance by “feigners” on the pro-saccade task. It would be interesting, and clinically relevant, to find out if people who (are instructed to or willing to) feign get carried away in their attempts and exaggerate on proximal or unrelated measures. Nonetheless, this finding substantiates the rationale on which many feigning detection tests are based – that is, exaggerated effort or performance points to an increased likelihood of an invalid presentation.

All eye-tracking variables showed notably higher variability within the feigner sample than within the nonclinical controls. This suggests that different “feigners” may be using different strategies to feign: some are choosing to show the effects mentioned in their instructions strongly and other only weakly. This may be a deliberate choice or a consequence of lack of knowledge on how to feign most convincingly. From an applied perspective, one might anticipate that different people instructed to feign will behave differently from each other on eye-tracking tasks, which makes it harder to identify an eye-movement behavior “prototypical” of schizophrenia “feigners”.

When considering the implications of these findings, however, it should also be noted that our study had some important limitations. First, our sample did not include a direct comparison with bona fide schizophrenia patients. Although we made an effort to compare our findings against findings in previously published literature, future studies should include clinical comparison samples to evaluate the ability level of “feigners” in fabricating and/or exaggerating psychiatric symptoms. Second, the external validity of our study may be limited, given the fact that our experimental participants were instructed to feign, rather to generate spontaneous or natural behavior. Although they were engaged in the tasks in many ways (e.g., through financial incentives, relevant vignette scenario), they may differ from real-life “feigners”. For instance, in high-stakes evaluations, the time spent on eye-tracking tasks may be influenced by other factors like preoccupation, anxiety, and such. Finally, our experimental participants instructed to feign had a relatively high education level, which may limit the generalizability of our findings, given that real-life patients with schizophrenia are usually less academically accomplished compared to their peer group (Frissen et al., 2015). Similarly, other demographic characteristics (e.g., ethnicity) may differ from our sample to that of true patients affected by schizophrenia.

Despite all these limitations, this study showed that the smooth pursuit task, and especially the RMSE measure, is a behavioral marker of schizophrenia that is difficult to feign. The anti-saccade task was feigned more successfully by participants, but response latency may be a more reliable cue to feigning during this task. Finally, the pro-saccade task detected participants instructed to feign well, although as a standalone measure, it is not a biological marker of schizophrenia.

Practical Implications and Conclusion

Research has consistently found that psychologists are poor at identifying invalid symptom or performance presentations through mere interaction with the examinee, such as an interview (Boone, 2013; Dandachi-FitzGerald et al., 2017). In fact, as evidenced by the development of consensus statements from national neuropsychological organizations (e.g., Bush et al., 2005; Heilbronner et al., 2009), in the last two decades, PVTs have become standard of practice in medicolegal contexts and in forensic settings, and are becoming routinely used in clinical settings as well. Moreover, recent studies demonstrated that integrating SVTs and PVTs yields incremental validity compared to using SVTs only or PVTs only (Banovic et al., 2021; Giromini et al., 2020; Šömen et al., 2021).

The current study adds to the growing literature which demonstrates the value of PVTs and specifically, the use of eye tracking technology as a means of reducing error in making determinations about the attempt of feigning in the performance of people affected by schizophrenia. From a clinical and forensic perspective, eye tracking is becoming increasingly useful for assessing symptom validity, especially in schizophrenia. However, it is important to consider comprehensive patient profiles, as the effects of symptoms do not manifest uniformly, and psychological injuries can involve schizophrenic-like symptoms, without necessarily being attributed to schizophrenia. In our study, differences in eye behavior between honest and feigning respondents were found to be useful in identifying faking behavior. Therefore, it may be beneficial to examine whether identification of feigning can be improved by developing specific eye tracking measures able to strongly differentiate between feigners and honest responders.

Determining symptom validity as it relates to a schizophrenia diagnosis has important medicolegal and clinical implications. First, relative to other common psychopathological complaints involved in medicolegal dispute (e.g., PTSD, TBI), schizophrenia tends to develop rather gradually, and often without a single salient trigger. Absent a full psychiatric history, collateral information from others’ reports, and verifiable information reported by the patient should be considered alongside prolonged multiple observations and assessments over time, as current research on eye tracking and feigning schizophrenia has focused on bona fide patients, as opposed to suspected patients. For example, in cases of suspected early psychosis, poor quality of life scores (e.g., assessed by Global Assessment Functioning and Quality of Life Scale) as contrasted with prior patient history, should warrant (eye-tracking and other) assessments for schizophrenia (Hosseini & Yousefi, 2011), as lower GAF scores in early diagnosis of schizophrenia tend to point to poor prognosis (Köhler et al., 2016), and thus, eye tracking and other feigning measures may be clinically useful in these cases.

Second, someone with schizophrenia could suffer a psychological injury and this comorbidity could complicate presentation of both schizophrenia and psychological effects of the injury. Despite unclear findings on the relationship of childhood trauma to schizophrenia (Morgan & Fisher, 2007; Ruby et al., 2017; Cancel et al., 2019), recent research has shown that the lack of clinical attention to various negative life experiences (specifically, traumatic loss) that have contributed to trauma exasperates experienced psychotic symptoms through maintenance factors like cognitive distortions and attenuated affective responses (Vallath et al., 2020). Therefore, thorough history, including pre-existing psychological trauma should be taken into consideration when assessing suspected schizophrenia following psychological injury which points to traumatic loss (e.g., bereavement, job loss). Further, schizophrenic-like symptoms, such as delusions and hallucinations can emerge following a stressful event, perhaps due to an underlying vulnerability, such as a personality disorder (D’Agostino et al., 2019). In forensic cases, stressful events that cause psychological injury (e.g., arrest, incarceration) can be the catalyst for first-episode psychosis to appear, and this may significantly complicate the outcome of the forensic evaluation.

Third, eye movement abnormalities occur in other poorly understood psychological conditions, like Capgras delusion. Delusional misidentification syndromes likely share an organic basis of structural organ damage, which contributes to schizophrenic-like symptoms (Josephs, 2007; Young et al., 1993), including ocular abnormalities (Brighetti et al., 2007). While Capgras and other misidentification syndromes most commonly occur during the course of comorbid schizophrenia, they can also be attributable to or aggravated by head injury, physiological illness, and other neurodegenerative diseases, which might impact prognosis and presentation. In forensic cases, it may not be easy to discern the veracity of the defendant’s performance and consequently, assess his/her ability to stand trial.

Fourth, schizophrenic-like symptoms could emerge after the event for which the defendant is being tested, perhaps because suffering from an underlying vulnerability. Such cases could also significantly complicate the outcome of the forensic evaluation.

Finally, there are conditions that could further complicate the differential diagnosis, such as schizoaffective disorder beforehand or depression with psychotic features after the event.

Without valid test results, clinicians may reach incorrect conclusions regarding the diagnosis and severity of a patient’s difficulties or symptoms and may subsequently initiate a number of inappropriate referrals and treatments, thus, increasing iatrogenic effects for the client and limiting resources for other patients.

Analysis of eye movements is, therefore, a worthwhile PVT measure for future research investigating the feigning of cognitive impairments. This study represents the beginning of a systematic effort to utilize eye movement responses as a performance validity measure. Given the high rate of feigning of psychiatric illnesses and the increasing availability of low-cost and portable eye tracking devices, eye tracking likely has the potential to become a crucial part of PVTs.