Article Text

Download PDFPDF

Improved antisaccade performance with risperidone in schizophrenia
  1. J G Burke,
  2. M A Reveley
  1. Neuropsychopharmacology Unit, University Department of Psychiatry, Leicester General Hospital, Gwendolen Road, Leicester LE5 4PW, UK
  1. Correspondence to: 
 Professor M A Reveley, Neuropsychopharmacology Unit, University Department of Psychiatry, Leicester General Hospital, Gwendolen Road, Leicester LE5 4PW, UK; 


Objective: Antisaccade errors are consistently increased in schizophrenia. As they have been demonstrated only in cross sectional studies, it is unclear how they vary longitudinally or with different medications. In a previous cross sectional study, we reported a trend towards a reduction in error rates in a patient group treated with risperidone, compared with clozapine and sulpiride treated groups.

Methods: Gap random and antisaccade paradigms were performed on two occasions in the same sample of DSM-IV schizophrenic patients (n=12) in transition between conventional antipsychotic drugs and risperidone. A cross over design was used with six patients switching from risperidone to conventional (group I) and six in the opposite direction (group II). A control sample (n=12) was also tested on two occasions and their performance compared. The effects of practice between first and second testing and of switching between conventional antipsychotic drugs and risperidone and vice versa was also evaluated.

Results: A significant reduction in error rate was demonstrated during risperidone treatment (n=12), compared with conventional APD treatment. Switching from conventional to risperidone produced a reduction in errors, and vice versa.

Conclusions: Treatment with risperidone was associated with improvement in antisaccade errors.

  • eye movements
  • schizophrenia
  • atypical antipsychotic drugs
  • SEMs saccadic eye movements
  • DLPFC, dorsolateral prefrontal cortex
  • APDs, antipsychotic drugs
  • SAPS, scale for the assessment of positive symptoms
  • SANS, scale for the assessment of negative symptoms

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Saccadic eye movements (SEMs) are rapid eye movements which redirect gaze between different objects of interest. They consist of a large primary saccade which covers most of the target distance in one movement, usually but not always followed by smaller corrective saccades which allow accurate fixation on the target. Most SEMs are visually guided, performed reflexively, and controlled by midbrain and brain stem structures, with also higher cortical (neocortex) involvement.1 In schizophrenic patients there are no impairments of these visually guided SEMs.2,3 The consistently reported SEM abnormalities in schizophrenia occur in the non-visually guided paradigms—for example, the antisaccade task.2,4 In the antisaccade task the subject is instructed to generate eye movements in equal and opposite directions from a series of peripheral targets. If correctly performed this requires the initial suppression of a powerful response to look reflexively towards the novel peripheral targets and the volitional generation of a movement as accurately as possible away from the target and towards its mirror image, in the absence of visually guiding cues. Conscious effort and higher cortical control are therefore required.

The main SEM abnormality reported in schizophrenia is of high rates of antisaccade distractibility errors due to a failure to carry out the initial suppression, thereby resulting in incorrectly directioned primary saccades towards the target. Animal and human lesion studies have implicated dysfunction of the inhibitory pathways from the dorsolateral prefrontal cortex (DLPFC), via the basal ganglia to the brain stem, as responsible for these abnormalities.1 This has been supported by the failure of drug free schizophrenic patients to activate the left DLPFC compared with normal controls, as shown by PET while undertaking an antisaccade task.5 However, another study failed to show DLPFC activation during antisaccade tasks in normal subjects.6

In addition to the well replicated finding of antisaccade errors in schizophrenia, some studies also show increased latency of response in correctly performed antisaccades.3,4,7 Some studies measured the positions of the primary saccade and the final eye position and compared these with the target position. The primary saccade tends to grossly undershoot the target distance, and there is a tendency for failure of the usual corrective saccades to correct this shortfall relative to the target in the final eye position.2,8

These abnormalities are part of a group of neurocognitive deficits associated with schizophrenia. So far, saccadic abnormalities have been demonstrated only in cross sectional studies, and it remains to be determined how they vary over the longitudinal course of the disorder or with different types of medication. In an earlier cross sectional study we have reported that patients treated with risperidone showed a trend towards a reduction in error rates, compared with clozapine and sulpiride treated patients.9 To investigate this finding further, we compared SEM performance on two occasions (during each of two treatment phases: (1) risperidone and (2) conventional antipsychotic drugs (APDs)) in the same group of DSM-IV schizophrenic patients.

We predict an improvement in antisaccade errors with risperidone because of its site of action in the frontotemporal regions and relative sparing of the basal ganglia and extrapyramidal tracts especially in lower dosages.



Schizophrenic patients (DSM-IV) in Leicestershire hospitals who were about to undergo treatment change, for clinical reasons, from conventional APDs to risperidone or vice versa, were identified by reviewing hospital records and by direct contact with clinicians after a specific request. Exclusion criteria included significant substance misuse or medical or neurological disorder as well as the use of anticholinergic medication or more than one antipsychotic agent. The reasons for patients switching from risperidone to conventional APDs included poor patient compliance with oral medications, side effects, perceived lack of treatment response, and financial constraints on prescribing by some general practitioners.

A control sample of healthy volunteers was also recruited and tested on the same SEM paradigms on two occasions to determine the effects of repeated testing on SEM performance in normal individuals.

Twelve patients were included (six in the process of changing from risperidone to conventional APD and six from conventional APD to risperidone). The sample consisted of three women and nine men with a mean age of 37 years (SD 8.64, range 28–55). Mean age at onset was 24.83 years (SD 7.3, range 17–37) and mean duration of illness was 136.83 months (SD 78.75, range 50–288) (table 1). Conventional APDs comprised sulpiride (n=3), chlorpromazine (n=3), trifluoperazine (n=2), haloperidol (n=2), flupenthixol (n=1), and clopenthixol (n=1). Patients were selected on single APD therapy and were not on concurrent anticholinergic treatment, as this may affect SEM performance.

Table 1

Means (SD) for demographic and medication details for subjects treated with either risperidone or conventional APDs first

The use of benzodiazepines and hypnotic drugs were excluded in the 4 weeks before study. Tea, coffee, and tobacco were allowed up to 1 hour before testing. The identified patients who were suitable and consenting underwent testing on two occasions: (1) while still on the original APD; and (2) after becoming established on the alternative APD. We adopted a prospective cross over study design, with equivalent numbers of patients switching medications from conventional APD to risperidone, as in the opposite direction of risperidone to conventional APD. This counterbalanced treatment sequence was used to offset any effects of repeat testing on test performance.

Those six patients (group I) who were tested on risperidone first (table 1) had undergone this treatment for a mean of 5.67 weeks (SD 3.01; range 3–10) and the mean dose of risperidone was 7.67 mg/day (SD 1.97; range 6–10). They were retested after a mean of 10.83 weeks on conventional APD (SD 2.86; range 8–16) and the mean dose of conventional APD (CPZ equivalent) was 325 mg/day (SD 160.47, range 150–600).

The other six patients who were tested on conventional APD first (group II) had a mean duration of treatment of 77.17 weeks on the same antipsychotic drug (monotherapy) (SD 66.75, range 24 -200), and mean dose of conventional APD (CPZ equivalent) was 558.33 mg/day (SD 500.42, range 150–1200). In addition they were on a stable dose for a mean duration of 54 weeks. During the risperidone treatment phase, the mean dose was 6.83 mg/day (SD 1.83, range 5–10), and mean duration of treatment on testing was 9.67 weeks (SD 5.43, range 4–16).

The duration of conventional APD treatment was significantly longer for those who were tested on conventional APD first (mean 77.1 weeks (SD 66.7), v mean 10.83 weeks (SD 2.8); Mann-Whitney test z value −2.9, p<0.01) There was no similar effect for sequence of testing on dose or duration of treatment with risperidone. Thus there was a possibility that such prolonged exposure to conventional APD for those patients tested on conventional APD first may have had an effect on their test performance and we undertook a two way analysis of variance (ANOVA) to determine whether there was any significant effect of drug sequence.

Clinical evaluations

The hospital case notes were reviewed and patients underwent personal clinical assessments on both test occasions using the scale for the assessment of positive symptoms (SAPS),10 and the scale for the assessment of negative symptoms (SANS).11 These SANS and SAPS ratings were performed by a single trained rater (JGB) not specifically blind to medication status (but the ratings were initially conducted as part of a study on a much large number of other subjects and not with this particular study in mind so that any potential bias was therefore reduced) but blind to SEM performance.

Measurement of SEMs


Each subject was positioned 1.5 m in front of a display containing a central light emitting diode and four peripheral light emitting diodes located at ±7.5° and ±15°. The test was conducted in a darkened room which rendered the diodes visible only when illuminated. Eye movements were recorded using an infrared limbus reflection device (Skalar IRIS), with a linearity range of ±15°. The stimulus display and data sampling were controlled by a 486 PC computer using software written specifically for the purpose (AMTech GmbH). The frequency of data point acquisition by the apparatus for the measurement of saccadic eye movements was 400 Hz.

Saccadic analysis was conducted off line using interactive software which enabled the rejection of artefacts due, for example, to blinks. The SEM testing performed on each subject consisted of a visually guided task (gap random paradigm) and a non-visually guided task (antisaccade paradigm).

Gap random paradigm

The subjects were instructed to fixate on the central light emitting diode. After 2 seconds this diode was extinguished followed by a 200 ms gap before one of four peripheral target diodes were illuminated. Simultaneously a 200 ms buzzer sounded. The subject was instructed to look as quickly and accurately as possible towards the target light emitting diode. After 2 seconds the peripheral light emitting diode was extinguished and the process repeated. There were 48 trials (administered in four blocks of 12) and each diode was illuminated 12 times in pseudorandom order.

Antisaccade paradigm

The subject again fixated on the central light emitting diode until it was extinguished after 3 seconds and immediately replaced by a peripheral target diode together with a 200 ms buzzer. The use of a simultaneous buzzer has become a fairly standard element in such experiments.2,12 The speaker is positioned on the front panel at ground level directly below the centre light.

Here the subject was instructed to look to an equivalent distance in space but in the opposite direction from the target light emitting diode. After 2 seconds the peripheral diode was extinguished and immediately its mirror image diode in the opposite hemifield was illuminated for 1 second to assist the subject in focusing exactly on the appropriate position. Then the sequence was repeated with a pseudorandom variation of the four peripheral target diodes (in 48 trials each was illuminated 12 times). The individual trials were administered in six blocks of eight

Eye position gain was calculated for correctly performed SEMs in both paradigms as the proportion of the target distance covered by the primary saccade only. Final eye position gain was calculated for correctly performed SEMs in both paradigms as the proportion of the target distance covered by the primary saccade and all subsequent corrective saccades (performed before the appearance of any mirror image guiding light).

The scoring of eye movements was conducted by a research assistant completely blind to the medication status. Antisaccade error rates were calculated as the number of primary SEMs incorrectly directioned towards the target, as a percentage of the total analyzable number of trials.

The first set of testing was conducted just before the medication change. The second set was performed after becoming established on the alternative medication after a varied interval (mean 10.4 weeks (SD 4.18), range 4–16)

Statistical analysis

Due to the small sample size, the data were non-parametric in form and the Wilcoxon signed ranks test was used to compare the following SEM measures, under both test conditions (risperidone treatment and conventional APD treatment): the mean latency, peak velocity, eye position gain, and final eye position gain of correctly performed SEMs in both the gap random and antisaccade conditions. The error rates were included for the antisaccade condition only.

The effects of both (1) each individual drug and (2) the sequence of drug treatment, on any significant findings were then evaluated by means of the two way ANOVA. The total SAPS and total SANS scores were computed for each subject (by adding the values of each of the five subscale scores) and compared by means of the Wilcoxon matched pairs signed ranks test.

To evaluate the possible effects of clinical symptom change on any significant SEM measure changes between the two test conditions, the change in SAPS and SANS scores between the two test conditions were computed and entered as covariates in a simultaneous multiple regression analysis.

The SEM performance of the control group was also compared on both occasions in order to determine the extent of the practice effect. A multivariate ANOVA (MANOVA) was used to examine the effects of group (risperidone, conventional APD, and control) and time (test 1 and test 2) on error rates. Multivariate analysis of covariance (MANCOVA) was then used to evaluate the effects of the possibly confounding covariates (interval between test 1 and test 2, dose of risperidone, duration of risperidone treatment, dose of conventional APD, and duration of conventional APD treatment). The Wilcoxon signed ranks test was used to compare SEM performance at both test intervals in group I and group II.


Gap random paradigm

In the gap random paradigm there were no significant differences between the two groups (table 2). Both treatment groups tended to overshoot the target but managed to use correctional saccades to achieve impressively accurate mean final eye positions.

Table 2

Mean (SD) of saccadic eye movement measures during risperidone and conventional APD conditions of study

Antisaccade paradigm

There was a significantly lower antisaccade error rate (Wilcoxon signed ranks test z value −2.82, p=0.005) in the risperidone treatment test condition (mean 46.4%) compared with the conventional APD condition (mean 64.9%). By means of two way ANOVA we showed that there was a significant main effect on the error rate by the individual drug (F=5.54, df=1, p=0.029), but no significant effect by drug sequence (F=0.223, df=1, p=0.642). However, there is also a significant two way interaction between drug and drug sequence (F=4.615, df=1, p=0.044). Therefore, as a result of this interaction the significantly prolonged exposure to conventional APD in those tested on conventional APD first (group II) may exert an effect on the error rate.

The Wilcoxon matched pairs signed ranks test showed a significant improvement in total SAPS scores with risperidone. In six cases, total SAPS scores were reduced during the period of treatment with risperidone, with five cases tied, and one case greater (Z=−2.13; p=0.033). For negative symptoms (SANS ratings) there was no significant change over the treatment period. It was therefore considered helpful to determine how much, if any, influence symptomatic change could have on the antisaccade error rate improvement. We calculated the change in percentage antisaccade error rates and also the change in total SAPS and total SANS scores between the two conditions of testing. We entered the latter two scores as covariates in a simultaneous multiple regression analysis together with other variables which could feasibly be considered to exert an influence on the error rate improvement. None of these variables (change in SAPS scores, change in SANS scores, age, age at onset, duration of illness, and duration of risperidone treatment) showed any significant effect on the error rate change.


The control sample consisted of 12 healthy volunteers who underwent SEM testing on two occasions. There were four men and eight women. The mean age was 34.08 and the interval between testing was 9.2 weeks. The only SEM measure (table 3) that altered significantly on repeat testing was antisaccade latency, which showed a reduced interval (t value=2.52, p=0.029), between first testing (mean 513.1) and repeat testing (mean 480.5). The antisaccade error rate did show an improvement but was not significant (p=0.13) (table 3).

Table 3

Mean (SD) of saccadic eye movement measures for control group during test 1 and test 2

These results indicate the effects of repeat testing on SEM performance (table 4). Here the control group shows a trend towards reduction in error rate on retesting. This is taken as a normal practice effect.

Table 4

SEM error rates and corresponding clinical ratings

Group effects on error rates

We carried out a MANOVA comparing the effects of group (risperidone, conventional, or control) and test sequence on SEM error rate. The results showed a significant time×group interaction (p=0.009) but no significant effect of TIME alone.

We then carried out a MANCOVA to determine the effects of the covariates (test interval, dose of risperidone, duration of risperidone treatment, dose of conventional APD, and duration of conventional APD) on the error rates on risperidone in group II. None of the covariates were found to have a significant effect on the error rates.

For the patient groups there was a highly significant difference (p<0.01, Wilcoxon signed ranks test) on first testing between group I (mean=39.7) where the error rates approach those of the control group, and group II (mean=75.3). On retesting, group II improved but only to a degree than would be predicted by practice effect alone, (mean 53, p=0.046). Although this improved error rate of group II on risperidone was less than that seen with group I on risperidone (mean 39.7), the difference between the mean scores (p=0.319) does not reach statistical significance. Likewise, the impressive performance of those on risperidone on first testing in group I (mean 39.7) decreases markedly on switching to conventional APD (mean=54.6, p=0.028) and clearly goes against the trend expected of a practice effect and therefore is a more compelling argument in favour of an initial risperidone induced improvement.

Although the error rates were exceptionally high in group II on first testing (while on conventional APD for a considerable period of treatment) the primary error was corrected in 87% of cases suggesting that the instructions were understood and that errors were due to initial failures of suppression which were then able to be corrected.

The clinical symptom ratings were lower than in the other group so that the error rate does not seem to be a feature related to illness severity. One difference between the two groups is sex as group I contained equal numbers of men and women whereas group II consisted entirely of men. This raises the possibility of an effect of sex ratio upon performance as men tend to have earlier ages of onset and more progressive course as well as more neurocognitive deficits.


During risperidone treatment a significant reduction in antisaccade error rate was shown. This error rate improvement was independent of the improvement in positive psychotic symptoms which occurred during the risperidone treatment phase. This suggests that risperidone was causing an improvement in the neurophysiological deficits of schizophrenia, and that this effect was not mediated indirectly via symptomatic improvement.

As the switch from conventional medication to risperidone in one patient group was associated with improved antisaccade performance and the medication switch in the opposite direction in the other group was associated with a worsening of antisaccade performance it suggests that any improvement brought about by risperidone may be temporary, at least for the short duration of exposure in this sample.

The error rate during risperidone treatment was lowest in group I and here the error rate approached that of the control group on first testing. In absolute terms the error rate of 39.7 for group I on risperidone is comparable with some previous findings for conventionally medicated schizophrenic samples: 38.9%,13 33.6%,14 and 30%15. On that basis the present sample is not in itself unrepresentative of the area of study. In group II also a clear improvement was shown in the error rate after switching to risperidone from conventional APD.

Risperidone is an atypical antipsychotic drug (benzisoxazole derivative) with strong affinities for D2 and 5-HT2a receptor sites, and lower affinities for cholinergic receptors than conventional APDs.16 It has much reduced potential for causing extrapyramidal side effects compared with conventional APDs.17 Risperidone shows few extrapyramidal side effects in the dose range of 2–6 mg/day, but then shows a dose related increase of extrapyramidal side effects similar to typical medicationss18 Risperidone is the longest established of the new generation of atypical APDs and its clinical use is not restricted to refractory schizophrenia; hence it attracts a wider spectrum of severity of illness.

Recent studies have highlighted the beneficial effects of risperidone on other neurocognitive deficits in schizophrenia, with significant improvements occurring even in the early weeks of treatment.19–23 Our current understanding of the biochemical actions of risperidone suggests that any unique effect is the result of the blocking of 5-HT2a receptors, for which risperidone has the highest affinity among all available APDs.24 In addition, the distribution of this receptor subtype shows high concentrations in the prefrontal cortex,24 an area which has been identified as the focus of antisaccade distractibility errors,1 which could help to explain the effectiveness of risperidone in correcting such antisaccade abnormalities. L-tryptophan, the precursor of serotonin, causes saccadic disinhibition,25–27 which may help explain the effectiveness of risperidone in counteracting the disinhibition and thereby correcting error rates.

The only other published study which examined the effects of risperidone on SEM performance28 compared reflexive saccade performance in a before and after treatment study of risperidone and haloperidol on previously antipsychotic naïve first episode schizophrenic patients. They found widespread and significant worsening of SEM performance with risperidone compared with haloperidol, with a significantly prolonged latency, decreased peak velocity, and reduced accuracy with the risperidone treated group after 4 weeks of treatment. The suggested mechanism was lack of development of acute tolerance to risperidone's powerful serotonergic antagonism, which could lead to disruption of brain stem physiology in regions controlling SEMs. The comparable arm of our study was the gap random paradigm, which failed to replicate any of these effects with risperidone. This difference may be due to any of various factors which vary between the two studies. These include the course of the disorder as patients in the early course of schizophrenia may respond differently than more chronically ill patients. Duration of exposure to medication differs greatly in the two studies. First exposure to antipsychotic drugs may produce different effects on SEM mechanisms than in patient groups who have been chronically medicated. Other variables which differ and may be responsible for the different findings in the two studies include the dose of risperidone (higher in our study) and the sex difference (9:1 in favour of males in the Sweeney study).

Our study provides further support for recent evidence that atypical antipsychotic drugs including risperidone have a direct beneficial effect on neuropsychological and psychophysiological deficits in chronic schizophrenia compared with conventional drugs. This may be partly explained by the relatively low sedative potential of risperidone and also low intrinsic anticholinergic effects as well as reduced need for adjunctive anticholinergic treatments.


The findings of this study are limited by several factors.

  1. The sample sizes are small.

  2. Although the clinical rater was not formally blind to medication status, the ratings were initially conducted as part of a data collection of a much larger number of subjects and not with this particular study in mind, so that any potential bias was therefore greatly reduced.

  3. Drug assignment, dosage, and cross over were non-random

  4. There were varying starting points in exposure to medications before treatment.

  5. Although only one atypical APD was examined, there were various conventional APDs grouped together. Although broadly classified as a single group, these drugs do have important pharmacologial differences and this is an important limitation of this study.


We are grateful to the staff and patients of the Leicestershire and Rutland Healthcare NHS Trust for their support and cooperation, and to Penny Morris for technical assistance with SEM testing.