Metric issues in the study of eye movements in psychiatry

Viewer
Transcript

ARTICLE IN PRESS Brain and Cognition xxx (2008) xxx–xxx

Contents lists available at ScienceDirect

Brain and Cognition journal homepage: www.elsevier.com/locate/b&c

Metric issues in the study of eye movements in psychiatry Nikolaos Smyrnis * Psychiatry Department, National and Kapodistrian University of Athens, Medical School, Eginition Hospital, 72 Vas Soﬁas Avenue, Athens 11528, Greece

a r t i c l e

i n f o

Article history: Accepted 26 August 2008 Available online xxxx Keywords: Smooth eye pursuit Saccade Antisaccade Reliability Consistency Stability Standardization

a b s t r a c t This review provides a description of the measurement methods, task deﬁnitions and measurement parameters in the study of smooth eye pursuit and saccade–antisaccade tasks in psychiatry. The large heterogeneity in task deﬁnitions and deﬁnitions of parameters and its potential impact on the large variability of the parameter measures is presented. Finally issues relating to data aggregation, practice effects and reliability of these smooth eye pursuit and saccade–antisaccade task performance measures are also discussed. The main conclusion of this review is that oculomotor function testing is still lacking standardization of methods, tasks and parameters affecting its usefulness in certain areas of psychiatric research. A future research program could derive a set of tentative recommendations for test standardization. Ó 2008 Elsevier Inc. All rights reserved.

1. Introduction The use of oculomotor function tests in psychiatry research has been growing steadily in the last 35 years since the pioneering work of Holzman, Proctor, and Hughes (1973). The disturbances in oculomotor function tests observed in patients suffering from certain psychiatric disorders such as schizophrenia open the possibility to investigate the pathophysiological substrate of these complex disorders and link behavior to brain function, especially with the application of the new functional brain imaging technologies. Moreover the oculomotor function deﬁcits in these patients could be used as endophenotypes (see glossary) that is intermediate variables measuring one aspect of the complex disorder and linking the phenotype of the disorder to the corresponding genotype. These intermediate variables then that can be objectively measured could be used in genetic studies in search of the complex and probably polygenic substrate of psychiatric disorders (Calkins & Iacono, 2000; see also Calkins et al., in this issue). Finally oculomotor function variables could serve as biomarkers (see glossary) of the disorder that could be used in the evaluation of treatment response and the development of new treatments such as new pharmacological agents (see Reilly et al., in this issue). Is it important then to standardize the oculomotor function measurements in psychiatric research? The answer to this question depends on the speciﬁc research goal. If the goal would be to explore the range of the eye movement function deﬁcits in * Fax: +30 2107293245. E-mail address: [email protected] URL: http://www.cag.eginitio.uoa.gr.

schizophrenia in different conditions probably involving different brain areas then we should strive for diversiﬁcation instead of standardization. If on the other hand the goal would be to use the eye movement function deﬁcits as endophenotypes of the disorder in order to study its complex genetic substrate or develop new treatment strategies then we should strive for the best possible standardization of these measurements. Table 1 presents a list of speciﬁc objectives for measurement standardization. The ﬁrst four objectives relate to metric issues in the application of these tests and will be the focus of this review. The ﬁrst objective is to standardize the methods of measuring oculomotor function variables in psychiatric research and this issue will be addressed separately in the following chapter of this review. Then focusing on the three most used oculomotor paradigms in psychiatry research, namely the smooth eye pursuit task, the visually guided saccade (saccade) task and the antisaccade task (while also brieﬂy mentioning the memory saccade and predictive saccade task) the test procedures used in psychiatric research will be reviewed addressing the objective on standardization of test procedures. The different parameters used in assessing performance in each task will be reviewed focusing on different deﬁnitions of parameters, their objectivity and the effects of practice on each one of them to address the objective of parameter standardization. Finally for each parameter in each oculomotor task practice effects and reliability and will be discussed. Reliability is deﬁned as the consistency of individual differences in a measured variable over repeated measures for the same sample of individuals and can be divided into measures of test–retest reliability and measures of internal consistency (Box).

0278-2626/$ - see front matter Ó 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.bandc.2008.08.022

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS 2

N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

Box 1. Reliability criteria of test quality

Deﬁnition of reliability: Test theory assumes that the expected value of the true score for a particular individual on a test does not change with repeated applications of the same test paradigm but due to random error in each speciﬁc measurement the observed scores can differ. The greater the random error in a particular test the less reliable will be the test because the observed scores will vary more or equivalently will be less consistent. Reliability is deﬁned as the ratio of true score variance to observed score variance. Data aggregation and reliability: An increased number of measurements, called data aggregation, would in general lead to increased reliability (Spearman–Brown prophecy formula). Test–retest reliability: Compares samples of measurements for the same individuals for different sessions at different points in time using correlation analysis. In practice either the Pearson correlation coefﬁcient or the more appropriate Intra Class Correlation coefﬁcient (Shrout & Fleiss, 1979) is used. This type of reliability estimate is suitable only under the assumption that the true score does not change in time. If the true score changes over time and this change is different for each individual then test retest reliability will be affected by the differential time domain change in the true score as well as the change in the observed scores due to measurement error. Internal consistency: Divides the total number of available measurements in a single test session into sub groups and tests their correlation. This is usually done by the split-half method that is by dividing the data set of measurements into halves. The easiest way to do this is to take the ﬁrst half and compare that with the second half. The problem with this splitting is that it is again sensitive to differences in performance of the test with time. Thus it is considered more appropriate to use the odd–even system: the ﬁrst half group comprises of the odd-numbered data and the second of the even-numbered data. Then the correlation coefﬁcient between the two halves is computed corrected for the half length by the Spearman–Brown correction (Kaplan & Saccuzzo, 1989). Finally by assigning observations randomly for each half the Cronbach alpha (a) coefﬁcient can be computed which is a general reliability coefﬁcient ( Cronbach, 1951). Further reading on test theory: Lord and Novick (1968) Kaplan and Saccuzzo (1989).

The last objective in Table 1 refers to the assessment of the validity of oculomotor function tests. Validity can be deﬁned as the level of agreement between a particular test measure and the

Table 1 Objectives for the use of oculomotor function testing in psychiatry 1. 2. 3. 4. 5.

Standardize measurement procedure Standardize test procedure Standardize speciﬁc objectively measured outcome variables Test for reliability of outcome variables Test for validity of outcome variables

quality it is believed to measure (Kaplan & Saccuzzo, 1989). There are four types of validity evidence: content, predictive and concurrent which together form a criterion validation and ﬁnally construct validity (Cronbach & Meehl, 1955). A particular test or measure is supposed to have content validity if it samples adequately the universe that the investigator is interested in. For example a test of smooth eye pursuit should cover adequately the universe of pursuit function that we are interested at. This question is relevant to the variations of the different pursuit tasks and how relevant they are at exploring the pursuit behavior. Criterion validity refers to how good the particular measure correlates with a criterion measure that is the ultimate measure of interest for which the particular measure is a substitute. There are two types of criterion validity the predictive and the concurrent validity. Predictive validity refers to the forecasting function of the test. For example we might be interested at using the performance in the antisaccade task for predicting functional recovery in schizophrenia or response to certain treatment. Concurrent validity refers to the relation of the test performance, for example smooth eye pursuit performance with a criterion such as for example the speciﬁc psychiatric diagnosis of schizophrenia. Finally construct validity is relevant when the test is interpreted as a measure of some attribute or quality that is not operationally deﬁned. For example the antisaccade test is supposed to measure the cognitive function of inhibition (see glossary) or smooth eye pursuit is supposed to measure attention. Construct validation then is related to the process by which we accumulate evidence for building a theory that explains the test results in a theoretical framework. This is usually done by comparing this test with other tests and measures and building on the information about these relationships. Two types of such evidence can be acquired, convergent and discriminant. Convergent evidence for validity is obtained when the particular measure correlates well with other tests that are believed to measure the same construct. Discriminant evidence for validity is obtained when the particular measure does not correlate with other tests that are believed to measure different constructs. In the former example the antisaccade test performance would not be expected to correlate with tests that are supposed to measure cognitive functions entirely unrelated to inhibition. In conclusion the objective of validity for oculomotor function tests in psychiatry spans many different domains that will be covered in other reviews included in this special issue. 2. Methods for measuring eye movements The initial clinical studies on eye movements relied on clinical inspection of the eyes. Modern studies have introduced objective methods for measuring eye movements. The ﬁrst such method employed was electrooculography (EOG; see glossary) that uses skin electrodes placed around the eye and observed that a chance in the position of the eye resulted in electrical potential differences. The rotation of the eye in the orbit results in a change in the electrostatic ﬁeld that rotates with the eye. Conjugate horizontal eye movements can be measured by using two skin electrodes placed over the outer canthi of both eyes. Separate recordings from each eye can be measured by placing a third common electrode over the bridge of the nose. Vertical eye movements can be measured by placing electrodes over the eyelid and under the eye at the maxilla. The potential differences are measured using a dc ampliﬁer or an ac ampliﬁer with a large time constant (>0.1 s) (Young & Sheena, 1975). The obvious advantage of using EOG or any other quantitative method in the measurement of eye movements is that an objective measurement is acquired. EOG can measure lateral displacement of the eyes up to ±70 deg with an accuracy of 1.5–2 deg. The disad-

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

vantages of the method are that it is prone to large artifacts from the movements of the muscles in the face area. The measurement is also vulnerable to variation of the potential attributable to light adaptation, diurnal variations and state of alertness. Finally EOG as well as all the other measurements of eye movements that will be discussed below, except the search coil, are susceptible to blink artifacts. The major problem though with the EOG method is the non-linearity of the measurement. The position of the eye is measured by a change in the potential due to rotation of the electrostatic ﬁled of the eye. The potential theoretically varies with the sine of the angle of eye deviation but practically the values largely depart from the theoretical ones due to the fact that the conducting medium is non homogeneous. Due to the non-linearity problem EOG is not the preferred method today to study the metrics of eye movements. Some authors propose the concurrent use of EOG with other methods of eye movement measurement in order to accurately detect blinks and other muscle artifacts (Calkins, Katsanis, Hammer, & Iacono, 2001). The most widely used method to study eye movements is the infra red limbus or pupil detection method. Infrared light illumination of the eye is used and the reﬂection of the light by the border of the sclera and the pupil (called the ‘‘limbus”) or the reﬂection from within the pupil is detected. Most modern systems incorporate both the infrared illumination and the photodetectors into goggles worn by the subject (Young & Sheena, 1975). Depending on the orientation of the illuminating source and the photodetectors horizontal and vertical eye movements can be measured. The precision of the system can be very high in the order of a few minutes of arc especially if the head is restrained using a chin rest or a bite board. The infrared method is not affected by muscle artifacts as the EOG. The linearity of the measurement is also very good for eye movements up to ±30 deg. A disadvantage of the infrared method is that large eye displacements cannot be measured due the fact that the limbus is covered by the eye lid. This problem is especially manifested when trying to measure vertical eye movements. Measuring the pupil instead of the limbus can offer a wider range of movement but the signal is not as strong (Young & Sheena, 1975). A good practice when using these systems is to restrain the head and measure preferably horizontal eye movements up to a range of ±15 deg where most systems have very good linearity and sensitivity. Another method for measuring eye movements is the use of a video camera to capture the eye movement. Some systems use the corneal reﬂection of bright light to measure the movement of the eye (Young & Sheena, 1975) while more modern systems use complex pattern recognition algorithms to detect the pupil or the iris or both. Modern technological advancements have allowed the manufacturing of miniature video cameras that can be incorporated into goggles that the subject can wear. These systems allow for very accurate eye movement position measurement up to 0.01 deg in both X and Y axes. The video based systems can also use table mounted cameras that can measure both the eye movements and the head movements thus allowing for the measurement of gaze. The linearity of the systems is very good for lateral displacements of 30 or even 40 deg. The video camera systems have obvious advantages over the other methods for measuring eye movements in two dimensions with the disadvantage of the larger cost for these systems. The other potentially very serious problem with these systems is the sampling frequency of the video camera. Most commercially available video camera systems have sampling frequencies below 100 Hz. Using such a system for measuring eye movements will be detrimental to the quality of results. Finally the most accurate method for measuring eye movements is the use of a device embedded in a contact lens that is ﬁt-

3

ted tightly to the eye of the subject. One such method that offers excellent spatial and temporal resolution in eye movement measurement is the ‘‘search coil” technique introduced by Robinson (1963). The technique uses two small wire coils embedded in the contact lens worn by the subject. These wires pick up an induced voltage that is created by two large electromagnetic coils surrounding the subject. This method can accurately measure three dimensional movements of the eye in the orbit and is not vulnerable to blink artifacts. The major disadvantages of the method are the use of a large and difﬁcult to operate device and most importantly that the subject has to wear a tight contact lens that is causing distress and even pain after some time. Also the contact lens should be kept and applied in aseptic conditions or a new lens should be used for each subject in order to avoid infections. The invasiveness then of this method has precluded its use in psychiatric research. 3. Smooth eye pursuit task: Test procedure deﬁnition Table 2 presents a series of publications reporting on smooth eye pursuit in patients suffering from schizophrenia or in individuals within the spectrum of schizophrenia and compare their performance with that of normal controls. Some studies also report on ﬁrst degree relative of schizophrenia. These studies were found by a search in the Medline using keywords ‘‘schizophrenia” and ‘‘smooth eye pursuit” from 1990 until the summer of 2007. One study was included that reported on pursuit function in obsessive compulsive disorder (Nickoloff, Radant, Reichler, & Hommer, 1991) and one that reported on pursuit function in bipolar patients and controls (Holzman, O’Brian, & Waternaux, 1991). A few studies were added that report on pursuit performance in healthy adult samples (Ettinger et al., 2003; Green, King, & Trimble, 2000; Katsanis, Iacono, & Harris, 1998; Roy-Byrne, Radant, Wingerson, & Cowley, 1995; Versino et al., 1993). The last entry in the table (Smyrnis et al., 2007) is our sample of the ASPIS study (Athens Study for Psychosis Proneness and Incidence of Schizophrenia) that involved the collection of eye movement and other variables in a sample of 2130 young men enlisted in the Greek Air Force in the years 1999–2000. Since the focus of this review will be on standardization of measurement the selection of studies was such as to offer the possibility of comparing the values of different parameters across studies for normal controls. Thus studies that did not include a control sample, studies that used very different task procedures from what is commonly used as test of smooth eye pursuit function were excluded. Studies that used only qualitative pursuit assessment or did not report the values of parameters measured for the control group as well as studies that reported new analyses on the same dataset were also excluded. The studies that were selected cover the large heterogeneity of measurement methods, task deﬁnitions and parameters measured in the smooth eye pursuit task that were used in the psychiatric literature in the last two decades. The third column of Table 2 lists the number of normal control participants tested in each study, the fourth column presents the type of stimulus motion used, the ﬁfth and sixth columns present the target speed and the amplitude from the center covered by the target stimulus and ﬁnally the last column presents the number of pursuit cycles performed by each subject. The obvious observation that can be made looking at the table is that all these task parameters vary considerably from study to study and this variation might result in differences in the parameters measured. All studies used a predictable motion pattern (except for the study of Hong, Avila, and Thaker (2005) where there were unexpected changes in target direction of motion) but this pattern differed among different groups of studies. Some studies used a sinusoid target motion pattern in which the target speed varies

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS 4

N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

Table 2 List of publications on smooth eye pursuit performance Year

Authors

N of normal control subjects

Stimulus

Speed (deg/s)

Amplitude

N cycles

1990 1991 1991 1991 1991 1991 1992 1992 1993 1993 1994 1994 1994 1995 1996 1996 1996 1997 1997 1998 1998 1998 1998 1999 2000 2000 2000 2000 2000 2000 2000 2001 2001 2001 2001 2001 2002 2002 2002 2003 2003 2003 2003 2003 2003 2004 2005 2005 2005 2005 2006 2006 2006 2007 2007

Clementz et al. Litman et al. Holzman et al. Nickoloff et al. Friedman et al. Abel et al. Radant and Homer Campion et al. Sweeney et al. Versino et al. Sweeney et al. Siever et al. Gooding et al. Roy-Byrne et al. Ross et al. Ross et al. Thaker et al. Ross et al. Litman et al. Sweeney et al. Katsanis et al. Ross et al. Thaker et al. Ross et al. Ross et al. Green et al. Lencer et al. Katsanis et al. Gooding et al. Levy et al. Hutton et al. Ross et al. Karoumi et al. Nikam et al. Calkins et al. Hutton et al. Ross et al. Flechtner et al. Avila et al. Lencer et al. Hong et al. Cerbone et al. Olincy et al. Kathman et al. Ettinger et al. Ettinger et al. Hong et al. Holahan, O’Driscoll Sporn et al. Boudet et al. Hong et al. Spengler et al. Lenzenweger and O’Driscoll Hong et al. Smyrnis et al. ASPIS

38 12 13 12 45 21 17 23 55 20 52 37 49 8 21 24 11 25 24 10 39 adolescents, 36 older 18, 26 older 42 64 37 20 77 58 Monozygotic, 40 dizygotic twins 94 42 23 29 21 34 19 54 69 42 12 20 young, 15 middle, 12 older 37 15 15 84 31 24 22 29 21 young, 78 middle, 47 older 21 young, 31 older 90 24 311 22 1854

3/4 Sinusoid 1/4 triangular Triangular 3/4 Sinusoid 1/4 triangular Trapezoid Trapezoid Trapezoid Triangular Sinusoid 2/3 Sinusoid 1/3 triangular Trapezoid 3/4 Sinusoid 1/4 triangular 2/3 Sinusoid 1/3 triangular Sinusoid Trapezoid Sinusoid Step ramp Step ramp Sinusoid Triangular Step ramp 2 types Sinusoid Trapezoid Triangular with masks Trapezoid Trapezoid Sinusoid Triangular Sinusoid Sinusoid Sinusoid Triangular Trapezoid Sinusoid Sinusoid Sinusoid Triangular Trapezoid Sinusoid Step-ramp, triangular with masks Triangular Trapezoid Sinusoid Trapezoid Trapezoid Triangular Trapezoid Triangular Sinusoid Trapezoid Sinusoid Trapezoid Constant Sinusoid Step ramp, triangular, foveal stabilization Triangular

16 16.67 16, 16.67 11 5 5 16.7 18 and 24 16 10–50 16 19.2 16 10 12 15 15 12 16.67 8, 16, 24, 32 16 16.7 9.4, 14, 18 16.7 16.7 13.2 15, 30 16 16 16 10, 20, 30, 36.5 16.7 16 24 16 10, 20, 30, 36.5 16.7 16 9.4, 18.7 15 9.4, 14, 18 0.3–1.1 Hz 16.7 17.13 12, 24, 36, 48 10, 20 9.4, 14, 18.7 8 15 24 18.7 15, 30 16 10, 18.7 10, 20, 30, 40, 50

10 15 10 15 10 ? 15.5 15, 20 10 30 10 12 10 15 10 12 14,15,19 10 15 15 10 15 12 15 15 20 15 10 10 10 11.25 15 10 15 10 11.25 15 10 12 15 12 ? 15 11 12 15 10 10 ? ? 12 15 20 12 10

48 8 48 ? 15 4 33 60 36 29 48 ? 16 7 4 6 30–45 10 33 32, 24 12 6.5 12 ? ? 15 30, 15 12 60 12 24 ? 48 24 12 24 ? ? ? 5 12, 30, 30 25 ? 37.5 8 6.5,7 60 36 5 5–10 48 5 36 ? 5

continuously in a sinusoid fashion determined by a single frequency. This type of stimulus is useful when the objective is to determine the acceleration saturation of smooth pursuit to predictable target motion (Leigh & Zee, 1991) and is also useful for measuring overall pursuit performance using global measures as we shall discuss later on. Other studies used a triangular velocity proﬁle meaning that the target moves with constant speed and changes direction abruptly at the end of each run. Another group of studies used a variation of this type of stimulus called the trapezoid pattern of target motion. In this pattern the target moves with constant speed as in the triangular pattern but stops at each of the two end locations to the right or to the left of the target motion range. At these locations the target remains stationary for a period of time (usually between 1 and 2 s) and then starts to move again in the opposite direction. The triangular and trapezoid pat-

terns are useful for determining the steady state pursuit velocity and the velocity saturation of pursuit (Leigh & Zee, 1991). They are not though suitable for the measurement of global pursuit performance measures since the performance at the stops or direction reversals can vary considerably from subject to subject based on the particular strategy. In any case the eye cannot completely follow the target motion pattern (the eye cannot stop instantly or begin moving at a constant speed instantly that is with inﬁnite acceleration). A few studies used a combination of sinusoid and triangular or trapezoid stimulus motion patterns (Clementz, Sweeney, Hirt, & Haas, 1990; Holzman, O’Brian, & Waternaux, 1991; Sweeney et al., 1993). These differences in the pattern of stimulus used could lead to different interpretations of the results. If for example one study would ﬁnd a relation of pursuit performance (phenotype) to some genetic locus (genotype) in schizophrenia

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

using a sinusoid pursuit stimulus and a second one does not replicate this result using a triangular pursuit stimulus it would not be known if the failure of replication was due to the lack of a true correlation between the phenotype and the genotype or due to the difference in the phenotype measured. This problem could be even larger if the stimulus pattern for measuring pursuit performance is known to measure a different type of pursuit performance. A few studies used the step ramp task in which the stimulus makes a sudden step to a peripheral location and then moves with constant speed in the opposite direction and stops (Avila, Weiler, Lahti, Tamminga, & Thaker, 2002; Ross, Ochs, Pandurangi, Thacker, & Kendler, 1996; Thaker, Cassady, Adami, Moran, & Ross, 1996; Thaker et al., 1996). This stimulus is useful in determining the so called open loop pursuit that is the initial acceleration of the eye to catch up with the stimulus after the step when the ramp begins (Leigh & Zee, 1991). A few studies have used very different stimulus motion paradigms to measure other aspects of the pursuit function. Table 2 lists two studies that used the mask pursuit stimulus in which the visual target stimulus temporarily disappears (mask) to reappear again later on in the course of a triangular pursuit pattern (Hong, Avila, Adami, Elliot, & Thaker, 2003; Thaker et al., 1998) and one study that used a foveal stabilization of the target on the retina during the course of a triangular pursuit pattern (Hong et al., 2007). These three studies used these manipulations of the stimulus in order to determine the so called predictive pursuit that is the pursuit based on extra retinal cues. Another study by Hutton et al. (2001) used a texture background in one condition that decreased pursuit performance. Finally another task manipulation used in some studies was the introduction of a dual task condition called the monitoring task (Clementz et al., 1990; Holahan & O’Driscoll, 2005; Holzman, O’Brian, & Waternaux, 1991; Kathmann, Hochrein, Uwer, & Bondy, 2003; Lenzenweger & O’Driscoll, 2006; Sweeney et al., 1993). In this condition the pursuit stimulus changed shape unexpectedly during performance of the task and the individual was required to monitor such changes. This manipulation was introduced in order to facilitate performance of the task by driving the individual’s attention to the stimulus. It is rather obvious that applying these highly diverse stimulus manipulations from study to study makes even more remote the possibility to reduce the phenotypic complexity of the measured variable in order for it to be a useful endophenotype in genetic studies. Another parameter that varied considerably from study to study was the speed of the visual target that the subject was instructed to pursue. Table 2 lists average target speeds that range from 5 to 50 deg/s. It should be also made clear that for sinusoid patterns of target motion the target speed varies continuously and Table 2 presents the average speed based on the frequency and the amplitude of target motion. This speed is not directly comparable to the speed for triangular or trapezoid target motion since for example a mean speed of 16 deg/s in a sinusoid pattern means that the speed ranges from 0 to 22 deg/s. Also many studies using a sinusoid pattern of target motion report the frequency of the sinusoid instead of the average target speed. The average speed can be calculated if the amplitude of target motion is reported. In the study of Cerbone et al. (2003) the authors report the use of sinusoid stimuli at different frequencies from 0.3 Hz up to 1.1 Hz but do not give the amplitude of the target motion so it is impossible to estimate the average speed of the target stimulus. A related problem concerns the comparison in terms of average target speed of studies that used sinusoid stimuli at 0.4 Hz. Most of them (Clementz et al., 1990; Sweeney et al., 1993, etc) used target amplitudes of 10 deg but for example Boudet et al. (2005) used target amplitude of 15 deg resulting in a different mean speed and speed variation range. All pursuit performance parameters are affected by target

5

speed as we will discuss later on and in general pursuit performance deteriorates with increasing speed (Cerbone et al., 2003; Ettinger et al., 2003; Hong et al., 2005; Hutton et al., 2001; Smyrnis et al., 2007; Sweeney et al., 1998; Versino et al., 1993). Another task parameter in the pursuit task is the amplitude in degrees covered by the moving target to the right and left of the central ﬁxation (see glossary) point. Looking at Table 2 it is clear that the target amplitude varied considerably among different studies ranging from to 10 to 30 deg. There has not been to my knowledge a systematic exploration of the question of whether such large changes in target motion eccentricity play a role in pursuit performance. The last column in Table 2 presents the overall task duration in cycles. A complete cycle represents the motion of the target towards one direction and then back again towards the opposite direction until to the same starting point is reached. Again this parameter varied greatly from study to study ranging from 4 to 60 to cycles of pursuit. These differences are also of importance in terms of the number of available measurements of pursuit performance. An increase in the number of cycles of pursuit performed will result in the increase in the number of available measurements of pursuit performance thus increasing the reliability of the measured parameter (larger data aggregation) as I have already discussed in the introduction but on the other hand the repeated measurements might be prone to time dependent changes in the true scores as for example practice effects or effects of fatigue. I will discuss these effects for each pursuit parameter later on. 4. Smooth eye pursuit task: Global measures, practice effects and reliability In the early years of psychiatric eye movement research the golden standard for measuring pursuit performance was a qualitative assessment by two or more researchers, categorizing global pursuit quality in a scale ranging from excellent to poor (Levy, Holzman, Matthysse, & Mendell, 1994). A problem with the use of qualitative indexes in measuring performance is the objectivity in the measured outcome variable. One way of dealing with the objectivity problem is by measuring inter-rater reliability and assessing how similar the ratings are among different raters. In order to achieve reliability among different raters from different research groups a standard set of pursuit records and a standard set of rules for rating them would have to be agreed upon. Then all different raters should be trained using this tool. These issues introduce difﬁculties in the effort for standardization of this measure. A more fundamental problem though with the qualitative global assessment of pursuit performance is that it adds up measures of different phenomena. It is known from the neurophysiology of the pursuit system that successful smooth eye pursuit involves the integrated function of both the pursuit and the saccadic system (Leigh & Zee, 1991). A decrease in the overall quality of pursuit could thus be the product of two qualitatively different effects, the inability of the individual to pursuit the target at the same speed or the inability of the individual to suppress the saccadic system function during pursuit resulting in unwanted intrusive saccades during pursuit performance such as anticipatory saccades or square wave jerks (Abel & Ziegler, 1988). Thus a global assessment of pursuit performance is ambiguous. The introduction of quantitative measures of global pursuit performance solved the problem of objectivity and led to the abandonment of global qualitative ratings for smooth eye pursuit in modern pursuit research. Still quantitative global measures of pursuit performance suffer from the same fundamental problem of ambiguity adding up the qualitatively different measurements of pursuit performance and saccade intrusion into one.

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS 6

N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

Abel and Ziegler (1988) made a crucial observation in favor of the ambiguity argument for global pursuit measures. They simulated pursuit records for a sinusoid stimulus pattern that produced the same decrease in a ﬁnal global measurement of pursuit performance. The measurement was the natural logarithm of the ratio of the power of the signal at target frequency to the power at a higher spectral range (0.8–8 Hz or 1.2–12 Hz). The smaller the result of the logarithm function the higher would be the pursuit deﬁcit. In each of these simulated records though the decrease of the logarithm was the result of a very different process. In one record the decrease was the result of lower pursuit speed that resulted in increased number of small saccades in order to realign the eyes to the target (catch up saccades) while in the next it was the result of a large number of saccades intruding into normal speed pursuit. It could be inferred that the same problem would be encountered with any global measure of pursuit performance such as for example the root mean square error (Fig. 1). This last measure corresponds

to the sum of the squared differences between eye and target location at each point in the pursuit record. The square root of this sum divided by the number of points of measurement gives the ﬁnal global measure. Many researchers still use global measures such as the root mean square error for assessing pursuit performance (Campion et al., 1992; Gooding, Iacono, & Beiser, 1994; Gooding, Miller, & Kwapil, 2000; Ross et al., 1996; Ross et al., 1997; Smyrnis et al., 2007; Sporn et al., 2005; Thaker et al., 1998). It has also been shown that the root mean square error has excellent test–retest reliability and thus it is very useful for dissociating normal from abnormal pursuit over time (Campion et al., 1992; Gooding et al., 1994). Other researches though argue for the abandonment of the root mean square error since it adds indiscriminately different pursuit abnormalities such as the intrusion of anticipatory saccades and low pursuit speed leading to corrective catch up saccades (Clementz et al., 1990). A consensus for the use of global measures of pursuit performance has yet to be reached.

Fig. 1. (A) Record of smooth eye pursuit, represented as the position of the eye (heavy black line) following a target that moves in a triangular motion pattern (thin black line). Three blink artifacts are present in the record (marked with arrows). (B) A smooth eye pursuit record of an individual with congenital nystagmus (see glossary). The eyes do follow the target but the nystagmus superimposed clearly distorts the smooth eye motion. (C) Calculation of pursuit gain: the selected pursuit record (Box) is broken into parts of smooth pursuit (red) and parts were saccades are present (blue). After exclusion of the saccades the ratio of eye speed to target speed is computed (a1, a2) which is the gain. In this record the gain is lower than target speed thus the eyes lag behind the target resulting in catch up saccades (blue). (D) In record I a large anticipatory saccade (AS; see glossary) is depicted. A smaller one is also present at the ﬁnal direction shift in the record. The gray area in record I represents graphically the difference between the eye position and the target position that is used to calculate the root mean square error (RMSE). In record II the gray area also represents the error in position. The RMSE1 calculated for record I and the RMSE2 calculated for record II are very similar while the two records differ substantially. Record I represents an excellent pursuit (with a gain of 1) interrupted by a large AS while in record II the pursuit gain is low followed by a large number of catch up saccades. This example shows the ambiguity of the RMSE. (E) An example of two square wave jerks (SWJ) in the pursuit record marked in blue. The square wave jerk consists of two small saccades in opposite directions with an intervening normal pursuit of 50–500 ms. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this paper.)

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

Maybe global measures could be useful in a screening situation where the speciﬁc deﬁcits could be postponed for later phase studies but then again one would eventually reach the problem of what speciﬁc abnormality in smooth eye pursuit one is measuring. Current research in this ﬁeld aiming at establishing clear cut endophenotypes should aim at speciﬁc quantitative measures of pursuit performance. In this framework I would favor the view that global measures should be abandoned. 5. Smooth eye pursuit task: Speciﬁc measures, practice effects and reliability Gain (see glossary) as originally deﬁned in the neurophysiological literature is the most widely acceptable measure of pursuit performance (Leigh & Zee, 1991). Gain is deﬁned as the ratio of eye velocity to target velocity and if pursuit performance is perfect this value should be equal to one (Fig. 1). The type of stimulus used for smooth eye pursuit is of great importance in the deﬁnition of the gain. When a sinusoid stimulus is used the target velocity is constantly changing so the deﬁnition of the denominator in the ratio eye velocity/target velocity becomes more complicated. In contrast gain can be easily measured for constant speed stimulus patterns such as the triangular and trapezoid patterns that are commonly used in the literature. In that case saccades and artifacts are ﬁrst removed from the pursuit record and then the eye velocity is computed for each saccade free segment simply by dividing distance traveled by time and then this velocity is divided by the constant target velocity (Fig. 1). The mean of these gain values is the mean gain of pursuit and this deﬁnition was used in many studies in Table 2. Two studies used the median instead of the mean that is less inﬂuenced by outliers (Lencer, Trillenberg-Krecker, Schwinger, & Arolt, 2003; Smyrnis et al., 2007). In a number of studies the gain values at each segment were multiplied by the corresponding time and then summed. The sum of these products was then divided by the total time duration of these segments to derive a time weighted pursuit gain. In some studies that used a combination of sinusoid and constant speed stimuli the gain was computed only for the constant speed stimulus (Clementz et al., 1990; Siever et al., 1994; Sweeney, Hass, Li, & Weiden, 1994; Sweeney et al., 1993). In all these studies then pursuit gain deﬁnition was similar. In contrast studies of sinusoid target motion use deﬁnitions of pursuit gain that vary dramatically and the use of the term gain in these studies is rather diverse. Some studies have used a small window at the peak of target velocity in order to deﬁne a peak eye velocity that they then divided by peak target velocity to produce a gain value. The mean of these values was then measured (Flechtner, Steinacher, Sauer, & Mackert, 2002; Gooding et al., 2000; Holahan & O’Driscoll, 2005; Lenzenweger & O’Driscoll, 2006; Levy et al., 2000). Some studies have used the time domain gain. First the saccades and blink artifacts were removed from the signal and it was corrected for phase lag differences. Then the ratio of the eye velocity to target velocity at each point of the record was computed excluding all points with target velocity below 0.25 deg/s. ﬁnally the median of these ratios was computed (Katsanis, Taylor, Iacono, & Hammer, 2000; Katsanis et al., 1998). Others have used even more diverse deﬁnitions of pursuit gain for sinusoid stimuli like for example a model of pursuit velocity and gain (Karoumi et al., 2001; Nkam et al., 2001) or a ratio of the fundamental frequency of the eye movement signal to the frequency of the target signal after a transformation of the pursuit signal to exclude saccades and artifacts (Ross, Thaker et al., 1996; Ross et al., 1996) or a ratio of the amplitude of a ﬁtted sinusoid corresponding to the eye motion to the target sinusoid (Boudet et al., 2005). Another complication for deﬁning a common ground on the computation of the pursuit gain is the deﬁnition of saccadic eye movements and artifacts in the pursuit record so that they would

7

be removed. In this respect no study uses the same exact criteria as the other. In order to deﬁne saccades some studies have used visual inspection (Calkins et al., 2001; Clementz et al., 1990; Green et al., 2000; Levy et al., 2000; Siever et al., 1994; Versino et al., 1993). Others have used absolute velocity criteria, for example eye movement velocity exceeding 35 deg/s (Hong et al., 2003) or another such absolute value (Flechtner et al., 2002; Holzman, O’Brian, & Waternaux, 1991; Litman et al., 1991). The problem with such a deﬁnition is that it is difﬁcult to apply it in cases were pursuit at higher speeds of 30–50 deg/s is included in the task because eye velocity for pursuit would be equal or larger than the exclusion velocity for saccades. Some studies used absolute acceleration instead of velocity (Sweeney et al., 1993; Sweeney et al., 1994; Thaker, Ross et al., 1996; Thaker et al., 1996). Others have used more complex criteria such as both absolute velocity and acceleration (Avila et al., 2002) or amplitude and absolute velocity (Ettinger et al., 2003) or duration, velocity and acceleration (Nickoloff et al., 1991; Olincy, Johnson, & Ross, 2003; Ross, Olincy, Harris, Sullivan, & Radant, 2000; Ross, Olincy, Zerbe, & Radant, 2001; Ross et al., 1999). Finally others have used relative criteria such as increases in velocity or acceleration or both above a previous level (Smyrnis et al., 2007). Also in many studies exclusion of segments of pursuit was performed before the calculation of mean gain. For example a common practice for studies that used triangular or trapezoid signals was to exclude the ends of the trapezoid or triangular motion were the target either stopped (trapezoid) or changed direction (triangular). The segments that were excluded varied considerably from study to study ranging from a small period of time excluded at the end points to large periods excluded in order for a small period of pursuit around the central position of the target to be included in the computation. These differences could result in large variation of the measured gain since it is well known that close to the directional shifts the pursuit of the target becomes more erratic and intrusion saccades are more frequent (Leigh & Zee, 1991). Finally as was mentioned before, the number of cycles of continuous pursuit used to calculate the mean gain varied considerably among different studies as shown in Table 2 introducing the problem of data aggregation. One could still argue that the obvious lack of standardization in pursuit test design and measurement procedures has little importance for the outcome. Thus the gain of pursuit could be a robust measurement that does not have a signiﬁcant variation thus standardization of test and parameter is really not important. Fig. 3A shows a plot of the mean gain for the groups of normal individuals included in all studies of Table 2 that used a trapezoid, triangular or step ramp stimulus and target speed range of 5–30 deg/s. The plot presents pursuit gain as a function of target speed and the regression line of gain versus target speed. The ﬁlled circles show the corresponding gain values from the ASPIS sample (Smyrnis et al., 2007). A very large variation in mean gain from study to study is evident in this ﬁgure that is unrelated to differences in target speed. Indeed although the regression of gain to target speed was signiﬁcant (F1,33 = 4.2, P = .048) the proportion of variance explained by the regression was 11% (r2 = .11). The large variability that remains unexplained could be the result of lack of standardization or could reﬂect an inherent inter-subject variability for this parameter. This issue could be resolved if the lack of standardization would be eliminated as a possible cause of variability. The appearance of saccadic eye movements during pursuit has led to the introduction of saccadic frequency as another parameter measuring pursuit performance. Many studies have used the frequency of all saccadic eye movements during pursuit as another indicator of pursuit integrity. This parameter though presents the same major conceptual problem as the global parameters measuring pursuit performance. Eye movements during pursuit can be

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS 8

N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

Fig. 2. (A) Saccade (red line) towards a visual target (target displacement shown in dotted black line). The latency (SL) is the time from target displacement to saccade onset. The amplitude of the initial saccade is marked with A1 and the amplitude of a secondary saccade is marked with A2. The gain is an accuracy measure deﬁned as the ratio of the saccade amplitude to the amplitude of the target. The thin black line corresponds to the instantaneous speed record which was derived from the position record by numeric differentiation. The maximum speed for the ﬁrst saccade (S1) and the secondary saccade (S2) are marked on the instantaneous speed record. The duration of the ﬁrst saccade (D1) and the secondary saccade (D2) are also marked on the instantaneous speed record. (B) Correct antisaccade (red line) that brings the eye in the opposite direction from the direction of target displacement (black dotted line). The latency (AL) and the amplitude of the antisaccade are marked on the position record while the maximum speed (S) and duration (D) are marked on the instantaneous speed record (thin black line). (C) Error pro target saccade (blue line) followed by a correct antisaccade (red line). The latency of the error prosaccade (EL), the amplitude of the error prosaccade A1, the latency of the correction antisaccade after the error (CL) and the amplitude of the antisaccade A2 are marked. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this paper.)

very different in nature and purpose (Abel & Ziegler, 1988). The catch up saccades are small amplitude eye movements in the direction of target motion that bring the eye close to the target when pursuit gain is low. These are very different in nature and purpose from the intrusive saccades that are performed when pursuit gain is normal and are related to an intrusion of the saccadic eye movement system into pursuit. Thus measuring the total saccade frequency is putting together qualitatively different parameters. Most studies in Table 2 report on different types of saccadic eye movements during smooth pursuit. A detailed analysis of these deﬁnitions leads to the disturbing conclusion that there is a large heterogeneity among different studies. Actually this heterogeneity is so large that virtually every study differs from all others in respect to one or more of these deﬁnitions.

Fig. 3. (A) Plot of the mean gain for the groups of normal individuals included in all studies of Table 2 that used a trapezoid, triangular or step ramp stimulus and target speed range of 5–30 deg/s. The plot presents pursuit gain as a function of target speed and the regression line (black heavy line) of gain versus target speed. The ﬁlled circles show the corresponding gain values from the ASPIS sample (Smyrnis et al., 2007). (B) Plot of the mean CUS frequency of groups of normal individuals that was reported in some of the studies of Table 2 versus target speed ranging from 5 to 30 deg/s. The black heavy line shows the regression of CUS frequency versus target speed. (C) Plot of the mean AS frequency of groups of normal individuals that was reported in some of the studies of Table 2 versus target speed ranging from 5 to 30 deg/s. The black heavy line shows the regression of AS frequency versus target speed.

The catch up saccades (CUS) are deﬁned by most studies as saccades in the direction of target motion that start behind the target and bring the eyes close to the target. Some studies also deﬁne quantitatively how close the eye should get to the target for example reducing the positional error to less than 1 deg (Radant & Hommer, 1992) or reducing the positional error by more than 50% (Ross et al., 1998; Ross et al., 1999; Ross et al., 2000; Ross et al., 2002). Others do not add the criterion of positional error reduction but other criteria like for example the criterion that the saccade should be less than 5 deg in amplitude and should be preceded and followed by pursuit gain greater than 0 (Calkins et al., 2001; Friedman, Jesberger, & Meltzer, 1991; Katsanis et al., 1998). Some studies use unique criteria like for example Campion et al. (1992) who deﬁned CUS as all saccades that are not deﬁned as square wave jerks or Roy-Byrne et al. (1995) who deﬁne CUS as all saccades that decrease positional error to less than 1 deg, or even Len-

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

zenwenger and O’Driscoll (2006) who deﬁne CUS as all saccades that are less than 5 deg in amplitude and are preceded and followed by pursuit. In a very detailed study on saccadic eye movements in pursuit Ross et al. (2001) showed that different deﬁnitions of CUS in pursuit resulted in very different estimates of their frequency. They deﬁned CUS as those saccades that start behind the target and reduce positional error by more than 50%. If they added a criterion of pre- and post-saccadic slowing for 40 ms, the frequency of CUS in their normal control group dropped from 0.8 to 0.08 Hz. If instead they used a criterion of 300–400 ms of pre- and post-saccadic slowing the frequency dropped to 0.01 Hz. This study then showed convincingly that different deﬁnitions of CUS result in very different estimates of their frequency. In Fig. 3B the mean CUS frequency of groups of normal individuals that was reported in some of the studies of Table 2 is plotted for target speeds ranging from 5 to 30 deg/s. The regression of target speed to CUS frequency showed that a signiﬁcant but small proportion of the variability in these means was explained by the different target speeds used (r2 = .13, F1,30 = 4.48, P < .05). The major part of variability though in CUS frequency remains unexplained. This very large variability could be the result of true variability between subjects and populations or could be the result of differences in the task and parameter deﬁnition as previously described, pointing again to my previous argument for the lack of standardization. The same disturbing conclusion of heterogeneity of deﬁnitions can be drawn for another type of saccadic eye movements observed during smooth eye pursuit that are named anticipatory saccades (AS). Abel and Ziegler (1988) deﬁned AS those eye movements that are in the direction of motion, take the eyes ahead of target, have amplitudes of 5 deg or more and are followed by a time interval during which the eyes virtually remain stationary or go back to the target with a saccade in the opposite direction. This last saccade was named by some authors as a back up saccade. Some authors reﬁned these criteria even more. For example Clementz et al. (1990) added to the previous deﬁnition the prerequisite that the AS must travel at least half of this distance ahead of target and that after the saccade the eyes must have a slow velocity for at least 250 ms. Holzman, O’Brian, and Waternaux (1991) required that post-saccadic velocity is zero for 250 ms while Radant and Hommer (1992) required a 50 ms interval with a velocity of less than 5 deg/s. Other studies do not use the amplitude or the post-saccadic slowing criteria for AS. Nickoloff et al. (1991) deﬁned AS the saccades that start and end ahead of target or increase positional error by 100% or more. Sporn et al. (2005) deﬁned AS those saccades that increase positional error by more than 4 deg and Spengler et al. (2006) as those saccades that move the eyes ahead of target. Others combine all the previously mentioned criteria together (Olincy et al., 2003). Finally Ross et al. (2001) computed AS frequency using three different criteria. The ﬁrst criterion was that AS should start and end ahead of target or start behind and end at least twice the distance from the target and be larger than 1 deg. This deﬁnition resulted in AS frequency of 0.4 Hz for the normal control group. The second deﬁnition added to the previous one the requirement of a post-saccadic slowing of 50 ms at a velocity equal or less than 50% of target velocity. This deﬁnition resulted in AS frequency of 0.11 Hz in the control group. Finally the last definition increased post-saccadic slowing duration to 300–400 ms. The resulting AS frequency dropped to 0.03 Hz. It should be noted here that this study as well as the study on normative data of Ettinger et al. (2003) deﬁned AS using an amplitude criterion of 1 deg instead of the 5 deg criterion introduced by Abel and Ziegler (1988). Finally Sweeney et al. (1993) have shown that the frequency of AS decreased signiﬁcantly when the attention was manipulated in the monitoring task.

9

In Fig. 3C the mean AS frequency of normal individuals that was reported in some of the studies of Table 2 is plotted for target speeds ranging from 5 to 30 deg/s. The regression of target speed to AS frequency showed that this variability was not explained by the different target speeds used (r2 = .05, F1,29 = 1.49, P = .23). Again this large inter-study variability points in my opinion to the need for measurement standardization as a possible source of variation that could be eliminated. The same heterogeneity in the literature can be found for the deﬁnition of square wave jerks (SWJ). Abel and Ziegler (1988) deﬁned SWJ as pairs of small saccades 0.5–3 deg in amplitude separated by a 200–400 ms interval during which smooth eye pursuit is not interrupted. Other authors added that SWJ should be of opposite direction. Others do not use the criterion of uninterrupted pursuit in between the SWJ, or preceding and following the SWJ. Different studies used different time intervals between the two saccades of the SWJ pair. The mean SWJ frequency of normal individuals that was reported in some of the studies of Table 2 ranged from 0 to 5.16 Hz and this variability was not explained by the different target speeds used (regression of target speed on SWJ frequency: r2 = .001, F1 = 0.13, P = .9). Finally some studies measured the frequency of occurrence of other types of saccades during pursuit such as back up saccades (Calkins et al., 2001) or large amplitude SWJ (Clementz et al., 1990). Others used completely different deﬁnitions of saccadic types in pursuit (Siever et al., 1994). A small number of studies have investigated time dependent changes of speciﬁc pursuit performance measures. Holzman, O’Brian, and Waternaux (1991) used a sinusoid plus triangular stimulus to measure pursuit gain and saccade frequency in patients and controls in two time points 1–3 weeks apart. They found no difference for the gain and frequency of total saccades, CUS, AS and SWJ in the control group at the two time points. Ettinger et al. (2003) tested for within session and between session practice effects in normal volunteers by comparing the mean scores between sessions or within a single session of pursuit performance. They used a triangular stimulus at four different speeds (see Table 2) and measured mean pursuit gain CUS and AS. Between-session effects were not signiﬁcant for all variables except for the frequency of AS at 36 deg/s where participants made fewer AS at retest. Also within session effects were not signiﬁcant except for a linear reduction of CUS frequency art 12 and 36 deg/s and a linear increase of AS frequency at 36 deg/s. These few studies then suggest that there might not be signiﬁcant practice effects for most smooth eye pursuit parameters although a rigorous treatment of the issue is not present in the literature. A few studies have measured test–retest reliability of smooth eye pursuit performance and one study has measured internal consistency (Ettinger et al., 2003). Versino et al. (1993) measured test– retest reliability of pursuit gain in a group of normal individuals tested at three different time points the ﬁrst two differing by 1week and the second from the third differing by 4–7 months. The authors used the intra class correlation coefﬁcient for measuring reliability. The problem with this study is the deﬁnition of the gain measure that differs from the one that is most widely used in the literature. Gain was deﬁned as the ratio of amplitude of smooth eye tracking after excluding saccades to amplitude of total tracking while the most widely used deﬁnition of gain is the mean ratio of eye speed to target speed from all segments of pure pursuit after excluding saccades (see Fig. 1). The authors claim that their gain measurement was different from the most widely used one because their deﬁnition corresponds to the integral of eye and target velocity signals instead of a mean of selected values. Furthermore the authors used a target velocity gain relationship expressed as an equation that was ﬁtted to each subject’s gain data. The slope and intercept coefﬁcients of this equation were then the variables used

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS 10

N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

for each subject to test reliability that was 0.71 for the slope and 0.87 for the intercept. Gooding et al. (1994) used Pearson correlation coefﬁcients and reported signiﬁcant (P < .05) test–retest reliability for the root mean square error in pursuit in groups of schizophrenia patients (0.68) bipolar patients (0.51), depressive patients (0.48) and controls (0.57) that was measured in two time points separated by 9.5 months. Roy-Byrne et al. (1995) reported on test–retest reliability of smooth eye pursuit in a sample of normal volunteers. All subjects were tested four times at weekly to bimonthly intervals twice in the morning and twice in the afternoon. Intra class correlation coefﬁcients were computed for morning and afternoon pairs of measurements as well as for the average of morning and afternoon pairs to assess diurnal variation. The stimulus was a trapezoid and the speed was 10 deg/s. The deﬁnition of gain was the most widely used one of mean gain weighted for time. The deﬁnition of catch up saccades was all saccades in the direction of target motion that decreased positional error. The authors used a deﬁnition of intrusive saccades that was close to the deﬁnition of AS used in the literature. These were all saccades in the direction of target motion that increased positional error. Finally they deﬁned SWJ as a pair of saccades of the same amplitude with an intersaccadic interval of ongoing pursuit of 50–500 ms. The gain for the intervening pursuit should be less than 0.6. All intra class correlation coefﬁcients measuring test–retest reliability for pursuit mean gain and frequency of total saccades, CUS, AS and SWJ were signiﬁcant (P < .05) and ranged from 0.71 to 0.97. Flechtner et al. (2002) used a sinusoid pursuit stimulus and computed mean gain in a center window of 800 ms for a group of schizophrenia patients and a control group. They also measured CUS, AS and SWJ as well as back up saccades using deﬁnitions very close to those of Abel and Ziegler (1988). A retest was performed in the control group after 28 days. The test–retest reliability using the intra class correlation coefﬁcient was signiﬁcant for all measures (P < .05) and the values ranged from 0.71 to 0.9. Ettinger et al. (2003) performed the most comprehensive study for reliability of smooth eye pursuit measures in normal volunteers. Their deﬁnition of CUS was similar to that of Flechtner et al. (2002) but they used a very different amplitude criterion for AS that is saccades greater than 1.5 deg. They tested individuals at two time points 2 months apart on average (range 38–105 days) and measured test–retest reliability using Pearson and Intra Class Correlation coefﬁcient. They also divided each session into 4 segments and measured internal consistency among the different section segments using the Cronbach a coefﬁcient. They found non-signiﬁcant test–retest reliability for pursuit gain both using Pearson correlation coefﬁcient and Intra Class correlation (ICC) coefﬁcient for pursuit at 12 deg/s (Pearson r = .11, P = .64, ICC = 0.1, P > .1) and 24 deg/s (Pearson r = .31, P = .17, ICC = 0.31, P > .1) while for higher speeds of 36 and 48 deg/s both coefﬁcients were signiﬁcant (for 36 deg/s: Pearson r = .81, P < .001, ICC = 0.77, P = .01; for 48 deg/s: Pearson r = .71, P < .001, ICC = 0.71, P < .01). The coefﬁcients for AS were all signiﬁcant (P < .05) ranging from 0.56 to 0.79.The coefﬁcients for CUS were not signiﬁcant for the target speed of 12 deg/s (Pearson r = .42, P = .07, ICC = 0.34, P > .1) but were signiﬁcant for higher speeds (P < .05) ranging from 0.58 to 0.64. Internal consistency was very high for all measures (a P 0.73) except for CUS at 12 deg/s and AS at 48 deg/s. Finally Lenzenweger and O’Driscoll (2006) compared the peak pursuit gain and CUS (deﬁned as saccades in the same direction as the target preceded and followed by pursuit and with amplitudes of less than 5 deg) among three sessions of sinusoid pursuit in normal volunteers and found highly signiﬁcant (P < .05) Pearson correlation coefﬁcients among all three sessions (gain: 0.7, 0.71, CUS: 0.84, 0.98). A ﬁnal question concerns the inter correlations among different measurement parameters of smooth eye pursuit. Friedman et al. (1991) used a very elegant model that quantitatively associates

the number and amplitude of catch up saccades to pursuit gain to conﬁrm that gain is inversely analogous to the catch up saccade frequency and amplitude. Other studies have also explored the correlations between different pursuit measures of performance. Clementz et al. (1990) found a non-signiﬁcant correlation of pursuit gain with the frequency of SWJ but they found a signiﬁcant correlation of gain and AS frequency. Abel, Friedman, Jesberger, Malki, and Meltzer (1991) found a signiﬁcant correlation of pursuit gain with CUS at both 5 deg/s and at 20 deg/s. Holzman, O’Brian, and Waternaux (1991) found no signiﬁcant correlation of gain to CUS frequency. Radant and Hommer (1992) found a non-signiﬁcant correlation of CUS frequency with pursuit gain and a signiﬁcant correlation of AS frequency with gain. They also found non-significant correlation of SWJ with gain. Sweeney et al. (1993) found non-signiﬁcant correlation of pursuit gain with AS frequency. Siever et al. (1994) used another way of dividing saccades (large– small saccades in the direction of the target, large–small saccades in the opposite direction of the target) in smooth pursuit and found no signiﬁcant correlation among any type of these saccades and pursuit gain. Katsanis et al. (1998) found a signiﬁcant correlation between saccade frequency and time domain gain but they did not dissociate among different types of saccades. Overall then these results are rather inconclusive with some studies reporting signiﬁcant correlations between different parameters while others reporting non-signiﬁcant correlations between the same parameters. In summary there is little information in the literature concerning the test–retest reliability of speciﬁc quantitative pursuit performance measures as well as information regarding within and between session practice effects and even less information regarding internal consistency of these measures. There is also some evidence for signiﬁcant correlations among certain pursuit parameters especially gain and catch up saccades.

6. Saccade, antisaccade task: Test procedure deﬁnition Table 3 presents a series of publications reporting on saccade, antisaccade task performance in patients suffering from schizophrenia or in individuals within the spectrum of schizophrenia and compares their performance with that of normal controls. Some studies also report on the performance of ﬁrst degree relatives of schizophrenia. These studies were found by a search in the Medline using keywords ‘‘schizophrenia”, ‘‘saccade” and ‘‘antisaccade” from 1990 until the summer of 2007. One study reporting on antisaccade performance in the social-emotional disorder was also included (Manoach, Weintraub, Daffner, & Scinto, 1997). A few studies were added that report on saccade, antisaccade task performance in healthy adult samples (Ettinger et al., 2003; Fischer, Biscaldi, and Gezeck (1997); Green et al., 2000; Roy-Byrne et al., 1995). The last entry in the table is our sample of the Athens Study for Psychosis proneness and Incidence of Schizophrenia (ASPIS). The results concerning the antisaccade task in this sample have been published in Evdokimidis et al. (2002). The saccade task was performed by 1100 individuals in this sample and the results were published in Constantinidis et al. (2003). Studies that did not include a control sample and studies that used very different task procedures from what is commonly used as tests of saccade–antisaccade function were excluded. Studies where no values of the outcome parameters were reported for the control group as well as studies that reported new results on the same data set were also excluded. The studies that were selected cover the large heterogeneity of measurement methods, task deﬁnitions and parameters measured in the saccade and antisaccade task that were used in the psychiatric literature in the last two decades.

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS 11

N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx Table 3 List of publications on saccade, antisaccade performance Year

Authors

No of subjects

Center dur

Overlap (ms)

Target dur

Feedback

Amplitude

Mirror

No trials

1990 1993 1994 1994 1995 1995 1995 1996 1997a 1997 1997

Fukushima et al. Rosse et al. Fukushima et al. Clementz et al. Roy-Byrne et al. Sereno and Holzman* Crawford et al. Thaker et al. Fischer et al.* Manoach et al. McDowell and Clementz*

24 12 9 Young, 21 older 27 8 14 31 51 289 6 12 15 50 15 14 38 17 94 89 37 14 20 109 38 25 20 30 21 34 30 16 23 12 49 31 17 18 24 8 29 21 Young, 31 older 15 41 39 41 15 22 115 14 24 15 195 2006

4–6 1.50 4–6 2–3 2 0.8 0.8 0.5–2 1 ? 2–2.5 2–2.5 ? 2–2.5 1.4–2.4 0.8 1.93 2–3 0.6 1.5–2.5 3.5 3–3.5 2–3 2–3 2–2.5 0.8 0.6 1.4–2.4 2–4 2.5 2 0.8 3 2 1–2 1–2 1 1 0.8–1.2 0.8–1.2 2–4 2–3 0.8 0.6 3–5 1 1–2 1–2 1–1.5 0.4–1.6 2–3 2.4–3.6 1–2

0

0.5 Until EM 0.5 Until EM 1 Until EM 1 2 1 0.2 1 1 0.1 3.5 0.75 1 0.1 0.5 0.9 2 1 Until EM 1.5 2 1.3 2 0.9 0.75 0.5 1 Until EM 2 2 2 1 2 1 1 2 ? 0.5 Until EM 3 0.9 1.5 1 1 1 Until EM 0.5–1 1.2 0.8 1.5

No No No No Yes No No No No No Yes Yes No No Yes No Yes No No No No Yes No No Yes No No Yes No No No No Yes No No No No No No No No No No No Yes No No No No No No Yes No

12 7.5 8, 12, 24 15

Yes Yes Yes No Yes Yes Yes No No No Yes Yes No Yes Yes Yes Yes Yes No Yes Yes Yes Yes No Yes No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No No Yes Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes No

20 50 3 20 20 20 40 48 ? 200 Mixed pro and anti for 2 min 6 20 4 20 45 40 60 24 60 s 60 48 60 2 80 54 20 20 60 48 48 60 50 24 52 mixed with Pro and 2 26 48 4 12 14 60 90 16 16 12 72 3 30 60 80 24 48 36 16 60 60 4 26 20 3 81 3 20 90

1998 1998 1998 1998 1998 1999 1999 2000 2000a 2000 2001a 2001b 2001 2001 2001 2001 2001 2002 2002 2002 2002 2003 2003 2004 2004a 2004b 2004 2005 2005 2005 2005 2005 2005 2005 2005 2006 2006 2006 2007 2002

*

O’Driscoll et al. Maruff et al.* Karoumi et al. Crawford et al. Ross et al. McDowell et al.* Gooding* Thaker et al. Klein et al.* Green et al. Curtis et al. Curtis et al. Brenner et al.* Broerse et al.* Gooding and Tallent* Karoumi et al. Nkam et al. Hutton et al. Manoach et al. Broerse et al. Burke and Reveley Brownstein et al.* Ettinger et al.* Smyrnis et al.* Ettinger et al.* Ettinger et al.* Larrison-Faucher et al. Holahan and O’Driscoll Boudet et al. Reuter B et al.* MacCabe et al.* Gooding et al.* Harris et al. Kumari et al.* Ettinger et al.* Polli et al.* Spengler et al.* Reuter et al.* Radant et al.* Evdokimidis et al. ASPIS*

500 0 0 0 0 0 0 200 0 200, 0, 200 0, 200 0 0 0 0 75 200 0 0 0 0 0 200, 0 200 0 0 0 0 0 0 0 0 0 0 200, 2000 0 0 160, overlap 200, 0, overlap 0 * 0 0 0 0 0 0 0 0 0 200 200 0

The stimuli for the saccade and antisaccade tasks were identical in most studies except for a few studies for which the saccade task involved random lateral target displacements to the right or to the left of the previous target position and thus the target did not always step from the central position to the periphery as is the case for the antisaccade task (Clementz, McDowell, & Zisook, 1994; Constantinidis et al., 2003; Maruff, Danckert, Pantelis, & Currie, 1998; Roy-Byrne et al., 1995). Finally some studies also used a memory saccade task (Broerse, Crawford, & den Boer, 2002; Crawford, Haeger, Kennard, Reveley, & Henderson, 1995; Fukushima, Fukushima, Morita, & Yamashita, 1990; Hutton, Joyce, Barnes, & Kennard, 2002; Ross et al., 1998; Ross et al., 1998) and some used a predictive saccade task (Crawford et al., 1995; Spengler et al., 2006). The third column of Table 3 shows the number of normal controls that participated in each study. The other columns in the table present the different task parameters that deﬁned the antisaccade

12 7.5, 15 ? 4 8 10, 20 8, 16 12 7.5, 10, 15 10, 20, 30 7.5, 15 5, 15 8, 16 4, 8, 12 5, 15 5 12, 15, 18 10 10 8, 16 7.5, 15 4, 8, 12 5, 15, 25 15 7.5, 15 20 7.5, 15 7.5, 15 8 6,12 2, 4, 6, 8, 10 15 15 7.5 10 15 8 15 4, 8, 12 8, 16, 24 15 6, 12 6, 12 20 12 12.5 10, 15 2, 4, 6, 8, 10

task in these studies such as the central ﬁxation duration (column 3) and the peripheral target duration (column 5). The fourth column in Table 3 shows whether the peripheral target presentation onset occurred before the central ﬁxation offset introducing a gap (negative values) or synchronously with the central ﬁxation offset (zero value) or ﬁnally after the central ﬁxation offset introducing an overlap between the two (marked with positive values). The sixth column shows whether a target appeared at the correct antisaccade location for error feedback and the seventh presents the amplitudes in degrees of the locations for the peripheral target presentation. The eighth column shows whether the instruction to the subjects was to perform an antisaccade to the exact mirror location from the peripheral target or just to look in the other direction from that of the target. The last column shows the number of antisaccade trials presented. A few task parameters were common among different studies, for example the variable central ﬁxation duration in order to avoid time prediction (see glossary)

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS 12

N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

and the unpredictable direction of the peripheral target in order to avoid prediction of the metrics of the upcoming eye movement. Also most studies used a block design for administering saccades and antisaccades. Still a large variation in task design is obvious in this table. In the following sections I will discuss the potential impact of these task design variations on the main antisaccade task parameters measured. 7. Saccade, antisaccade task: Performance measures, practice effects and reliability The main outcome parameter in the antisaccade task is the percentage of error prosaccades (PE; see glossary) indicating the efﬁciency of the ability of the individual to inhibit a visually triggered saccade to the target in order to perform a correct antisaccade (Massen, 2004; see Hutton in this issue). The same measure of inhibition has also been used in the memory saccade task and the main arguments that will be made for PE in the antisaccade task are also valid for the PE in the memory saccade task. Although the operational deﬁnition of PE in the antisaccade task seems straight forward there is a crucial difference in deﬁnition among different studies that could result in very different estimates of PE. Twenty-two studies in Table 3 include to the count of erroneous prosaccades in the antisaccade task all saccades that start after the peripheral target presentation and are directed towards the target. The other twenty-nine studies in Table 3 marked with an asterisk next to the authors introduce the additional prerequisite that these saccades start after a minimum latency from target presentation. This additional criterion actually excludes from the calculation of antisaccade errors all anticipatory saccades that are saccades that are not visually guided but preprogrammed. Does this criterion have any effect on PE? One could ﬁrst argue that the two deﬁnitions differ conceptually since we know that saccades made to a predetermined target location (in this case the subject decides to look in one direction or the other at chance) and visually guided saccades do not share the same neurophysiological mechanisms for their production (Leigh & Zee, 1991). Moreover the inclusion of predictive saccades (see glossary) has a signiﬁcant effect on PE by increasing the absolute number of errors counted since most errors in this task have fast latencies. We have shown in a model of the effects of different parameters on PE in the antisaccade task that the latency of the ﬁrst saccade after target presentation is the single most signiﬁcant predictor of whether an error will occur and that there are signiﬁcant changes in the predicted percentage of errors in the latency range of 100– 180 ms (Smyrnis et al., 2002). Variations in the reported PE can also be introduced by differences in the cut off point for predictive saccades that ranges from 70 to 130 ms for different studies. Klein and Fischer (2005b) showed that the percentage of errors in the antisaccade task with latencies at the express range (80–120 ms) can be dissociated from the percentage of errors with latencies above the express range. The ﬁrst variable was subject to only minor developmental changes from childhood to young adulthood while the second showed substantial developmental changes. Klein, Heinks, Andresen, Berg, and Moritz (2000) used three error rates in their study, one for anticipatory error saccades (latency <80 ms), another for express error saccades (80–130 ms) and ﬁnally one for regular error saccades (>130 ms). Finally variations in the percentage of directional errors in the antisaccade task can be introduced by the deﬁnition of minimum amplitude required to detect an error prosaccade (Mokler & Fischer, 1999). I will now turn to differences in task parameters that could have an effect on the PE in the antisaccade task. The fourth column of Table 3 presents the duration of central ﬁxation before the target appearance in seconds that varied considerably from study to

study. We have examined the effect of the duration of ﬁxation interval before the execution of antisaccades and found that it had a signiﬁcant effect for all task parameters measured (Smyrnis et al., 2002). In particular the PE increased with decreasing ﬁxation interval from 2 to 1 s. The manipulation of central ﬁxation offset is known to result in large changes in all behavioral outcomes of antisaccade performance. Fischer and Weber (1997) varied the time of the gap between central ﬁxation offset and target appearance and showed that the PE increased with decreasing gap interval up to a maximum when the gap was 200 ms and then decreased again until an overlap occurred. McDowell and Clementz (1997) conﬁrmed this observation using three conditions of 200 ms gap, zero gap and overlap of 200 ms and suggested the use of the overlap condition in studies with patients suffering from schizophrenia. We also observed the same effect using a 200 ms gap and overlap condition Smyrnis et al. (2004) and Holahan and O’Driscoll (2005) using a 200 ms gap, zero gap and overlap conditions. A number of studies as shown in column 5 of Table 3 have used either a gap (with different gap intervals ranging from 500 to 75 ms) or an overlap paradigm, thus affecting the resulting PE. Larrison-Faucher, Matorin, and Sereno (2004) used a task design of interleaved gap of 160 ms and overlap trials and report on the average PE for both types of trials. Reuter, Rakusan, and Kathmanna (2005) used another task design where the center ﬁxation offset was not followed by the presence of the target (zero gap) but the subjects had to wait for a tone presented either simultaneously with the peripheral target or after a delay of 400–600 ms or after a delay of 900– 1100 ms and these increasing delays resulted in a reduction of the PE. Two studies in Table 3 have also used spatial cues that were presented during central ﬁxation before the appearance of the peripheral target (Polli et al., 2006; Rosse, Schwartz, Kim, & Deutsch, 1993). The appearance of a valid spatial cue indicating the location of the upcoming target 100 ms before the execution of an antisaccade was shown to increase the PE (Fischer & Weber, 1996). Column 6 of Table 3 shows that the time of presentation of the peripheral target also varied among different studies. To my knowledge it is not known whether there is an effect of target presentation interval on antisaccade task parameters. Column 7 of Table 3 shows that some studies used a feedback procedure in which after an amount of time the correct target location for the antisaccade was turned on so that subjects could correct the amplitude of their antisaccade. Although it is known already from Hallet (1978) that feedback does not preclude subjects from making an error in the antisaccades, I know of no systematic study that assesses whether antisaccades parameters such as PE or others differ quantitatively in a condition with feedback compared to a condition with no feedback. Column 8 of Table 3 shows the different target amplitudes used in the different studies that range from 2 to 30 deg. Fischer and Weber (1997) examined the effect of target amplitude on PE and reported that increasing amplitude from 1 to 12 deg resulted in an increase in PE. We observed a decrease in PE with increasing amplitude from 2 to 10 deg (Smyrnis et al., 2002). McDowell, Myles-Worsley, Coon, Byerley, and Clementz (1999) also found a signiﬁcant decrease in PE when amplitude increased from 8 to 16 deg. Finally Thaker et al. (2000) present their results on PE for different target amplitudes were the same trend of decrease in PE with increasing amplitude is present although the authors do not report on statistical tests of these differences. Column 9 of Table 3 presents another task difference among the different studies namely the use of the instruction to perform a mirror antisaccade or to perform a saccade in the opposite direction from the stimulus. We investigated the effect of this instruction in a recent study on normal volunteers and showed that the

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

PE for mirror antisaccades was not different from that of simply directional antisaccades (Evdokimidis, Tsekou, & Smyrnis, 2006). Another parameter that also does not seem to have an effect on PE is the direction of the target displacement. We have investigated the effect of direction on PE in the antisaccade task in our ASPIS sample and found that there was no signiﬁcant difference of PE for targets appearing to the left versus targets appearing to the right visual ﬁeld (Constantinidis et al., 2003). The same conclusion was also reported by Clementz et al. (1994) while Thaker et al. (2000) report separately the PE for left and right target displacements and do not report a test for their differences. Finally some studies in Table 3 used a mixed saccade and antisaccade task design instead of administering saccades and antisaccades in separate blocks of trials (Manoach et al., 1997, 2002). The effect of this manipulation on PE is discussed by Hutton in his review on cognitive processes involved in reﬂexive and volitional saccades. The different task designs and deﬁnitions of error prosaccades among different studies might or might not result in the introduction of variation. Thus the same question can be asked as was done for the review of the smooth eye pursuit studies, namely how variable is the mean error rate from study to study? This is graphically shown in Fig. 4A that presents the mean PE for normal controls for the subset of studies in Table 3 that did not include anticipatory saccades in the PE and used a zero gap paradigm. The black circle represents the PE for the ASPIS sample (Evdokimidis et al., 2002). There is a huge variability of mean error rate among these different studies reporting on normal individuals. It is very interesting to contrast this ﬁgure with the ﬁgure showing the different PE values observed in a multi center measurement of antisaccade task performance reported by Radant et al. (2007). This study used an overlap antisaccade task (thus it is excluded from Fig. 4A) that was administered in seven different sites by different research groups. An effort was made to standardize task and recording procedures as well as measurement and analysis among the different sites. This resulted in a very different picture as shown in their Fig. 1 where all reported PEs for normal individuals varied very little around a mean of 18%. This difference in variability suggests that indeed task standardization might have a large impact on the variability of PE from study to study. After a directional error in the antisaccade task a correction antisaccade is performed in the vast majority of trials (see Fig. 2). This type of error has been termed a corrected error to distinguish it from the rare case where a correction antisaccade is not performed resulting in a uncorrected directional error (Fischer, Gezeck, & Hartnegg, 1997). Some studies have also measured the proportion of corrected errors in the antisaccade task (Ettinger et al., 2004; Klein & Fischer, 2005a; Spengler et al., 2006). These studies conﬁrmed that the vast majority of errors for all subjects in the antisaccade task are corrected. In the psychiatric literature the most widely used measure after the PE in the antisaccade task is the latency of antisaccades (AL). Other latencies measured in the antisaccade task are the latency of erroneous prosaccades as well as the latency of correction after an error has occurred (Fischer, Gezeck et al., 1997, see Fig. 2). Few studies reported on error prosaccade latencies (Boudet et al., 2005; Broerse et al., 2002; Clementz et al., 1994; Curtis, Calkins, & Iacono, 2001; Ettinger et al., 2003; Evdokimidis et al., 2002; Gooding, 1999; Gooding, Shea, & Matts, 2005; Gooding & Tallent, 2001; Klein & Fischer, 2005a; MacCabe et al., 2005; McDowell & Clementz, 1997; O’Driscoll, Lenzenweger, and Holzman, 1998; Radant et al., 2007). Other measures include the standard deviations of the mean for all latencies in the saccade and antisaccade task (Evdokimidis et al., 2002; Fischer, Gezeck et al., 1997; Klein & Fischer, 2005a). Fischer, et al. (1997) also used a division of the number of saccades according to the latency range to measure

13

Fig. 4. (A) Plot of the mean PE for normal controls for the subset of studies in Table 3 that did not include anticipatory saccades in the PE and used a zero gap paradigm. The black circle represents the PE for the ASPIS sample (Evdokimidis et al., 2002). (B) Plot of the mean AL for normal controls for the subset of studies in Table 3 that did not include anticipatory saccades in the AL and used a zero gap paradigm. The black circle represents the AL for the ASPIS sample (Evdokimidis et al., 2002). (C) Plot of the mean SL for normal controls for the subset of studies in Table 3 and two studies not included in Table 3. In all studies anticipatory saccades were not included in the SL and all used a zero gap paradigm. The black circle represents the SL for the ASPIS sample (Constantinidis et al., 2003).

the percentage of anticipatory saccades (latency <80 ms), the percentage of express saccades (see glossary) (latency >80 ms and <120 ms), the percentage of fast regular saccades (latency 135– 179 ms), the percentage of slow regular saccades (180–399 ms) and the percentage of late saccades (400–699 ms). In the following section I will concentrate on AL as well as on the latency of saccades in the saccade task (SL) but many of the arguments that I make could be also valid for the other latency measures. The main concern made for the operational deﬁnition of the PE in the antisaccade task could be made for SL and AL too. Thus the inclusion of low latency cut off in the deﬁnition of a saccade or antisaccade would result in different mean antisaccade or saccade latencies computed. Moreover both these populations would be contaminated by anticipatory saccades as explained previously. In the same fashion that task parameters affect the PE in the antisaccade task they are expected to have an effect on SL and AL. Thus we have shown that the increase in ﬁxation interval duration results in a decrease in AL (Smyrnis et al., 2002).

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS 14

N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

The manipulation of ﬁxation offset is also known to affect SL and AL. It has been repeatedly documented in the literature that the introduction of a gap between the offset of the central target and the onset of the peripheral target results in a great reduction of response latency (Fischer et al., 1993). Actually the use of the gap paradigm has led to the identiﬁcation of a separate population of very fast latency saccades (80–120 ms) called the express saccades (Fischer et al., 1993). The same effect of gap has also been demonstrated for the AL in the antisaccade task (Fischer & Weber, 1997; Klein, Bruegner, Foerster, Mueller, & Schweickhardt, 2000; Reuter-Lorenz, Hughes, & Fendrich, 1991; Reuter-Lorenz, Oonk, Barnes, & Hughes, 1995). These studies have also showed that the gap effect for AL is signiﬁcantly smaller in magnitude than the gap effect for SL. In the studies of Table 3 the effects of ﬁxation offset manipulation were found to be signiﬁcant on AL (McDowell & Clementz, 1997; Smyrnis et al., 2004) and SL (Smyrnis et al., 2004). Since different studies in Table 3 have used different gap conditions it should be assumed that these task differences have an effect on SL and AL. In addition to the effects of ﬁxation manipulation prior cueing of the target location might also have an effect on AL and SL. Of the two studies that used prior cues in Table 3 one does not report on either SL or AL (Rosse et al., 1993) and the other used only an antisaccade task and reported AL (Polli et al., 2006). The appearance of a valid spatial cue indicating the location of the upcoming target 100 ms before the execution of an antisaccade was shown to increase the AL (Fischer & Weber, 1996). As mentioned before for PE I know of no studies reporting the effect of manipulation of target presentation interval and presence of feedback on SL and AL. There is though an effect of target amplitude on both AL and SL. In our study it was observed that AL decreased with increasing target amplitude from 2 to 8 deg and then increased again for amplitudes of 9 and 10 deg (Smyrnis et al., 2002). Fischer and Weber (1997) also showed a decrease with increasing target amplitudes (from 1 to 12 deg). Looking at the data of Reilly, Harris, Keshavan, and Sweeney (2005) one can also see a decrease in SL with increasing target amplitude from 10 to 30 deg although they do not report on this difference. We also observed a signiﬁcant decrease in SL with target amplitude in our ASPIS sample (data not published). As far as the effect of direction of target on SL is concerned we have shown in that both hemispace differences and directional differences target presentation have no signiﬁcant effect on SL (Constantinidis et al., 2003). In the same study we have also shown no differences in SL for saccades starting from a peripheral position and directed towards the center (centrifugal saccades) versus the SL for saccade starting from the center and directed towards a peripheral position (centripetal saccades). Similarly we found no directional differences on AL for left versus right directed antisaccades. The effect of the instruction to perform a mirror antisaccade versus simply an antisaccade to the other direction (direction only antisaccade) of target stimulus on AL has been examined in our recent study comparing these two task conditions (Evdokimidis et al., 2006). We have shown in that study that accurate mirror antisaccades had signiﬁcantly slower AL than direction only antisaccades while inaccurate mirror antisaccade AL did not differ from that of direction only antisaccades. The increase though in AL for correct mirror antisaccades was small (19 ms average). Finally some studies in Table 3 use a design were saccade and antisaccade trials are mixed in the same block (Manoach et al., 1997, 2002). The AL for antisaccades is increased if these are mixed with prosaccades in the same block and the more the percentage of prosaccades versus antisaccades in a block the larger the AL for antisaccades (Massen, 2004; see also Hutton in this issue). In conclusion there are many task related differences that could have a signiﬁcant effect on AL and SL just as was the case with PE.

Fig 2B and C shows the mean AL and SL, respectively, for a subset of studies in Table 3 that used a zero gap task and a latency cutoff for anticipatory saccades. For the SL I used data from two more studies reporting only on saccades that are not included in Table 3 (Evans & Schwartz, 1997; Reilly et al., 2005). The black dot in both ﬁgures shows the mean AL and SL for the ASPIS sample (Constantinidis et al., 2003; Evdokimidis et al., 2002). It is obvious looking at these ﬁgures that a large variation of mean latency is present between studies. Radant et al. (2007) reported in their multi center study no difference in group mean AL or SL among the seven different sites where the antisaccade and saccade task performance was measured providing further evidence to my conclusion that task and administration procedure differences shown in Table 3 are a major source of variation in these measures among different studies. Some studies in Table 3 have also measured the accuracy of saccades and antisaccades (Clementz et al., 1994; Crawford et al., 1995; Crawford et al., 1998; Curtis, Calkins, Iacono et al., 2001; Ettinger et al., 2003, 2004; Gooding & Tallent, 2001; Karoumi et al., 2001; Manoach et al., 1997; Maruff et al., 1998; McDowell & Clementz, 1997; McDowell et al., 1998; Nkam et al., 2001; Ross et al., 1998; Ross, Olincy et al., 1998; Spengler et al., 2006; Thaker et al., 2000) and few of those have also measured the accuracy of the memory saccades in the memory saccade task (Brenner, McDowell, Cadenhead, & Clementz, 2001; Broerse et al., 2002; Fukushima et al., 1990; Karoumi, Ventre-Dominey, Vighetto, Dalery, & d’Amato, 1998; McDowell & Clementz, 2001; Ross, Olincy et al., 1998; Ross et al., 1998). The most frequently used measures of accuracy in these studies is the gain deﬁned as the ratio of the saccade amplitude to the target amplitude. It should be emphasized here that at least for the antisaccade task and the memory saccade task the presence of a feedback presentation of the correct target location should have an effect on the accuracy of antisaccades and saccades. Another important parameter inﬂuencing saccade accuracy is amplitude. We have shown using amplitudes that ranged from 2 to 10 deg that accuracy decreased for both saccades and much more for antisaccades when the target amplitude was small (the gain increased to more than 1) and when target amplitude was large (the gain decreased to less than 1) while the best accuracy was for the middle range of target amplitudes (Evdokimidis et al., 2006). Other task related factors might also have an effect on accuracy and, of course, accuracy might be affected by practice. Finally some studies report on the maximum speed and duration of saccades for saccades and antisaccades as deﬁned in Fig. 2 (Burke & Reveley, 2002; Clementz et al., 1994; Fukushima, Fukushima, Miyasaka, & Yamashita, 1994; Fukushima et al., 1990; Thaker et al., 2000). The last column of Table 3 shows that different studies of antisaccades and saccades have used very different numbers of trials ranging from 14 to 864. In order to get reliable measurements on all task parameters a sufﬁcient number of data are required. A very small number of trials used in many studies of Table 3 (<20 trials) might result in large errors in the estimation of the true value of the parameter due to insufﬁcient data aggregation. The studies in Table 3 also used different protocols with some studies administering blocks of trials while others administering all trials in a single block. It is important then to know how different numbers of trials within a session and different numbers of sessions affect these parameters. In other words the question is whether these parameters are prone to practice and/or fatigue related effects. Fischer and Weber (1992) reported that AL decreased with practice over a period of 12–15 days. They also reported that subjects have been trained in the saccade task in order to produce a clear population of express saccades with latencies of 80–120 ms indicating that with training SL becomes faster. In that same study they showed that the PE in the antisaccade task was also prone to

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

practice effects but the effect was small (PE decreased only by 2– 3% over 12–15 sessions). In contrast Green et al. (2000) in their study on normal subjects performed six practice sessions of the antisaccade task and showed that the PE dropped from 21% at the ﬁrst to 14% at the second session and then leveled off at around 10% for the subsequent sessions. They neither report on practice effects nor on other measures of saccades and antisaccades nor for within session practice effects. Ettinger et al. (2003) in their study of stability of oculomotor variables reported no signiﬁcant within session effects for PE, AL and SL neither for saccade gain and antisaccade gain measures. They did ﬁnd though a signiﬁcant reduction of PE, a decrease in saccade gain and an increase in antisaccade gain at retest while there was no test–retest effect on AL and SL. Larrison-Faucher et al. (2004) examined practice effects among 12 sessions of performance of a saccade and antisaccade task. They found no practice effects for either PE or latency variables for both tasks. Harris, Reilly, Keshavan, and Sweeney (2005) also measured antisaccade performance in four time intervals spanning a period of one year. They found no differences in the PE of normal controls over this time span while they did ﬁnd a signiﬁcant reduction of AL by 8.8%. Reilly et al. (2005) measured saccades at four time intervals spanning 52 weeks and report no signiﬁcant effect of time on SL. Radant et al. (2007) compared performance in three blocks of 30 trials in the antisaccade task and found a signiﬁcant reduction of PE from the ﬁrst to the second block while there was no difference between the second and the third block. Finally we have studied the effect of the number of trials in the task on different antisaccade parameters in our ASPIS sample (Smyrnis et al., 2002) and showed that PE followed a U shape curve with an initial reduction of PE in the ﬁrst 10 trials (learning effect), a subsequent stabilization and ﬁnally an increase in PE after the ﬁrst 60 trials up to the ﬁnal number of 90 trials in the task (fatigue effect). We also observed a decrease in AL in the antisaccade task as the number of trials increased. In conclusion then there is discrepancy among different studies as to whether there are within and between session effects for PE in the antisaccade task. The same is true concerning latencies with some studies showing a modest decrease in AL and SL within and between sessions while others reporting no change. A study with multiple administrations of blocks of trials with varying numbers of trials within block could help to quantify within and between session practice effects and suggest a recommendation for standardization of this parameter. A few studies report on test–retest reliability of saccade and antisaccade measures of performance. Versino et al. (1993) reported test–retest reliability of 0.91 for SL using the intra class correlation coefﬁcient. They also reported good test–retest reliabilities for the relationship between saccade amplitude and duration and the relation of amplitude to peak velocity (range 0.63–0.87). In their study of diurnal variation Roy-Byrne et al. (1995) used intra class correlation coefﬁcients and reported signiﬁcant test–retest reliability (P < .05) for SL (0.69) and AL (0.78) but not for PE (0.22) in the antisaccade task. In the same study the authors measured performance in a memory saccade task and they observed signiﬁcant (P < .05) test–retest reliability for the PE in the memory saccade task for the same individuals (0.86). Ettinger et al. (2003) reported signiﬁcant test–retest reliability (P < .05) for PE (Pearson r: .89, ICC: 0.79) and AL (Pearson r: .69, ICC: 0.65) in the antisaccade task but not for gain (Pearson r: .51, ICC: 0.35) in the antisaccade task. They also reported signiﬁcant test–retest reliability for SL (Pearson r: .79, ICC: 0.76) and gain (Pearson r: .67, ICC: 0.59) in the saccade task. This was also the ﬁrst study to measure internal consistency in these tasks and they reported that all antisaccade measures (PE, AL and gain) as well as saccade measures (SL, gain) had signiﬁcant (P < .05) internal consistency (Cronbach a > 0.7). These authors though measured internal consistency be-

15

tween time segments by dividing each session in four segments. The most elaborate study on internal consistency for the saccade and antisaccade tasks was performed by Klein and Fischer (2005a). These authors measured both test–retest reliability and internal consistency using both the time dependent split-half and the odd–even methods. For the internal consistency calculations they used a sample of 327 normal subjects. For the test–retest reliability they used a sample of 117 subjects that were tested twice with an inter session interval of 17–21 months. They found good to excellent internal consistency using both the split-half and the odd–even method for SL, AL and PE as well as for error latency and correction latency and proportion of error corrections in the antisaccade task (range for split-half reliability: 0.67–0.92, range for odd–even reliability: 0.65–0.96). The test–retest reliability was also good to excellent for most of these variables (0.68–0.77) except for the latency of error saccades (0.48), the latency of correction antisaccades (0.57) and the proportion of error corrections in the antisaccade task (0.55). These lower reliabilities for these measures could be explained by the lower data aggregation since they are based not on the full data set but on a much smaller subset of directional errors. In conclusion the relevant literature on saccades and antisaccades provides support for good reliability of these measures but there is still no conclusive evidence regarding the effects of practice on these measures. 8. Epilogue The main focus of this review on metric issues for the most widely used oculomotor paradigms in the psychiatric literature was to summarize the information on measurement procedures, test deﬁnitions, parameters measured and the effects of practice and reliability on these parameters. The basic message conveyed is that the use of oculomotor tasks in psychiatric research after 35 years of application has yielded a wealth of studies using a large variation of task procedures and outcome measurements that may probe different aspects of cognition in these disorders. One the other hand this review also showed that the notion that eye movement dysfunction is measured with a few well-deﬁned parameters that can be repeatedly used in genetic and clinical studies is misleading. If the current research goal would be to deﬁne such parameters, then oculomotor function test procedures and measurements are in desperate need of standardization. A very powerful tool in validating tests for clinical research is meta-analysis but in order for it to produce meaningful results, the studies that are to be entered should have comparable test application and measurement procedures. Unfortunately, as I tried to show in this review, every single group doing research in this ﬁeld has applied oculomotor tasks with differences in either measurement or testing procedures, or parameter deﬁnitions making it an almost impossible task to ﬁnd a corpus of studies that can be safely grouped together in a meta-analysis. On the other hand in the same growing literature there were studies that addressed methodological issues such as reliability or temporal stability of oculomotor function parameters and studies that addressed the issue of parameter deﬁnition and measurement in this ﬁeld (Fischer, et al., 1997; McDowell & Clementz, 1997). It could then be the focus of a future research program in the ﬁeld to work out the many details of a ﬁnal synthesis of recommendations in an effort to achieve the standardization of oculomotor function test procedures and outcome measurements. Table 4 could be an example of how the outcome of this research program might look like. Studies based on these recommendations could then form the basic corpus for the validation of these tests. These

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS

9

4 5 6 7 8

antisaccade Use the infrared or video camera method at a sampling frequency P200 Hz Use preferably a zero gap or overlap paradigm. Avoid introducing a gap or precues that change the performance. Avoid complex visual stimuli Use two directions (right, left) for stimulus presentation and, preferably, more than one amplitudes within the range of 10 deg. Avoid close stimuli to the central ﬁxation target (62 deg) and stimuli very far from the central ﬁxation target (>15 deg) Use a sufﬁcient number of trials but not too large to avoid practice and fatigue effects (>30 and 6100). Include a block design for saccades and antisaccades to measure the performance Reject artifacts by visual inspection or if using a pattern recognition program use visual inspection to validate the program decisions Use combined criteria (for example amplitude and velocity and acceleration) to detect the onset and duration of saccadic eye movements in the saccade record. A simple absolute speed increase is not enough Exclude from analysis or analyze separately predictive saccades (reaction time 680 ms) and very slow onset saccades (reaction time >500 or 600 ms) Use as primary performance outcomes in the saccade task the speed of performance measured as the mean or preferably the median reaction time and the accuracy of performance measured as the gain (the mean or preferably the median ratio of saccade amplitude to target amplitude). Use the ﬁrst saccade from target onset to measure accuracy Report the primary performance outcome parameter for the antisaccade task which is the percentage of error prosaccades. Report the speed of performance measured as the mean or preferably the median reaction time for correct antisaccades. Secondary outcome parameters that could also be reported include the accuracy of antisaccades measured as gain (the mean or preferably the median ratio of antisaccade amplitude to target amplitude) and the speed and accuracy for the erroneous prosaccade in error trials Saccade, 1 2 3

Smooth 1 2 3 4 5 6 7 8

eye pursuit Use the infrared or video camera method at a sampling frequency P200 Hz Include a closed loop condition of continuous pursuit. Preferably use a trapezoid or triangular signal. Use at least two different speeds covering both the low end of 610 deg/s and the high end of 20–30 deg/s Use a sufﬁcient number of cycles for each stimulus speed (at least ﬁve but it is unclear how many would be enough or if there are practice effects) Discard the pursuit record close to the direction shifts thus keeping only a central window around the primary eye position for analysis in order to avoid different strategy effects in the points of a direction shift Reject artifacts by visual inspection or if using a pattern recognition program use visual inspection to validate the program decisions Use combined criteria (for example amplitude and velocity and acceleration) to detect the onset and duration of saccadic eye movements in the pursuit record. A simple absolute speed increase is not enough Use the pursuit gain as the primary outcome parameter. Deﬁne gain as the ratio of eye speed to target speed preferably time weighted. Report the mean or preferably the median value of pursuit gain for each target speed Other parameters such as frequency of CUS, AS and SWJ can be used as secondary outcome measures but a difference in pursuit function should be corroborated by a difference in the primary outcome variable, such as the gain

N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

Table 4 Tentative recommendations for standardization of oculomotor function tests in psychiatric research

16

future instruments might allow oculomotor research to fulﬁll even greater aspirations in psychiatric research and clinical practice. Acknowledgments I thank my colleague Professor Ioannis Evdokimidis for his help and support in the preparation of this manuscript and for assistance in preparing the ﬁgures. I also thank Dr. Klein and Dr. Ettinger for their invitation to contribute this manuscript and their valuable comments for improving it. Finally I would like to thank two anonymous reviewers for their valuable comments during the revision of this manuscript. References Abel, L. A., Friedman, L., Jesberger, J., Malki, A., & Meltzer, H. Y. (1991). Quantitative assessment of smooth pursuit gain and catch-up saccades in schizophrenia and affective disorders. Biological Psychiatry, 29, 1063–1072. Abel, L. A., & Ziegler, A. S. (1988). Smooth pursuit eye movements in schizophrenics—what constitutes quantitative assessment? Biological Psychiatry, 24, 747–761. Avila, M. T., Weiler, M. A., Lahti, A. C., Tamminga, C. A., & Thaker, G. K. (2002). Effects of ketamine on leading saccades during smooth-pursuit eye movements may implicate cerebellar dysfunction in schizophrenia. American Journal of Psychiatry, 159, 1490–1496. Boudet, C., Bocca, M. L., Chabot, B., Delamillieure, P., Brazo, P., Denise, P., et al. (2005). Are eye movement abnormalities indicators of genetic vulnerability to schizophrenia? European Psychiatry, 20, 339–345. Brenner, C. A., McDowell, J. E., Cadenhead, K. S., & Clementz, B. A. (2001). Saccadic inhibition among schizotypal personality disorder subjects. Psychophysiology, 38, 399–403. Broerse, A., Crawford, T. J., & den Boer, J. A. (2002). Differential effects of olanzapine and risperidone on cognition in schizophrenia? A saccadic eye movement study. Journal of Neuropsychiatry and Clinical Neurosciences, 14, 454–460. Broerse, A., Holthausen, E. A., van den Bosch, R. J., & den Boer, J. A. (2001). Does frontal normality exist in schizophrenia? A saccadic eye movement study. Psychiatry Research, 103, 167–178. Brownstein, J., Krastoshevsky, O., McCollum, C., Kundamal, S., Matthysse, S., Holzman, P. S., et al. (2003). Antisaccade performance is abnormal in schizophrenia patients but not in their biological relatives. Schizophrenia Research, 63, 13–25. Burke, J. G., & Reveley, M. A. (2002). Improved antisaccade performance with risperidone in schizophrenia. Journal of Neurology Neurosurgury and Psychiatry, 72, 449–454. Calkins, M. E., & Iacono, W. G. (2000). Eye movement dysfunction in schizophrenia: A heritable characteristic for enhancing phenotype deﬁnition. American Journal of Medical Genetics, 97, 72–76. Calkins, M. E., Katsanis, J., Hammer, M. A., & Iacono, W. G. (2001). The misclassiﬁcation of blinks as saccades: Implications for investigations of eye movement dysfunction in schizophrenia. Psychophysiology, 38, 761–767. Campion, D., Thibaut, F., Denise, P., Courtin, P., Pottier, M., & Levillain, D. (1992). SPEM impairment in drug-naive schizophrenic patients: Evidence for a trait marker. Biological Psychiatry, 32, 891–902. Cerbone, A., Sautter, F. J., Manguno-Mire, G., Evans, W. E., Tomlin, H., Schwartz, B., et al. (2003). Differences in smooth pursuit eye movement between posttraumatic stress disorder with secondary psychotic symptoms and schizophrenia. Schizophrenia Research, 63, 59–62. Clementz, B. A., McDowell, J. E., & Zisook, S. (1994). Saccadic system functioning among schizophrenia patients and their ﬁrst-degree biological relatives. Journal of Abnormal Psychology, 103, 277–287. Clementz, B. A., Sweeney, J. A., Hirt, M., & Haas, G. (1990). Pursuit gain and saccadic intrusions in ﬁrst-degree relatives of probands with schizophrenia. Journal of Abnormal Psychology, 99, 327–335. Constantinidis, T. S., Smyrnis, N., Evdokimidis, I., Stefanis, N. C., Avramopoulos, D., Giouzelis, I., et al. (2003). Effects of direction on saccadic performance in relation to lateral preferences. Experimental Brain Research, 150, 443–448. Crawford, T. J., Haeger, B., Kennard, C., Reveley, M. A., & Henderson, L. (1995). Saccadic abnormalities in psychotic patients. I. Neuroleptic-free psychotic patients. Psychological Medicine, 25, 461–471. Crawford, T. J., Sharma, T., Puri, B. K., Murray, R. M., Berridge, D. M., & Lewis, S. W. (1998). Saccadic eye movements in families multiply affected with schizophrenia: The Maudsley family study. American Journal of Psychiatry, 155, 1703–1710. Cronbach, L. J. (1951). Coefﬁcient alpha and the internal structure of tests. Psychometrika, 16, 297–334. Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bulletin, 52, 281–302. Curtis, C. E., Calkins, M. E., Grove, W. M., Feil, K. J., & Iacono, W. G. (2001). Saccadic disinhibition in patients with acute and remitted schizophrenia and their ﬁrst-degree biological relatives. American Journal of Psychiatry, 158, 100–106.

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx Curtis, C. E., Calkins, M. E., & Iacono, W. G. (2001). Saccadic disinhibition in schizophrenia patients and their ﬁrst-degree biological relatives. A parametric study of the effects of increasing inhibitory load. Experimental Brain Research, 137, 228–236. Ettinger, U., Kumari, V., Chitnis, X. A., Corr, P. J., Crawford, T. J., Fannon, D. G., et al. (2004). Volumetric neural correlates of antisaccade eye movements in ﬁrstepisode psychosis. American Journal of Psychiatry, 161, 1918–1921. Ettinger, U., Kumari, V., Crawford, T. J., Corr, P. J., Das, M., Zachariah, E., et al. (2004). Smooth pursuit and antisaccade eye movements in siblings discordant for schizophrenia. Journal of Psychiatric Research, 38, 177–184. Ettinger, U., Kumari, V., Crawford, T. J., Davis, R. E., Sharma, T., & Corr, P. J. (2003). Reliability of smooth pursuit, ﬁxation, and saccadic eye movements. Psychophysiology, 40, 620–628. Ettinger, U., Kumari, V., Crawford, T. J., Flak, V., Sharma, T., Davis, R. E., et al. (2005). Saccadic eye movements, schizotypy, and the role of neuroticism. Biological Psychology, 68, 61–78. Evans, W. J., & Schwartz, B. D. (1997). Attentional mechanisms of saccadic eye movements in schizophrenia. Neuropsychiatry Neuropsychology and Behavioral Neurology, 10, 17–24. Evdokimidis, I., Smyrnis, N., Constantinidis, T. S., Stefanis, N. C., Avramopoulos, D., Paximadis, C., et al. (2002). The antisaccade task in a sample of 2,006 young men. I. Normal population characteristics. Experimental Brain Research, 147, 45–52. Evdokimidis, I., Tsekou, H., & Smyrnis, N. (2006). The mirror antisaccade task: Direction-amplitude interaction and spatial accuracy characteristics. Experimental Brain Research, 174, 304–311. Fischer, B., Biscaldi, M., & Gezeck, S. (1997). On the development of voluntary and reﬂexive components in human saccade generation. Brain Research, 754, 285–297. Fischer, B., Gezeck, S., & Hartnegg, K. (1997). The analysis of saccadic eye movements from gap and overlap paradigms. Brain Research Protocols, 2, 47–52. Fischer, B., & Weber, H. (1992). Characteristics of ‘‘anti” saccades in man. Experimental Brain Research, 89, 415–424. Fischer, B., & Weber, H. (1996). Effects of procues on error rate and reaction times of antisaccades in human subjects. Experimental Brain Research, 109, 507–512. Fischer, B., & Weber, H. (1997). Effects of stimulus conditions on the performance of antisaccades in man. Experimental Brain Research, 116, 191–200. Fischer, B., Weber, H., Biscaldi, M., Aiple, F., Otto, P., & Stuhr, V. (1993). Separate populations of visually guided saccades in humans: Reaction times and amplitudes. Experimental Brain Research, 92, 528–541. Flechtner, K. M., Steinacher, B., Sauer, R., & Mackert, A. (2002). Smooth pursuit eye movements of patients with schizophrenia and affective disorder during clinical treatment. European Archives of Psychiatry and Clinical Neuroscience, 252, 49–53. Friedman, L., Jesberger, J. A., & Meltzer, H. Y. (1991). A model of smooth pursuit performance illustrates the relationship between gain, catch-up saccade rate, and catch-up saccade amplitude in normal controls and patients with schizophrenia. Biological Psychiatry, 30, 537–556. Fukushima, J., Fukushima, K., Miyasaka, K., & Yamashita, I. (1994). Voluntary control of saccadic eye movement in patients with frontal cortical lesions and parkinsonian patients in comparison with that in schizophrenics. Biological Psychiatry, 36, 21–30. Fukushima, J., Fukushima, K., Morita, N., & Yamashita, I. (1990). Further analysis of the control of voluntary saccadic eye movements in schizophrenic patients. Biological Psychiatry, 28, 943–958. Gooding, D. C. (1999). Antisaccade task performance in questionnaire-identiﬁed schizotypes. Schizophrenia Research, 35, 157–166. Gooding, D. C., Iacono, W. G., & Beiser, M. (1994). Temporal stability of smoothpursuit eye tracking in ﬁrst-episode psychosis. Psychophysiology, 31, 62–67. Gooding, D. C., Miller, M. D., & Kwapil, T. R. (2000). Smooth pursuit eye tracking and visual ﬁxation in psychosis-prone individuals. Psychiatry Research, 93, 41–54. Gooding, D. C., Shea, H. B., & Matts, C. W. (2005). Saccadic performance in questionnaire-identiﬁed schizotypes over time. Psychiatry Research, 133, 173–186. Gooding, D. C., & Tallent, K. A. (2001). The association between antisaccade task and working memory task performance in schizophrenia and bipolar disorder. The Journal of Nervous and Mental Disease, 189, 8–16. Green, J. F., King, D. J., & Trimble, K. M. (2000). Antisaccade and smooth pursuit eye movements in healthy subjects receiving sertraline and lorazepam. Journal of Psychopharmacology, 14, 30–36. Harris, M. S., Reilly, J. L., Keshavan, M. S., & Sweeney, J. A. (2005). Longitudinal studies of antisaccades in antipsychotic-naive ﬁrst-episode schizophrenia. Psychological Medicine, 36, 485–494. Holahan, A. L., & O’Driscoll, G. A. (2005). Antisaccade and smooth pursuit performance in positive- and negative-symptom schizotypy. Schizophrenia Research, 76, 43–54. Holzman, P. S., O’Brian, C., & Waternaux, C. (1991). Effects of lithium treatment on eye movements. Biological Psychiatry, 29, 1001–1015. Holzman, P. S., Proctor, L. R., & Hughes, D. W. (1973). Eye-tracking patterns in schizophrenia. Science, 181, 179–181. Hong, L. E., Avila, M. T., Adami, H., Elliot, A., & Thaker, G. K. (2003). Components of the smooth pursuit function in deﬁcit and nondeﬁcit schizophrenia. Schizophrenia Research, 63, 39–48. Hong, L. E., Avila, M. T., & Thaker, G. K. (2005). Response to unexpected target changes during sustained visual tracking in schizophrenic patients. Experimental Brain Research, 165, 125–131.

17

Hong, L. E., Mitchell, B. D., Avila, M. T., Adami, H., McMahon, R. P., & Thaker, G. K. (2006). Familial aggregation of eye-tracking endophenotypes in families of schizophrenic patients. Archives of General Psychiatry, 63, 259–264. Hong, L. E., Turano, K. A., O’neill, H., Hao, L., Wonodi, I., McMahon, R. P., et al. (2007). Reﬁning the predictive pursuit endophenotype in schizophrenia. Biological Psychiatry [Epub ahead of print]. Hutton, S. B., Crawford, T. J., Gibbins, H., Cuthbert, I., Barnes, T. R., Kennard, C., et al. (2001). Short and long term effects of antipsychotic medication on smooth pursuit eye tracking in schizophrenia. Psychopharmacology, 157, 284–291. Hutton, S. B., Crawford, T. J., Kennard, C., Barnes, T. R., & Joyce, E. M. (2000). Smooth pursuit eye tracking over a structured background in ﬁrst-episode schizophrenic patients. European Archives of Psychiatry and Clinical Neuroscience, 250, 221–225. Hutton, S. B., Joyce, E. M., Barnes, T. R., & Kennard, C. (2002). Saccadic distractibility in ﬁrst-episode schizophrenia. Neuropsychologia, 40, 1729–1736. Kaplan, R. M., & Saccuzzo, D. P. (1989). Psychological testing. Principles, applications and issues (2nd ed.). Paciﬁc Grove, CA: Brooks/Cole Publishing Company. Karoumi, B., Saoud, M., Amato, T., Rosenfeld, F., Denise, P., Gutknecht, C., et al. (2001). Poor performance in smooth pursuit and antisaccadic eye-movement tasks in healthy siblings of patients with schizophrenia. Psychiatry Research, 101, 209–219. Karoumi, B., Ventre-Dominey, J., Vighetto, A., Dalery, J., & d’Amato, T. (1998). Saccadic eye movements in schizophrenic patients. Psychiatry Research, 77, 9–19. Kathmann, N., Hochrein, A., Uwer, R., & Bondy, B. (2003). Deﬁcits in gain of smooth pursuit eye movements in schizophrenia and affective disorder patients and their unaffected relatives. American Journal of Psychiatry, 160, 696–702. Katsanis, J., Iacono, W. G., & Harris, M. (1998). Development of oculomotor functioning in preadolescence, adolescence, and adulthood. Psychophysiology, 35, 64–72. Katsanis, J., Taylor, J., Iacono, W. G., & Hammer, M. A. (2000). Heritability of different measures of smooth pursuit eye tracking dysfunction: A study of normal twins. Psychophysiology, 37, 724–730. Klein, C., Bruegner, G., Foerster, F., Mueller, W., & Schweickhardt, A. (2000). The gap effect in prosaccades and antisaccades in psychometric schizotypes. Biological Psychology, 55, 25–39. Klein, C., & Fischer, B. (2005a). Instrumental and test–retest reliability of saccadic measures. Biological Psychology, 68, 201–213. Klein, C., & Fischer, B. (2005b). Developmental fractionation and differential discrimination of the anti-saccadic direction error. Experimental Brain Research, 165, 132–138. Klein, C., Heinks, T., Andresen, B., Berg, P., & Moritz, S. (2000). Impaired modulation of the saccadic contingent negative variation preceding antisaccades in schizophrenia. Biological Psychiatry, 47, 978–990. Kumari, V., Ettinger, U., Crawford, T. J., Zachariah, E., & Sharma, T. (2005). Lack of association between prepulse inhibition and antisaccadic deﬁcits in chronic schizophrenia: Implications for identiﬁcation of schizophrenia endophenotypes. Journal of Psychiatric Research, 39, 227–240. Larrison-Faucher, A. L., Matorin, A. A., & Sereno, A. B. (2004). Nicotine reduces antisaccade errors in task impaired schizophrenic subjects. Progress in Neuropsychopharmacology & Biological Psychiatry, 28, 505–516. Leigh, J. R., & Zee, D. S. (1991). The neurology of eye movements (2nd ed.). Philadelphia: F.A. Davis Company. Lencer, R., Malchow, C. P., Trillenberg-Krecker, K., Schwinger, E., & Arolt, V. (2000). Eye-tracking dysfunction (ETD) in families with sporadic and familial schizophrenia. Biological Psychiatry, 47, 391–401. Lencer, R., Trillenberg-Krecker, K., Schwinger, E., & Arolt, V. (2003). Schizophrenia spectrum disorders and eye tracking dysfunction in singleton and multiplex schizophrenia families. Schizophrenia Research, 60, 33–45. Lenzenweger, M. F., & O’Driscoll, G. A. (2006). Smooth pursuit eye movement and schizotypy in the community. Journal of Abnormal Psychology, 115, 779– 786. Levy, D. L., Holzman, P. S., Matthysse, S., & Mendell, N. R. (1994). Eye tracking and schizophrenia: A selective review. Schizophrenia Bulletin, 20, 47–62. Levy, D. L., Lajonchere, C. M., Dorogusker, B., Min, D., Lee, S., Tartaglini, A., et al. (2000). Quantitative characterization of eye tracking dysfunction in schizophrenia. Schizophrenia Research, 42, 171–185. Litman, R. E., Hommer, D. W., Clem, T., Ornsteen, M. L., Ollo, C., & Pickar, D. (1991). Correlation of Wisconsin card sorting test performance with eye tracking in schizophrenia. American Journal of Psychiatry, 148, 1580–1582. Litman, R. E., Torrey, E. F., Hommer, D. W., Radant, A. R., Pickar, D., & Weinberger, D. R. (1997). A quantitative analysis of smooth pursuit eye tracking in monozygotic twins discordant for schizophrenia. Archives of General Psychiatry, 54, 417–426. Lord, F. M., & Novick, M. R. (1968). Statistical theories of mental test scores. MA: Addison-Wesley. Maccabe, J. H., Simon, H., Zanelli, J. W., Walwyn, R., McDonald, C. D., & Murray, R. M. (2005). Saccadic distractibility is elevated in schizophrenia patients, but not in their unaffected relatives. Psychological Medicine, 35, 1727–1736. Manoach, D. S., Lindgren, K. A., Cherkasova, M. V., Goff, D. C., Halpern, E. F., Intriligator, J., et al. (2002). Schizophrenic subjects show deﬁcient inhibition but intact task switching on saccadic tasks. Biological Psychiatry, 51, 816–826. Manoach, D. S., Weintraub, S., Daffner, K. R., & Scinto, L. F. (1997). Deﬁcient antisaccades in the social–emotional processing disorder. Neuroreport, 8, 901–905.

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

ARTICLE IN PRESS 18

N. Smyrnis / Brain and Cognition xxx (2008) xxx–xxx

Maruff, P., Danckert, J., Pantelis, C., & Currie, J. (1998). Saccadic and attentional abnormalities in patients with schizophrenia. Psychological Medicine, 28, 1091–1100. Massen, C. (2004). Parallel programming of exogenous and endogenous components in the antisaccade. Quarterly Journal of Experimental Psychology A, 57, 475–498. McDowell, J. E., & Clementz, B. A. (1997). The effect of ﬁxation condition manipulations on antisaccade performance in schizophrenia: Studies of diagnostic speciﬁcity. Experimental Brain Research, 115, 333–344. McDowell, J. E., & Clementz, B. A. (2001). Behavioral and brain imaging studies of saccadic performance in schizophrenia. Biological Psychology, 57, 5–22. McDowell, J. E., Myles-Worsley, M., Coon, H., Byerley, W., & Clementz, B. A. (1999). Measuring liability for schizophrenia using optimized antisaccade stimulus parameters. Psychophysiology, 36, 138–141. Mokler, A., & Fischer, B. (1999). The recognition and correction of involuntary prosaccades in the antisaccade task. Experimental Brain Research, 125, 511– 516. Nickoloff, S. E., Radant, A. D., Reichler, R., & Hommer, D. W. (1991). Smooth pursuit and saccadic eye movements and neurological soft signs in obsessive– compulsive disorder. Psychiatry Research, 38, 173–185. Nkam, I., Thibaut, F., Denise, P., Van Der, E. A., Segard, L., Brazo, P., et al. (2001). Saccadic and smooth-pursuit eye movements in deﬁcit and non-deﬁcit schizophrenia. Schizophrenia Research, 48, 145–153. O’Driscoll, G. A., Lenzenweger, M. F., & Holzman, P. S. (1998). Antisaccades and smooth pursuit eye tracking and schizotypy. Archives of General Psychiatry, 55, 837–843. Olincy, A., Johnson, L. L., & Ross, R. G. (2003). Differential effects of cigarette smoking on performance of a smooth eye pursuit and a saccadic eye movement task in schizophrenia. Psychiatry Research, 117, 223–236. Polli, F. E., Barton, J. J., Vangel, M., Goff, D. C., Iguchi, L., & Manoach, D. S. (2006). Schizophrenia patients show intact immediate error-related performance adjustments on an antisaccade task. Schizophrenia Research, 82, 191–201. Radant, A. D., Dobie, D. J., Calkins, M. E., Olincy, A., Braff, D. L., Cadenhead, K. S., et al. (2007). Successful multi-site measurement of antisaccade performance deﬁcits in schizophrenia. Schizophrenia Research, 89, 320–329. Radant, A. D., & Hommer, D. W. (1992). A quantitative analysis of saccades and smooth pursuit during visual pursuit tracking. A comparison of schizophrenics with normals and substance abusing controls. Schizophrenia Research, 6, 225–235. Reilly, J. L., Harris, M. S., Keshavan, M. S., & Sweeney, J. A. (2005). Abnormalities in visually guided saccades suggest corticofugal dysregulation in never-treated schizophrenia. Biological Psychiatry, 57, 145–154. Reuter, B., Herzog, E., & Kathmann, N. (2006). Antisaccade performance of schizophrenia patients: Evidence of reduced task-set activation and impaired error detection. Journal of Psychiatric Research, 40, 122–130. Reuter, B., Rakusan, L., & Kathmanna, N. (2005). Poor antisaccade performance in schizophrenia: An inhibition deﬁcit? Psychiatry Research, 135, 1–10. Reuter-Lorenz, P. A., Hughes, H. C., & Fendrich, R. (1991). The reduction of saccadic latency by prior offset of the ﬁxation point: An analysis of the gap effect. Perception & Psychophysics, 49, 167–175. Reuter-Lorenz, P. A., Oonk, H. M., Barnes, L. L., & Hughes, H. C. (1995). Effects of warning signals and ﬁxation point offsets on the latencies of pro- versus antisaccades. Experimental Brain Research, 103, 287–293. Robinson, D. A. (1963). A method of measuring movement using a scleral search coil in a magnetic ﬁeld. IEEE Transactions on Biomedical Electronics, BME, 1, 137–145. Ross, R. G., Harris, J. G., Olincy, A., Radant, A., Adler, L. E., & Freedman, R. (1998). Familial transmission of two independent saccadic abnormalities in schizophrenia. Schizophrenia Research, 30, 59–70. Ross, D. E., Ochs, A. L., Pandurangi, A. K., Thacker, L. R., & Kendler, K. S. (1996). Mixture analysis of smooth pursuit eye movements in schizophrenia. Psychophysiology, 33, 390–397. Ross, R. G., Olincy, A., Harris, J. G., Radant, A., Adler, L. E., Compagnon, N., et al. (1999). The effects of age on a smooth pursuit tracking task in adults with schizophrenia and normal subjects. Biological Psychiatry, 46, 383–391. Ross, R. G., Olincy, A., Harris, J. G., Radant, A., Adler, L. E., & Freedman, R. (1998). Anticipatory saccades during smooth pursuit eye movements and familial transmission of schizophrenia. Biological Psychiatry, 44, 690–697. Ross, R. G., Olincy, A., Harris, J. G., Sullivan, B., & Radant, A. (2000). Smooth pursuit eye movements in schizophrenia and attentional dysfunction: Adults with schizophrenia, ADHD, and a normal comparison group. Biological Psychiatry, 48, 197–203.

Ross, R. G., Olincy, A., Mikulich, S. K., Radant, A. D., Harris, J. G., Waldo, M., et al. (2002). Admixture analysis of smooth pursuit eye movements in probands with schizophrenia and their relatives suggests gain and leading saccades are potential endophenotypes. Psychophysiology, 39, 809–819. Ross, R. G., Olincy, A., Zerbe, G., & Radant, A. (2001). Which duration of postsaccadic slowing identiﬁes anticipatory saccades during smooth pursuit eye movements? Psychophysiology, 38, 325–333. Ross, D. E., Thaker, G. K., Buchanan, R. W., Kirkpatrick, B., Lahti, A. C., Medoff, D., et al. (1997). Eye tracking disorder in schizophrenia is characterized by speciﬁc ocular motor defects and is associated with the deﬁcit syndrome. Biological Psychiatry, 42, 781–796. Ross, D. E., Thaker, G. K., Buchanan, R. W., Lahti, A. C., Medoff, D., Bartko, J. J., et al. (1996). Association of abnormal smooth pursuit eye movements with the deﬁcit syndrome in schizophrenic patients. American Journal of Psychiatry, 153, 1158–1165. Rosse, R. B., Schwartz, B. L., Kim, S. Y., & Deutsch, S. I. (1993). Correlation between antisaccade and Wisconsin card sorting test performance in schizophrenia. American Journal of Psychiatry, 150, 333–335. Roy-Byrne, P., Radant, A., Wingerson, D., & Cowley, D. S. (1995). Human oculomotor function: Reliability and diurnal variation. Biological Psychiatry, 38, 92–97. Sereno, A. B., & Holzman, P. S. (1995). Antisaccades and smooth pursuit eye movements in schizophrenia. Biological Psychiatry, 37, 394–401. Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations: Uses in assessing rater reliability. Psychological Bulletin, 86, 420–428. Siever, L. J., Friedman, L., Moskowitz, J., Mitropoulou, V., Keefe, R., Roitman, S. L., et al. (1994). Eye movement impairment and schizotypal psychopathology. American Journal of Psychiatry, 151, 1209–1215. Smyrnis, N., Evdokimidis, I., Mantas, A., Kattoulas, E., Stefanis, N. C., Constantinidis, T. S., et al. (2007). Smooth pursuit eye movements in 1,087 men: Effects of schizotypy, anxiety, and depression. Experimental Brain Research, 179, 397–408. Smyrnis, N., Evdokimidis, I., Stefanis, N. C., Constantinidis, T. S., Avramopoulos, D., Theleritis, C., et al. (2002). The antisaccade task in a sample of 2,006 young males. II. Effects of task parameters. Experimental Brain Research, 147, 53–63. Smyrnis, N., Malogiannis, I. A., Evdokimidis, I., Stefanis, N. C., Theleritis, C., Vaidakis, A., et al. (2004). Attentional facilitation of response is impaired for antisaccades but not for saccades in patients with schizophrenia: Implications for cortical dysfunction. Experimental Brain Research, 159, 47–54. Spengler, D., Trillenberg, P., Spenger, A., Nagel, M., Kordon, A., Junghanns, K., et al. (2006). Evidence from increased anticipation of predictive saccades for a dysfunction of fronto-striatal circuits in obsessive-compulsive disorder. Psychiatry Research, 143, 77–88. Sporn, A., Greenstein, D., Gogtay, N., Sailer, F., Hommer, D. W., Rawlings, R., et al. (2005). Childhood-onset schizophrenia: Smooth pursuit eye-tracking dysfunction in family members. Schizophrenia Research, 73, 243–252. Sweeney, J. A., Clementz, B. A., Escobar, M. D., Li, S., Pauler, D. K., & Haas, G. L. (1993). Mixture analysis of pursuit eye-tracking dysfunction in schizophrenia. Biological Psychiatry, 34, 331–340. Sweeney, J. A., Hass, G. L., Li, S., & Weiden, P. J. (1994). Selective effects of antipsychotic medications on eye-tracking performance in schizophrenia. Psychiatry Research, 54, 185–198. Sweeney, J. A., Luna, B., Srinivasagam, N. M., Keshavan, M. S., Schooler, N. R., Haas, G. L., et al. (1998). Eye tracking abnormalities in schizophrenia: Evidence for dysfunction in the frontal eye ﬁelds. Biological Psychiatry, 44, 698–708. Thaker, G. K., Cassady, S., Adami, H., Moran, M., & Ross, D. E. (1996). Eye movements in spectrum personality disorders: Comparison of community subjects and relatives of schizophrenic patients. American Journal of Psychiatry, 153, 362–368. Thaker, G. K., Ross, D. E., Buchanan, R. W., Moran, M. J., Lahti, A., Kim, C., et al. (1996). Does pursuit abnormality in schizophrenia represent a deﬁcit in the predictive mechanism? Psychiatry Research, 59, 221–237. Thaker, G. K., Ross, D. E., Cassady, S. L., Adami, H. M., LaPorte, D., Medoff, D. R., et al. (1998). Smooth pursuit eye movements to extraretinal motion signals: Deﬁcits in relatives of patients with schizophrenia. Archives of General Psychiatry, 55, 830–836. Thaker, G. K., Ross, D. E., Cassady, S. L., Adami, H. M., Medoff, D. R., & Sherr, J. (2000). Saccadic eye movement abnormalities in relatives of patients with schizophrenia. Schizophrenia Research, 45, 235–244. Versino, M., Castelnovo, G., Bergamaschi, R., Romani, A., Beltrami, G., Zambarbieri, D., et al. (1993). Quantitative evaluation of saccadic and smooth pursuit eye movements. Is it reliable? Investigative Ophthalmology & Visual Science, 34, 1702–1709. Young, L. R., & Sheena, D. (1975). Eye movement measurement techniques. The American Psychologist, 315–330.

Please cite this article in press as: Smyrnis, N. Metric issues in the study of eye movements in psychiatry. Brain and Cognition (2008), doi:10.1016/j.bandc.2008.08.022

Visuomotor characterization of eye movements in a ...

Smooth pursuit eye movements in 1087 men: eVects of ...

Smooth pursuit eye movements in 1087 men: eVects of ... - Springer Link

Using Eye Movements for Online Evaluation of Speech ...

Redox-Linked Domain Movements in the Catalytic Cycle of ...

Networks of Outrage and Hope - Social Movements in the Internet ...

The application of eye-tracking technology in the study ...

Culture and (in)dependence Issues of Independence in the ... - Calenda

LÃ©vy-like diffusion in eye movements during spoken ... - Dan Mirman

Teamwork in the Eye of the Beholder

Looking for meaning: Eye movements are sensitive to ...

control of upper extremity movements in expert pianists ...

Eye Movements to Pictures Reveal Transient Semantic ...

Movements in the Equity Premium: Evidence from a ...

Oxford Textbooks in Psychiatry

Redox-Linked Domain Movements in the Catalytic ...

Fifty years of the metric system in India and its adoption in our daily ...

Role of dynamic group therapy in psychiatry

Complementary medicines in psychiatry Review of ...

Issues in Minority Languages in India - University of Graz