The latter is particularly important as a number of studies ( including that of Oborne & Holton ) appear to overestimate the sensitivity of global measures such as reading comprehension or measures of speed based on sentences or even whole texts .sx On balance , the evidence points to a genuine disadvantage in processing text on a VDU , particularly when task demands are high ( Wright & Lickorish , 1983 ; Zwahlen , Hartman , Ranjarajula , & Escontrela , 1985) .sx However , there is no generally accepted explanation of the effects .sx Visual discomfort and headache appear to be a consequence of staring at a flickering striped surface ( Wilkins , 1985a) .sx Changes in reading speed must obviously relate in some way to the distribution of eye movements over the screen or page , involving , for example , changes in the frequency and/or duration of fixations and the extent of saccades .sx However , these variables have not previously been systematically examined using a task approaching normal reading .sx An important observation regarding the effects of flicker on eye movement control was made by Wilkins ( 1985b) .sx He examined eye movements as subjects made repeated saccades between two letters embedded in a piece of text .sx The display was illuminated at 50 Hz or 100 Hz .sx The results showed reliably longer saccades at 50 Hz than at 100 Hz , the size of the effect being about one character position .sx A further , similar experiment contrasted 100-Hz with 20-KHz flicker , and again saccades were longer at the lower rate - although the effect was only reliable when the target letters were wide apart .sx Wilkins is tempted to interpret this finding as indicating an increase in saccade length in flicker , but it is important to note that absolute accuracy differed in the two conditions .sx At 50 Hz saccades overshot the target slightly , and at 100 Hz slightly undershot it .sx Without a baseline condition of steady illumination it is not possible to interpret the results in terms of a flicker - induced extension of saccade length .sx Further , since Wilkins employed an external event ( a tone ) as a signal for a saccade to take place , the possibility arises that the obtained inaccuracy could relate to an increased proportion of prematurely triggered saccades ( this is discussed in greater detail below) .sx There are , in fact , two sources of evidence converging on the prediction that interfering transient events , of the kind that may arise in flicker , may disturb eye movement control .sx First , a study by Becker , Kieffer , and Jurgens ( 1987 ) showed that both auditory and visual signals delivered after the onset of a saccade may modify its extent and velocity .sx The effect of the transient signal varied somewhat between subjects but led overall to a change in the final target location , shortening the saccade's length .sx Similarly , Neary and Wilkins ( 1989 ) showed that when saccades are made between two points on a raster display with a short-persistence phosphor , transient 'ghost images' may be seen in the vicinity of the target point .sx When lines rather than points are used , the fixated target may appear to tilt momentarily as a result of the top-to-bottom scan of the raster display .sx Clearly , these data also show that brief stimulation during a saccade can interfere with the normal suppression processes .sx The second line of evidence derives from the possibility that saccades may be triggered by transient external events that are independent of the process of saccade computation itself .sx We assume that when such external triggers occur , saccades are launched towards their destination on the basis of the current ( incomplete ) state of the on-going computation ( O'Regan , 1989) .sx For each saccade , target location is continually updated on the basis of available visual information .sx This becomes more refined with time and is modulated by attentional processes ( Deubel , Wolf , & Hauske , 1984) .sx Thus , high spatial frequency information becomes available later than low-frequency information .sx A 'premature' trigger will result in a saccade launched effectively to the available centre of gravity of the target region as a whole ( Findlay , 1981 , 1982) .sx In principle , such an effect may result in either undershoot or overshoot , depending on the particular visual configuration ( see O'Regan , 1989 , for an interesting discussion of this point ) , but it is important to note that in normal reading , for the majority of cases , large saccades over the page or screen surface are likely to be directed towards a boundary , in which case the 'centre of gravity' effect will lead to a preponderance of undershoots .sx It is possible that both mechanisms operate while a person is reading flickering text .sx By definition , saccades interrupted in flight will be of reduced extent .sx Similarly , but for a quite different reason , large saccades prematurely triggered will tend to land short .sx On the other hand , if the effects of premature triggering on saccades of normal extent is to increase the frequency of both under- and over-shoot , the overall result will be a change in the variability in landing position .sx To this will be added the possible effect of interference during a saccade , although this may be more likely to affect those saccades directed to remote targets , which will be longer in duration .sx The key question is whether an increase in the number of mislocated saccades produced in this way has important consequences for the reader ?sx The results from a number of recent eye movement studies suggest a positive answer .sx It is now well established that the eye movement control system is relatively intolerant of deviations from what is variously termed the 'convenient' or 'preferred' viewing position for a given word ( O'Regan & Levy-Schoen , 1987 ; Rayner , 1979) .sx Under laboratory conditions , involving the inspection of words in isolation , the 'convenient viewing position' for the majority of words of average length lies slightly to the left of the centre .sx The penalty for landing outside this point is approximately 20 msec per letter .sx However , this figure is a statistical average , primarily resulting from increases in the probability of making additional corrective saccades when the eye falls outside the convenient viewing position .sx The degree to which these effects are found in normal reading has recently been examined by McConkie , Kerr , Reddix and Zola ( 1988) .sx The 'preferred viewing position' ( i.e. where the eye actually falls , as distinct from what may be the best place for it to be ) shows a similar curvilinear relationship with letter position , although the curve is much flatter ( see also Vitu & O'Regan , 1989) .sx When account is taken of the 'launch site' ( the distance from the target word ) from which a saccade is made , the probability of landing on a particular letter , for words of all lengths , is a roughly normal function about the word centre .sx McConkie et al. ( 1988 ) identify a number of visual and oculomotor processes that might underlie the obtained variability in landing position , but for present purposes it is only necessary to note ( a ) that a 'preferred viewing position' ( i.e. the word centre ) exists for normal reading ; and ( b ) that an increase in variability about this position or a systematic shift away from it will change the probability of additional corrective saccades ( McConkie , Kerr , Reddix , Zola , & Jacobs , 1989) .sx In previous work we have shown that one class of eye movements distinguishes a successful reading style .sx This involves the use of large saccades ( e.g. spanning more than 40 character positions ) to reinspect selected parts of previously read text .sx The reader can , by this means , locate antecedents of anaphoric expressions ( Murray & Kennedy , 1988 ) ; deal with local syntactic ambiguity ( Kennedy & Murray , 1984 ; Kennedy , Murray , Jennings , & Reid , 1989 ) ; or reanalyse an initially incorrect semantic reading ( Pynte , Kennedy , Murray , & Courrieu , 1988) .sx Since the targets of such reinspections are remote , their control cannot be based on physical identification of the target :sx it must , rather , depend on the reader maintaining a form of spatially coded information ( Kennedy , 1987) .sx Obviously , if flicker acts to perturb saccade control , this class of saccades provides an ideal litmus test of its effect since the saccades in question are large , increasing the probability of disturbance ; they have a distinct role to play in the reader's process of comprehension ; and their use effectively characterizes an efficient reading style .sx The experiment reported here examined the effects of flicker on landing position and on a number of other eye movement measures under experimental conditions that called for the execution of a number of large reinspecting saccades .sx Following Wilkins , two flicker frequencies were used ( 50 Hz and 100 Hz ) , but in the present study a control condition was introduced using steady illumination .sx Two populations were studied - a group of typists who reported relatively severe symptoms of headache and visual fatigue associated with using word-processors ; and a group of students with little experience of VDUs .sx The procedure employed was based on that of Kennedy and Murray ( 1987) .sx Subjects read a sentence displayed on a single horizontal line , and this was then followed by the presentation of a stimulus word to the right of the sentence on the same line .sx The subject's task was to indicate whether or not the word was present in the displayed sentence .sx This task produces a relatively high number of large saccades , involving both reinspections of the target word and returns to the stimulus .sx It is a situation more closely approximating normal reading than that used by Wilkins ( 1985b ) in that it encouraged , but did not demand , that specific reinspections took place .sx The task is thus a good analogue of some aspects of normal reading since when subjects do reinspect , it is to a variable target position that is not pre-defined ( i.e. not involving a visual search ) and yet is specifiable for each item .sx Four hypotheses were tested :sx ( 1 ) that flicker will act to alter the distribution of landing positions within the word in the 'first pass' ( i.e. in the phase of the experimental procedure before presentation of the stimulus word ) ; ( 2 ) that flicker will act to alter the accuracy of large saccades directed towards a remote target , causing an increased proportion to fall short ; ( 3 ) that the effects of an increased probability of mislocated saccades will be an increase in the number of small 'corrective' eye movements ; ( 4 ) that the disruptive effects of flicker will be greater in a sample of typists reporting symptoms of visual discomfort .sx The first three hypotheses arise naturally from the earlier discussion .sx The fourth requires some additional justification .sx Whether or not an increase in the proportion of mislocated saccades seriously disrupts the reader may relate to both subject and task demands .sx There is evidence that professional typists adopt specific reading strategies to optimize performance of a complex motor skill ( Shaffer , 1975) .sx Typists pay more attention to local orthographic features of the text and exhibit an 'eye - hand' span that is much shorter than the normal eye-voice span ( Butsch , 1932) .sx Average fixation duration during transcription typing is much longer than normal , and average saccade extent is much shorter ( Inhoff , Morris , & Calabrese , 1986) .sx We hypothesize that these habits will carry over to unconstrained reading - in which case , small deviations from intended eye position will be particularly disruptive for this group of subjects .sx The choice of an appropriate comparison groups is , however , problematic .sx Completely asymptomatic typists were extremely rare in the population sampled .sx In any case , it is not clear that such a group would comprise and effective control if , as both Shaffer and Inhoff et al. suggest , typists , in general , exhibit a characteristic reading style .sx In the present study , therefore , comparisons were drawn between a group of typists and a group of normal readers ( students) .sx The latter group permits direct comparisons to be made between performance on the task employed in the present study and the extensive literature on eye movement control , which has almost exclusively made use of this population of subjects .sx Analysis of the results will focus on differences in reading strategy between the two groups and on the effects of flicker for each group .sx Method Materials and Design .sx A set of 60 items was constructed , 30 of the form shown in ( 1 ) , where the stimulus word had occurred in the sentence , and 30 like ( 2 ) , where the sentence contained a close synonym of the target .sx 1 .sx The novels in the library had started to go mouldy with the damp .sx novels .sx