88 Experiment E0: Location Assessment in Peripheral Vision
eccentricity
Euklidian deviation DXY (degrees)
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
I II III IV
Figure 5.12: Euclidean deviation DXY of the comparison marker position from the target marker po-sition as a function of eccentricity.
5.3 Discussion and Conclusions 89
saccades within the allowed range of 1ofrom the central fixation point is almost impossible, particularly when the task requires “covert” attention to a peripheral location. Indeed, the eye-movement data shows that in most cases only a single fixation was recorded, so that for the given scenario the fixation duration coincides with the reaction time RT.
When we now take into account that, under normal viewing conditions when the gaze is not restricted, eye fixations approximately last between 150 and 400 ms, it is quite
“natural” that the values measured here for RT are quite low. Even without the special requirements of the task here, it is a rather difficult task for subjects to accomplish extremely long fixations. Consequently, only very few values of RT exceed 1350 ms and were measured on the rare occasions when subjects managed to execute more than just one fixation within the allowed range around the fixation point – although it is known from aftereffect experiments that subjects are able to fixate a particular point for about 30 seconds. For the present task, however, subjects obviously extend fixation durations to only such a degree that still does not present too great a (concentration) effort and, on the other hand, does not – at least in their own belief – compromise too much their performance in the assessment of the target marker location.
Certainly, the further the target marker is moved out to the periphery of the fixation point, the more difficult the assessment of the marker position becomes. Although subjects obviously do not feel very comfortable executing long fixations, they try to compensate for the increased difficulty of the task by an increase of RT. This may be understood when discussed in the context of mental rotation experiments where reaction time was found to increase for larger angles between two (target and comparison) stimuli. In the present scenario, the task may be considered a “mental translation”. In analogy to men-tal rotation, longer translation distances may require more time to accomplish stimulus assessment. This could possibly explain the increase of RT in more peripheral regions in this experiment.
An interesting finding, which the analyses of the other dependent variables – at least partially – support, is that the difference between RT for Eccentricity II and Eccentric-ity III is considerably smaller than the differences between the other adjacent eccentricEccentric-ity regions, although the boundaries between the eccentricity regions are all spaced at equally distant intervals. We recall the original definition of both the Eccentricities II and III as belonging to the parafoveal region, as the range of this region (from 3o to 10o) appeared to be too large to be accounted for by only a single level of the factor eccentricity in these studies. Although the marker position in Eccentricity III still takes longer to be assessed than in Eccentricity II, the respective RTs are not found to differ significantly.
This observations could be interpreted as supporting the classification of peripheral view-ing in foveal, parafoveal and peripheral perception regions. In addition, it could give rise to the hypothesis that the perception effort (in terms of RT) within each eccentricity region does not vary considerably or is possibly even invariant and that differences are most pro-nounced around the designated boundaries of these eccentricity regions. However, if we consider a physiological approach that attributes the differences to a decrease in receptor density in the retina, this density drops rapidly from the fovea to the periphery, but does not show obvious plateaux that could account for three explicit eccentricity levels. Maybe
90 Experiment E0: Location Assessment in Peripheral Vision
a detailed review of the finer structures in the topology of the retinal receptors or the respective cortical areas could produce evidence for or against such a link.
What these physiological properties and the findings regarding RT certainly do not imply, is, of course, that sharp boundaries exist to separate the foveal from the parafoveal, and the parafoveal from the peripheral region. On the basis of the analysis of RT alone, a distinction between three levels (foveal, parafoveal and peripheral) and the choice of boundaries between them as accomplished in our studies appear nevertheless sensible.
In contrast to the effect of eccentricity on reaction time, the lack of significant differ-ences in RT for the two meridial positions (horizontal vs. vertical, relative to the fixation point) can be considered a first indicator that no such differences exist in the other de-pendent variables either – which is indeed the case. Alternatively, we could also have concluded that subjects were just not aware that one of the meridial configurations would be more difficult to assess than the other, but then find significant differences in the positional deviations DX(p), DY(p) and/or DXY(p).
However, the analyses of the deviations suggest no influence of the meridial position on those variables. Although we noticed this effect – or rather that such an effect does not exist – for RT, it is again surprising to find no such effect for the positional deviations either. Whether the marker’s position has to be assessed when it is presented in proximity to the horizontal or to the vertical axis does not obviously result in significant differences within either DX, DY or DXY. Specifically, targets’ positions along the vertical meridian can be as accurately judged as those along the horizontal meridian.
We would certainly not have expected this observation initially. First, we must con-sider that the retina is not a true hemisphere, but ellipsoidal in shape with the horizontal axis being the longer. Furthermore, with binocular viewing, the visual field is ovoid, its horizontal axis being approximately 0.5 times greater than its vertical axis (Prinzmetal &
Gettleman, 1993). Taking this asymmetry into account, the preferred horizontal orienta-tion of the visual field and the further range of receptors along this axis could be well expected to lead to a better position assessment accuracy along the horizontal than along the vertical meridian. Instead, the tendencies towards an interaction in the analyses of DX, DY and DXY in Experiment E0 only slightly hint at such a dependence. It must be taken into account, however, that the assessment is probably based on the cortical rather than the retinal representation. The retinal information may thus be corrected in subse-quent processing steps as to compensate for the retinal distortion. This apparently yields cortical position representations which are equally accurate for all meridial positions of the stimulus marker.
Contrary to the missing meridial position effect, the significantly increasing assessment error with increasing eccentricity could be expected. The rise of all absolute deviations DXp, DYp and DXYp the further the target is being presented from the fixation point can be attributed to the poorer spatial resolution in the peripheral visual field (e.g. Thomp-son & Fowler, 1980). This again is an integral part of Tsal’s (1999) concept of different metrics for attended and unattended stimuli. Even though advanced by Tsal for length estimation (cf. Section 2.2), the idea of metrics composed of fine units for attended judge-ment – resembling high spatial resolution in foveal viewing due to the retinal anatomy
5.3 Discussion and Conclusions 91
with high central receptor density – and of coarse units for unattended judgement – re-sembling low(er) spatial resolution in peripheral viewing – appears appealing for visual localisation as well. Targets that are presented foveally or in near-foveal regions, for ex-ample in eccentricity region I, can be explicitly visually attended. This can be achieved without the execution of eye movements, which are not permitted in this experiment, and thus allows for fine judgements with respect to the target position. In consequence, this leads to a fairly accurate position assessment. In contrast, targets presented in more eccentric regions, such as in the eccentricity regions II, III and IV, cannot be directly visually attended. The then only coarse position judgement of these unattended stimuli produces less accurate results with a greater variance or standard deviation – as observed in Experiment E0.
Some specific observations require clarification in this discussion with regard to the separate analyses for the radial deviation DX(p), the tangential deviation DY(p) and the Euclidean deviation DXY(p). Thus, the comparison between the absolute deviations DXp and DYp reveals that the tangential position of the target marker can be far better assessed than the radial position. In other words, it appears to be much easier for subjects to correctly judge the direction of the marker relative to the fixation point – which can be seen as corresponding with the tangential position – than to correctly judge the distance between the fixation point and the marker – corresponding with the radial position. This is shown in DX and DY – the two variables that take into account the direction of the respective deviations – even more explicitly than in the absolute deviations DXp and DYp. Here it can be clearly seen that, on average, there is hardly any systematic deviation of the comparison from the target position in the tangential direction. Furthermore, this judgement is achieved more or less independently of the eccentric position. However, the increase in standard deviations with increased peripheral presentation indicates that the tangential position can be judged with less precision under eccentric viewing conditions, although no specific directional effect, i.e. clockwise or counter-clockwise from the original tangential position, prevails.
On the other hand, the radial deviation DX exhibits a directional effect, the distance between fixation point and target marker position is increasingly underestimated with increasing eccentricity. The uncertainty along the radial axis, which might be termed “ec-centricity axis” as well, seems to be greater than along the tangential axis and it appears that with increasing eccentricity the axis“contracts” so thatdistances appear closer than they actually are. This could possibly be caused by mapping the same distance on fewer receptors in the peripheral than in the more central visual field. If this is not compensated for by other mechanisms – which is obviously not the case – the distance perceived in the periphery will be judged shorter. The fact that the distance judgement along the eccentric-ity axis is characterised by constantly changing receptor densities – which decrease with increasing eccentricity – does not facilitate the computation either. (It must be noted, however, that the receptor density decreases much stronger than the observed underes-timation effect and can thus probably not alone account for this effect.) The fact that the tangential (“direction”) assessment of the target marker does not have to account for such changes – as being computed along the axis perpendicular to the eccentricity axis –
92 Experiment E0: Location Assessment in Peripheral Vision
might motivate the better performance in tangential position judgement.
These results suggest that the assessment of position is governed by two distinct pro-cesses. The first process is responsible for the assessment of the direction in which the target in question is situated, a process that obviously works quiteaccurately and more or lessindependently of the peripheral position of the target. The second process determines the distance between fixation point and target. This process yields less accurate judge-ments as theradial position of the target is significantlyunderestimated. Furthermore, this process is eccentricity dependent. On aggregate, the combination of these two processes yields positional judgements that are dominated by the distance component and results in a perceived position of the target marker that is shifted towards the fixation point, but shows very little directional divergence.
The following chapters will now address the perception of the next more complex, higher dimensional stimulus type, namelyline segments, under the same peripheral view-ing conditions as in Experiment E0. First, the assessment of line segment lengths will be discussed.
Chapter 6
Experiment E1: Length Assessment in Peripheral Vision
In the second experiment (Experiment E1) in this series of three experiments that aim to establish and quantify the effects of eccentric stimulus presentation on the perception and assessment of specific stimulus dimensions, we investigate line segment length. The
“reference” data obtained in this experiment allows for the testing of the existence of cor-relations between the assessment error of peripherally perceived lengths of line segments and the mislocation of marker positions. If a correspondence can be found, this would indicate that observations in this experiment can possibly be attributed to mechanisms that were identified to influence location assessment. The peripheral assessment of line segment length might thus be decomposed and could be accomplished by peripherally assessing the locations of the end points of the line segment. The distance between the end points then yields the line segment length.
The following sections describe the experimental method for Experiment E1, based on the methodological preliminaries that were established in Chapters 3 and 4. Subsequently, the results will be presented and discussed which allows us to draw conclusions about the mechanisms that may govern peripheral length perception.
6.1 Method
6.1.1 Subjects
The subjects were twelve experimentally naive students – six male and six female – from the University of Bielefeld. Their average age was 27.5 years. All subjects had normal or corrected-to-normal vision and no pupil anomalies. The subjects were paid for their participation in the experiment.
6.1.2 Stimuli
As in Experiment E0, the stimulus images were displayed on a computer screen with a spatial resolution of 1280x1024 pixels. At the center of each picture a fixation point
94 Experiment E1: Length Assessment in Peripheral Vision
Figure 6.1: Typical stimulus picture in Experiment E1. Subjects had to assess the length of the target line segment while observing the central fixation point.
was displayed with 0.2o of visual angle in diameter. The stimulus target line segment was presented at a pseudo-random location so that it lay entirely within one of the eccentricity regions I–IV (see Section 4.1.2). The target line segment had a thickness of one pixel (0.03o), an orientation of either 0o ± 22.5o (“horizontal”), 45o ±22.5o or 135o ±22.5o (“oblique”) or 90o±22.5o (“vertical”) and a length of either 1o±0.3o (“short”), 4o±0.3o (“intermediate”) or 7o±0.3o (“long”) of visual angle (see Section 4.1). The fixation point, the target line segment and the background were set to the same dark and light grey colours, respectively, as in Experiment E0. Figure 6.1 shows a typical stimulus picture.
6.1.3 Apparatus
The apparatus used was the same as in Experiment E0.
6.1.4 Procedure
The procedure varied from that implemented in Experiment E0 only insofar, as subjects were asked to assess thelength of the target line segment presented in one of the eccentric-ity regions I-IV without foveally looking at it. Again, the gaze was restricted to a region of 1o around the central fixation point (Frame 1) and this contingency was monitored by the eye tracker. When subjects had successfully finished the assessment task and memo-rised the perceived length of the target line segment, they pressed the left button of the computer mouse (Frame 2). Subsequently, a blank screen was shown for 500 ms (Frame 3) and the fixation point reappeared (Frame 4) for 300 ms.
The fixation point was then replaced by the comparison line segment (Frame 5), the line segment center being located at the center of the screen. The comparison line segment had the same dimensions as the previously shown target line segment with respect to