There were two main purposes of the present experiment. The first was to compare spatial discrimination perfomiance between identical cyclopean and luminance defined stimuli at a frequency close to the peak sensitivity of the DSF. The second was to compare the cyclopean analogues of the three spatial discrimination tasks reviewed above with the disparity increment discrimination function. Aside from the comparison between luminance and disparity domains, this experiment is hence the direct cyclopean analogue o f the study o f Bowne (1990), except that vernier acuity was also studied and temporal frequency discrimination was not. The logic of the experiment is also similar: if it is the case that cyclopean spatial discrimination is not limited by the same source of noise as disparity increment discrimination, then it is clear that specialized cyclopean mechanisms exist, which process the location, orientation or size of three-dimensional objects.
Methods
Observers had to perform 4 tasks with cyclopean corrugations: disparity in c re m en t discrim inatio n, o rien tation d iscrim in a tio n , co rru g a tio n frequency discrimination and vernier acuity. The frequency of the corrugations was 0.6 c/deg. Five values of peak-trough disparity were tested: 1,2, 4, 8 and 16 min arc. In all the tasks the corrugations were horizontal.
Three of the tasks followed an identical 2ATFC procedure consisting of two 1000 msec intervals separated by 500 msec. The exception was the vernier acuity task which consisted of a single presentation lasting 1000 msec. Due to the limit of the number of images which could be stored in RAM at any one time, the phase of the stimuli was not altered between trials. This introduces potential cues into spatial discrimination tasks, such as the distance between a feature and the edge of the stimulus. To remove these possible artifacts, the size of each stimulus was jittered (as outlined below) and the fixation spot was also shifted from presentation to presentation as outlined in Experiment 3.
In addition, thresholds for spatial frequency discrim ination, orientation discrimination and vernier acuity were collected using sinusoidal luminance gratings at 50% contrast, with the same mean luminance as the RDS images (as described in Chapter 2)28. Thresholds for luminance-defined stimuli were estimated using exactly the same equipment and procedure, including being displayed via the stereoscope.
There were some slight differences between the stimuli and procedures for each of the tasks, due to the need to eliminate unwanted cues. More precise descriptions of the individual cyclopean tasks are as follows:
(1) D is p a r ity Increm ent D iscrim ination . The observer had to indicate on which trial the corrugations had more depth. Apart from the fact that the stimuli were 8x8 deg RDS, this condition was exactly as outlined in Experiment 3.
(2) Co rru g a tio n Frequency D iscrim ina tio n . The observer had to indicate on which trial the corrugations had the higher frequency. Stimuli were 8 deg in width but the height of the stimulus was chosen randomly from a uniform distribution ranging from 6.66 to 8 deg. This ensured that neither the phase at the edge, nor the distance between the edge and a stimulus feature, nor the number of cycles could be used as a cue to solve the task. The other details were identical to (1) above.
28w iih the equipm ent employed in the cxperimcm.s o f this thesis it is im possible to obtain the contrast resolution necessary to measure contrast increm ent discrim ination thresholds. As it was desirable to com pare perform ance in the different tasks under identical conditions, contrast increm ent discrim ination data were not collected.
(3) Orientation D iscrim ination. The observer had to indicate whether the second stimulus was rotated clockwise or anti-clockwise of the first. Stimuli were confined to a circular region the diameter of which varied according to a uniform distribution between 7 and 8 deg. APE selected from a stored set o f 19 images with orientations centred on horizontal. Each of these images was randomly assigned one of 4 random-dot patterns. The reference stimulus could be chosen from any of the stimuli whose orientation was between ±2 deg of horizontal. This perturbation of the reference orientation ensured that all of the dot-patterns were associated with both responses and that the observer had to compare the two stimuli on every trial, rather than with reference to an internal standard, such as horizontal.
(4) Vernier a c u ity . The observer had to state whether the right-hand corrugations were higher or lower than the left-hand corrugations. The stimulus consisted of two 4x8 deg half images. The half-images could not be perfectly abutting because the outline of the three-dimensional corrugations is not vertical. Therefore the centre region was ramped to the vertical mid-line using a half-cycle of a cosine over a region of 16 min arc. This meant that there was a ramped gap between the corrugations at peak disparity of 32 min arc. APE selected stimuli from a set of 18 stored half images which could appear on the left or the right and were made up of 4 random-dot patterns. Each dot-pattern was thus associated with both responses.
The luminance stimuli at 50% contrast were constructed in precisely the same manner as is described above for the cyclopean stimuli, except that the ramps to zero disparity were replaced by ramps to the mean luminance. The same computer program was used to generate and present the cyclopean and the luminance stimuli.
Runs of 64 were carried out in blocks of 4, comprising one run for each type of task. Runs employing different peak-trough disparities were interleaved with the 50% contrast runs and were randomized and counter-balanced across blocks.
There were two observers in this experiment, DD and PDL. DD was naive to the purposes of the experiment, had no experience of viewing RDSs, had corrected-to-normal vision and had never participated in a psychophysical experiment. Prior to data-collection she was shown several RDS's depicting a variety of surfaces and had no problem in determining the forms which they contained.
Results
In general there was very little learning observed, even for DD, who had no previous experience of viewing RDSs. The thresholds reported below are therefore mostly the r.m.s. of the first 3 estimates taken for each condition (entirely so for PDL). Slight exceptions to this pattern are given below for DD. Where a power function is fitted, the same method of linear regression was used as in Experiment 3. Thus, the figures in brackets represent the 95% confidence interval.
"O < 1 1 e —DD — 0 “ - P D L 0.1
0.01
10 1Peak-trough disparity, d (min arc)
F i g u r e 5 .4 . D i s p a r i t y i n c r e m e n t d i s c r i m i n a ti o n f u n c t i o n s f o r tw o