Stimuli were displayed on a Manitron monochrome monitor driven by a Matrox IM-1280 graphics card (see Chapter 2 for details). The screen was viewed from a distance of 1 metre, mean luminance was 14.36 cd/m^. Stimuli were displayed in a rectangular area located in the middle of the screen. From the viewing distance of 1 metre the screen had a width of 20.80° and a height of 16.67°. The height of the patches was equal to the height of the rectangular
Standard Binary Standard Binary Standard Binary Quaternary
0 ^memi 0 ^mean ^mean ^^2
1 hnean “ ^mean ~ ^2 1 ^mean T / ^mean ^2 2 ^mean ~ ^mean " ^2 3 hnean "" '^7 ^mean ~ ^^2 Standard Binary Top biased Binary Standard Binary Bottom biased Binary
0 hnean ^mean ^^2 0 ^mean -^7 ^mean ^2 1 ^mean ^mean ^^2 1 ^mean ^ "^7 ^mean ^2 2 ^mean ^mean ~ ^2 2 ^mean ^1 ^mean4* ^2
3 ^meun ~ ^mean ~ ^2 3 ^mean ~ ^mean4" ^2 4 ^mean ~ ^1 ^mean ~ ^2 4 ^mean ~ ^1 ^mean ~ ^^2 5 ^mean ~ ^1 ^mean ~ ^2 5 ^mean ' hnean ~ ^^2
Standard Binary Top biased Tertiary Standard Binary Bottom biased Tertiary
0 ^mean+ hnean+ •^•^2 0 ^niean^ ^7 ^mean4" ^^2 1 ^mean ^mean '^2 1 ^mean -^7 ^mean4" 3%2 2 ^mean ' ^mcan ^^2 ^mean ~ ^1 hnean " 4-2
3 ^ntean ~ ^mean ~ ^^2 3 ^mean “^1 hnean ~ ^^2
Table 6.1: Mappings between luminance values of noise types when they are replaced by one another in stimuli containing static noise patterns. Numbers in columns 1 and 4 refer to the values that the notional underlying noise may take. The luminance values in columns 2 & 3 and 5 & 6 show the luminance values that are mapped over the underlying noise. Note that
Xj > 0 and ^ 0-
areas within which the stimuli were displayed. The direction of motion of the modulant was horizontal.
In experiments 6.1 and 6.3 the rectangular area was 12.57° wide and 7.86° high. The width of noise patches was 1.58° giving a spatial frequency of 0.32 cycles per degree (cpd) where the length of the cycle is the width of two adjacent noise patches. Noise check size was 2.96 arc minutes. Stimuli were presented for 200 ms which allowed time for two of the quarter cycle steps. The final position of the bars was therefore shifted half a cycle from the start. Subjects could not therefore derive the direction of motion from a knowledge of the start and end positions of the motion sequence.
In experiments 6.2 and 6.4 the width of the height of the rectangular area was increased to 10.46°. The width of the noise patches was 2.10° giving a modulant spatial frequency of 0.24 cpd. Noise check size was 1.97 arc minutes. The stimuli were presented for 600 ms, allowing for 8 quarter cycle steps. As these two experiments utilised static noise, the start and end frames of the sequences were identical.
One potential problem with measuring contrast modulation tuning curves is that the results can easily become contaminated if there is any consistent luminance modulation in the signal. The Matrox IM-1280 graphics card only has 8 bit output, giving a total of 256 discrete luminance values. For a particular luminance level, the usual procedure is to choose the pixel value that gives the closest luminance value to that which is required. Clearly, there may well be a difference between the value required and the value produced. If a number of luminance levels are required to produce a certain contrast then the value of all the luminance levels needed to produce that contrast may vary by some small amount. The actual expected mean luminance level of the noise pattern may therefore differ from the required expected mean luminance level of the noise pattern.
The motion stimuli in this chapter consist of alternating noise patterns (see Figures 6.1 and 6.2). If there is a consistent difference between the actual expected mean luminance levels of the target and conparison patterns, then luminance defined motion will be introduced into the signal. To minimise the possibility of this occurring, the difference between the actual expected mean levels of the two patterns needed to be minimised. This was achieved using
a simple nearest neighbours search procedure (in terms of the luminance values that could be produced) so that the actual expected mean luminance of the comparison pattern would be close to the actual expected mean luminance of the target pattern. The use of this procedure meant that there was some jitter in the conçarison contrast which carries through into the modulation depths. Therefore in the results detailed below, the modulation depths plotted are those calculated from the actual discrete values used in the stimuli.
6.2.4.2. Procedures.
Experiments 6.1 to 6.4 each consist of one or more sets of trials. Each set of trials contains two runs of a particular stimulus type where a stimulus type is defined by the two noise types that make up the stimuli. A run consists of a single stimulus type at a number of predetermined levels, a level being defined by the modulation depth and the contrast of the target pattern. Within a set, the target patterns of the two runs have the same contrasts as one another. The members of each of the sets are repeated a number of times so that the time taken to run a whole set takes about 20 minutes. Each set was repeated until 100 trials per level per stimulus type had been completed. The trials within a set were stacked and chosen at random from that stack. For each trial the stimulus was generated online. Starting position and direction of motion were randomized. Subjects were asked to fixate in the centre of the screen and judge whether the stimuli were moving to the right or to the left. Subjects responded by pressing the left/right arrow keys of a PC keyboard. Feedback was provided in all four experiments for both correct and incorrect responses. There was a minimum gap of 1 second between trials.
6.2.4.3. Data collection and analysis.
Data consisted of a number of scores out of 100 which indicate the number of correct responses for a particular level on a particular stimulus type. From this the proportion of correct responses can be calculated. In the definition of modulation depth given in equation
6.2, modulation depth is negative when the comparison contrast is greater than the target contrast and positive when the comparison contrast is less than the target contrast. As the absolute value of modulation depth increases, so too should the number of correct responses made by the subjects. On a graph showing proportion of correct responses over modulation