The first section of this paper describes in detail the computational model and its processing assumptions.
We conducted two new experiments for the explicit purpose of testing the predictions made by the computational model. In both experiment we investigated the time course of S-R and S-S consistency. In both experiments there was one condition in which there was no DO – that was the neutral condition (Type 1). In experiment 1, the irrelevant stimulus overlapped with either the relevant stimulus (Type 4) or with the response (type 3), but not with both. The delays occurred in a relatively narrow time range: (SOA = 0 – 200 ms). In the second experiment the relevant stimulus overlapped with the irrelevant stimulus, as well as with the response (Type 7). The range of delays was extended: (SOA = 0 – 800 ms).
EXPERIMENTS 1 & 2
Stimuli and Responses
The responses are left or right key presses.
The stimuli consisted of a rectangle (1.5 cm. x 3.3cm.) in which three attributes could be presented: color, position, and word. Color was the relevant attribute; position and word were both irrelevant. The color was either blue or green. “Position” refers to the particular part of the rectangle in which the color was presented: upper/lower half – left/right half. “Word” refers to a word which was presented in the middle of the rectangle. The word was either “blue”, “green”, “detail”, or “novel”.
Fig.1 from SK, GTS, AW, JR, (1999) (P. 689)
Each trial began with the presentation of the four corners of the rectangle. This also served as the beginning of the warning interval which was of random duration. At the end of the warning period, the rectangle was completed by lines joining the four corners. In one condition, the zero delay condition, color (the relevant stimulus), word, and position (the irrelevant stimuli) were all presented in the rectangle simultaneously. In the non-zero delay condition, only the irrelevant stimuli were presented (position was indicated with gray). Following a short delay of between 50 and 800 ms. the gray in the rectangle was replaced by either the color blue, or the color green. The rectangle display was terminated by the subjects’ response.
The stimuli: Type 1, 3, and 4.
Delays: 0, 50, 100, and 200.
Each stimulus Type was run in a separate block at one delay. The individual stimuli within each type were randomized within that block.
(see Fig. x )
1. There was no effect of delay in either the neutral or the Type 3 conditions, whether consistent or inconsistent.
2. The largest effect of delay was for the Type 4, (S-S overlap), inconsistent condition: as delay increased, RT increased steeply. However, for the consistent condition there was no effect of delay.
3. The effect of delay on consistency effects differed greatly between Type 3 (S-R) and Type 4 (S-S) conditions. As delay increased, the Type 3 SR consistency effect decreased, whereas the SS consistency effects increased. These data are in accord with those reported in a similar experiment ( Kornblum, 1994).
4. The RT for Type 3 (S-R overlap) was faster for consistent than for inconsistent trials.
here I suggest including the table on p. 691, as well as inserting a simple graph for these mean data. It would just make the results so much easier to see.
The stimuli: Types 1, and 7.
Delays: 0, 100, 200, 400, 800.
Each stimulus Type was run in a separate block, at one delay. Within each block, one third of the trials were neutral wrt both the relevant stimuli and the responses, the other two third overlapped with both the stimuli and the responses: one sixth were doubly consistent (c/c), one sixth were doubly inconsistent (i/i), one sixth were S-S consistent and S-R inconsistent (c/i), and the remaining sixth were S-S inconsistent and S-R consistent (i/c).
Note: the value of S-S consistency is specified before the value of S-R consistency (e.g. c/i for S-S consistency, etc.)
1. RT’s for doubly consistent (c/c) and doubly inconsistent (i/i) conditions were the fastest and slowest respectively at all delays.
For the mixed consistency conditions at zero delay:
2. The RT for c/i, was indistinguishable from the RT at doubly inconsistent (i/i) condition.
3. Similarly, the RT for i/c was indistinguishable from doubly consistent (c/c) condition.
? Table 3, on p. 694; SK, GTS, AW, JR, (1999).
Because in Experiment 2, the effects of both S-S and S-R overlap were examined in a broad range of SOA values, the results of Exp. 2 were simulated first. The parameters of the model were therefore set to provide the closest fit to the empirical data of Exp. 2. In order to demonstrate the robustness of the model the parameters used for the simulation of Exp. 2, were then changed, in a theoretically appropriate way, to simulate the data of Exp. 1.
In addition, the data of two comparable experiments in the literature were simulated as well.
The correlation between the empirical and simulated mean RT difference (i.e. RT – neutral) for the four experimental conditions was .9283.
For Type 3, the correlation between the empirical and the simulated data was .9877, for Type4 it was .9460
For the two studies in the literature:
1. Hommel (1997), Experiment 2.
For the two conditions of the experiment, the correlation between the empirical and the simulated data was .9968 for condition 1 (the horizontal condition), and .9999 for condition 2 (the vertical condition).
2. De Jong et al. (1994), Experiment 3.
The correlation between the empirical data and the simulation was .9565