On the Time Course of Exogenous Cueing Effects: A commentary on Tassinari et al. (1994).
In their classic paper, Posner and Cohen (1984) studied the time course of the attentional effect of uninformative peripheral cues. In the first experiment they describe, one of three horizontally aligned boxes (each 1deg. x 1deg. of visual angle in size) was cued with a 150 ms brightening of its outline. At a variable SOA of 0, 50, 100, 200, 300 or 500 ms the target (a 0.1deg. filled square) was displayed inside one of the boxes and subjects were to detect it as quickly as possible by pressing a single key. Responses were faster when cue and target appeared in the same location than when they appeared in opposite locations, but only if the cue-target SOA was less than 200 ms. With longer SOAs, the pattern was reversed: That is, responses were slower when cue and target appeared in the same location. Posner and Cohen argued that the early facilitation for cued targets was automatic in nature, and that after a short interval it was counteracted by an inhibitory effect. This inhibitory effect was due to attention being inhibited from returning to previously explored locations --hence the name “inhibition of return” (IOR)-- and “evolved to maximise sampling of the visual environment” (p. 550). Similar results have been reported by Maylor (1985; Maylor and Hockey, 1985).
Recently, this time course of the attentional effect of uninformative cues (i.e. first facilitation and then inhibition) has been challenged. According to Tassinari, Aglioti, Chelazzi, Peru, and Berlucchi (1994), “inhibition does not follow facilitation, as envisioned by Maylor (1985), but precedes and overlaps with it” (p.187). There are two parts to this statement: First, it is claimed that the inhibitory effect precedes the facilitatory effect; and second, that the inhibitory and the facilitatory effects overlap in time. We do not dispute the second of these claims. Posner and Cohen (1984) were first to suggest that peripheral cues have both facilitatory and inhibitory effects, and that these effects overlap in time. Tipper et al (in press) have made the same suggestion, and we concur with this view. We do not agree with the first claim, however, and will attempt to explain why not in what follows.
Tassinari et al. (1994) base their claims, especially the first one, on the results they obtained in a series of four experiments that included a 0 ms cue-target SOA condition. We contend that the results of this 0 SOA condition suffer from very serious methodological problems, and that the authors are in no position to support their claim that inhibition precedes facilitation.
In Tassinari et al.’s (1994) paper the target to be detected was a 0.5deg. “greenish” filled square, which could be displayed inside one of four 1.2deg. (or 2deg.) boxes aligned on the horizontal axis, two to the right and two to the left of fixation (at 4deg. and 12deg.). Before the target was displayed, one of the boxes was cued 0, 65, 130, 300 or 900 ms before target appearance. Apart from SOA, the other independent variable in the experiments was Cueing, with three levels: SP (cue and target appeared at the same point), SF (they appeared in the same field), and OF (they appeared in opposite fields). The target was always displayed for 16 ms, and the duration of the cue was manipulated between experiments. (Cue durations were 16 ms and 300 ms in Experiments 1 and 2 respectively; and 130 ms in Experiments 3 and 4.)
The main result was that in all experiments, 0 ms SOA responses were slower in the SP condition than in OF. Depending on cue duration, this negative effect (inhibition?) either disappeared or turned into a positive effect (facilitation) at an intermediate SOA. The negative result in the 0 ms SOA led the authors to conclude that the inhibitory effect (IOR) is present from the beginning and only later can be overridden or counteracted by facilitation. However, it is very likely that the negative effect they obtained was due to masking or other problems we describe below.
In Tassinari et al’s (1994) experiments a detection task was used. By definition, a detection task requires no discrimination. In the experiments of Tassinari et al (1994), when the cue-target asynchrony was long enough, the task was to detect the second increase in luminance (i.e. the target). However, with 0 ms SOA, the task was not in fact a detection task. Rather, it was a discrimination task: Subjects had to discriminate between a simple increase in luminance (cue alone) and two increases in luminance (cue and target). In the first case no response was to be made, and in the second case subjects had to press the response key. It is clear then that in the 0 ms SOA condition, subjects in fact performed a Go-NoGo task. Furthermore, in the context of all the other SOA conditions, this Go-NoGo task had a very large number of “catch trials”. (That is, in addition to the “real” catch trials in the 0 SOA condition, all trials with SOAs greater than 0 ms were effectively catch trials for the 0 SOA condition.)
We believe that this kind of two vs. one discrimination can occur much more quickly when the two events appear in opposite hemifields than when they appear at the same point. It follows that if the two events appeared at different points but within the same hemifield, response time would be intermediate. This is exactly the pattern of results reported by Tassinari et al. (1994): They observed longer RT in SP than in SF and OF, and slightly longer RT in SF than in OF. (Note however that the difference between SF and OF was never significant at the 0 ms SOA).
Therefore, we think that Tassinari et al.’s (1994) argument that IOR appears with 0 ms SOA is simply not valid. The negative effect they obtain occurs because with 0 ms SOA the task is automatically converted into a discrimination task, and the discrimination is much harder in cued than in uncued trials, for obvious reasons.
There are several points in Tassinari et al.’s (1994) work that support our argument. First, among the first three experiments (which use four locations), they obtained the largest negative effect at 0 ms SOA in Experiment 1 (-56 ms), in which cue and target were displayed and removed simultaneously; and the smallest negative effect (-12.6 ms) in Experiment 2, in which they used the longest cue duration (300 ms). In the first experiment, both cue and target were displayed for 16 ms, so in the 0 ms SOA condition, the target appeared and disappeared simultaneously with the cue. In this case there was no way to detect the target. The appearance of cue and target together had to be discriminated from the cue alone. Therefore the task was a pure Go-NoGo discrimination task, and logically should have been much more difficult in SP than in OF.
In Experiment 2, the target was again displayed for 16 ms (as always), but the cue was displayed for 300 ms. This asynchrony was enough to allow temporal segregation of cue and target. However, given that both stimuli appeared at the same time, the onset of the target could not be detected: As in Experiment 1, the onset of cue and target could only be discriminated from the onset of cue alone. But because of the different durations for cue and target, the offset of the target could be detected. Note as well that the duration of the observed negative effect in Experiment 2 is similar to the duration of the target. This suggests that in the SP condition subjects were responding at least in part to detection of the offset of the target.
In Experiments 3 and 4 the cue duration was 130 ms and the 0-SOA negative effects were -40 and -13.4 ms respectively. Again, target onset could not be detected, but had to be discriminated from the onset of cue and target. And the short duration of the cue made temporal segregation of cue and target more difficult, so that the one vs. two discrimination process may have been necessary, rather than detection of target offset. In experiment 4 stimuli were at 4Ż from fixation, and the negative effect was much smaller (-13.4 ms) than in exp. 3 with targets at 4deg. and 12deg. (-40 ms). This suggests that the negative effect observed at 0 ms SOA was mainly due to the target appearing at 12Ż of eccentricity. Imagine discriminating one vs. two increases in luminance when both appear in the same location 12deg. from fixation, compared with two increases that appear 24deg. from each other, to the left and right of fixation. When the target appears at 4deg. the differences in the discrimination are smaller and so is the observed negative effect.
Further support for our account of Tassinari et al.’s (1994) data is provided by the way in which RT in the SP condition varies across the first 3 experiments. Note that RT in the OF condition is always right around 320 ms in these experiments. In the SP condition, on the other hand, RT increases proportionally to the cue-target duration similarity: RT is lowest in Experiment 2, where the cue and target are most dissimilar, and highest in Experiment 1, where cue and target have equal durations. This increase is due to increasing problems with discrimination of cue alone from cue and target.
Finally, note that masking could also account in part for the results at short SOAs, especially with similar cue and target durations. Note that the ratio of cue size to target size was 1/10 (0.1deg. and 1deg. for cue and target respectively) in Posner and Cohen’s (1984) experiments, but only 1/4 in the best case of Tassinari et al. (1994) (0.5deg. and 2deg. for cue and target respectively). This could have been enough to produce some masking, especially with 12deg. eccentricity. (In Posner and Cohen, 1984, peripheral targets appeared at 8deg.)
In summary, Tassinari at al. (1994) argued, contrary to the commonly accepted time course of the exogenous attentional cueing effect (i.e. first facilitation then IOR) that the inhibitory effect appears first and only later can be overridden by facilitation. They based their argument on the negative effect they observed with 0 ms cue-target SOA. However, as we have shown above, with simultaneous cue-target presentation (i.e., 0 ms SOA), the task is no longer detection, but a Go-NoGo task, in which the discrimination process is much harder when both cue and target appear in the same location. In conclusion, Tassinari et al.’s results in no way prove the existence of IOR with 0 ms SOA. The traditional time course, as shown by Posner and Cohen (1984), seems to be the most reliable when appropriate experimental procedures are used, and so should be maintained.
(1997). In press, Vision Research