The present study investigated whether the sequence effect of cueing paradigm could be triggered by non-predictive arrow cues. The results showed that the sequence effect of arrow cueing could be observed when voluntary control was not required to detect the target (i.e., the arrow cue did not predict the target location). Additionally, when the previous SOA is short, no sequence effect was observed; however, when the previous SOA is long, the sequence effect was shown both at the short and long current SOAs. Furthermore, though both the study of Jongen and Smulders (2006) and the present study found that cueing effects were not influenced by a preceding catch trial, interestingly, we observed that following a catch trial, the overall RTs were facilitated, rather than slowed.
Sequence effects of cueing paradigm have been reported by several studies. Some of them (Dodd & Pratt, 2007; Mordkoff, Halterman, & Chen, 2008) have shown the sequence effect by using non-predictive peripheral cues. These results support the automatic memory check hypothesis (Logan, 1988), which suggests that when performing a task, participants are highly likely to automatically and unintentionally retrieve information from memory in order to facilitate current task performance. Specifically, when the previous trial type (cued or uncued) is consistent with the current trial type, performance will be facilitated, whereas when the previous and current trial types differ, performance is slowed due to the conflict between the two trial types. As a result, the magnitude of cueing effects (i.e., RT facilitation effect) was reduced during short SOAs and the magnitude of IOR was increased during long SOAs after an uncued trial compared with a cued trial. Similar phenomena of automatic memory mechanisms, such as priming of pop out (e.g., Lamy, Amunts, & Bar-Haim, 2008; Maljkovic & Nakayama, 2000; Kristjansson, 2006), and negative priming (e.g., Neill & Valdes, 1992; Egner & Hirsch, 2005), have also been reported by using other paradigms. All of these studies suggested that the sequential processes were afforded by implicit visual memory mechanisms, which operated in an automatic way without conscious intervention. A different hypothesis was proposed when arrow cues were tested by Jongen and Smulders (2006). They argued that the sequence effect was due to some strategies under the voluntary control of the participants. However, because the arrow cues predicted the target location in most of their experimental trials, their explanation may have confounded the voluntary cueing effect within one trial and the automatic sequence effect between trials. The present study extends the findings of Jongen and Smulders (2006) by demonstrating that sequence effects can be observed even when arrow cues are non-predictive to the actual target location. The strategy adjustment hypothesis will predict either no sequence effects or reversed sequence effects with non-predictive arrow cues. Therefore, the present results suggest that sequence effects of arrow cueing are not attributed to the voluntary control or explicit strategies of participants, but attributed to memory retrieval mechanisms, as suggested by the automatic memory check hypothesis.
Although the automatic memory check hypothesis may have revealed the nature of memory under the sequence effect, it does not explain the details of the sequential processes, such as what exactly happens within a spatial cueing task and how the information of previous trials is processed. Some recent studies by Hommel and his colleagues (Hommel, Proctor, & Vu, 2004; Hommel, 2004) proposed a feature-integration account, which tried to explain the sequence effects in spatial attention tasks. The basic idea is that co-occurrence of a cue and a target leads to a transient representation of the relation in which their features (at least the features related to task) are spontaneously integrated without need for voluntary control. This relation would be reactivated in the next trial, and good performance would be expected if the same relation is repeated but interference would occur if it were alternated. According to this feature-integration account, the spatial meaning of the arrow cues and the spatial location of the targets in the present experiment were integrated to form a relation (either cued or uncued). This relation was retrieved in the next trial, and faster response was conducted when the same relation is repeated than when it is alternated.
One thing we need to point out is that the magnitude of the sequence effect observed in present study (19 ms) is very close to the results of previous studies (15 ms at Dodd and Pratt (2007); 17 ms at Mordkoff et al. (2008); around 20 ms at Jongen and Smulders (2006), perceived from their Fig. 4). The stable magnitude of sequence effects across very different experiments provided further evidence to support the automatic memory check hypothesis. In addition, considering the weak average cueing effect in the present study (only 8 ms), it is not difficult to explain why the cueing effect of trials was completely lost when the previous trial was uncued with a 700 ms SOA. The answer is probably that the cueing effect was overpowered by the sequence effect.
In addition, we investigated the influence of previous SOAs and current SOAs on the sequence effect. It was found that when the previous SOA was short, no sequence effect was observed; but when the previous SOA was long, sequence effects were shown at both short and long current SOAs. This is a novel finding in the investigation of sequence effects of cueing paradigm. As mentioned in the introduction, the result can be explained by the different time course of two phases (i.e., initial encoding phase in previous trials and later retrieval phase in current trials) in the sequential processes. However, there are still some issues that need to be considered. First, the impairment of sequence effects when previous trials had a short SOA apparently contradicts the results of Mordkoff et al. (2008), in which the SOA was also very short but resembling sequence effects were observed. One critical difference between the two studies is the different attentional cues. The arrow cue involved in present study is perceptually different but spatially similar whereas the peripheral cue in their study is perceptual identical but the spatial location differs. Therefore, it is easy to integrate a peripheral cue with a target directly based on their spatial locations. On the contrary, arrow cues need to be discriminated before the spatial meaning of them can be acquired. It is widely accepted that though both peripheral cueing and arrow cueing can orient attention reflexively, their relative time courses are very different. In the same way, it is possible that though both peripheral cues and arrow cues could induce sequence effects automatically, some different processes have been involved, like different processing routes and different information that are encoded. This assumption is to some extent supported by the results of several pilot experiments, which are in preparation for a new research in our laboratory. The results showed that alternation of cue types (peripheral onset vs. central arrow) abolished overall sequence effects, whereas alternation of cue types (central gaze vs. central arrow) did not.
Second, we suggest that the influence of previous SOA may reflect a difficulty in encoding the relation between an arrow and a target with a short SOA relative to a long SOA. One may argue that the time interval between trials when a fixation point was presented for a full 2000 ms should be sufficient to let the relation be encoded. However, this view ignores the important fact that the cue and the target have disappeared before the 2000 ms inter-trial interval. Automatic processing is usually transient and stimulus-driven, so it is unlikely that the automatic encoding of the trial could occur without stimulus inputs during the inter-trial interval.
Third issue is whether the influence of previous SOAs can be explained by the strategy adjustment hypothesis. In our opinion, the answer is probably no. Although similar explanation can be made, i.e., that a short perceiving time of the arrow may not be sufficient to enable participants to perceive the trial types on an initial trial, this notion faces the same question as why the perceiving cannot be done during a full 2000-ms inter-trial interval. This period of time should be enough for participants to discriminate between cued and uncued trials voluntarily. Another explanation could be that participants formed the expectation on an initial trial based on not only trial types, but also cue-target SOAs of that trial. Consequently, the participants adapted their utilization of the cue depending on if it correctly or wrongly directed their attention to a location on the previous trial, only when the cue-target interval of the previous trial was long enough. However, it is hard to believe that such a complex and resource-consuming strategy was maintained by participants across the whole experiment in spite of the fact that they explicitly knew the arrow cue was uninformative and SOAs were chosen randomly. In addition, the strategy explanation mentioned above will face many new questions. For example, how participants perceive the length of cue-target SOA as long or as short; is there a certain threshold or is it a relative adjustment? Therefore, at this stage, the influence of previous trial SOA cannot be used to discriminate between automatic and strategy hypotheses, we would like to consider this effect as originating from the different spatial representations between peripheral and central symbolic cues. On the other hand, the automatic memory check hypothesis is supported by the other results of present study, such as significant sequence effects by non-predictive arrow cues and the stable magnitude of the sequence effects across different studies. In all, though more systematic investigations are needed to reveal the precise mechanisms under the present results, our results are more consistent with the automatic memory check hypothesis and might reflect some different temporal characteristics of sequential memory mechanisms between peripheral cues and arrow cues.
Another effect that was examined in this experiment was the influence of preceding catch trials. Consistent with the findings of Jongen and Smulders (2006), we found that though the overall RTs were influenced following a catch trial, it did not influence the cueing effect. This observation supports the distinction between orienting and alerting processes of attention (e.g., Fernandez-Duque & Posner, 1997; Posner & Petersen, 1990). However, contrary to the findings of the present study, Jongen and Smulders found that overall RTs were delayed, rather than facilitated, after a catch trial. Besides their study, the overall delay in RTs after a catch trial has been reported by several other studies (Alegria, 1978; Correa, Lupianez, & Tudela, 2004), and it was attributed to a decrease in preparation for the target. The preparation refers to the general readiness to respond to an anticipated target stimulus after the occurrence of a warning cue. Therefore, if catch trials were considered as trials with extended cue-target SOA, a previous catch trial will reduce the target expectation of participants, resulting in a delayed RT at other SOAs.
Depending on the preparation account, arrow cues need to be utilized under some degree of strategy control to form expectancies about the target appearance. Therefore, it is not surprising to find that the RT delay effect of catch trials was not shown in the present experiment when voluntary control was not required and participants were encouraged to ignore the central cues. Another difference between the experiment of Jongen and Smulders and ours is the cue-target SOA; while the single SOA of their experiment was relatively long, the present experiment used two SOAs with relatively short lengths. This setting may have increased the temporal uncertainty of the target appearance, which in turn reduced the influence of the attention preparation effect. In an exogenous cueing study, Los (2004) reported that target detection was slower when the cue-target SOA of the preceding trial was longer than the SOA of the current trial. However, at the shortest SOA (100 ms) of the two experiments that he conducted, he observed that responses after a preceding catch trial were faster, rather than slower, than that after a preceding long SOA. This observation is very similar to the finding of the present experiment. In all, these results suggest that a catch trial cannot be simply considered as a trial that extended cue-target interval, and it may have a complex influence on the RTs depending on experimental contexts. Further investigation is needed to reveal the precise mechanisms under the RT effect of preceding catch trials.
The present study also has some implications on current and future investigations that involved cueing paradigm. As mentioned previously, a traditional way for measuring attention orienting is to calculate the difference between the mean RTs to detect targets at cued and uncued trials. This manipulation ignored the potential influence of trial-by-trial effects. Though most cueing experiments included an equal number of cued and uncued trials, some researchers used a different proportion of cued trials relative to uncued trials in their experiments to investigate the influence of voluntary control on attention orienting (e.g., Driver et al., 1999; Friesen, Ristic, & Kingstone, 2004). Sequence effects may have influenced their results. For example, when the cue predicts the target location with a rate of 80 percent, there will be more pre-cued trials than pre-uncued trials. As a result, larger average cueing effects for predictive cues than for non-predictive or counter-predictive cues are due in part to sequence effects, not only due to the voluntary control of participants. It is clearly important for future studies to take the influence of sequence effects into account when results are evaluated.
Finally, though both previous studies and our study focused on the sequence effect by traditional cues, such as peripheral cues or arrow cues, the sequential processing is not necessarily limited to these cue types. In another study (Qian, Shinomori, & Song, 2011, in submission), we found significant sequence effects when a face stimulus looking left or right was used as a central cue. Another person’s gaze has been considered as a special attentional cue for its biological significance (e.g., Friesen & Kingstone, 1998, 2003). The findings that sequence effects could occur among very different attentional cues may suggest that sequential processing is a common phenomenon in daily life and the investigation into it will provide better understanding of human cognition systems.
In summary, the present experiment mainly demonstrated that sequence effects of cueing paradigm could be observed for non-predictive arrow cues. In addition, the sequence effects are influenced by the SOA of previous trials. Although the precise mechanisms under the different influence of previous SOAs between peripheral cues and arrow cues need further investigations, overall, our results support the automatic memory check hypothesis for the sequence effects of cueing paradigm more than the strategy adjustment hypothesis.