Abstract: Return inhibition refers to presenting clues in a certain position in space. After a certain period of time (usually about 300 milliseconds after the cue is presented), the response of the stimulus reappearing at the cue position is slower than that at the non-cue position, which is exactly the opposite of the facilitation caused by the cue in the initial stage (about 100 milliseconds). This paper introduces the main research results of the stability, capacity and mechanism of return inhibition, as well as the arguments in the research of object return inhibition and discrimination task return inhibition, and puts forward the prospect of further research.

Attention, return inhibition, clue, goal, SOA

Classification number: B842.3 The research on inhibition of return (IOR) in the field of attention began in the mid-1980s, and gradually became a hot topic in the 1990s. This phenomenon was first reported by Posner and Cohen[ 1]. In the experiment, they investigated the attention orientation effect caused by clues, and the experimental arrangement is shown in figure 1. In the experiment, the subjects were asked to keep their eyes fixed on the box in the middle. First, a small box around it lights up. Secondly, the middle box lights up after a period of time (distracting attention from the first clue position). Finally, the target appears in the surrounding small boxes (small boxes with clues or small boxes without clues). After the target appears, the subject is required to make a key response as soon as possible. The results show that when the time interval (SOA) from the first clue to the target is about 100 milliseconds, the response time of the target at the clue position is shorter than that of the target at the non-clue position, that is, facilitation; When the SOA reaches 200-300 milliseconds, facilitation is replaced by inhibition, that is, the response time of the target at the prompt position is longer than that at the non-prompt position. Posner and Cohen call this inhibition "return inhibition". It is believed that the inhibition of return reflects the flexibility and adaptability of people's psychological mechanism, which makes attention quickly turn to the position that has not been noticed before, which is conducive to rapid search activities.

Figure 1 Posner and Cohen (1984) Typical return inhibition paradigm

Note: The stimulus pictures are presented from left to right. Among the three small boxes of each stimulus picture, the middle small box is the box where the gaze point is located; The small gray boxes in the big boxes A and B represent that the small boxes light up, that is, clues; The asterisks in the big box C 1 and C2 represent the targets. C 1 indicates that the target appears in the clue position, and C2 indicates that the target appears in the non-clue position. Berger[2] discussed the stability of return inhibition by comparing the return inhibition of attention position and neglect position. He first presents an arrow in the center, indicating where the target will appear (the probability of the target appearing in this position is 80%). Participants were asked to pay attention to the small peripheral box indicated by the arrow and ignore another small peripheral box. 500 milliseconds later, a small box around it lit up. Then the target appeared. The results show that when the target appears in the small box of attention, the response time is shorter when the subject uses the information prompted by the arrow. The reaction time when the target appears in the ignore box. However, from the prompt point of view (that is, the surrounding small boxes become bright), no matter in the box pointed by the arrow or in the opposite box pointed by the arrow, the return inhibition appears consistently. Therefore, the return suppression seems to be automatic, and even at the marked position, the return suppression cannot be actively suppressed. However, some studies [2] have found that when the time when the indicator arrow appears is shorter than the time when the target appears, the return inhibition decreases. Therefore, although return inhibition is a strong automatic phenomenon, it can also be actively suppressed to some extent.

Other studies have discussed whether the return inhibition will be suppressed when the position where the target appears is fixed [2]. In the experiment, the experimental condition of half of the subjects is that the target appears on the left side, and the experimental condition of the other half is that the target appears on the right side. At the beginning of the experiment, the first clue appears in the outer box where the target will appear, or in the center box. 500 milliseconds after the first clue, prompt the central box. After 500- 1000 milliseconds, the target appears. The results show that the reaction time of the first clue in the outer box is slower than that of the first clue in the center box. It shows that the return inhibition still exists even if the target position is fixed. It can be seen that the return inhibition does have quite strong stability. Pratt and Abrams [3] first investigated the ability of return inhibition. In the experiment, they successively used the method of prompting two small boxes around them. In one case, two clues appear in the same box, and in another case, two clues appear in different boxes. The results show that there is inhibition of return in the overlapping position of two clues and the non-overlapping subsequent clues. But when the two clues do not coincide, the reaction time of the first clue position is the same as that of the non-clue position when they coincide. This shows that the first clue has no effect on the reaction time, that is, there is no inhibition of return. Therefore, it is considered that only the recent clues cause the inhibition of return; When prompts appear in different positions behind, the inhibition caused by the prompts in the previous position disappears automatically.

Tipper et al. [4] think that it is limited for Pratt and Abrams to use only two peripheral boxes when discussing the inhibition ability of return. Tipper et al. used four small peripheral frames and continuously prompted three of them in an experiment. It was found that there was inhibition of return in all three consecutive prompt positions. Therefore, it is considered that the inhibition ability of return is not one, but at least three.

Abrams and Pratt[5] experiment 1 repeated the experiment of Tipper et al. and got the same result. However, in Experiment 2, when two adjacent or separated small boxes in four small boxes were prompted in turn, it was found that under the adjacent condition, the two prompted positions were inhibited from returning; Under the condition of separation, only the second clue position appears return inhibition. In experiment 3, they further investigated that when the target has six possible positions, they continuously prompted three non-adjacent positions, and it was also found that the return inhibition only appeared in the last prompted position. Therefore, they suggest that there may be two different inhibition of return, one is diffuse, and its capacity is very large; One is centralized, with small capacity, such as only one. When using simple conscious tasks (as shown in figure 1), almost all the studies have found the inhibition of return. However, when the discrimination task is used in the experiment, many experiments have not found the inhibition of return. For example, Terry et al [6]. In the experiment, they took letters A and B as targets, and C and D as interference terms. When the target letter appears (no matter which letter), press the space bar to respond. If interference letters appear, they don't press the key (experiment 1). The target letter and the interference letter are presented in two surrounding boxes respectively, and the corresponding keys are pressed according to the position (left and right) of the target letter to react (Experiment 2). It was found that the inhibition of return did not appear under the above two conditions. In addition, other experiments on color discrimination, size discrimination, brightness discrimination and pattern discrimination have not found return inhibition [2,7].

However, some studies have found that there is inhibition of return in the task of distinguishing reactions. Pratt [8] presents the target and interference items at two peripheral positions (prompt position and non-prompt position) at the same time, requiring the subjects to move their eyes to the target as soon as possible. When recording the subjects' eye movement reaction, it was found that there was inhibition of return. Lupianez et al. [9] applied five-level SOA in color identification task, namely100/400/700/1000/1300ms. It is found that when SOA is 100 and 400ms, clue location becomes easy. Return suppression occurs when the SOA is 700 milliseconds and disappears at 1300 milliseconds. Therefore, Lupianez thinks that there is return inhibition in the discrimination task, but it appears later than 300 milliseconds in the general detection task and disappears quickly. Return inhibition can be divided into position-based return inhibition and object-based return inhibition. The regression inhibition we mentioned earlier is generally position-based regression inhibition, that is, when the follow-up target is in the same position as the previous clue, the reaction time is longer; Object-based return inhibition means that when the object in the target position is the same as the object in the previous clue, the reaction time is longer.

Tipper et al. [10, 1 1] first reported the existence of object return inhibition. In their initial experiment, they turned the static peripheral box into motion, as shown in Figure 2. Three small black squares appear on the screen, and the middle square is the location of the gaze point. At the beginning, one square around is higher than the gaze point and the other is lower than the gaze point. After 500 milliseconds, the surrounding squares began to rotate. When it turns to the horizontal position, any surrounding squares are prompted with 100 milliseconds. After the interval of 100 milliseconds, the center square is prompted with 100 milliseconds in the same way, and the squares continue to move in the original direction while being prompted. When the box moves to the position of 90 or 180, the target appears in any surrounding box, and the subject is required to press the key to respond as soon as possible. The results show that there is inhibition of return at both 90 and 180 positions, that is, the reaction time of the target in the original clue box is longer than that in the original non-clue box. Therefore, this study supports object-based return inhibition.

Fig. 2 shows the experimental method of object-based return inhibition of Tipper et al. (199 1, 1994).

Note: In the figure, A is the starting position of the square movement, the white square in B is the peripheral clue, the white square in C is the central clue, D 1 is the target on the original clue square at 90, and D2 is the target on the original clue square at 180. The arrow in the figure indicates the moving direction of the square. The figure does not show that the target appears in the original clue-free box.

Abrams and Dobkin [12] found that if the object appeared in the prompt box after moving 90 degrees, the eye saccade movement time would be longer than that in the non-prompt box. The results are consistent with those of Tipper et al. (199 1, 1994). Law et al. [13] also found that there is color-based return inhibition. However, the experiments of Muller and Muhlenen[ 14] do not support object return inhibition. They adopted a research method similar to that of Tipper et al. (199 1, 1994). As a result, they found no inhibition on the return of objects, but there was a kind of attention tracking from left to right. Maylor and Hockey[ 15] discussed several reasons of return inhibition. First of all, in the experiment 1, they found that when the n+1 target appeared in the position of the nth target, the reaction time was longer than that of the opposite nth target through the target-target experimental paradigm (corresponding to the clue-target paradigm introduced in figure1). So the inhibition of return is ruled out because there is no response to clues [15]. In experiment 3, when the gaze point when the target appeared shifted below the gaze point in the first clue position stage, it was found that the inhibition of return still existed. Therefore, it is considered that the inhibition of regression cannot be caused by emotional habituation caused by clues. In addition, in the experiment 1-3, it was found that the duration of return inhibition exceeded 1000 milliseconds, which ruled out that return inhibition was caused by the pre-masking effect caused by clues.

At present, the discussion on the mechanism of return inhibition mainly focuses on whether there is attentional inhibition in return inhibition. Posner and Cohen[ 1] think that return inhibition means that attention is inhibited when returning to the previously noticed position, so it is vividly called return inhibition. This assumption has caused two different arguments. A hypothesis supporting Posner and Cohen is called the attention view of return inhibition, that is, return inhibition is caused by the inhibition of attention; Another view is that the inhibition of return is caused by reaction-related inhibition, and there is no attentional inhibition. This view is called reaction view.

The research of Abrams and Dobkin[ 12] supports the attention view of return inhibition. After giving clues to a small box around them, they put forward two different requirements. One is to present the target in a small box and ask the subjects' eyes to move in the direction where the target appears; One is to present an arrow at the central fixation point, indicating that the eyes of the subject move in the direction indicated by the arrow. The results show that the former condition (exogenous cue) has greater inhibitory effect than the latter condition (endogenous cue). They believe that this is because there is not only reaction inhibition, but also perceptual inhibition under exogenous conditions; However, under endogenous conditions, there is only reactive inhibition of return, and there is no perceptual inhibition of return, so the former has greater inhibition on return than the latter. It can be seen that this study believes that the inhibition of return is partly due to the inhibition of perception.

The research based on perceptual methods supports the reaction view of return inhibition. The hypothesis based on perceptual method is that if the inhibition of return is caused by the inhibition of attention or perception, then this perceptual inhibition should also be manifested in a single perceptual task. In the study of regression suppression, the perceptual methods mainly include time series judgment and line segment similarity. The operation of time sequence judgment method is to present two signals that are very close in time in different spatial positions (clue position and non-clue position), so that subjects can judge which signal appears first [16]. The operation of line segment similarity method is to present a line segment, and the two ends of the line segment are in two different spatial positions (clue position and non-clue position), so that the subjects can report from which position the line segment moved to which position [17]. Both perceptual methods assume that the stimulus at the location where attention is directed is perceived first. Perception-based method corresponds to the method of investigating rapid response (or reaction time method), which is called response-based method and is a general method of return inhibition research.

Maylor[ 16] Experiment 4 adopted the method of time series judgment for the first time. It is found that although return inhibition usually appears in the clue position when SOA exceeds 300 milliseconds, it is not perceived as backward in timing judgment, but is preferred. Therefore, the inhibition of return is manifested in reaction, not perception (attention). Posner et al. [18] also adopted the research method of time series judgment, and the results were similar to those of Maylo's research, which supported the reaction viewpoint of return inhibition. Schmidt[ 17] adopted the method of line segment movement, and found that the reported line segment moved from the clue position to the non-clue position, that is, the clue position had attention priority, which lasted until 900 milliseconds after the clue was presented. Therefore, it is considered that the inhibition of return is not obvious.

Part of the research on the inhibition mechanism of return comes from neurophysiology. Posner et al [18] found that the inhibition of return disappeared in patients with moderate brain injury (progressive advanced nuclear paralysis). However, patients with Parkinson's disease and patients with frontal lobe, temporal lobe and parietal lobe injuries did not show the disappearance of return inhibition. Therefore, it is considered that the inhibition of return is caused by the activation of the greater omental tectum channel (the visual motor center of the midbrain). Rafal et al. [19] found that there was a higher inhibition of return in the temporal visual field when comparing the inhibition of return in the temporal visual field and the nasal visual field of normal subjects. In addition, newborns and adult patients also showed the temporal visual field advantage of return inhibition.