Recent attention models have proposed that, in addition to bottom-up (or stimulus-driven) and top-down (or goal-driven) controls (e.g., Corbetta & Shulman, 2002), a third mechanism, termed “history,” can serve as a key concept for understanding various attention phenomena (see Awh et al., 2012; Failing & Theeuwes, 2018; Theeuwes, 2018, 2019, for reviews). The history consists of the “selection history” (e.g., a priming effect; Tulving & Schacter, 1990) and the “reward history” (e.g., reward associative learning; e.g., Anderson et al., 2011a, 2011b). Studies concerning reward history have shown that rewarding experiences modulate attentional selectivity by enhancing visual processing to the reward-associated stimuli (see Anderson, 2013, 2016; Chelazzi et al., 2013; Le Pelley et al., 2016, for reviews). These reward effects on attentional selection have been reported to arise from various features, such as color (e.g., Anderson et al., 2011a), shape (Della Libera & Chelazzi, 2009; Wang et al., 2013), orientation (Laurent et al., 2015; Lee & Shomstein, 2014), and object category (e.g., Failing & Theeuwes, 2015; Yokoyama et al., 2015).

While a great deal of research on reward history has focused on the association between rewards and stimulus features (i.e., color, shape, orientation, and object category; see Anderson, 2013, 2016), less is known about how experiences of the association between rewards and spatial locations of stimuli may modulate attentional selection. Unlike the feature–reward associations, most studies exploring the effect of location–reward associations have failed to show a significant bias in attentional selection (e.g., Jiang et al., 2015; Won & Leber, 2016, 2018; but see Chelazzi et al., 2014). For example, in a visual search task, participants received a reward depending on a quadrant of the display where a target appeared; the target could be presented in any quadrant with an equal probability, but the expected rewards were different across four quadrants (Won & Leber, 2016, Experiment 1a). As a result, the reaction times (RTs) for a target appearing at the high-reward quadrant were equivalent to those for a target appearing at another quadrant; that is, the manipulation of the location–reward association did not affect attentional selection. Similar results were also reported in other studies (e.g., Jiang et al., 2015, Experiments 2 and 3).

Although most studies have reported the lack of the effect of location–reward associations, a few studies have indicated the significant effect of location–reward associations in limited situations (e.g., Anderson & Kim, 2018a, 2018b; Chelazzi et al., 2014; Jiang et al., 2015, Experiment 4; Sisk et al., 2020; Won & Leber, 2016, Experiment 4b, 2018). The literature suggests that the prioritization of highly rewarded locations depends on the nature of the task. No evidence of the location–reward effect has been found for visual search tasks involving a single target (e.g., Jiang et al., 2015; Won & Leber, 2016), whereas the effect was found for visual search tasks with multiple targets (e.g., Chelazzi et al., 2014), or for visual choice tasks in which participants were asked to click on any item on the display (e.g., Won & Leber, 2016, 2018). These results suggest that the location–reward association effect is evident only when a task induces competition for attention selection between potential targets.

The implicitness or explicitness of associative learning is controversial; however, a recent study has demonstrated that explicit knowledge of the reward structure is associated with spatial biases toward high-reward locations (Sisk et al., 2020). Sisk et al. (2020) used a hybrid search and choice task with multiple targets among multiple distractors and demonstrated that only participants aware of the reward structure showed spatial biases toward high-reward locations. This finding suggests that attentional competition is insufficient for the location–reward association effect, and awareness of the reward structure is necessary even for tasks that might cause attentional competition between targets.

Jiang et al. (2015) reported the location–reward association effect in a visual search task involving a single target when participants were explicitly informed about the association. Before the experiment, their participants were told that rewards would appear more often in one quadrant than in the other three quadrants. Also, they were instructed to identify a high-reward quadrant during the visual search task and report it after each block. Results revealed that faster RTs were observed in the high-reward quadrant than in the low-reward quadrant only for participants who correctly reported the association between rewards and the high-reward quadrant. The authors suggested the necessity of explicit instructions concerning the location–reward association for attentional biases in the visual search to emerge.

Jiang et al. (2015) demonstrated the effect of the location–reward association on attentional selection when participants were required to intentionally look for a high-reward quadrant concurrently with a visual search task involving a single target. However, the findings of Jiang et al. have two possible explanations. One explanation is that goal-driven manipulation is necessary for attentional biases to occur, and the other explanation is that the awareness of the association, without any goal-driven manipulation, is sufficient for this effect. Therefore, it remains unclear whether instructions and exploring the reward structure are necessary to induce the spatial bias to high-reward quadrants. Here, we used the term “goal-driven manipulation” to denote the instruction to look for an association between rewards and target locations intentionally. When participants’ task includes identifying a high-reward quadrant, then, predictably, the participants would intentionally search for targets among high-reward quadrants because they would focus on the location–reward association and conduct the task while remembering the high-reward quadrant to report after each block. The strategy resulting from the goal-driven manipulation could contribute to the formation of location–reward associations in the former case. In contrast, the awareness of the location–reward association itself is critical, and the association’s spontaneous awareness is sufficient for attentional biases to occur in the latter case. It is considered that the effects of the feature–reward association (e.g., the color–reward association) on attentional selection could occur regardless of observers’ current goals (e.g., Anderson et al., 2011a; Failing et al., 2015; Le Pelley et al., 2015). Therefore, the necessity of goal-driven manipulations is essential for understanding the differences between the effects of the feature–reward and the location–reward associations.

We designed this study to clarify the role of goal-driven manipulations in a visual search task involving a single target. We investigated whether the effect of location–reward associations on attentional selection would be observed in a visual search task, even if the participants were not instructed to look for a high-reward location. In other words, Jiang et al. (2015) explicitly instructed their participants to be aware of the high-reward quadrant. In contrast, we examined whether spontaneous awareness of the location–reward association would elicit a spatial bias to a location signaling a high reward. We conducted three experiments using a visual search task in which participants were required to identify a line orientation (vertical or horizontal) inside circles. The location of each circle predicted the magnitude of the reward that participants would receive for correct responses. In Experiment 1, a target line segment was presented among distractor lines, and participants had to look for a target and report its orientation. However, this manipulation might allow participants to use specific search strategies that might affect the awareness of location–reward associations. Therefore, we made the target more salient in Experiments 2 and 3 such that it would capture attention and reduce the possibility of top-down control strategies for finding the target. If goal-driven manipulations (i.e., instructions to look for an association between rewards and target locations) were necessary to elicit the location–reward effect, the association’s awareness would have no effect because participants were not encouraged to look for an association. In contrast, if spontaneous awareness of the association were sufficient to elicit the location–reward effect, the effect would be observed in participants who were aware of the association.

Experiment 1

Method

Participants

Young adults (N = 60, 23 women, 19–33 years of age, mean age = 22.6 years, SD = 2.5 years) who reported normal or corrected-to-normal visual acuity participated in Experiment 1. Participants earned up to 700 JPY (about US$7; mean payout = 675 JPY, SD = 50.8 JPY) dependent on performance and received monetary compensation for their participation (i.e., 1,250 JPY per hour). This study was approved by the institutional review board of the National Institute of Advanced Industrial Science and Technology (AIST), and all participants gave written informed consent before the experiment. Note that we did not perform a power analysis to determine a sample size before collecting data. Since the main purpose of this study was to examine the effect of spontaneous awareness on reward biases to spatial locations, the proportion of participants that would be aware of the location–reward association was unknown prior to the experiment. Therefore, a larger number of participants was needed. Additionally, the number of participants must be multiples of 12 because the task in the current study had 12 combinations of high/low-reward locations (see below). Based on these considerations, we predetermined the number of participants (N = 60).

Apparatus and stimuli

Participants were tested individually; each was seated in a dark room; their heads rested on chin rest for a viewing distance of approximately 57 cm from the computer display. Stimuli were generated using MATLAB software (R2018b, The MathWorks) with the Psychophysics Toolbox extensions (Version 3.0.15; Brainard, 1997) and presented on a BenQ ZOWIE XL2540 monitor (1,920 × 1,080 resolution, 60-Hz refresh rate). Participants entered responses manually on a keyboard (“z” and “m” keys).

Experimental stimuli consisted of a series of black displays, as shown in Fig. 1. A white fixation cross (0.5° × 0.5° of visual angle) was presented at the center of the initial black display. The search display had outlines of four gray circles (2.3° × 2.3° of visual angle) with a gray line segment (0.8° of visual angle) within each circle (see Fig. 1). The circles were presented at the top-left, top-right, bottom-left, and bottom-right locations of the display. The distance between the fixation cross and the center of each circle was 5° of visual angle. A target was defined as a vertically or horizontally oriented line segment. In each trial, one of the circles had either a vertical or a horizontal line segment (i.e., target), and the other three circles had a line segment that was randomly tilted by 45 degrees left or right (i.e., distractor). The feedback display indicated reward points received on a current trial.

Fig. 1
figure 1

The trial sequence of Experiment 1. The task was to report a line orientation of a target (vertical or horizontal) as quickly and accurately as possible. The reward feedback was presented only for correct responses (10 points for a high reward/1 point for a low reward/0 point for no reward)

Procedure

Each trial began with the fixation display presented for a randomly determined period of 400–600 ms. As suggested in Fig. 1, the fixation display was followed by the search display for 1,200 ms or until response. Participants were instructed to report a target orientation by pressing one of two keys with their left or right hand; the “z” key for the horizontal target and the “m” key for the vertical target, respectively. Participants were required to make a response as quickly and accurately as possible. Following the blank display for 500 ms, the feedback display appeared for 1,250 ms. Correct responses resulted in the reward feedback showing either a high reward (“+ 10 points”), low reward (“+ 1 point”), or no reward (“+ 0 point”), while error responses were followed by the “WRONG” feedback. Also, “TOO SLOW” was presented when no response was made within the search time window (i.e., 1,200 ms). Participants were instructed that reward points would be presented only for correct responses. The next trial started after a blank display for 750 ms. The experiment started with a practice session (24 trials) with no reward feedback, followed by 10 experimental blocks of 48 trials each, yielding a total of 480 experimental trials. The target locations (top-left/top-right/bottom-left/bottom-right) and the target orientations (vertical/horizontal) were balanced in each block, and trials were presented in a random order. Participants could take a break and see the cumulative total reward points after each block ended.

We manipulated the association between the target location and the magnitude of reward. One of the four possible target locations (e.g., top-right) was associated with a high reward (“+ 10 points”), and another location (e.g., bottom-left) was associated with a low reward (“+ 1 point”). The remaining two locations (e.g., top-left and bottom-right) did not lead to any rewarding results regardless of performance (“+ 0 point”). Since the target appeared at one of four locations with an equal probability in each trial block, one-fourth of trials could be a high-reward condition, and another one-fourth of trials could be a low-reward condition. For each participant, the specific association between the target location and the magnitude of reward was constant throughout the experiment.

There were 12 combinations of high/low-reward locations (see Fig. 2 for all combinations). Sixty participants were assigned across the 12 combinations (i.e., five participants for each). For the analysis, these combinations were categorized according to the following three types (i.e., categories); diagonal, horizontal, and vertical location types. In the diagonal type, high-reward and low-reward locations were positioned diagonally (e.g., top-left for a high reward and bottom-right for a low reward). In the horizontal type, the reward locations were positioned horizontally (e.g., top-left for a high reward and top-right for a low reward). In the vertical type, the reward locations were positioned vertically (e.g., top-left for a high reward and bottom-left for a low reward). Unlike the previous study (Jiang et al., 2015), participants were not given any information regarding the location–reward association.

Fig. 2
figure 2

The location–reward associations from diagonal, horizontal, and vertical location types (H, L, and N represent high, low, and no reward, respectively)

After the experiment, all the participants made written responses to a questionnaire on how they earned rewards during the task. They responded by choosing a “Yes” or “No” option to the question, “Was there any regularity about the magnitude of the reward points that you won on correct trials?” Then, the participants that responded “Yes” to the first question responded freely to the question “What type of regularity did you observe? Please describe the details.” We used these responses as an index of participants’ awareness of the location–reward association.

Results

Overall reward effects

Accuracy was calculated as the percentage correct of all trials. The overall accuracy was sufficiently high (92.8%; see Appendix Tables 1, 2 and 3). However, the accuracy was not affected by the reward. Therefore, we focused and reported only the RT data in the following analysis. The trials in which no response was given within 1,200 ms (2.28% of trials) and trials with incorrect responses were excluded from the RT analysis.

To investigate whether the location–reward association affected attentional selection in visual search, we first compared RTs between the reward conditions (see Fig. 3). A two-way repeated-measures analysis of variance (ANOVA), with both block (1–10) and reward (high/low/no) as within-participant factors. This analysis revealed a significant main effect of reward, F(2, 118) = 16.75, p < .001, ηp2 = .221, but no main effect of block, F(9, 531) = 0.96, p =. 474, ηp2 = .016. The Block × Reward interaction also reached statistical significance, F(18, 1062) = 4.99, p < .001, ηp2 = .078. Post hoc comparisons using Shaffer’s (Modified Sequentially Rejective Bonferroni) procedure showed significant differences between the high-reward and low-reward conditions and between the high-reward and no-reward conditions in Blocks 4–10 (ps < .01). Furthermore, the differences between the low and no-reward conditions were also significant in Blocks 8–10 (ps < .04).

Fig. 3
figure 3

Mean reaction times by reward conditions over Blocks 1 to 10 in Experiment 1. Error bars indicated one standard error of the mean

Awareness

As mentioned in the Introduction, a previous study demonstrated that the effect of the location–reward association appeared only for participants who could correctly report the association (Jiang et al., 2015). To examine the influence of awareness of the association, we classified our participants into two groups based on their reported awareness revealed in response to questions after the experiment. Twenty-five out of 60 participants reported the correct location–reward associations, whereas the remaining 35 participants could not report the rule at all or misreported the rule (e.g., reward points were given depending on reaction times). No participants reported the association partially.

We performed a mixed ANOVA on mean RTs, with awareness (aware/unaware) as a between-participants factor and block (1–10) and reward (high/low/no) as within-participants factors. As a result, we obtained a significant Awareness × Block × Reward interaction, F(18, 1044) = 7.31, p < .001, ηp2 = .112, suggesting that awareness of the location–reward association influenced the magnitude of the Block × Reward effect. However, we interpreted the interaction with awareness with caution because the participants’ numbers were different between the aware and unaware groups. To further investigate how awareness affected the reward effect, we submitted RT data to two-way ANOVAs, with block (1–10) and reward (high/low/no) as within-participants factors separately for the aware and unaware groups (see Fig. 4). For the aware group (N = 25), the ANOVA revealed a significant main effect of reward, F(2, 48) = 37.11, p < .001, ηp2 = .607, but no main effect of block, F(9, 216) = 1.20, p = .299, ηp2 = .048. The Block × Reward interaction was also significant, F(18, 432) = 9.37, p < .001, ηp2 = .281. Post hoc comparisons using Shaffer’s procedure showed significant differences between the high-reward and low-reward conditions and between the high-reward and no-reward conditions in Blocks 3–10 (ps < .04). Furthermore, the differences between the low-reward and no-reward conditions were significant in Blocks 5–10 (ps < .05). For the unaware group (N = 35), on the other hand, the same two-way ANOVA showed neither a main effect of reward, F(2, 68) = 0.53, p = .590, ηp2 = .015, nor an interaction, F(18, 612) = 0.65, p = .864, ηp2 = .019. In sum, these findings indicated that awareness of the location–reward association was necessary for the reward effect.

Fig. 4
figure 4

Mean reaction times by reward conditions over Blocks 1 to 10 separately for the awareness groups in Experiment 1. Error bars show one standard error of the mean

Location types

We were also interested in the difference in performance between the location types (diagonal/horizontal/vertical, see the Method section). Since the reward effects were observed only when participants were aware of the location–reward association, we used the data from the aware group for this analysis. Note that it is hard to provide conclusive evidence in this analysis because the sample sizes were relatively small due to the awareness issue; 10, eight, and seven participants could report the association correctly in the diagonal, horizontal, and vertical location types, respectively. Thus, we need to interpret the following results with some caution.

To assess whether the reward effect was affected by the location type, a mixed ANOVA with location type (diagonal/horizontal/vertical) as a between-participants factor and block (1–10) and reward (high/low/no) as within-participants factors was performed for the aware group (see Fig. 5). The main effect of location type, F(2, 22) = 0.65, p = .530, ηp2 = .056, the Location type × Block interaction, F(18, 198) = 0.77, p = .729, ηp2 = .066, the Location Type × Reward interaction, F(4, 44) = 0.87, p = .491, ηp2 = .073, and the Location Type × Block × Reward interaction, F(36, 396) = 1.26, p = .147, ηp2 = .103, all were not significant, suggesting no sign of the effect of the location types.

Fig. 5
figure 5

Mean reaction times by reward conditions over Blocks 1 to 10 for the aware group separately for the location types in Experiment 1. Error bars show one standard error of the mean

Discussion

In Experiment 1, we observed the effect of the location–reward associations when one of the possible target locations was associated with a high reward, and another location was associated with a low reward. RTs in the high-reward condition were faster than those in the low-reward and no-reward conditions, and these reward effects were mainly observed in participants that could correctly report the association between locations and rewards, which was consistent with the finding of previous studies on location–reward association (e.g., Jiang et al., 2015).

In Experiment 1, targets and distractors were presented simultaneously. Therefore, a target line segment might not have been salient because participants had to search for a target after the search items appeared in each trial. As a result, the participants might have deployed attention to each item in a serial manner. Therefore, the order of examination (i.e., the strategy) could significantly impact search performance. One possible explanation of the observed reward effect is that the reward manipulation affected the search strategy, such that when participants noticed the association, they intentionally started looking for a target among high-reward locations, making it easier and faster to identify the target orientation when a target appeared in high-reward than in low-reward or no-reward locations. Another possibility is that the reward modulated attentional allocation and/or the examination of the target, regardless of the top-down search strategy, such that knowledge about location–reward associations made the locations associated with a high reward more salient, and more likely to capture attention than locations that were associated with low or no reward.

To test these possibilities, in Experiment 2, we removed line segments except for the target circle in the search display. Also, the placeholders were presented in the fixation display before the target line segment was shown so that a target captured attention by abrupt onset (Yantis & Jonides, 1984). These manipulations allowed us to examine the effect of the strategy that participants used when looking for a target in a high-reward location because a target would pop out when it appeared in the search display. If the search strategy caused the reward effect observed in Experiment 1, then that effect would be eliminated in Experiment 2. On the other hand, if the knowledge of the association between locations and rewards modulated attentional allocation and/or target examination, then we would observe the reward effect in Experiment 2.

Experiment 2

Method

Participants

Young adults (N = 60, 27 women, 18–34 years of age, mean age = 22.3 years, SD = 2.8 years) who did not participate in Experiment 1 were recruited for Experiment 2. As in Experiment 1, although a power analysis was not conducted, we predetermined the number of participants before the data collection. All of the participants reported normal or corrected-to-normal visual acuity. Similar to Experiment 1, the participants could earn up to 700 JPY (mean payout = 687 JPY, SD = 34.3 JPY) based on their performance, in addition to the monetary compensation for their participation (i.e., 1,250 JPY per hour). All participants gave their written informed consent before participating in the experiment.

Apparatus, stimuli, and procedure

The apparatus, stimuli, and procedure were identical to Experiment 1, with certain exceptions (see Fig. 6). In Experiment 2, four gray circles (2.3° × 2.3° of visual angle) were presented at the top-left, the top-right, the bottom-left, and the bottom-right locations of the fixation display. In the search display, one of the four gray circles had a vertically or horizontally oriented line segment (i.e., the target), whereas the line segments inside the other three circles were removed. The participants’ task was to report the orientation of the target, similar to Experiment 1.

Fig. 6
figure 6

The trial sequence of Experiment 2

Results

Overall reward effects

Similar to Experiment 1, Experiment 2 focused on RTs because we did not observe reward effects in the data on accuracy (94.9%; see Appendix Tables 1, 2 and 3). We excluded trials with no responses within 1,200 ms (0.93% of trials) and trials with incorrect responses from the RT analyses. We conducted a two-way repeated-measures ANOVA, with block (1–10) and reward (high/low/no) as within-participants factors to examine whether the association between locations and rewards affected the search RTs (see Fig. 7). The result showed significant main effects of block, F(9, 531) = 7.29, p < .001, ηp2 = .110, and reward, F(2, 118) = 26.25, p < .001, ηp2 = .308. We also observed a significant Block × Reward interaction, F(18, 1062) = 1.94, p = .011, ηp2 = .032. Post hoc comparisons using Shaffer’s procedure were performed for the observed interaction. The differences between the high-reward and low-reward conditions were significant in Blocks 4–10 (ps < .03), except for Block 8 (p = .62), and the comparisons between the high-reward and no-reward conditions showed a significant difference in Blocks 2–10 (ps < .04). Moreover, significant differences between the low-reward and no-reward conditions were observed in Blocks 2, 4, 6, and 8 (ps < .04).

Fig. 7
figure 7

Mean reaction times by reward conditions over Blocks 1 to 10 in Experiment 2. Error bars indicated one standard error of the mean

Awareness

Similar to Experiment 1, we classified the participants into two groups based on their reported awareness and examined the influence of awareness about location–reward associations. Of the 60 participants, 40 reported correct location–reward associations, whereas 19 participants could not report any association. Due to an experimenter error, we could not collect the awareness questions results from one participant, and the data from this participant was excluded from the analysis of awareness.

A mixed ANOVA conducted on mean RTs, with awareness (aware/unaware) as a between-participants factor and block (1–10) and reward (high/low/no) as within-participants factors revealed a significant Awareness × Block × Reward interaction, F(18, 1026) = 1.95, p = .010, ηp2 = .033. This three-way interaction suggested that awareness of location–reward associations modulated the magnitude of the Block × Reward effect. However, because the numbers of participants were different between the aware and unaware groups, the interactions that included awareness should be interpreted with caution. We conducted two-way ANOVAs, with block (1-10) and reward (high/low/no) as within-participant factors (see Fig. 8) separately for the aware and the unaware groups. The aware group (N = 40) indicated significant main effects of block, F(9, 351) = 6.36, p < .001, ηp2 = .140, and reward, F(2, 78) = 29.22, p < .001, ηp2 = .428. Furthermore, the Block × Reward interaction was significant, F(18, 702) = 1.85, p = .017, ηp2 = .045. Post hoc comparisons using Shaffer’s procedure showed significant differences between the high-reward and low-reward conditions in Blocks 4, 5, 7, 9, and 10 (ps < .04) and differences between the high-reward and no-reward conditions in Blocks 2–10 (ps < .05). Moreover, the differences between the low and no-reward conditions were significant in Blocks 2–10 (ps < .05), except for Blocks 5 and 10 (ps > .07).

Fig. 8
figure 8

Mean reaction times by reward conditions over Blocks 1 to 10 separately for the awareness groups in Experiment 2. Error bars show one standard error of the mean

The identical two-way ANOVA in the unaware group (N = 19) revealed neither a main effect of the block, F(9, 162) = 1.54, p = .139, ηp2 = .079, nor the reward, F(2, 36) = 1.59, p = .218, ηp2 = .081. However, unlike in Experiment 1, there was a significant Block × Reward interaction, F(18, 324) = 1.96, p = .012, ηp2 = .098. Post hoc comparisons revealed significant differences between the high-reward and low-reward conditions in Block 7 (p = .01) and between the high-reward and no-reward conditions in Blocks 6 and 7 (ps < .02), suggesting that the location–reward effect did not completely disappear in participants that did not report an association. However, contrary to the aware group, this effect was neither systematic nor permanent in the unaware group. These results suggested that awareness of location–reward associations played an important role in the observed effects of rewards.

Location types

Similar to Experiment 1, we compared performances based on location types in the aware group. We assigned 11 (diagonal), 17 (horizontal), and 12 (vertical) participants to the aware group. Then, we conducted a mixed ANOVA, with location type (diagonal/horizontal/vertical) as a between-participants factor and block (1–10) and reward (high/low/no) as within-participants factors (Fig. 9). The results showed that the Location Type × Block interaction was significant, F(18, 333) = 2.46, p < .001, ηp2 = .117. However, neither the main effect of location type, F(2, 37) = 0.53, p = .594, ηp2 = .028, nor the Location Type × Reward interaction, F(4, 74) = 0.16, p = .959, ηp2 = .008, nor the Location Type × Block × Reward interaction, F(36, 666) = 0.96, p = .532, ηp2 = .050, were significant. Therefore, we did not find clear evidence that location–reward mappings modulated the prioritization of attention.

Fig. 9
figure 9

Mean reaction times by reward conditions over Blocks 1 to 10 in the aware group separately for location types in Experiment 2. Error bars show one standard error of the mean

Discussion

In Experiment 2, we replicated the main finding of Experiment 1, even when the target on the search display captured attention. In the aware group, RTs were faster in the high-reward than in the low-reward and no-reward conditions. However, such reward effects were not systematically observed in the unaware group. Also, similar to Experiment 1, there was no evidence of any location–reward mapping effect on attentional allocation to a high-reward location. These results suggest that awareness of location–reward associations plays a vital role in inducing attentional allocation based on reward associations, rather than the search strategy.

Nevertheless, it remains possible that participants aware of the location–reward association fixated their eyes on a high-reward location before the onset of the search display to respond to high-reward targets quickly and accurately. More specifically, although there was no advantage of a bias to a specific location because a target could appear in any quadrant with an equal probability, faster response to a target could be expected if participants were already looking at a high-reward location. To test this possibility, we replicated Experiment 2 and monitored the participants’ eye position during the task in Experiment 3. If the eye position at the onset of the search display (i.e., the initial eye position) resulted in faster RTs in the high-reward condition, the reward effect would disappear if participants had to fixate at the center of the screen until the search display appeared. In contrast, if the reward effect emerged irrespective of the initial eye position, then similar reward effects would be observed even if the initial eye position was restricted.

Experiment 3 was similar to Experiment 2, except that we measured the participants’ initial eye positions and repeatedly instructed them to fixate their eyes on the fixation cross until the onset of the search display. We only analyzed the data of participants who were aware of the location–reward associations in Experiment 3 because Experiments 1 and 2 showed that location-based reward effects were observed only in the location–reward association aware group.

Experiment 3

Method

Participants

As in the previous experiments, because there were 12 combinations of high/low-reward locations, the number of participants must be multiples of 12. Also, Experiment 2 had revealed large effect sizes (f = 0.87) in the difference between the location–reward conditions for participants (n = 40) that could report the location–reward association. The proportion of participants that would be aware of the location–reward association was unknown prior to running the experiment. Therefore, we assumed that half of the participants would be aware of the association. In this case, a power analysis conducted with G*Power (Faul et al., 2007) indicated that a sample size of n = 12 would give a power of .95 to detect an effect size of f = 0.50 for the main effect of reward in a repeated-measures ANOVA. We collected data from 24 young adults (10 women, 18–27 years of age, mean age = 22.1 years, SD = 2.7 years) that did not participate in the previous experiments. All participants reported normal or corrected-to-normal visual acuity. The participants could earn up to 700 JPY (mean payout = 696 JPY, SD = 20.4 JPY) based on their performance, in addition to the monetary compensation of 1,250 JPY per hour for participation in the study. All the participants provided their written, informed consent before participating in the experiment.

Apparatus, stimuli, and procedure

The apparatus, stimuli, and procedure were identical to Experiment 2, with the following exceptions. Participants held their chins on a chin rest and viewed the monitor from a distance of 70 cm. An EyeLink 1000 eye tracker (SR Research Inc.) was used to monitor each participant’s eye position. The sampling frequency was set at 1000 Hz. We used nine-point calibration, and the calibration was checked at the beginning and the middle of each block (i.e., every 24 trials throughout the experiment). In Experiment 3, the distance between the fixation cross and the center of each circle was changed from 5° to 10° of visual angle to enable us to specify whether a participant was looking at the fixation cross or one of the circles.

A self-paced drift correction was performed at the beginning of each trial. After the drift correction, a fixation cross appeared against a black background for 500 ms (without placeholders). The cross was followed by presenting the placeholders at the top-left, the top-right, the bottom-left, and the bottom-right locations for a randomly determined period between 400 and 600 ms (see Fig. 6). Participants were required to remain fixated on the fixation cross until the search display appeared. Eye movements were not restricted after the target appearance. The participants made a manual response to the target’s orientation as quickly and accurately as possible, similar to the previous experiments.

Results

The awareness of location–reward associations indicated that 19 out of the 24 participants reported high-reward and low-reward locations correctly, and one participant only reported a high-reward location. The remaining four participants reported no position regularities related to the magnitude of reward points they received on correct trials. We have reported the data of 20 participants that reported the high-reward locations in the following analysis. The main findings were significant even after one participant, who only reported a high-reward location, was excluded. We discarded trials from accuracy and RT analyses if participants’ gaze deviated more than 2° from the fixation cross between search display onset and 100 ms after onset (mean = 3.02%, range: 0%–11.3%). This criterion of 2° was determined based on previous studies that confirmed participants’ fixation during visual search (e.g., Wang & Theeuwes, 2018a, 2018b). The results indicated that accuracy was not affected by the reward manipulation (95.8%; see Appendix Tables 1, 2 and 3). Trials with no responses within 1,200 ms (0.22% of trials) and trials with incorrect responses were also excluded for the RTs analyses.

A two-way repeated-measures ANOVA, with block (1–10) and reward (high/low/no) as within-participants factors, with RTs as the dependent variable showed, a significant main effect of reward, F(2, 38) = 7.23, p = .002, ηp2 = .276, and a marginally significant main effect of block, F(9, 171) = 1.90, p = .055, ηp2 = .091 (see Fig. 10). Post hoc comparisons using Shaffer’s procedure indicated faster responses in the high-reward condition than in the no-reward condition, t(19) = 3.16, p = .015. Moreover, RTs were faster in the high-reward condition relative to the low-reward condition, t(19) = 2.52, p = .021. However, unlike in Experiments 1 and 2, the differences between low-reward and no-reward conditions were not significant, t(19) = 1.13, p = .272. Also, there was no Block × Reward interaction, F(18, 342) = 1.42, p = .118, ηp2 = .070.

Fig. 10
figure 10

Mean reaction times by reward conditions over Blocks 1 to 10 for the aware group in Experiment 3. Error bars indicated one standard error of the mean

Discussion

Consistent with Experiments 1 and 2, we observed faster RTs in Experiment 3 when a target appeared at a high-reward location than in other locations, even when participants remained fixated on the fixation cross for 100 ms after the search display’s onset. These results indicated that the reward effects observed in Experiments 1 and 2 could not be attributed to preparatory eye movements. However, unlike in the previous experiments, there was no significant difference in RTs between low-reward and no-reward conditions, even in participants who correctly reported the location–reward association. Although the reason why we did not observe the difference between low-reward and no-reward conditions is unclear, one of the possible reasons for the lack of the difference could be the long distance between possible target locations and the repeated instructions to fixate the eyes on the fixation cross before starting a trial. As we mentioned in the Introduction, previous studies suggested that the effect of location–reward associations was more evident in a task that induced competition for attention selection between potential targets (tasks with multiple targets, e.g., Chelazzi et al., 2014). Therefore, the long distance between possible target locations with the eye movement restrictions might weaken the reward effect in Experiment 3 because it would cause less attentional competition between stimuli and make the search easier. Another possible reason for the reduced reward effect could be the longer duration of Experiment 3, which may have led to participant fatigue. Nevertheless, the results of Experiment 3 clearly showed a prioritization of high-reward locations, which suggests that preparatory eye movements did not cause the prioritization.

Also, one may consider that the data of saccade latency could be interesting because many previous studies have shown the effects of reward on eye movements (e.g., Failing et al., 2015; Theeuwes & Belopolsky, 2012). Since we instructed our participants to remain fixated on the fixation cross until the search display appeared, a majority of participants (12 out of 20 aware participants) showed few saccadic eye movements (less than 2% of trials), even after the onset of the search display (when we examined saccades within the time epoch between 100 ms and 300 ms from the onset of the display). As for the remaining eight participants, the result of the first saccade latencies for each reward location condition showed no signs of reward effects (201 ms, 203 ms, and 206 ms for the high-reward, low-reward, and no-reward conditions, respectively), and it is therefore unlikely that the reward effect on reaction times observed in the current study could be explained by first saccade latency. Because the analysis of the first saccade latencies in the current study is based on the data of only eight participants, further studies are needed to clarify this issue.

General discussion

As discussed in the Introduction, Jiang et al. (2015) reported a significant location–reward effect only when participants were instructed to intentionally look for a high-reward quadrant and were made aware of the reward structure through these instructions. Therefore, it was unclear whether such goal-driven manipulations are necessary, or spontaneous awareness of the association is sufficient for the location–reward effect to occur in a task involving a single target. The present study investigated whether spontaneous awareness could also induce the location–reward association effect, even if participants were not instructed to look for the high-reward location. The results of three experiments indicated that reward effects could be observed without goal-driven manipulations. Moreover, consistent with previous studies (e.g., Jiang et al., 2015; Sisk et al., 2020), the reward effect was observed only for participants that were aware of the location–reward association. Furthermore, we demonstrated that the reward effects could be observed even when the target was a pop-out stimulus that could capture attention (Experiments 2) and when the observed effect could not be attributed to preparatory eye movements (Experiment 3). These findings suggest that goal-driven manipulations to look for a high-reward location are unnecessary for the reward effect to occur. On the contrary, awareness of the location–reward association is essential for the reward effect.

Why did most previous studies (e.g., Jiang et al., 2015, Experiment 2) fail to observe the location–reward effects? The present study provides clear evidence that awareness of such an association is important for the occurrence of the location–reward effect. Based on this notion, it is possible that the lack of the effect in most previous studies may be due to a small proportion of participants who were aware of the association. Moreover, the significant effect observed in Jiang et al. (2015, Experiment 4) might be due to the instructions, which can help gain awareness of the association. Related to this issue, in the present study, we used a search display containing only four items, whereas the previous studies used a search display containing more than 10 items (Jiang et al., 2015; Sisk et al., 2020; Won & Leber, 2016). The smaller number of items in a visual search task used in this study might facilitate awareness of the association between specific location and reward. Also, a larger number of participants (N = 60) in Experiments 1 and 2 allowed us to investigate the effect of awareness systematically. Overall, it is plausible that the number (or population) of participants who are aware of the association is critical for the appearance/disappearance of the location–reward effect.

Previous studies investigating the feature–reward effect (e.g., the attentional biases induced by the color–reward association) have demonstrated that the effect could occur regardless of any awareness of the association (e.g., Anderson, 2015; Bucker & Theeuwes, 2018; Le Pelley et al., 2015). However, under certain conditions, awareness could play an important role in forming feature–reward associations. For instance, Failing and Theeuwes (2017, see Experiment 6) examined attentional bias by the feature–reward association in terms of awareness of the association using a visual search task in which the color of task-irrelevant and nonsalient distractors signaled the magnitude of the reward. Participants were not given information about color–reward associations, and they completed an awareness test after the search task. The results indicated that the reward effect was observed only in participants that were aware of the color–reward association. Based on this finding, Failing and Theeuwes (2017) suggested necessary conditions for establishing feature–reward associations; a reward-associated feature should be task-relevant (e.g., target), or physically salient, or participants should know the association and trigger the process of reward learning. The reward was associated with a target location in all experiments of the present study, and the target was also salient in Experiment 2. However, we could not observe systematic reward effects when participants were not aware of the association even under these conditions. Therefore, the location–reward effect seems to depend on different learning processes from the feature–reward effect, at least partly. Nevertheless, these effects are consistent with each other regarding the importance of awareness.

The main task of the present study was a target orientation identification task. Therefore, looking for a high-reward location was unnecessary for performing the task. However, the general goal of the participants might have been reward outcome maximization, without concern for awareness of the association. Even though the participants realized that prioritizing a high-reward location was not the best search strategy because the target appeared in all the quadrants with an equal probability, they prioritized high-reward locations so as not to miss any opportunities to receive a high reward. Therefore, attentional prioritization to high-reward locations could be consistent with the participants’ goal of earning as many reward points as possible. The issues related to these findings have been investigated in previous studies, especially studies on feature–reward associations. For example, attentional prioritization to stimuli signaling the availability of reward was observed not only when the reward was no longer available (e.g., Anderson et al., 2011a, 2011b) but even when it was counterproductive and attending to reward-associated stimuli resulted in missing an opportunity to obtain rewards (e.g., Failing et al., 2015; Le Pelley et al., 2015; Pearson et al., 2015; Watson et al., 2019; Watson et al., 2020). Furthermore, Jiang et al. (2015) showed a persistent prioritization of attention to the high-reward quadrant in the test phase, in which rewards were randomly distributed in all quadrants. However, it remains unknown whether a similar effect would be observed for location–reward associations based on spontaneous awareness. Future studies are needed to investigate the relationship between observers’ goals and attentional prioritization to reward-associated locations, as well as counterproductive effects of location–reward associations.

Although the present study demonstrated that awareness of the association between locations and rewards is essential for the occurrence of the reward effect, it is still unclear what process can be modulated by awareness. Unlike the study of Jiang et al. (2015), participants in this study were not required to find a high-reward location. Additionally, a target could appear at four possible locations with an equal probability. Therefore, attentional biases favoring a high-reward location are not necessary to perform the task efficiently. Moreover, the results of Experiment 2 suggested that the reward effects were not caused by the search strategy (i.e., the order of examination) even after participants realized the location–reward associations. A possible explanation for the modulation by awareness could be that awareness facilitates examining the target (e.g., orientation judgment) such that participants try to strategically identify high-reward targets more quickly than low-reward or no-reward targets when they know of the high-reward location. Another possibility is that awareness modulates attentional allocation such that a high-reward location becomes subjectively salient and captures attention when participants know the association. Further studies are needed to clarify the exact process that is modulated by the awareness of the association.

After the experiment, we asked the participants about the association because we did not know when the participants realized the association. Some participants possibly realized the association from the first block, whereas others did not recognize it until the last half block. What we found was that the location–reward effect occurred when participants had knowledge of the association at the end of the task. It was technically possible to ask our participants about their awareness in the middle of an experiment; however, this manipulation might have encouraged them to look for the reward association. If awareness triggers reward learning, attentional biases caused by the reward should be observed on the trials in which participants have realized the association. This possibility needs to be clarified in further investigations on the direct relationship between awareness and location–reward attentional bias.

Another unsolved question is the issue of individual differences in the awareness of location–reward associations. It is known that attentional biases related to feature–reward associations are affected by individual differences, and specific personality traits might predict the magnitude of reward-learning effects on attentional selection (e.g., Anderson et al., 2020; Jahfari & Theeuwes, 2016). Specific studies have shown that attentional capture by reward-signaling stimuli are modulated by personal traits, such as impulsiveness, although they remain controversial (e.g., Anderson et al., 2011a), depression (Anderson et al., 2014; Anderson et al., 2017), addiction (Anderson et al., 2013), visual working memory capacity (Anderson et al., 2011a), the pursuit of desired goals (e.g., Hickey et al., 2010), and gender (Anderson et al., 2013). Similarly, a study on location–reward associations has reported that men are more sensitive to reward outcomes than women are (Della Libera et al., 2017). Such personal traits might possibly explain why specific participants in the present study become aware of the association, whereas others remained unaware. For example, highly impulsive or highly motivated people for obtaining their desired goals (i.e., reward outcomes) might become more acutely aware of reward structures. Unfortunately, the relationship between personality traits and awareness of location–reward associations remains unknown because the present study did not acquire information on personal traits. It is suggested that future studies must clarify individual differences in the location reward effect to explain the characteristic of attentional biases to rewarded locations.

In the present study, to examine the influence of the location–reward mapping, participants were assigned into the diagonal, horizontal, and vertical location types. As far as we know, there have been no studies examining the effect of a location–reward mapping systematically. Furthermore, a recent eye-tracking study on reward learning proposed the possible influence of stimulus competitions and/or integrations when the high-reward and low-reward locations are placed within the same visual hemifield (see McCoy & Theeuwes, 2018). Therefore, it was possible that the location–reward effect could be modulated by location–reward mapping. In contrast to this prediction, the present results showed no significant difference in the location–reward effect between location types. Although the effects of stimulus competition and/or integration may be observed at the level of eye movements, the results of this study suggested that location–reward mapping is less important in eliciting a reward. It should be noted that, because the sample sizes were relatively small due to the awareness issue, we need to interpret the results with some caution.

In conclusion, the present study investigated whether the effect of the location–reward association can appear regardless of goal-driven manipulation. The results from three experiments showed that the reward effects appeared even when participants were not asked to intentionally look for the location–reward association. Moreover, it was demonstrated that the effects were especially pronounced in the aware group, and that the search strategy did not influence the reward effect caused by awareness. These results suggest that awareness of the association is essential for the location–reward effect.