Negative and Positive Templates: Two Forms of Cued Attentional Control

Abstract

Our ability to control our attention to focus on goal-relevant information is critical for functioning in daily life. In addition to the typical attentional control driven by target enhancement described in most theories of attention, recent research has focused on our ability to use information about distractions maintained in working memory to direct our attention away from known distractors. Using these negative templates can improve the efficiency of attention, much in the same way as enhancing information matching search targets. However, these effects only occur for specific tasks or in specific circumstances. In this review, I will focus on our emerging understanding of the relationship between distractor ignoring from negative templates and target enhancement from positive templates. I will also highlight key remaining questions for further study.

While most theories of attention describe the influence of top-down control as directing attention toward target items (Bundesen, 1990; Desimone & Duncan, 1995; Duncan & Humphreys, 1989; Wolfe, 1994; 2007), recent research has also examined how we are able to use top-down control to direct our attention away from distractors (Carlisle, 2019; Geng, et al., 2019). In one common design for cued ignoring studies, participants perform a search task for a shape-defined target (Figure 1b). Half of the items appear in the same color as the target (target color), and the other half appear in a different color (distractor color). Prior to the onset of the search array, participants see color cues with differing meanings. Positive cues indicate the target color and if participants use the cue it could effectively reduce the set size by half. Negative cues indicate the distractor color, and participants know the target would not appear in this color. Once again using this informative cue could reduce the set size by half. Neutral color cues are non-informative, providing a baseline reaction time and attention measure when participants must rely on target shape alone to perform the search task. Importantly, the colors presented on each trial changes randomly, so participants needed to maintain the cues in working memory (Carlisle, et al., 2011; Woodman, et al., 2013) in order to exert attentional control. I will refer to this as active control, because any attentional effects should be driven by top-down control settings instead of learned implicit association between target and distractor responses¹.Evidence that participants can direct attention away from negatively cued colors or toward positively cued colors focuses on benefits to attentional processing for these informative cues compared to the neutral cue baseline. Some typical dependent variables indicating a benefit include faster RTs, higher accuracy, or overt or covert attentional measures of greater attention to target-colored items than distractor-colored items.

Figure 1.

Examples of search arrays used in negative cue experiments. Arrows indicate search target. A) Three cue types. A negative cue indicates color of upcoming distractor. A positive cue indicates color of upcoming target. A neutral cue is uninformative for search. B) Search array example from Arita, et al. (2012). The target is a shape-defined item, the Landolt-C with a gap at the top or bottom. C) Search array example from Moher & Egeth (2012). The target is a capital B or F. D) Search array example from difficult condition in Conci, et al. (2019). The target is a letter T rotated 90 or 180. Note: Examples are for illustrative purposes only. Colors are not representative of those used in individual studies, and stimulus sizes are not representative.

In the first study directly designed to examine negative cues, Arita and colleagues (2012) found positive cues led to RT benefits compared to neutral cues, in line with the major theories of attention. Critically, negative cues also led to significant RT benefits, although not as large as the positive cues. In additional experiments, Arita and colleagues probed the negative cue benefit. They ensured that participants were not recoding the negative cue into a positive cue prior to the onset of the search array by increasing the number of potential colors in the array from 3 to 7, which would be above the capacity of visual working memory (Vogel & Awh, 2008). The benefits of negative and positive cues were stable across cue-to-search delays of 100-1250 ms. A final experiment showed the benefits for both positive and negative cues decreased as set size was reduced, with non-significant negative cue benefits when the search task became easier with set size 4 arrays. These results highlighted that attentional control from working memory can also be used to guide attention away from distractors using negative templates (also called templates for rejection) in addition to the typical use of positive templates to enhance items matching the search target. While Arita and colleagues (2012) were agnostic as to the underlying mechanism of negative templates, much subsequent research has been focused on examining alternative hypotheses for the behavioral benefits from negative cues..

In this brief review, I will discuss our emerging understanding of the relationship between distractor ignoring from negative templates and target enhancement from positive templates. I will begin by showing that the use of positive and negative templates is strongly related across individuals, opening questions about potential shared mechanisms underlying the two types of control. I will discuss the multiple hypothesized mechanisms for negative templates, and describe which hypothesized mechanism currently has the strongest support. Through this discussion, I will highlight evidence that guidance from negative templates is consistently less robust than guidance from positive templates, both in terms of reduced or delayed benefits from negative templates as well clear evidence that negative templates use seems to be dependent on task design. Finally, I will also highlight key remaining questions for further study.

Relationship Between Positive and Negative Templates

While decades of research has focused on positive templates, we know much less about negative templates and the mechanisms underlying their use. Given that both positive and negative templates can guide attention (Carlisle, 2019), one key question is how these two processes are related. One possibility is that both positive and negative templates rely on the same underlying mechanism, but positive templates enhance while negative templates suppress certain features (Carlisle & Nitka, 2019). Alternatively, the two forms of attentional control could be distinct and rely on separate mechanisms (see Noonan, et al.).

If positive and negative templates share underlying mechanisms, we would expect to see a positive correlation between the use of the two cues across individuals. Previous reports (Beck & Hollingworth, 2015; Becker, et al., 2015) have shown large individual differences in the use of positive and negative cues. Beck and Hollingworth (2015) reported a strong positive correlation (r = .67) between benefits from negative and positive cues (e.g. neutral cue RT- negative cue RT), however their sample size was relatively small with only 29 participants. In addition, benefits are derived in comparison to a neutral baseline, meaning that it is possible the correlation is driven by differences in this baseline alone.

To confirm the relationship between positive and negative cue benefits, I examined search RT benefits and accuracy benefits in search trials for 92 participants² collapsed across 4 previously reported experiments from my lab (Zhang, Gaspelin, & Carlisle, 2020). The search trials were structured like those of Arita et al. (2012), see Figure 1a-b (details can be found in Zhang, et al. 2020). The previous report focused on the analysis of participant reports of occasional letter probes (⅓ of 288 trials) presented on the target and distractor items, which varied in probe timing across experiments. However, all participants performed the same basic search task with negative, positive, or neutral cues on the majority of trials (⅔ of all trials, 64 trials for each cue condition), creating a large participant set to verify the correlation of benefits from negative and positive cues. Like Beck and Hollingworth (2015), our data showed a significant correlation between positive and negative RT benefits, r (90) = .63, p < .001 and accuracy benefits, r (90) = .701, p < .001 (Figure 2). The positive cue benefits were significantly larger than negative cue benefits for both RT, t (91) = 14.74, p < .001, and accuracy, t (91) = 6.96, p < .001. The strong positive correlation in RT benefits replicate the relationship reported in Beck and Hollingworth (2015) in an independent sample with a larger set of participants.

Figure 2.

Correlations between positive and negative cue benefits for A) reaction time and B) accuracy. Data from cued search trials of 92 participants in Zhang, Gaspelin, & Carlisle (2020).

Because RT benefits are computed in comparison to the neutral condition, I wanted to ensure that the correlation between positive and negative cue benefits was not driven by differences in neutral RT (e.g. people with longer RTs overall show larger benefits). To do so, I performed a hierarchical multiple regression to evaluate the prediction of negative RT from neutral RT and positive RT. For the first block analysis, the predictor neutral RT was analyzed. The result of the first block hierarchical linear regression analysis revealed a significant model, R² = .25, p < .001. For the second block analysis, I added positive RT as a predictor to the analysis. If active control from positive and negative cues relies on shared underlying mechanisms, I would expect the addition of positive RT to lead to better prediction of negative RT compared to neutral RT alone. This is what was found. The model for the second block was significant, R² = .43, p < .001, but more importantly the R² change of .182 was significant, p < .001. Controlling for positive RT, the regression coefficient for neutral RT was significant, b = .23, 95% C.I. (.028, .425), β = .22, t (89)= 2.27, p = .026. Controlling for neutral RT, the regression coefficient for positive RT was significant, b = .625, 95% C.I. (.393, .858), β = .51, t (89) = 5.35, p < .001. The larger beta value for positive RT than neutral RT highlights the importance of active control in predicting the variation in negative RT. This analysis gives us confidence that significant positive correlation between positive and negative RT benefits cannot be explained by differences in neutral RT, and confirms the relationship between the use of positive templates and negative templates.

These analyses confirm a strong relationship between the use of positive and negative cues. If no such relationship was found, it would be clear that positive and negative templates rely on separate mechanisms. The strong relationship suggests some shared underlying mechanisms, which could be due to either state-level factors (e.g. motivation) contributing to task performance or more intriguing trait-level individual differences related to underlying attentional mechanisms. Although the correlation does not tell us what mechanisms might be shared between positive and negative templates, it opens the stage for the exploration of shared mechanisms. Next, we turn to how the two types of attentional control may be related.

Negative Templates as a Special Use Case of Positive Templates?

The relationship between positive and negative cue benefits suggests the two types of cued suppression rely on the same underlying mechanisms. Three hypotheses have been proposed suggesting that the RT benefits following negative cues are actually a special use case of positive templates, rather than a separate mechanism of distractor ignoring. In this section, I describe these hypotheses and review the data related to each.

Search and Destroy

The search and destroy hypothesis (Moher & Egeth, 2012, see Figure 2b) suggests that RT benefits from negative cues are actually a result of participants first attending the negatively cued feature during search before disengaging and ignoring. This hypothesis was generated largely based on data from a probe-dot study which found that participants were faster to report probe dots appearing on items matching negative cue in contrast to a non-cued distractor item at 117ms and slower to report probes appearing on items matching a negative cue at 167ms. According to this hypothesis, participants will begin by utilizing the negative cue information to create a positive template. After attention is directed toward the distractor color, attention can be moved away. Given long enough search times or difficult enough search tasks, this process could lead to RT benefits. Although there are similarities between this hypothesis and the idea that distractors can be rapidly rejected or reactively suppressed if they are attended (Geng, 2014, see also Aron, 2011), the search and destroy hypothesis specifically states that individuals will actively search for the negatively cued item whereas rapid rejection does not propose an initial guidance toward the distractors.

There are two main predictions from search and destroy: better distractor suppression later in a search trial (destroy) and evidence of initial attention directed toward the negatively cued items (search). In order to assess these predictions, studies must have a way to assess attention at multiple timepoints during search. Eye tracking studies (Beck, et al., 2018; Kugler, et al., 2015; Lu, et al., 2017; Zhang, et al., 2022), attentional probe studies (Zhang, et al., 2020), and an ERP study (Carlisle & Nitka, 2019) have shown better distractor ignoring during the later portions of search trials, in line with the ‘destroy’ prediction of search and destroy.

However, assessments of the ‘search’ prediction from search and destroy, initial attention towards the negatively cued distractors, has not been well supported. An ERP study examining whether early covert attention was directed toward negatively cued colors before attention shifted towards a target found no evidence of an initial N2pc towards negatively cued distractor items (Carlisle & Nitka, 2019), but attention was directed towards the possible targets following the negative cues. This indicated directing covert attention based on negative cues did not require an initial covert search for the negatively cued distractors. Beck, et al. (2018) measured eye movements to assess where attention is directed during negatively cued search. They found initial fixations were more likely to go to a negatively cued distractor than expected by chance, but determined initial attention toward the cued distractor was not necessary for suppression of the cued distractor on later fixations (see also, Zhang & Carlisle, in press). In fact, Beck and colleagues (2018) found trials where attention did not go to a negatively cued item on an early fixations showed significant suppression of search items matching the negative cue by the third fixation, whereas trials where negatively cued distractors were attended on early fixations did not lead to significant suppression on the third fixation. This suggests the ‘destroy’ for overt attention did not require search and initial overt attention toward a negatively cued distractor may actually interfere with later suppression. Finally, a study examining reports of letter probes presented at 25, 100, 250, and 400 ms into search (Zhang, et al., 2020) looked for evidence of early attention to negatively cued distractors followed by later suppression, as measured by the number of probes reported. The timing of enhancement and suppression described in Moher and Egeth (2012) led to the prediction that the 25 and 100 ms probe timepoints should show attentional enhancement of distractors matching the negative cue (more letter probes reported on negatively cued distractor items), while the 250 and 400 ms timepoint should show ignoring of the negative cue matches (fewer letter probes reported on negatively cued distractor items). At the 25 ms timepoint, Zhang, et al. (2020) found equivalent probe reports on negatively cued and neutral cued distractors. For probes at 100, 250, and 400 ms, fewer letter probes were reported on negatively cued distractors than neutral cued distractors, indicating cued distractor ignoring. Given that the search and destroy suggests that initial attention to the negatively cued items is what leads to later avoidance of target items, the lack of support for initial ‘search’ from ERP, eye tracking, and probe studies findings do not provide strong support for the search and destroy hypothesis.

Location recoding

Another hypothesis that proposes negative cue benefits can be explained by positive templates is the location recoding hypothesis. Noticing that the original Arita et al. (2012) arrays had potential targets and distractors separated by hemifield, Beck and Hollingworth (2015) proposed that participant may wait until the search array is presented, identify the location of the negatively cued items, and develop a positive spatial template for the other hemifield. They found that while some participants could benefit from negative cues when the targets and distractors were spatially separated, there was no evidence of a negative cue benefit when the target and distractor colors were spatially intermixed. However, in some blocks participants were explicitly instructed to use a location based cue (an arrow indicating the target hemifield) and both color and location cue trials were practiced at the beginning of the experiment, which may have biased participants to utilize a location based strategy. Carlisle & Nitka (2019) designed an experiment to examine whether the lack of benefits in Beck and Hollingworth’s study could be explained by strategy choice. In order to encourage participants to develop a strategy to use the negative cues, they had ⅔ of trials within the block appear as spatially separated arrays. Finding a negative cue benefit on these trials would ensure participants were using the negative cues. To test whether negative cue benefits were driven by location recoding, on a random ⅓ of trials participants would be shown a spatially intermixed array. Because participants did not know which type of trial would be presented, strategy choices leading to attentional control settings should be the same for both types of arrays. The location recoding hypothesis would predict significant benefits for the spatially separated arrays, and no benefits for the spatially intermixed arrays where location recoding is impossible. In contrast to the predictions of location recoding, negative cue benefits were equivalent in spatially separated (223 ms benefit) or spatially intermixed arrays (216 ms benefit). Multiple other studies have found benefits from negative templates when items were not spatially separated (Addleman & Störmer, 2022; Conci, et al., 2019; Kugler, et al., 2015; Reeder, et al., 2017; Tanda & Kawahara, 2020; 2022; Zhang, et al. 2022; Zhang & Carlisle, in press). Based on these findings, it does not seem that location recoding can explain the negative cue benefits.

Color recoding

The color recoding hypothesis of negative cue benefits was also generated in response to the specific design of two-color arrays separated by hemifield, which always contained two colors (Arita, et al, 2012). According to the color recoding hypothesis (Becker, et al., 2015), participants wait until the search array is presented, notice what color in the array does not match the negative cue and generate a positive template for this color. To test this hypothesis, Becker and colleagues generated arrays where the negative cue indicated the color of half of the search items, but where the other half each item was presented in a unique color. Their results were inconclusive, as they failed to find benefits for either two color or multicolor arrays, which could indicate participants generally ignored the negative cues. Multiple other studies which have failed to find negative cue benefits have used multicolor arrays (Cunningham & Egeth, 2016; Beck, et al., 2018; Stilwell & Vecera, 2019), but without demonstrating the participants are utilizing the negative cue to guide search. Testing the color recoding hypothesis requires participants are utilizing the negative cues to guide attention to determine whether they are engaging in color recoding. Zhang and Carlisle (in press included two color arrays in ⅔ of trials in a block, to ensure participants were utilizing the negative cues. The other third of trials in the block contained arrays with half of items in one color and the other half of items each containing a unique color. If negative cue benefits in the two color arrays were driven by participants engaging in color recoding, we would expect to see a significant negative cue benefit in two color arrays and no benefits for the multicolor arrays. However, there were similar negative cue benefits found for both two color separated arrays (200 ms) and multicolor arrays (220 ms), suggesting color recoding was not the mechanism behind the benefits. Similarly, Kerzel and Huynh Cong (2022) have recently reported contingent capture effects derived from negative cues prior to the onset of a search. Given that the contingent capture effects were demonstrated by showing a single item prior to search, it would be difficult to explain these effects from a color-based recoding account.

Three early hypotheses were generated suggesting that negative cue benefits could be explained as a special use case of typical positive templates either through attending then suppressing the negatively cued items (Moher & Egeth, 2012) or by recoding to a positive spatial (Beck & Hollingworth, 2015) or color (Becker, et al., 2015) template. However, the subsequent studies examining these hypotheses reviewed above have not consistently supported the idea that negative cue benefits are actually based on positive templates. This suggests that negative templates can lead to an ignoring of distractor information. Next, we will turn to the mechanisms underlying positive and negative templates.

Examining the Active Suppression Hypothesis

Positive color templates are typically viewed as leading to a global feature-based enhancement (Andersen, Hillyard, & Müller, 2008; 2013; Saenz, et al., 2002). While multiple previous studies have examined suppression that is an indirect effect of target enhancement (Andersen & Müller, 2010; Forschack, et al., 2017; Störmer & Alvarez, 2014), there has been a limited amount of data gathered on active suppression driven by attentional control from cued suppression tasks. In this section, we discuss evidence relevant to the hypothesis that benefits from negative templates are based on active suppression.

From a theoretical perspective, if positive templates lead to attentional enhancement through a feature-based upweighting of certain information (Wolfe, Bundesen, 1990), one obvious mechanism for negative templates would be a feature-based suppression (Carlisle, 2019). Addleman and Störmer (2022) used pre-search probes to examine whether negative cues led to proactive suppression of the distractor feature. They asked participants to identify whether the probe was colored or gray, and expected reduced accuracy on negative cue matching trials if participants were engaging a proactive feature-selective suppression. However, performance was equivalent between probe detection on negatively cued and neutral trials, suggesting participants were not engaged in a proactive feature-specific suppression of the negatively cued color (but see Kerzel & Huyhn Cong, 2022). Participants did show better probe detection performance following positive cues, consistent with a positive templates mechanism involving feature-specific proactive enhancement. In spite of the lack of evidence for proactive suppression, participants were faster to report the target on negative cue trials compared to neutral trials, demonstrating a benefit from negative templates.

Similar conclusions can be drawn from the studies examining the neural underpinnings of negative templates. Reeder and colleagues (2017; 2018) used fMRI to examine the proactive neural activity associated with positive and negative cues before visual search. While positive templates led to increased activity in the occipital pole compared to neutral trials, negative templates led to reduced activation in similar areas. To determine whether these differing levels of activation were driven by feature-specific suppression for negative templates and enhancement for positive templates, Reeder and colleagues (2018) performed a representational similarity analysis of the pre-search data. The positive cue condition showed a higher representational distinctiveness than neutral cue condition in line with feature-specific enhancement. However, negative cues were no more distinctive than neutral cues. To further assess whether negative cues and positive cues were leading to opposite patterns of activity, they directly contrasted the patterns in representational similarity between the positive and negative cue conditions. If participants were suppressing a negative cue color in the opposite way as they enhanced the positive cue color, the two sets of signals should be negatively correlated.

However, the positive and negative cue representational similarity measures from early visual cortex were significantly positively correlated, suggesting that the reduced activations in early visual areas (Reeder, et al., 2017) were a more general suppression rather than a feature-specific suppression. de Vries and colleagues (2019) found a similar pattern while measuring EEG by looking at alpha power. Following a negative cue, there was stronger bilateral posterior alpha power compared to a positive cue. Enhanced alpha power has been shown when participants are generally expecting a distractor to occur, and higher alpha power is associated with a reduced behavioral impact of distractors (Bonnefond & Jensen, 2012). There was no evidence of a lateralized alpha effect specific to the location of a lateralized negative cue- suggesting that the effect was not driven by a specific suppression of the negative cue itself. Instead, the larger non-lateralized alpha power can be viewed as evidence of perceptual gating. This bilateral alpha power increase has recently been replicated (Exp. 3; van Zoest, et al., 2021), with a negative cue also leading to a reduced Pd to distractor item following a negative cue compared to a neutral cue.

The general suppression of early visual areas demonstrated in Reeder, et al. (2017), de Vries, et al. (2019), and van Zoest (2021) may help to explain why negative templates seem to take longer to impact search performance than positive templates (Beck, et al, 2018; Carlisle & Nitka, 2019; Kugler, et al., 2015; Zhang, et al., 2020; 2022, see also Tanda & Kawahara, 2022)). This can also help to explain why negative templates are less effective than positive templates in terms of RT benefits, overall. However, it is difficult to see how a general suppression of incoming visual signals would be able to explain stronger attentional performance for negative cues compared to neutral cues (Beck, et al., 2018; Kugler, et al, 2015; Carlisle & Nitka, 2019; Zhang, et al., 2020; 2022; in press). One possibility is that general suppression could leave longer for other aspects of the target, such as shape, to accumulate evidence³. This possibility could explain RT benefits (Arita, et al., 2012) and lateralized ERP effects (Carlisle & Nitka, 2019). However, recent eye tracking evidence has suggested that negative cues lead to more attention to target-colored distractor objects (Zhang & Carlisle, in press; Zhang, et al., 2022), which could not be explained by accumulation of shape knowledge alone as the distractors do not match the target shape. Similarly, negative cues have been shown to lead to contingent capture for items that do not match the negatively cued color (Kerzel & Huynh Cong, 2022), an effect which is not dependent on match to targets. This suggests that while early visual areas do not show proactive feature-specific suppression, at some point in attentional processing, there must be a feature-specific impact of the negative cues in order to gain negative cue benefits.

This section has discussed research relevant to the active suppression hypothesis. While the results present in the literature may be in line with some form of active suppression (Zhang & Carlisle, in press), it is clear that any mechanisms active suppression from negative templates are distinct from the proactive feature-specific enhancement of positive templates (Addleman & Störmer, 2022; Reeder, et al., 2018). It is possible that distractor ignoring relies on a feature-specific equivalent of the signal suppression hypothesis from the learned suppression literature, which states that salient items can be suppressed before attention shifts toward them (Sawaki & Luck, 2011; Gaspelin, et al., 2015; 2017). This hypothesis would share some features with search and destroy (Moher & Egeth, 2012), in that the feature would need to be identified prior to suppression, however it does not require the negatively cued item be attended as proposed by search and destroy. In this way, it is possible that active suppression might be a form of control which is reactive to the onset of the search display, but proactive in relation to attentional shifts. Future research should pursue more direct evidence regarding the active suppression hypothesis, and/or find additional alternative hypotheses which can explain the full pattern of results in the literature.

Flexibility in Negative Template Use

While there is now a large body of work demonstrating the benefits of negative cues, there are also studies who have not found a negative cue benefit. Some recent work has begun to examine the factors that might be involved in when negative cue benefits will be demonstrated. One factor that seems important is how useful the cue is in performing the search task. Many of the studies that have failed to find a negative cue benefit have used cues which would only lead to elimination of less than half of the distractors (Beck, et al., 2018; Phelps, et al, 2022; Stilwell & Vecera, 2019a; 2019b), search arrays with a small set size (Berggren & Eimer, 2020), or both (Moher & Egeth, 2012; Williams, et al., 2020). Reducing the effectiveness of the cue or reducing the need to use the cue by making set sizes smaller may lead to little advantage for cue use. Indeed, the dependence of the negative cue effect on set size was shown in earlier work, where both positive and negative cue benefits were reduced at lower set sizes with no significant benefit for negative cues at set size 4 (Arita, et al., 2012). Given that the search task can be completed without utilization of the cues, it makes sense that participants would not utilize this additional information if the search task was easy enough to complete without them. However, it is important to recognize that these studies highlight another key difference between positive and negative cue use: Positive cues lead to benefits in a larger set of task designs than negative cues.

In line with this selective use of negative templates Kerzel and Huynh Cong (2022) found evidence from a contingent capture measure that negative templates were not used in a singleton search task where the target popped out and therefore using the templates was not mandatory. But negative cues were used in a very closely matched feature search task where negative templates were required to determine which item was the target during search. The utilization of negative templates may be more cognitively demanding than positive templates, as indicated by enhanced frontal theta in EEG (de Vries, et al., 2019) for negative compared to positive templates (and many, many anecdotal comments by participants and RAs to the author). Given that negative templates are cognitively demanding, participants may not utilize negative cues when there is an easier method for performing the search (Rajsic, et al., 2020).

Recent work by Conci and colleagues (2019) demonstrates how the utility of the cue for performing the task impacts the use of negative cues. In their study, participants who performed an easy search task where the shape-defined target was perceptually distinct from distractors showed no negative cue benefit (Figure 2c). Participants who performed a difficult search task where the target and distractors were more perceptually similar did show a negative cue benefit, suggesting participants will utilize the cue more as search becomes more difficult (see Lu, et al. 2017, for a similar shift based on distractor probability). The design of Conci and colleagues was very elegant, because the easy and difficult search arrays were matched in terms of cue-related features. The colors, size of cue-matching items, search set sizes were equated while search difficulty was manipulated.

Zhang et al. (2022) replicated and extended Conci, et al., (2019). In a first experiment, they ensured that the changing search difficulty would alter negative template use within the same participants to determine if strategy shifting was flexible. They had participants swap from the easy to difficult search tasks in the middle of a block following a short rest break, with no additional instructions. Replicating Conci, et al. (2019), participants showed larger negative cue benefits for the difficult search task, with no order effects based on whether participants had the easy or difficult search task first in the block. This indicates that negative template use can be flexibly adjusted, within participants, based on task demands. Next, Zhang et al. (2022) examined eye tracking while participants performed the easy or difficult search tasks. When the search task was difficult, participants exerted stronger attentional guidance as indicated by a larger proportion of fixation on potential targets compared to distractors for both the negative and positive cues. Eye tracking also allows for measures of reactive suppression, where attention can be quickly disengaged from known distractors when they are attended as indicated by reduced dwell time on distractors (Geng, 2014; Geng & DiQuattro, 2010). There were larger dwell time benefits for both negative and positive cues (e.g. reduced dwell time on distractors compared to neutral trials) when participants were engaged in the difficult search task. Both the enhanced guidance and increased dwell time benefits were correlated with overall search RTs, indicating that these shifts help drive overall performance. These results highlight RT benefits from positive and negative cues may come from both attentional guidance and reactive suppression benefits.

Looking across the studies reported in this section, it seems like there is flexibility in how strongly participants utilize negative templates, which is influenced by the task demands in addition to the individual differences in negative cue use reported above. However, it is important to remember that multiple studies have found weak, inconsistent, or no benefits for negative templates (Beck & Hollingworth, 2015; Beck, et al., 2018; Becker, et al, 2015; Berggren & Eimer, 2020; Moher & Egeth, 2012; Stilwell & Vecera, 2019a; 2019b; Williams, et al., 2020), and future research should focus on examining the specific task conditions that will lead to negative template benefits. Just as participants may utilize positive templates more when they are more beneficial for the task (Carlisle & Woodman, 2011), participants may choose whether or not to use negative templates (Rajsic, et al., 2019; Kerzel and Huyhn Cong, 2022), or how strongly to deploy negative templates (Zhang, et al., 2022; Arita, et al., 2012) as well. These results suggest that attentional control may be best characterized as a dial which can be turned up (enhancement) or down (suppression) to varying degrees (Carlisle, 2019).

Summary and Open Questions

It has been 10 years since the original study reported evidence of negative templates (Artia, et al., 2012), although hints were present in the earlier literature (Sawaki & Luck, 2011; Woodman & Luck, 2007). While we have learned much about negative templates in the last 10 years, there are still important remaining questions.

In this review, we have discussed active attentional control from negative and positive templates. While much of the research on top-down attentional control has focused on our ability to enhance potential targets with a positive template, recent research has begun to uncover how we can use active attentional control to suppress known distractors. The review began by confirming a previous report (Beck and Hollingworth, 2015) of a strong correlation between positive and negative cue benefits across individuals, which suggests a shared underlying mechanism for positive and negative templates. We discussed three theories which suggested negative cue benefits may be a special case of the use of positive templates (Moher & Egeth, 2012; Beck & Hollingworth, 2015; Becker, et al., 2015), but found that the existing literature does not support these hypotheses. This suggests that negative templates are an independent form of attentional control from positive templates, with some shared underlying mechanisms. We then discussed the existing data supporting a distinction between the feature-specific proactive mechanisms supporting positive template use, and the mechanisms of negative templates which incorporate aspects of perceptual gating (de Vries, et al, 2019; Reeder, et al., 2017; van Zoest, et al. 2021) but do not seem to rely on proactive feature-based suppression (Addleman & Störmer, 2022; Reeder, et al., 2018; but see Kerzel & Huynh Cong, 2022). However, all benefits from negative templates compared to neutral cues cannot depend on a general suppression mechanism because target-colored distractors are attended more frequently than cued distractors. And, unlike learned suppression of salient distractors (Gaspelin & Luck, 2018), cued suppression could not depend on a second order suppression mechanism (e.g. salience suppression, Won, et al., 2019). This indicates that determining more about the mechanisms underlying cued distractor suppression is a key area for further research.

If positive and negative templates rely on different underlying mechanisms, why is there a strong correlation between positive and negative cue benefits across individuals reported above (see also Beck & Hollingworth, 2015)? While the two types of templates may have differing underlying neural mechanisms (Carlisle & Nitka, 2019; de Vries, et al., 2019; Reeder, et al., 2017; 2018), they may lead to similar impacts on attention like increasing the likelihood of attending relevant information and decreasing the time spent processing known distractors (e.g. rapid rejection; Zhang, et al., 2022). Individual differences in template use may be due to stable differences across individuals, such as differing cognitive control abilities. These differences in template use could also be driven by state-level differences in individuals like overall motivation or arousal. While studies have shown differing levels of negative template use within individuals by altering task parameters (Conci, et al., 2019; Kerzel, et al., 2022, Zhang, et al., 2022), to date, no studies have addressed these questions. Examining whether the strong positive correlation between positive and negative template benefits is driven by state or trait level individual differences is another key area for future research.

Evidence in support of active attentional suppression is accumulating, although more direct evidence is a critical goal for future research studies. If such direct evidence could be uncovered, it would be an important piece in understanding the attentional control system. Attentional suppression might be important in tasks like helping to reduce the impact of emotional distractors (Kennedy, et al., 2018), where forewarning about an upcoming emotional distractor reduced emotion-induced blindness. Active suppression could similarly be useful in reducing persistent attentional biases to negative emotional faces in certain psychological disorders (Judah, et al., 2016) or to prevent attention toward items that are no longer relevant but which have strong selection histories due to previous reward associations (Gong, et al., 2016) or previous target status (Seidl, et al, 2012). Being able to actively suppress known distractors, as an additional mechanism to target enhancement, may be particularly useful in real-world situations, although there is little literature on this topic.

In this review, I have focused on cued suppression, but there has also been a large body of work generated over the same period examining how we are able to learn to ignore salient distractor features (see Gaspelin & Luck, 2018, for a review). Unlike cued suppression, which relies on explicit working memory, learned suppression is based on implicit memory mechanisms. One key question that has been raised in both literatures is whether the distractor suppression is proactive or reactive (Addleman & Störmer, 2022; Geng, 2014), or both (Geng & DiQuattro, 2010; Zhang, et al., 2022). Understanding how the mechanisms of cued attentional suppression are related to the mechanisms of learned attentional suppression (Geng & Duarte, 2021; Turatto, et al., 2018; Noonan, et al., 2018; Won & Geng, 2020) is another open question (Geng, et al., 2019).

Finally, it is important to recognize how active control from negative templates impacts our understanding of attentional control more broadly. Many theories of attention suggest that attention templates are maintained in working memory (Bundesen, 1990; Desimone & Duncan, 1995; Wolfe, 2007). Indeed, the mechanism of attentional control in Desimone & Duncan (1995) states that top-down templates are just the sustained activations from holding items in working memory, suggesting an automatic linkage between working memory and attentional enhancement (Soto, et al., 2008; Olivers, et al., 2011; Peters, et al., 2009). While this automatic linkage between positive templates and working memory has been challenged by evidence templates can be maintained in long-term memory (Woodman & Arita, 2011; Carlisle, et al., 2011; Grubert, et al., 2016), the cued suppression literature is also a very important challenge to an automatic linkage between working memory and attentional enhancement (Hollingworth, 2022). The trial-by-trial changes in cuing found in most studies examining negative templates ensures that the templates are maintained in working memory (Rajsic, et al., 2020). Utilizing information in working memory to ignore does not fit with an automatic relationship between working memory and attentional guidance (Carlisle, 2019; Hollingworth, 2022). In line with earlier work dissociating attention enhancement from working memory maintenance (Woodman & Luck, 2007; Sawaki & Luck, 2011; Carlisle & Woodman, 2011), it is now clear that attentional templates are distinct from WM representations themselves (Yu, et al., 2022). The literature on cued distractor suppression is therefore important for future theories of attention, which need to recognize and explain the importance of distractor processing in attentional control (Duncan & Humphreys, 1989; Tipper, et al., 1994).

In this review, I have covered our emerging understanding of the relationship between distractor ignoring from negative templates and target enhancement from positive templates. This work shows distractor ignoring is an important, but previously unrecognized, aspect of our attentional control system. There are many key remaining questions for further study. While the last 10 years has led to important insights, there is much left to learn.