Attention as a multi‐level system of weights and balances

Abstract

This opinion piece is part of a collection on the topic: “What is attention?” Despite the word's place in the common vernacular, a satisfying definition for “attention” remains elusive. Part of the challenge is there exist many different types of attention, which may or may not share common mechanisms. Here we review this literature and offer an intuitive definition that draws from aspects of prior theories and models of attention but is broad enough to recognize the various types of attention and modalities it acts upon: attention as a multi‐level system of weights and balances. While the specific mechanism(s) governing the weighting/balancing may vary across levels, the fundamental role of attention is to dynamically weigh and balance all signals—both externally‐generated and internally‐generated—such that the highest weighted signals are selected and enhanced. Top‐down, bottom‐up, and experience‐driven factors dynamically impact this balancing, and competition occurs both within and across multiple levels of processing. This idea of a multi‐level system of weights and balances is intended to incorporate both external and internal attention and capture their myriad of constantly interacting processes. We review key findings and open questions related to external attention guidance, internal attention and working memory, and broader attentional control (e.g., ongoing competition between external stimuli and internal thoughts) within the framework of this analogy. We also speculate about the implications of failures of attention in terms of weights and balances, ranging from momentary one‐off errors to clinical disorders, as well as attentional development and degradation across the lifespan.

This article is categorized under:

• Psychology > Attention

• Neuroscience > Cognition

Attention is analogous to a multi‐level system of weights and balances that help us make sense of the chaos of the external world and our internal thoughts.

1. INTRODUCTION: WHAT IS ATTENTION?

Cognitive scientists have a complicated relationship with attention. Depending on who you ask, attention may either be a gift that keeps on giving, with researchers devoting whole careers to understanding attention and its interactions, an inconvenient confound that must be considered when interpreting data, or an overused cliché.

With its near ubiquity, it can be somewhat surprising how difficult it is to pin down a consistent definition for this particular word. Over 130 years ago, William James (1890) published perhaps one of the first attempts to define attention as: “the taking possession by the mind… of one out of what seem several simultaneously possible objects or trains of thought.” During the 20th century, descriptions of attention as a spotlight (Posner et al., 1980), a zoom lens (Eriksen & St. James, 1986) or the “glue” that binds features to objects (Treisman & Gelade, 1980) were popularized and subsequently challenged (Cave & Bichot, 1999; Driver & Baylis, 1989; Wolfe et al., 1989). In addition to these debates surrounding what attention is, ample theories for how attention operates have also been posed, such as the biased competition model (Desimone & Duncan, 1995), filter model (Broadbent, 1958; McNab & Klingberg, 2008) attenuation models (Treisman, 1960), divisive normalization models (Heeger, 1992; Reynolds & Heeger, 2009), rhythmic oscillation models (Fiebelkorn & Kastner, 2019), etc. Modern introductory psychology textbooks tend to take a more nebulous approach towards defining this everyday concept, with one such textbook describing attention as simply, “a narrow focus of consciousness” (Cacioppo & Freberg, 2019).

Compounding this difficulty is the fact that several types of attention exist, resulting in differences in scope and focus across the different models and theories described above, as well as various attempts to organize attention into taxonomies. For example, studies of visual attention often differentiate between space‐, object‐, and feature‐based attention. Attention can be allocated across multiple modalities, for example, audio‐visual or spatio‐temporal. Attention can be allocated overtly (via eye movements) or covertly (via the “mind's eye”). A long‐standing dichotomy for how attention is controlled (top‐down vs. bottom‐up) has recently given way to a trichotomy that adds experience‐driven as its own branch (Awh et al., 2012), while others claim that even these three discrete categories are not sufficient to encapsulate all the unique avenues in which attention can be guided (Wolfe & Horowitz, 2017). Other distinctions have been raised between attention that is allocated externally compared to internally (Chun et al., 2011), or perceived versus nonperceived (Oberauer, 2019), while Hommel et al. (2019) defined over 10 separate forms of attention in an attempt to expose the uselessness of the term.

So, what is attention? In approaching this question, we suggest that it may be useful to have a simple, intuitive, conceptual‐level definition that embraces the presence and variability of these different types and taxonomies of attention. We propose a broad analogy: attention as a multi‐level system of weights and balances that weighs all possible inputs and influences to allow mental prioritization. As expanded in the sections that follow, this interpretation incorporates many of the previous ideas for what attention is and how it operates, but it is also general enough to encompass aspects of attention not yet resolved. We see particular value in conceptualizing attention as a system that monitors both externally‐generated and internally‐generated signals. In other words, attention enables us to manage the chaos of both the external world and our internal thoughts, balancing all of this information to streamline our experience in ways that are consistent with our goals while ensuring that potentially relevant information outside of our current focus of awareness is not missed. In this view, attention serves to weigh information both within and outside of current awareness and assign priority levels, with the weightings on the scales dynamically shifting from moment to moment, and the “winners” with the highest weightings on more local scales proceeding to compete against each other on progressively more global levels. By multi‐level, or multi‐tiered, we aim to capture the idea that there are many different types or targets of attention that may reflect different levels of processing involving competition within and across different brain regions. We specifically highlight that competition occurs both within the domains of internal and external attention and between them. The entire process—or set of processes—allows us to focus in on and prioritize relevant information while constantly considering new and competing signals to evaluate their importance, helping our brains maintain “balance” in the chaotic world.

Note that our goal here is not to propose a novel unifying model of attention, nor to imply that this loose high‐level definition should replace the important theoretical and mechanistic‐level models explaining various components of attention. Focusing theoretical models on specific subprocesses is useful for making testable predictions and proposing physical mechanisms. What we hope to capture in the review that follows is the sense that even if the specific mechanisms of this weighting‐and‐balancing may differ across levels, the different subtypes of attention may be thought of in a sense as part of the same grand system or goal, and perhaps this approach and analogy can help shine light on the different pieces that make up the puzzle of attention.

2. ATTENTION AS A SYSTEM OF WEIGHTS AND BALANCES

Many prominent theories of attention focus on attentional selection and modulation of external signals in the world. However, more recently, renewed focus has emerged on incorporating a wider range of attentional signals, including attention to internal mental representations. One proposed taxonomy of attention (Chun et al., 2011) describes attention based on the types of information that can be modulated, known as the targets of attention, which can be both external (i.e., sensory information such as locations, objects, sounds, etc.) and internal (i.e., internally represented perceptual information, memories, rules, etc.). Our analogy of attention as a multi‐level system of weights and balances is broad enough to encompass this full taxonomy of attentional targets, characterizing all these different types of attention as examples of signals subject to tradeoffs during processing. Moreover, it captures the idea that competition can occur between information within and across levels of the taxonomy, and on micro and macro scales. For example, attention not only resolves competition between sensory features of a given type (i.e., attend‐blue vs. attend‐red, attend to your friend's voice vs. another conversation in the room, attend to this location vs. that location), but also resolves competition across modalities (e.g., ignoring a noisy conversation when you are trying to read a book), among internal thoughts (e.g., a happy memory vs. a nagging worry), and between external and internal information (e.g., focusing on a class lecture vs. daydreaming about your lunch). “Attention” as a whole, then, is the system of neural mechanisms and mental processes that balances all of these competing influences and prioritizes the information with the heaviest weights at a given moment for selective processing. Here we expand on this analogy in the context of prior literature on different types of attention.

2.1. External attention: Balancing top‐down vs. bottom‐up influences

Information in our external environments constantly competes for awareness. A great deal of prior literature has focused on differentiating two distinct types of attention that help resolve this competition: top‐down, or goal‐oriented attention, which allows us to prioritize information consistent with a current behavioral goal, and bottom‐up, or stimulus‐driven attention, which instead biases selection towards inputs that are otherwise salient or unexpected (Beck & Kastner, 2009; Carrasco, 2011; Corbetta & Shulman, 2002; Desimone & Duncan, 1995; Kastner & Ungerleider, 2000). When searching for a friend wearing a blue jacket on a busy street, for instance, we can use top‐down attention to prioritize blue features, enhancing the efficiency of our search. This is well described by the biased competition account of attention, which suggests that the competition for attentional selection is biased in favor of objects that possess goal‐relevant features via enhanced activation in populations of neurons that specialize in their processing (Desimone & Duncan, 1995; Reynolds et al., 1999), like those that preferentially process the color blue during the search for your friend. However, if an ambulance unexpectedly appears in view during this search, this top‐down process might be interrupted, and your attention might instead be captured by its bright, flashing red light or loud siren in a bottom‐up manner.

A number of human behavioral studies have investigated how top‐down and bottom‐up attention differ in terms of temporal processing (Carrasco, 2011; Ling & Carrasco, 2006) and degree of automaticity/voluntary control (Folk et al., 1992; Pashler, 1988; Yantis & Jonides, 1984). The two systems are also thought to be subserved by distinct neural networks, with the dorsal attention network (DAN; including the intraparietal sulcus and frontal eye fields) active during sustained, goal‐based orienting of attention, and the ventral attention network (VAN; including the right temporal parietal junction and right ventral frontal cortex) acting as a “circuit breaker” to the dorsal network to quickly and efficiently reorient attention to salient events outside the current focus (Corbetta et al., 2000; Downar et al., 2001; Indovina & Macaluso, 2007; Serences et al., 2005; Shulman et al., 2009).

Neurophysiological studies in monkeys have identified neural priority maps in the lateral intraparietal area of posterior parietal cortex, the frontal eye field, the superior colliculus, and several visual cortical areas (Gottlieb et al., 2009; Li, 2002; Mazer & Gallant, 2003; Purcell et al., 2012; Zhang et al., 2012). The activity across these internal maps of the environment is determined by both goal‐relevance and physical salience of the objects and locations in a scene, where the most important information is represented by the greatest activity on the map. Attention is then oriented to the peak of this map, ensuring that the information deemed most relevant—whether determined via top‐down or bottom‐up attention—is efficiently selected for processing.

These findings and the biased competition theory are captured intuitively in our weights‐and‐balances account. External inputs are weighted based on both top‐down goal relevance and bottom‐up stimulus salience. Goal‐relevant information is weighted heavier than goal‐irrelevant information, tipping the scales in the competition for selection. At the same time, heavier weights may also be allocated to goal‐irrelevant but highly salient or novel signals that are potentially critical, causing them to incidentally “beat out” the goal‐relevant inputs for attentional selection and representation in the brain. These weights are constantly compared to determine when sustained focus on the goal should break in favor of attending an unrelated salient event; as discussed more in Section 3 below, this dynamic balance can interact with other factors such as cognitive state, context, and individual differences. From an ecological standpoint, this system balances information relevant to desires vs. survival: though the focus of awareness may largely be determined by a current behavioral goal, attention also ensures that awareness can quickly and efficiently shift to external signals that have exceeded some threshold for safety.

While this process is often beneficial at directing our attention to potentially relevant salient stimuli, sometimes higher weights are unfortunately assigned to salient stimuli that are totally irrelevant, like a flashing billboard, resulting in automatic attentional capture by distracting information that impedes goal‐driven behavior (i.e., Pashler, 1988; Yantis & Jonides, 1984). The debates surrounding attentional capture and distraction are numerous (Eimer & Kiss, 2008; Folk et al., 1992; Leber & Egeth, 2006; Luck et al., 2021; Theeuwes, 2004). Intriguingly, recent work suggests that sometimes salient distraction can even disrupt the weighting system all together—for example causing a temporary reversal in weights such that visual cortex briefly prioritizes task‐irrelevant features over task‐relevant features across the entire visual field (Dube et al., 2022).

2.2. Internal attention and the role of working memory

As summarized above, our behavioral interactions with the external environment require attention to operate over perceptual information. However, attention also operates over information represented internally in the absence of sensory input (Chun et al., 2011), and the efficiency of our behavioral interactions often relies on internal perceptual and mnemonic representations (i.e., Carlisle et al., 2011). Importantly, the targets of internal attention are also more flexible and varied than was once thought: Internal attention can operate over individual objects (Cowan, 2010; Luck & Vogel, 1997) and features (Serences et al., 2009) in memory, as well as nonsensory representations such as visuospatial and motor action plans (i.e., pragmatic representations) (van Ede, 2020; van Ede et al., 2019), and sensory‐motor rules (for a review see van Ede & Nobre, 2023). Effective behavior requires not only an internal representation of the relevant sensory information, but a plan for how to use that information and knowledge of the rules that govern your interactions with the environment. Internal attention can also involve competition between different task rules and response options (for a review, see Chun et al., 2011). For example, code switching—alternating between different languages, styles of speech, and/or behaviors based on social context (Auer, 2013)—is a prime example of internal attention in real‐world settings.

Working memory in particular is intricately linked with attention (Awh et al., 1998; Baddeley, 1993; Chun, 2011; Gazzaley & Nobre, 2012; Kane et al., 2001; Kiyonaga & Egner, 2014; Oberauer, 2009; Olivers, 2008; Wood & Cowan, 1995), though the exact nature of their relation is debated (for a review, see Oberauer, 2019). Some researchers suggest that “working memory” can simply be thought of as sustained internal attention to a set of features, objects, or events over time (i.e., Chun, 2011), and that attention and working memory share a single, limited resource (Kiyonaga & Egner, 2013). Other researchers suggest that working memory is instead a separate mechanism for active storage (Oberauer, 2019), and its interactions with perception and long‐term memory are governed by a set of attentional filters that select and restrict the flow of information based on relevance (Dube et al., 2017; Vogel et al., 2005). By both accounts, working memory serves as an interface between perception and long‐term memory that relies heavily on internal attention to store and evaluate incoming perceptual information, and to recall representations stored in long‐term memory to support behavior.

Not only does internal attention aid in the storage of information in working memory and long‐term memory, but it also allows for information to be dynamically reprioritized in working memory in service of current and prospective tasks. If a behavioral goal is updated following the encoding of information, for instance, the internal focus of attention can shift to the immediately goal‐relevant information, prioritizing it so that it exerts the greatest influence over performance (Garavan, 1998; Gehring et al., 2003; Griffin & Nobre, 2003; Landman et al., 2003; Oberauer, 2002), and shifting quickly and dynamically to another internal representation when required (van Moorselaar et al., 2015). As such, internal attention can operate flexibly over its targets stored in memory.

In an intuitive sense, internal attention is well described by the weights and balances analogy: Just as competing external inputs are assigned different weights and balances, the attentional system also monitors and weighs competing internal signals to control the deployment of internal attention within memory and decisions. This idea is common in the decision‐making literature, where both sensory‐based (e.g., this patch of dots is moving to the left or to the right: Shadlen & Newsome, 2001), and internal (e.g., do I want to eat this chocolate bar or the healthier salad: Krajbich et al., 2010; Krajbich & Rangel, 2011) decisions have been described as a neural evidence accumulation process, where attention can dynamically influence the weighting of different options. However, it is worth noting that, compared to external attention, less is known about the mechanisms of internal attention, in part because most internal targets of attention do not have an obvious spatial or map‐like representational structure, posing intriguing challenges to existing attention models such as salience maps or “spotlight” metaphors. The weights and balances metaphor is intended to be mechanistically‐agnostic (for now), so that it is broad enough to usefully describe what we do know about internal and external attention, while highlighting open questions for future research. Indeed, it is particularly intriguing to muse upon what such a system might actually look like neurally for an internal priority “map”. Our metaphor is of attention as a system of weights and balances, but that does not mean that the mechanism of weighting/balancing is the same across the whole system. What the different subprocesses have in common is that they are all doing some type of weighting/balancing. An ongoing goal of attention research is to uncover what these mechanisms and specifics are. Learning how the different subtypes of internal and external attention share similarities or differ in mechanisms and representational structures is an important direction for the field.

2.3. Balancing between internal and external influences

One limitation of a taxonomy classifying attentional targets as either external or internal (Chun et al., 2011) is that it does not explicitly account for what one could consider the meta‐level of attention: competitions between external and internal information. Conceptualizing attention as a broad system of weights and balances encompasses the idea that weights can also be dynamically assigned to external vs. internal information. In other words, there may be a set (or multiple sets) of weights/balances determining priority among external visual input, and another set(s) of weights/balances determining priority among internal thoughts, as well as a “meta” set of weights/balances determining whether more priority is given to the external stimuli vs. the internal thoughts (e.g., on‐task vs. mind‐wandering).

In one sense, this analogy could apply to the findings that people oscillate between states of “zoning in” vs. “zoning out” (Esterman et al., 2013), which can have consequences for both responsiveness to external stimuli and internal cognitive processing (de Bettencourt et al., 2018, 2019). In another sense, relative attentional weightings within one level could influence processing at the other level. Specifically, weights assigned to external sensory stimuli can become useful guides for internal attention, as sensory input that is deemed relevant is more likely to also become the focus of internal attention. For instance, visual information that is attended is more likely to be encoded into memory (Dudukovic & Wagner, 2006; Turk‐Browne et al., 2013) and, among attended and remembered information, the most relevant representations are encoded with the greatest precision (Dube et al., 2017; Emrich et al., 2017). Even external information that is erroneously assigned a heavy weight, like a physically salient but irrelevant visual distractor, has privileged access to memory (Dube & Golomb, 2021). Likewise, heavily weighted internal information is more likely to interact with perception. Assigning a heavy weight to the mental (memory) representation of your friend's blue jacket, for instance, “activates” that internal representation, allowing it to then bias external attention to a greater degree than other memory items (Olivers, 2008).

These varying weights may not only govern the information that flows between external and internal sources, but also the way working memory is used to interface between them. Searching for your friend based on the color of their jacket, for instance, relies on both external and internal attention: internal attention is used to retrieve a mental representation of that previously seen jacket from storage in long‐term memory, and to activate that representation in working memory (our “online” short‐term storage system for internal representations), so that it can be used to drive external attention to matching features in the environment (Oberauer, 2009). Much of the role of internal attention and working memory is to regulate the flow of information from the inside‐out and the outside‐in: Relevant internal information is represented in memory and used to exert control over behavior, and relevant external information is represented internally to be committed to long‐term memory or to aid in planning/future behavior. A large part of regulating such interactions, however, is closing the circuit between the internal and external environments so that they do not interact unnecessarily. When maintaining a representation in working memory for a future behavior, for instance, it may be unnecessary—and even harmful—if that representation guides behavior in a current task. On the other hand, a relevant or particularly important representation in working memory should be protected from irrelevant or distracting sensory input so that it is not forgotten or damaged. As such, in many contexts the broader attentional system must switch between external and internal foci, shielding external and internal information. Shielding can be evaluated via “switch costs” when alternating between external and internal attentional states: In experiments that require participants to perform a task using either information presented on‐screen or retrieved from memory, participants are slower and more error prone on “switch trials” that dynamically shift the focus from external to internal sources of information (or vice versa), demonstrating a shielding failure (Verschooren et al., 2019). These switch costs are asymmetric, suggesting that shielding internal representations from external interference is more efficient that the inverse (Verschooren et al., 2020), and recent work suggests that shielding failures can be mitigated by temporally predictable external interference (Gresch et al., 2021).

As such, the attentional system not only assigns relative weights to external sources and internal sources independently of each other, but it also consistently monitors these weights to compare external vs. internal attention to determine what should be prioritized in a given moment. This dynamic balancing of the various weights then gives working memory critical information about how to interface between external and internal attention from moment‐to‐moment to ensure that this circuit is operating effectively.

2.4. Balancing current, past, and future

In the sections above, we focused on ways the attentional system can weigh and balance signals to ensure that information that matters most in the current moment reaches awareness and prioritized processing. But attention is not just influenced by the current moment; attention is also biased by what tended to be most relevant in the past. In addition to top‐down and bottom‐up sources of attentional guidance, current theories widely support a third experience‐driven factor that describes the role of learned environmental regularities in driving attention (Anderson et al., 2021; Awh et al., 2012; Hutchinson & Turk‐Browne, 2012; Theeuwes, 2019). That is, information in the environment can be considered important if it is currently goal‐relevant (top‐down) or otherwise salient (bottom‐up), but we can also use experience to assign importance to inputs (locations, objects, sounds, etc.) that have been consistently useful. For instance, the weight assigned to a specific location may increase over time because a desired item appears there frequently (Geng & Behrmann, 2005), or one color may begin to elicit a stronger signal over another because it has been associated with a higher reward (Anderson et al., 2013). As such, experience can enhance the weights assigned to stimuli associated with relevance or reward. Experience can also decrease the weights assigned to stimuli that we learn are irrelevant: if a particularly salient signal repeatedly appears as a distractor that is deemed irrelevant by the observer, the attentional system can learn to ignore or suppress a consistent feature or location of that signal (Leber et al., 2016; Wang & Theeuwes, 2018). Interestingly, some studies have shown that experience‐driven learning over time can be more effective at de‐prioritizing information than explicit instructions about what, or where, to suppress (Beck et al., 2018; Moher & Egeth, 2012).

In addition to incorporating past experience into the system of weights and balances, attention may also incorporate future needs. We can, for instance, use internal attention to maintain information that is prospectively relevant, but in a state that does not allow it to interact with behavior in a current task (Olivers et al., 2006). For instance, when presented with a piece of visual information that will be useful to support an upcoming task, we can maintain this information in working memory in a passive, accessory state (Olivers et al., 2006). In this state, this representation is held “online” so that it can be activated when necessary, but is shielded from perception so that it cannot bias attention in the current task for which it is not relevant. This is well explained by the weighting system: a stronger weight is assigned to internal information that is currently relevant, and weaker weights assigned to information that will be relevant in the future, with these weights dynamically fluctuating with changes in behavioral goals and task demands.

3. A DYNAMICALLY CHANGING BALANCE

A key aspect of our definition of attention as a system of weights and balances is that the system is dynamic and fluid over time, an idea similar to the Dynamic Prioritization Approach proposed by Shomstein et al. (2022), also in this special issue. In the real world, there is rarely just one single “correct” target of attention. Externally, there are often multiple potentially important locations, objects, and features vying for attention, and the same is true for the vast number of possible internal targets of attention. As a result, sometimes multiple targets can be activated simultaneously or in quick succession, resulting in divided or sequential attention. The idea that time‐evolving weights can dictate the spatiotemporal path of attention is the foundation of dynamic salience maps within the realm of external attention (e.g., where a visual scene can be modeled as a map of continuous weights, and the location on the map with the highest weight at any given moment becomes the target of attention: Itti & Koch, 2000, see also Sprague & Serences, 2013; Zhang et al., 2012). An important aspect captured by the weights and balances analogy we propose here is that attention is selecting not simply the highest weighted external targets, but continually incorporating the relative weights of internal targets as well.

As a real‐world example of these dynamic interactions, say you have misplaced your car keys and are looking for them in your house. If it is getting near the time for you to leave for an important meeting, internal attention might prioritize a focused task state of searching for your keys (over, say, packing your lunch). A mental (working memory) target template may initially place more weight on top‐down, goal‐oriented information so that objects matching that template will attract your attention. Perhaps as you begin the search for your keys, your attention is also strongly influenced by experience‐driven information, drawing initially heavier weights to the nightstand where you have most commonly found your keys before. If that fails, your spatial attention may broaden (decreased location‐based weights), allowing for stronger weighting of bottom‐up stimulus‐driven information, and the shine of your keys against the dark floor then captures your attention and terminates your search, allowing you to shift internal focus to the task of packing your lunch.

3.1. Other factors influencing attentional weights

As illustrated above, attentional weights can change over time due to differences in top‐down goals, bottom‐up salience, and accumulating experience. Attentional weights can also be influenced by fluctuating internal factors such as intrinsic sampling rhythms (Fiebelkorn & Kastner, 2019), motivation (Engelmann & Pessoa, 2014), emotional state (Bar‐Haim et al., 2007), and external factors such as emotionally‐salient affective stimuli (Fenske & Eastwood, 2003) or self‐referential stimuli (the “cocktail party” effect is one such example of individual‐specific priorities, as it has been well‐established that we are quite adept at picking out the sound of our own name from within a noisy setting: Pollack & Pickett, 1957).

There also exist static traits that are known to lead to more stable differences in attention behaviors between individuals. Working memory capacity is one trait that has been found to impact attentional abilities (Kane et al., 2001), including findings that individuals with lower working memory capacity are less able to filter out irrelevant distractors (Vogel et al., 2005). Attentional filtering abilities and priorities can also vary across the lifespan, with young children exhibiting more distributed attention across both task‐relevant and irrelevant information (Plebanek & Sloutsky, 2019), and older adults exhibiting reduced attentional suppression (Campbell et al., 2012; Gazzaley et al., 2005); in the analogy of attentional weights and balances, one could describe young children as exhibiting immature attentional systems with relatively little variation in weights to different targets (too much balance, leading to less selectivity), and older adults' attentional systems as able to enhance relevant information with positive weights but not suppress irrelevant information with negative weights.

3.2. Failures of attention

Because the dynamic nature of the world necessitates a dynamic attentional system, it becomes inevitable that instances will arise in which the scales become imbalanced or inappropriate targets are erroneously assigned highest priority. What happens when things go wrong? In the lab, the consequences of not orienting sufficiently to the desired target of attention have been well‐documented. During search tasks, becoming distracted or misled by an invalid cue will slow reaction time and decrease simple response accuracy (Folk et al., 1992; Müller & Rabbitt, 1989). Moreover, because attention is also important for visual feature‐binding and object perception (Treisman & Gelade, 1980), imperfect attention can also result in distinct patterns of feature‐binding errors. For example, different signatures of feature errors have been found when attention is delayed in updating location‐based weights during dynamic shifts of attention versus when attentional weights are simultaneously split between two locations (Dowd & Golomb, 2019; Golomb et al., 2014), when attention is captured by a salient distractor (Chen et al., 2019), and when external spatial attention is temporarily in a “dual‐spotlight” state induced by remapping across eye movements (Golomb, 2019).

Outside the lab, attentional imbalances can result in various degrees of undesirable outcomes. Some may be isolated events, such as not noticing your oven timer sounding off or finding yourself unable to focus on a lecturer's lesson, while other impacts could be more chronic, particularly when it comes to determining priority at levels such as external stimuli vs. internal thoughts. When functioning properly, this system serves to ensure that a tenable balance between all the targets, goals, and influences on attention is maintained. Chronic instances of attentional imbalances—particularly those involving internal attention—can result in symptoms that lead to clinical diagnoses. One of the most prevalent examples is ADHD, the symptoms of which include difficulty sustaining focus and being easily distracted across multiple settings (American Psychiatric Association, 2013), and which has been associated with a delay in childhood development of some areas of prefrontal cortex known to be involved in attentional control (Shaw et al., 2007). In the context of weights and balances, this impaired attentional control may manifest as an overweighting of bottom‐up signals and/or novel information relative to sustained top‐down signals from longer‐term goals. Depression and anxiety may also be associated with attentional imbalances, particularly with too much attention being devoted to internal vs. external targets. Rumination, or the tendency to overly dwell on one's own negative thoughts, emotions, and behaviors, is particularly common among depressed individuals, and evidence has been found that this excessive internal focus can hinder executive functioning in depressed patients relative to when they were told to focus on more external items (Watkins & Brown, 2002). This may represent an overweighting of internal signals, particularly those coming from negative thoughts and emotions, and leading to a biased balancing at multiple levels that trap attention in a spiral of rumination, unable to appropriately weigh other signals, such as those from external sources, that would typically help guide attention to more positive life aspects.

4. CONCLUSIONS

The concept of attention is vast, and research in the field is constantly adding new knowledge and opening up new questions. In addition to open questions and debates within different subfields of attention, an unresolved question is whether—or to what extent—the mechanisms that we consider “attention” are common or distinct across levels. There will likely always be a tension between whether the field is better served by developing unified accounts of attention or focusing more on gaining more explicit understanding of the specific subprocesses, but we hope that this review and analogy of attention as a multi‐level system of weights and balances helps stimulate future discussion and research along these lines in putting the pieces together, providing a productive scaffolding to discuss both external and internal attention and how these different targets of attention can be delicately balanced—or dynamically imbalanced—during scenarios ranging from optimal behavior to various momentary or chronic failures of attention.

AUTHOR CONTRIBUTIONS

William Narhi‐Martinez: Conceptualization (equal); writing – original draft (lead); writing – review and editing (supporting). Blaire Dube: Conceptualization (equal); funding acquisition (supporting); writing – original draft (equal); writing – review and editing (equal). Julie D. Golomb: Conceptualization (equal); funding acquisition (lead); supervision (lead); writing – original draft (supporting); writing – review and editing (lead).

FUNDING INFORMATION

This work was supported by grants from the National Institutes of Health (2R01‐EY025648 to JG), National Science Foundation (NSF 1848939 to JG), and NSERC (PDF to BD).

CONFLICT OF INTEREST

The authors have declared no conflicts of interest for this article.

RELATED WIREs ARTICLES

Attention and platypuses

What is attention?

Narhi‐Martinez, W. , Dube, B. , & Golomb, J. D. (2023). Attention as a multi‐level system of weights and balances. WIREs Cognitive Science, 14(1), e1633. 10.1002/wcs.1633

DATA AVAILABILITY STATEMENT

Data sharing is not applicable to this article as no new data were created or analyzed in this study.