Enhanced but Indeterminate? How Attention Colors our World

Abstract

Attention makes things look brighter and more colorful. In light of these effects, representationalist philosophers propose that attentive experiences represent more determinate color properties than inattentive experiences. Although this claim is appealing, we argue that it does not hold for one of our best conceptualizations of content determinacy, according to which an experience has more determinate contents if it represents a narrower range of values within the relevant dimension. We argue that our current empirical evidence fails to show that attention has this kind of effect on color perception. We then offer an alternative, representationalist-friendly account of the attentional effects, as changes in vividness.

1 Introduction

This paper discusses a seemingly simple phenomenon. A blue car passes down the road and we see it out of the corner of our eye. We might pay attention to it, or we might not. Intuitively, paying attention to the car enables us to make more precise claims about its properties: for instance, we can say that the car is blue. Contrastingly, when we do not pay attention, the claims we can make become less precise: we might say that the car is dark, but we might not be able to give more information about its color. These observations seemingly support the following claim:

• (1) Attentive color experiences are more determinate than inattentive color experiences.¹

Claim (1) is often taken to be entailed by the more general claim that attentive perceptual experiences are more determinate than inattentive perceptual experiences, a claim often defended by representationalist philosophers (for instance, Nanay 2010; Stazicker 2011). Representationalist philosophers hold that all properties of experience are or supervene on representational properties, that is, properties about what is represented in experience.² Specifically, they hold that there is no phenomenal difference (i.e., a difference in what an experience feels like) without a representational difference (i.e., a difference in what the experience represents). The effects of attention on perceptual phenomenology have been often regarded as a pressing counterexample for representationalism, since attention seemingly elicits changes in what experiences feel like without a change in what is represented (Block 2010, 2015). In this view, attention only changes how we experience things, but not what is represented in our experience. In response to this worry, representationalists contend that attention in fact changes what properties are represented in experience: it makes the experience represent more determinate properties.

This idea is also expressed with the claim that attentive experiences have more determinate contents than inattentive experiences. For example, Nanay (2010) argues that when we are not attending to an object we visually perceive, our visual experience attributes to it highly indeterminate properties, which are very general and are compatible with several further specifications (e.g., “dark-colored”). In turn, attending to the object makes the attributed properties more determinate (e.g., “blue”).³ Recently, these ideas have been also cast in terms of precision: attention increases the precision of visual experience, by making its contents more precise.⁴

Philosophers on both sides of the dispute have used the notion of range content to cash out differences in the determinacy and precision of perceptual experiences (Block 2015; Boone 2020). A range content is an interval of values within a feature dimension attributed by a perceptual experience to an object. For example, your visual experience of a person can represent her as being between 160 and 170 cm tall, instead of attributing her a specific height value. In line with representationalism, the suggested idea is that attentive perceptual experiences attribute narrower value intervals than inattentive perceptual experiences. When you are not paying attention to the car, your experience attributes to it a range of color values that includes blue, gray and black shades. Contrastingly, paying attention to the car would reduce this value range, by excluding gray and black shades.⁵ Claim (2) summarizes this proposal:

• (2) Attentive color experiences have narrower range contents, along the color dimension, than inattentive color experiences.

Despite getting its initial footing from phenomenological intuitions (as in Nanay 2010), claim (2) is an empirical one. Plausibly, the contents of a visual experience E1 are a matter of the information represented by E1. If so, then the question of whether E1 has narrower range contents than another experience E2 is a question of how much information of the relevant property is represented by E1 and E2. And this question depends on facts about our visual system.

To date, several empirical arguments have been given in support of the general idea that attention makes the contents of visual experiences more determinate.⁶ These arguments often emphasize the influence of attention on properties like the spatial resolution of perception (Stazicker 2011) or contrast sensitivity (Brogaard 2015; Boone 2020). But notably, few of these arguments directly address the effects of attention on color contents.⁷ Our first goal in this paper is to start filling in this gap, by clarifying the kind of empirical evidence that could support or disprove claims (1) and (2). We suggest that findings usually invoked by representationalist philosophers may at best lend support to claim (1) but may not support (2). Thus, our second goal is to suggest an alternative interpretation of (1).

Here is the plan. In §2, we clarify the kind of “attentive” and “inattentive” experiences involved in claims (1) and (2), and we introduce some of the most important empirical findings evincing a difference between the two kinds of experiences. In §3, we clarify the notion of “the color dimension” from claim (2), in the light of an intuitive color system widely used in current vision science. In §4 and §5, we discuss different ways how the known effects of attention on color perception could be taken as evidence for (2), and conclude that this claim remains largely unsupported. Finally, in §6, we propose an alternative way of cashing out claim (1), which is compatible with both representationalist and anti-representationalist approaches.

2 The Attentional Difference

Characterizing an experience as “attentive” (or “inattentive”) requires specifying which of the many varieties of attention it recruits (or fails to recruit). Here, we restrict the claims to visual, external, spatial, covert attention. The external form of attention targets things in the perceptual environment as opposed to internal attention, which targets things within the subject’s mind (Chun et al. 2011). Spatial attention selects a spatial region and facilitates the processing of information at that region while suppressing information from other locations.⁸ It is overt when it follows the direction of gaze, and covert when it does not (for example while transferring attention between objects in the visual periphery).

This overt/covert distinction is important for disentangling the effects of attention from the effects of sensitivity differences throughout the retina. Photoreceptors are more highly concentrated at the fovea, the central retinal portion encoding the spatial region where the eyes are pointed at. Consequently, the fovea has higher spatial resolution and encodes more color information than the periphery, where receptors are sparser. When we deploy overt attention, the increased quality of our percept is due to a combination of attentional enhancement and these hardwired differences in photoreceptor distribution. By contrast, covert attention to a target in the visual periphery isolates the effect of attention.⁹

Spatial covert attention has a variety of measurable effects on visual perception. These effects are found regardless of whether attention is endogenously or exogenously cued,¹⁰ and are robust enough to be observed under a variety of control conditions. Thus, the phenomenological claim that there is a difference between attentive and inattentive experiences has an empirical basis. The open question is whether the attentional difference is a difference in determinacy and/or range contents.

Some of the attentional effects on perception often invoked in this debate include increasing spatial resolution (Yeshurun and Carrasco 1998; Carrasco and Yeshurun 2009; Gobell and Carrasco 2005), perceived size (Anton-Erxleben et al. 2007), perceived speed and flicker rate (Montagna and Carrasco 2006; Fuller et al. 2009), perceived contrast and, under some conditions, also perceived brightness (Carrasco et al. 2004; Tse 2005; Tsal et al. 1994 ).¹¹ Attention also increases perceived color saturation: an attended color patch looks more colorful (“greener”, “bluer”) than an unattended color patch with the same saturation value (Fuller and Carrasco 2006; Kim et al. 2014). Figure 1 illustrates some of these effects.

Fig. 1

Some effects of attention on visual perception. Attention makes us perceive finer details (a), increases the apparent size of objects (b) and gaps (c), increases apparent speed of moving objects (d), perceived contrast (e), and perceived saturation (f). Adapted from Carrasco and Barbot (2019)

Do any of these effects vindicate claims (1) and (2)? Prima facie, some effects seem more suitable than others. For example, brightness and saturation seem more directly related to the color contents of an experience than size or spatial resolution. Indeed, increased perceived brightness and saturation seem to be good evidence for (1), the claim that attentive color experiences are more determinate than inattentive color experiences. However, the case is not so clear for (2), the claim that attentive color experiences have narrower range contents than inattentive color experiences. To see why, we must first clarify what it means for the ranges of represented color values to be wider or narrower. This requires a clarification of what “the color dimension” is.

3 What is the Color Dimension?

In everyday language, we use the word “color” to refer to a property we take objects to have: tomatoes are red, limes are green, and oranges are orange. We use words like “red”, “green” and “orange” to name properties that most of us see similarly. These everyday words are sometimes ambiguous between an objective and a subjective sense: colors as properties of objects, and colors as properties that are in the eye of the perceiver.

In vision science, the term “color” is used to pick out at least three things: a physical stimulus, a neural state encoding that stimulus, and a conscious percept resulting from that encoding (Teller 2003). The physical stimulus is a surface reflectance profile or surface spectral reflectance, that is, the light wavelengths reflected and absorbed by an object’s surface. Surface spectral reflectance properties are plausibly the truth-makers for our talk of objects as having colors: for oranges to be orange is for their surface to absorb some wavelengths and reflect some others (Byrne and Hilbert 2003). However, our claims (1) and (2) are not about properties of physical objects in the world. They are rather claims about our experiences of these properties.¹² Thus, here we are primarily concerned with the third element in the vision scientist’s triad: subjective color perception.¹³

Vision science tells us that subjective color percepts result from an interaction between surface reflectance profiles and a source of light cast at an object. In addition, the colors we perceive are influenced by the quality of our photoreceptors, perceptual principles like color constancy, and even top-down factors like previous knowledge and expectations (Hughes et al. 2014). Attention is often included amongst these top-down factors.¹⁴

Claim (1) picks up one way how attention affects experienced colors, namely, by making them more determinate. This is not to say that this is the only influence of attention on perceived color. Other influences might be related to accuracy, quality of encoding or color memory. In turn, claim (2) says that, due to attention, our subjective color experiences represent narrower ranges of values, along the color dimension. But, how should this “color dimension” be characterized?

To get a sense of the kind of answer we are after, compare with other examples of range contents: height and temperature. Height is a dimension that can be measured in centimeters; to represent a narrower range of values along this dimension is to attribute a smaller number of centimeters or inches. Temperature is a dimension that can be measured in Fahrenheit degrees; to represent a narrower range of values along this dimension is to attribute a smaller number of Fahrenheit degrees. However, the answer for color does not seem so straightforward; especially, when one focuses on the subjective aspect of color that appears in our color experiences.¹⁵

We propose that a good initial grip on the relevant dimension is gained through the HSB color system used by vision scientists.¹⁶ This system construes color as a property constituted by three independent properties or dimensions: hue, saturation, and brightness ('H', 'S' and 'B', respectively).¹⁷ To illustrate, imagine your favorite shade of blue. Whatever this shade is, it has a hue property: “the human sensation according to which an area appears to be similar to one, or to proportions of two, of the perceived colors red, yellow, green, and blue” (Bora et al. 2015). Your chosen shade of blue also has some degree of purity: it can be mixed with different proportions of grey. This property is called saturation (Hughes et al. 2014, p. 748). Finally, this shade of blue has a luminance attribute: it can be lighter or darker.¹⁸ For illustration, see Fig. 2.

Fig. 2

A visualization of the HSB color space

The HSB conceptualization is the background hypothesis for our claims. Admittedly, this system does not capture every color we see in the world. Since HSB is designed for mapping the colors our devices can produce, it represents only a subset of colors defined by certain thresholds, which are determined by the physical constraints of our technologies (Hughes et al. 2014, p. 768). Thus, there are colors out there that evade HSB characterization. However, what is important for us is that the HSB conceptualization is familiar and intuitive. It breaks down perceived color into aspects we can easily discern in our phenomenology. Also, as we have seen, there is empirical research already at hand, measuring the attentional effects on hue, saturation and brightness. Hence, the hypothesis that the “color dimension” is constituted by the three-dimensional HSB space is at least a good starting point for giving claim (2) some empirical traction.¹⁹

According to (2), attentive color experiences have narrower range contents, along the color dimension, than inattentive color experiences. If the color dimension is a complex property constituted by sub-dimensions H, S and B, then one natural way of cashing out claim (2) is claim (3):

• (3) Attentive color experiences “carve out” a smaller region of the three-dimensional HSB space than inattentive color experiences.²⁰

We think that (3) lays down a plausible account of what it takes for a color experience to have a narrower range content: E1 has a narrower range content along the color dimension than another experience E2 if and when E1 attributes a smaller region of the HSB space than E2. In this way, E1 would attribute a smaller number of color values than E2 and would thereby exclude more values than E2. Unfortunately, however, our current evidence on the effects of attention on color experiences does not clearly show that attention can do this.

Conceptually, an experience E1 attributes a smaller region of the HSB space than another experience E2 if E1 attributes narrower value ranges along at least one of the sub-dimensions H, S or B. Thus, either of the following suffices as a truth-condition for claim (3):

• (3’) Attentive color experiences represent narrower intervals of values, along each of H, S and B, than inattentive color experiences.

• (3’’) Attentive color experiences represent narrower intervals of values, along some of H, S or B, than inattentive color experiences.

Since (3’) is a more ambitious claim than (3’’), we call the attempts to vindicate (3) with the aid of each of these the “ambitious” and the “modest” strategy, respectively. In the next two sections, we discuss each strategy and explain why it is insufficiently supported by our current evidence.

4 The Ambitious Strategy

The ambitious strategy requires defending (3’):

• (3’) Attentive color experiences represent narrower intervals of values, along each of H, S and B, than inattentive color experiences.

Claim (3’) entails that attention makes a difference to each of the H, S, and B components of a color percept. Thus, claim (3’) entails each of the following:

• (3’i) Attentive color experiences represent narrower intervals of values along H

• (3’ii) Attentive color experiences represent narrower intervals of values along S

• (3’iii) Attentive color experiences represent narrower intervals of values along B

Now, as we will explain shortly, our current evidence suggests that (3’i) might be false: attention might not change the range of hue values represented in color experiences. Furthermore, our current evidence is silent on whether attention changes the ranges of represented saturation or brightness values, so (3’ii) and (3’iii) remain unsupported. In the remainder of this section, we discuss (3’i). We come back to (3’ii) and (3’iii) in §5.

Fuller and Carrasco (2006) investigated whether attention affects how we perceive hue. Their answer was negative. In their study, participants were presented with tilted ovals that varied in hue from blue to purple, while maintaining constant saturation and luminance. Participants must report the orientation of the oval that looked bluer (or, in a control condition with reverse instructions, the oval that looked more purple). Attention was exogenously manipulated with a cue appearing at the location of either oval briefly before the oval was presented, or at the center of the screen for a neutral condition (i.e., where attention was not preferentially allocated to either oval). Fuller and Carrasco found no effects of attention on perceived hue. If attention caused participants to perceive stimuli as “bluer”, then they would have reported the orientation of cued ovals more frequently, but this did not happen.

Notably, Fuller and Carrasco did find an effect in the reverse instruction condition, where participants reported the orientation of the cued ovals more frequently. This suggests that cued ovals might have been perceived as “more purple”. However, the effect was not reliable, for it was not found in the “report bluer oval” condition (if attention was reliably causing stimuli to be seen as more purple, participants would have picked the cued ovals less frequently when asked to report the ones that looked “bluer”).

Fuller and Carrasco thus concluded that attention does not change apparent hue. Prima facie, this contradicts claim (3’i). Narrowing down the represented range of hue values entails making a change to represented hue, as per the following plausible conditional:

• (C1) If attention narrows down the range of hue values represented by a perceptual experience, then attention changes represented hue.

If C1 is true, then, if attention does not change represented hue, a fortiori it does not change the represented range of hue values. Now, to be sure, we must emphasize that Fuller and Carrasco’s experiment just tested for one specific way how attention might change perceived hue, namely, by shifting its value towards one or the other side of the spectrum. Evidence (or lack thereof) for this kind of change might well be orthogonal to the question of whether attention narrows down the range of represented hue values. For one thing, the following conditional is much more controversial than C1 (and it might even be false):

• (C2) If attention changes represented hue, then attention narrows down the range of hue values represented by perceptual experience.

Suppose that a reliable attentional effect was found in Fuller and Carrasco’s experiment, such that attended ovals were reliably perceived as, say, bluer. That would be evidence that attention changes perceived hue, but it would not yet be evidence that attention changes perceived hue in the required way, that is, by reducing the number of hue values represented by the experience. The way to do this would be to show that a color patch is perceived as having, say, 15 possible hue values when unattended, and perhaps only 1 or 2 values when attended (see Fig. 3).

Fig. 3

A visualization of the difference between narrow and wide ranges of hue values

To be sure, it remains possible that attention does reduce the values of hue represented within the portion of the spectrum that still constitutes the same hue value (as in Fig. 3). However, since the relevant evidence is not available, claim (3’i) remains unsupported: attentive color experiences might not attribute narrower ranges of values along all three dimensions of the HSB space. These considerations should undermine the ambitious strategy, even if they are not yet a knock-down argument.

Suppose, now, that claim (3’i) was in fact proved false. Suppose that Fuller and Carrasco’s results (namely, that attention does not affect how blue or how purple a stimulus appears) were just a manifestation of a more general fact, namely, that attention does not affect hue perception in any way. In that case, the ambitious strategy would be untenable, for it would be false that attention attributes narrower ranges of values along all three dimensions of the HSB space. And although the modest strategy (to be discussed in next section) would in principle remain available, one might worry that attention’s capacity to narrow down range contents along the color dimension in fact rises or falls with its capacity to narrow down hue contents.

Some authors emphasize that hue is the first characteristic of color we notice (Joblove and Greenberg 1978). And indeed, our most paradigmatic color terms (“red”, “green”, “blue”) individuate colors by hue. Accordingly, hue could be considered the primary or most important dimension of perceived color. Thus, if one wants to say that the color contents of E1 are more determinate than those of E2 (as per claim (1)), then one must be prepared to say that the hue contents of E1 are more determinate than those of E2. E1 must then represent a narrower range of hue values than E2, in order to represent a narrower range of values along the color dimension (as per claim (2)). Nothing short of this would seem to vindicate claims (1) and (2). That is to say: to vindicate (1) and (2), it is not sufficient for attentive experiences to just carve out a smaller region within the HSB space. They must carve a smaller region in the right way, that is, by carving down a smaller portion of the hue spectrum. We call this the hue-first worry. It is a worry for the proponents of claims (1)-(3), in the light of the (potential) falsity of (3’i).

We think that the hue-first worry might be on the right track. It seems hard to deny that hue has a special status in color perception. Besides its role in noticing and individuating colors, there is an intrinsic difference between hue and the other dimensions of the HSB space. Hue is a metathetic dimension. This means that it has no meaningful zero value, and that it cannot be increased or decreased: it makes no sense to say that a color patch is more hued than another. Saturation and brightness, on the other hand, are prothetic: it is possible to have zero values of saturation and brightness, and it is meaningful to say that a color patch is more or less saturated or bright.²¹

One can further argue that, being a metathetic dimension, hue can indeed be considered as the primary or more fundamental aspect of color, for example because it might constitute the main dimension of the color space, which is modulated, in a somewhat secondary sense, by saturation and brightness. However, even if hue is the primary dimension of color perception, this does not entail that carving out a smaller region of the color space requires representing narrower ranges of hue values.²²

Say your inattentive experience of blue represents 15 hue values, 15 saturation values, and 15 brightness values. It thus carves out a 15 × 15 × 15 region of the HSB space. Say then that, in virtue of directing your attention appropriately, your experience now represents fewer saturation and brightness values, say, 10 instead of 15. Your attentive experience then carves out a 10 × 10 × 15 region of the color space. Thus, your attentive experience does carve out a smaller region of the HSB space, even if hue is left untouched.

Along these lines, one might further argue that the notion of range contents does not really apply to all aspects of color experience. The view would be that an experience can only have range contents with respect to prothetic dimensions such as saturation and brightness. Hence, if determinacy is a matter of range contents, the way for color experiences to become more determinate would be for them to represent narrower ranges of values along the saturation and brightness dimensions. We will now consider the prospects for this strategy, and we will argue that it is also not supported by our current evidence.²³

5 The Modest Strategy

As suggested above, claim (3’) may be unnecessarily ambitious. Maybe one can carve out a smaller region of the HSB space by narrowing down the values represented along only a subset of HSB, and maybe that is sufficient for a color experience to have narrower range contents. Thus, one might replace (3’) with (3’’):

• (3’’) Attentive color experiences represent narrower intervals of values, along some of H, S or B, than inattentive color experiences.

The lack of attentional effects on represented hue values is not a problem for (3’’). This claim can still turn out true if either (or both) of the following are true:

• (3’ii) Attentive color experiences represent narrower intervals of values along S

• (3’iii) Attentive color experiences represent narrower intervals of values along B

One could characterize color range contents in this more modest way, regardless of their views on the relation between hue, saturation and brightness. The modest strategy is compatible with the proposal that hue is the primary dimension of color content. For example, one could argue that the phenomenology of hue is determined by represented values of saturation and brightness, maybe in this sense: narrower saturation and brightness ranges make hues more easily discriminable in our experience, even if the represented hue ranges remain unchanged.²⁴ On the face of it, this proposal seems supported by the observed effects of attention on perceived saturation (Fuller and Carrasco 2006; see also the discussion below). Increasing the saturation of a color patch plausibly facilitates discriminating its hue, while desaturating it makes discrimination harder (imagine desaturating a hot pink patch to make it grayish pink²⁵).

However, this strategy does not entail that the only roles for saturation and brightness in modulating the determinacy of color content are tied to hue discrimination. Think of our experiences of different shades of gray, which can be more or less determinate. Increasing the saturation of a gray color patch (for example, with attention) might enable us to see it as smoke gray rather than as the more indeterminate “light gray”. But it is not obvious that this phenomenal switch from light gray to smoke gray also improves our ability to pick the exact hue this shade belongs to (that is, its angle in the 360° hue space). So at least in gray perception, increases in determinacy need not be accompanied by better hue discriminability.

Thus, the modest strategy is compatible with different positions about the relations between hue, saturation and brightness and the determinacy of color content. But is the strategy sound? Are claims (3’ii) or (3’iii) supported by our current evidence about the effects of attention on visual perception? To assess this, we will now take a closer look at the studies we briefly introduced in §2 (Fuller and Carrasco 2006; Kim et al. 2014; Tsal et al. 1994; Tse 2005; see also Carrasco et al. 2004).

Let us start with (3’ii), the claim that attentive experiences represent narrower ranges of saturation values. Fuller and Carrasco (2006) presented participants with red, green, and blue tilted ovals, which varied in saturation while having constant hue and luminance. Participants fixated their gaze on a cross in the middle of a computer screen and were given either a neutral (central, on the fixation) or a peripheral cue. After another short fixation, two ovals appeared on either side of the screen: a standard one, with saturation in the middle of the saturation range, and a test one, with changing saturation. The two ovals were tilted in different directions. Participants must report the orientation of the more colorful oval (the one that looked “redder”, “greener” or “bluer”). Participants consistently reported the orientation of the attended oval, which indicates that it was perceived as being more colorful than the unattended one. Even if the attended oval was lower in saturation than the unattended one, the two seemed indistinguishable. These results were replicated in a later study with ADHD subjects, which further supports their robustness (Kim et al. 2014).

To be sure, it is plausible that these increases in perceived saturation underlie an increase in the determinacy of color perception, as per claim (1). However, this might not extend to the conceptualization of determinacy in terms of range contents proposed in (2), the claim that attentive color experiences have narrower range contents than inattentive color experiences, and its derived claims (3) and (3’). To see why, consider Fig. 4.

Fig. 4

A simplified visualization of the difference between changes in saturation and changes in range contents. Increasing saturation of a color could only move the same range of contents along the scale, without narrowing it down

The figure illustrates that increasing perceived saturation and narrowing down the range of saturation values represented are conceptually distinct. Increasing perceived saturation does not need to change the range of values represented. It could just move the range towards higher values in the scale. Think again of the blue car. Covertly directing our attention to it might increase its perceived saturation, resulting in a more saturated blue percept. But if we perceived the car as having, for example, 10 possible shades of blue, perceiving the car as more saturated need not narrow the range to, say, 3 possible shades. The range could be still 10 shades wide; but now, the shades are all a bit more saturated.

It is possible that attention does both things, that is, narrowing down the range of attributed saturation values, and shifting this range towards the more colorful end of the scale. But at this point our evidence is inconclusive. Hence, claim (3’ii) remains unsupported. The remaining possibility is now (3’iii), the claim that attention narrows down the ranges of values represented along the brightness dimension.

The brightness dimension (also called lightness, brightness, or value, with slightly different meanings) consists in the property of color related to luminance – that is, how light the color appears (as in light blue versus dark blue). Two studies by Tse (2005) and Tsal and colleagues (1994) explore the effects of attention on brightness perception.

Tse (2005) observed that covert attention to one of three overlapping gray disks, presented on a white background, makes the disk appear darker. The effect is as if this disk was popping out to a foreground plane, while the unattended disks seem to recede into a background plane. Although the effect disappears with a black background, it reappears in an inverse luminance version of the display. On this version, the attended disk looks brighter than the unattended ones. See Fig. 5 for illustration.

Fig. 5

Adapted from Tse (2005)

Attentional modulations of brightness. Left: Attended disc appears darker. Middle: This effect is not present with a dark background. Right: With dark background and reversed luminance, the attended disc appears brighter.

Tsal and colleagues (1994) further explored the interactions between the attentional effects on perceived brightness and stimulus background. They presented participants with gray squares on a white background and asked them to report the square’s brightness, by matching it to one of four squares, with Square 1 being the lighter one and Square 4 being the darker one. Attention was directed to either side of the display with endogenous and exogenous cues. They found that participants made more “dark” errors (i.e., reporting Square 2 as Square 3) for unattended squares (that is, squares from which attention was diverted away with an invalid cue). For attended squares, in turn, the proportion of “light” errors (i.e., reporting Square 2 as Square 1) was higher. These results suggested that participants were perceiving unattended squares as darker and attended squares as brighter.

Tsal and colleagues then conducted a follow-up study where they assessed whether attention was in fact “brightening” the stimuli or rather doing something else, like reducing the contrast between stimulus and background. To test this, they presented participants with the same set of squares, but this time on a black background. This condition had the opposite pattern of responses: participants made more “light” errors for unattended squares (which suggests that unattended squares were perceived as brighter) and made more “dark” errors for attended squares (which suggests that attended squares were perceived as darker). Due to these findings, the experimenters concluded that attention affects perceived brightness indirectly, that is, by removing processing resources from the background and concentrating them on the stimulus, so that the effect of the background was lessened (for example, a white background makes stimuli look darker, but if the effect of the background is lessened, the stimulus will look brighter).

Although the studies by Tse on the one hand and Tsal and colleagues on the other hand point in opposite directions, the common finding to the two of them is that attention modulates brightness perception.²⁶ Do these modulations support the view that attentive color experiences represent narrower ranges of values along the brightness dimension?

The short answer is, again: No, for familiar reasons. Evidence that attention shifts perceived brightness to the lighter or darker ends of the scale need not be evidence that a narrower range of brightness values is represented – although, as in the case of saturation, these shifts could still contribute to the representation of a narrower range of color values, in some other way.

Regarding the latter point, note that brightness is unlike saturation in one important aspect: whereas increased saturation may plausibly contribute to increasing the determinacy of color perception, increased brightness does not always do so. There is indeed an ideal level of brightness in which we are able to discern the greatest number of possible shades of color. However, that level is not towards the higher values on the brightness axis (see Fig. 3 again), but rather towards its middle portion. As an example, when we display a colored picture on the computer and increase or decrease the brightness of the display, at some point we become able to easily discriminate, for instance, orange from red. When we decrease the brightness substantially, it becomes impossible to discriminate between some colors. But the same would happen if we could increase the brightness of the display above the maximum brightness (which our displays do not allow). Too much brightness blurs the distinctions between colors. Some shades, for example of yellow, would be impossible to discriminate. Hence, if we want to say that attention increases determinacy by increasing brightness, we must show that attention helps shifting the values of perceived brightness closer to the ideal level where one value is more discriminable from another.²⁷

To be sure, the studies we just cited could indeed provide evidence that attention works in this way: depending on contextual conditions, like background color, attention might increase or decrease perceived brightness, which might in turn contribute to the representation of more determinate contents, including color contents. Still, as before, there is a conceptual gap between this kind of shift and a change in the width of the range of values represented.

If the considerations we offer in this section are correct, our current evidence on the effects of attention on perceived saturation and brightness does not unequivocally support the view that attentive experiences have narrower range contents along any of these dimensions. Hence, claims (3’ii) and (3’iii) remain unsupported: our current evidence does not make clear whether attentive color experiences represent narrower ranges of saturation or brightness values. Consequently, (3’’) also remains unsupported: it is not clear whether attentive color experiences represent narrower value ranges along a subset of the HSB space. Hence, the modest strategy also rests on shaky grounds.

To recap, the modest and the ambitious strategies are ways of defending (3), the claim that attentive color experiences carve out a smaller region of the three-dimensional HSB space. Since none of these two strategies is empirically well-supported, claim (3) remains unsupported as well: for all we know, attentive color experiences might not carve out a smaller region of the HSB space.

Now, recall that claim (3) was proposed as a plausible specification of (2), the claim that attentive color experiences have narrower range contents than inattentive color experiences. If carving out a smaller region of the HSB space is the best way (or the only way) to understand what it is to have narrower range contents along the color dimension, then claim (2) also remains unsupported: for all we know, attentive color experiences might not have narrower range contents along the color dimension.

Finally, recall that claim (2) is proposed as a plausible specification of (1), the claim that attentive color experiences have more determinate color contents. If having narrower range contents along the color dimension is the best way (or the only way) to understand what it is to have more determinate color contents, then claim (1) remains unsupported as well: for all we know, attentive color experiences might not have more determinate color contents than inattentive color experiences.

6 Enhancement, Vividness, and Phenomenal Precision

A blue car passing down the road and seen out of the corner of our eye elicits different visual experiences, depending on whether we pay attention to it or not. If we pay attention, our experience tells us that the car is blue, or at least bluish; if we do not pay attention, our experience tells us that the car is dark-colored, or even just colored. One plausible way of conceptualizing this difference is that the attentive experience represents a narrower range of color values than the inattentive experience. However, we have seen that our best evidence from current empirical research is perfectly compatible with the view that attentive and inattentive experiences do not differ in the ranges of values they represent along any of the dimensions of the color space.

We must emphasize that this upshot is contingent on the current state of our knowledge about the attentional effects on color perception. We have only been arguing that, contrary to what some representationalist philosophers suggest, current research does not really decide this question. But it is possible that new experiments will reveal that attention does narrow down the ranges of values represented along something like “the color dimension” (or, at least, along simpler dimensions like hue, saturation or brightness). We have found no principled reason for ruling out this possibility.

Note that one might still pursue different strategies for defending (2), the claim that attentive color experiences have narrower range contents than inattentive color experiences. For example, one might cash out this claim in different ways, so that having narrower range contents along the color dimension does not require “carving out” a smaller region in the HSB space, as stated by claim (3). But here we propose a different approach. Instead of finding alternative specifications of claim (2), we set this claim aside and endorse an alternative specification of claim (1).

We think that the attentional modulations in perceived saturation and brightness (even in the absence of effects for hue) undeniably show that attention enhances our color perception. As ordinarily used, the term enhancement has two meanings: bringing out and improving. We propose that the empirical findings discussed thus far show that attention does both things. Figure 6 illustrates this idea.

Fig. 6

Attentional enhancement of color perception

Increasing apparent saturation and modulating brightness so that a colored surface appears more prominent with respect to its background are ways of bringing out that surface in our perceptual experience. They are also ways of improving our experience of the surface: the experience feels of a better quality, and it puts us in a position to make more specific claims about the surface (“it is blue”). In that sense, enhanced experiences are more informative for the perceptual system and for the subject, so they can be flexibly used in more ways. Importantly, enhanced color perception need not result in color experiences with narrower range contents. Narrower range contents would be indeed one mark of enhanced color perception, but enhancement can also take other forms.

We propose that attention enhances color perception by increasing its vividness. Vividness is a term with a long history in the philosophy of mind and the philosophy of perception. It is traditionally associated with the “clearness” that differentiates the contents of perceptual experiences from the contents of, e.g., mental imagery and memories.²⁸ Here we conceive of vividness along these lines, too. We take vividness to be a qualitative property of perceptual experiences that comes in greater or lesser degrees, and that orders different types of experiences according to how well they exemplify this property. We will not attempt to offer a full account of vividness here, but we think it is illuminating to consider how this property orders types of experiences. Typically, perceptual experiences (understood as experiences that involve occurrent stimulation of sensory channels)²⁹ hold the greatest degrees of vividness. Dreams are typically less vivid than perceptual experiences (although there are some exceptions). On the other hand, some hallucinations can be more vivid than perceptual experiences. Thoughts, in turn (and supposing that they are also a kind of conscious experience), have perhaps the lesser degrees of vividness.

In contemporary vision science, vividness is often regarded as a characteristic of color, in line with brightness, purity, brilliance, and also saturation (Joblove and Greenberg 1978). According to Berns (2019, p.77), vividness is “an attribute of color used to indicate the degree of departure of the color from a neutral black color” . Finally, a recent memory experiment, (Rivera-Aparicio et al. 2021) describes an effect called “vividness extension”, which defines color vividness as increased saturation. In this way, vividness is associated with how bright and colorful surfaces appear. Therefore, it seems to be a suitable term for referring to the effects of attention on color perception.³⁰

Thus, we propose that (1), the claim that attentive color experiences are more determinate than inattentive color experiences, is better cashed out as (4):

• (4) Attentive color experiences are more vivid than inattentive color experiences.³¹

Claim (4) is neutral regarding representationalism. Attentive color experiences might well be more vivid because they have more vivid contents than inattentive color experiences, or they may be more vivid because they represent the same contents in a different way (e.g., more vividly).

In this way, our proposal is amenable to anti-representationalist approaches, such as Block’s (2015) phenomenal precision view. On Block’s view, attentive and inattentive experiences (not only of colors) differ with respect to their phenomenal, but not their representational precision. While representational precision is a matter of range contents and the properties represented in experience, phenomenal precision is a matter of the “crispness” with which these properties appear. Thus, two experiences E1 and E2 could have the same content, i.e., represent the same property or attribute the same range of values, yet this content could appear more “crisply” in E1 than in E2. All of this is compatible with claim (4), which allows for the following reading:

• (4’) Attentive color experiences represent their color contents more vividly than inattentive color experiences.

However, we think that the move from (4) to (4’) is unmotivated. For a more economical reading is (4’’):

• (4’’) Attentive color experiences have more vivid contents than inattentive color experiences.

We think that the discussed studies on the attentional effects on saturation and brightness show that there is indeed a change in the properties attributed by attentive and inattentive experiences. The former tend to attribute higher saturation values than the latter, and they also change the attributed brightness values, in either direction.³²Pace Block (2015), we see no reason for casting these effects in a non-representational light.

Think again of the blue car. If you do not pay attention to it, your experience might attribute it a lower saturation value (or range of values; see Fig. 6), and it might also attribute it a brightness value that makes it less prominent with respect to its background.³³ But if you do pay attention to it, the attributed values change. To our lights, this is sufficient for granting that attention brings about a change in the content of visual experiences.

To be fair, Block’s argument for phenomenal precision also draws on veridicality considerations that we unfortunately cannot discuss at length here. In a nutshell: Block (2015) argues that since there is no principled way of deciding whether attentive or inattentive percepts are more veridical, we have to say that either both are veridical, or none is. The latter option is inacceptable, Block argues, because then perception would become massively illusory, and this is highly implausible given the evolutionary advantageousness of having a capacity that represents our environment accurately. Thus, we must accept that neither attentive nor inattentive experiences misrepresent the world.

Now, if neither attentive nor inattentive experiences misrepresent the world, then what changes when we direct attention to a previously unattended object cannot be the content of the experience.³⁴ Thus, our account of the attentional effects as changes in content would seem to owe a story about veridicality. To close our discussion, we will mention two plausible options.

First, one might return to the hue-first view. One could argue that representing a color veridically is a matter of representing the right hue value, that is, the hue that corresponds to the reflectance properties of the relevant surface in the relevant situation. One could then argue that modulations of brightness and saturation aid in achieving a veridical representation of hue value. This could give a principled reason for considering attentive experiences to be the more veridical ones. For one thing, more saturated colors have less gray values and thus are more easily discriminable.

A second alternative is rejecting the veridicality constraints. The function of perception might well be to generate useful representations of the world, rather than accurate ones. Of course, accurate representations are useful. But usefulness might well triumph over accuracy. Thus, it would not be a problem if perception turns out to be “massively illusory”, so long as the “illusion” is useful for the goals of the organism and the species.

There is indeed an open debate about the objective existence of colors, as properties instantiated out there in the world, independently of the eye of the perceiver. If it turns out that there are no objective color facts for our representations to get right, then the question of whether attentive or inattentive color perception is more veridical loses its meaning. But the nonexistence of objective color facts does not preclude that some experiences represent more vivid color contents than others. All that is needed for this is that a relevant property is attributed, regardless of whether the property is or not instantiated in the world. Of course, it remains to be argued why experiences attributing more vivid color properties are also more useful for flexible purposes. This is a question we look forward to addressing in future works.

Appendices

Appendix 1: Inattentive color perception

Some proponents of claim (1) also defend a version of (1’)³⁵:

• (1’) Inattentive color experiences represent (only or mostly) indeterminate color properties.

This claim is appealing because color receptors are sparser in the visual periphery, and visual peripheral experiences are typically inattentive. Moreover, (1’) seemingly follows from the claim that attentive color experiences represent more determinate color properties. However, testing claim (1’) directly is notably difficult. In this appendix we discuss one recent study that might be brought to bear on this claim, and highlight some possible limitations.

Cohen et al. (2020) tested awareness of colors in the visual periphery during active real-world vision. Participants wore a virtual reality head-mounted system, which also tracked their eye movements. They were presented with a 360° video of an environment with sound and were told to explore it freely. After a few seconds, colors in their visual periphery started to desaturate until they all became shades of grey, and only the part of the scene where gaze was directed remained colored. After the trial, participants were asked whether they noticed something different about what they saw. Very often, they were not aware of any change.

This experiment suggests that in everyday perceptual experiences, the color properties of unattended objects in our visual periphery make no difference to our visual phenomenology – for, apparently, these properties can suffer big changes without us noticing. It might even be that detailed color properties are not represented at all. So, does this show that inattentive color perception is indeterminate?

Prima facie, the answer is No. The study shows that an absence of attention in the visual periphery impairs awareness (noticing) of big changes in color appearance. But this by itself need not tell us about the quality of color information represented in the unattended periphery.

Admittedly, Cohen and colleagues emphasize that the richness of peripheral perception is not fully fixed by the hardwired physical constraints of the visual system. In a second condition of their experiment, they told participants about the desaturation and instructed them to pay attention to the visual periphery, to report when the desaturation started. In this case, participants did notice the change. This might suggest that participants’ peripheral perception went from representing indeterminate properties like colored, in the unattended condition, to representing more determinate properties like gray, in the attended condition. In the first case, the change in saturation is not noticed because the property colored is compatible with the display having a wide variety of hues, with a wide range of saturation values. So, if the percept does not represent specific shades like sky-blue to begin with, it is not surprising that no change is noticed when the sky-blue portions of the display turn to gray. Contrastingly, in the attended condition, participants might notice the change because their initial percept attributes sky-blue to some display portions, and this is incompatible with attributing gray to the same display portions, as in their later percept.

Although we find these thoughts suggestive, we recognize that they still do not rule out an alternative interpretation of Cohen and colleagues’ (2020) findings, namely: that peripheral inattentive experiences do represent determinate color properties, but we fail to notice that they do.

It might be illuminating to compare this issue with a well-known discussion about auditory attention models, going way back into the historical debate about early versus late selection. Cohen and colleagues’ (2020) findings are reminiscent of the famous cocktail-party effect (Cherry 1953), insofar as they both suggest that attention selects some information for conscious awareness (and reportability), while filtering out some other information. The pressing question that remains, despite the present-day dismissal of the early versus late controversy, is: how is the unattended information impinging our sensory channels represented in conscious experience – if it is consciously represented at all? Specifically, we may ask: Are unattended color properties completely filtered out and not processed further? Or are they instead processed, but only to some extent?³⁶

Considering these possibilities, we can say the following: 1) unattended peripheral information is somehow filtered out, but 2) we do not know whether the visual system registers the color information to a point that legitimates calling this registering “indeterminate color perception”.

Appendix 2: Veridicality and color realism

Objective colors are at the center of a live debate, spanning philosophy and color science. Color realists defend their existence: they conceive of colors as properties that are instantiated in the world and not just in the eye of the perceiver. Anti-realists deny this: in their view, colors only have reality in the perceiver’s mind and in our linguistic practices. An aspect of this debate that is directly relevant for the discussion of the attentional effects concerns the relation between the precision of color perception and its veridicality.

A percept is veridical if and only if the perceived object has the property attributed by the percept. For example, in the broken pencil illusion, your percept represents a pencil as broken or bent. Since the pencil does not really have this property, your percept is not veridical. Accordingly, for color percepts to be veridical, surfaces represented as colored must have the color properties attributed to them. Hence, if color anti-realists are right, then color perception is never veridical: we never get color facts right, because there are no color facts to get right or wrong to begin with.

Suppose that color anti-realists are right, and there are no color facts to get right or wrong. In this scenario, is the discussion about the determinacy of color perception still meaningful?

As noted in the main text, philosophers often use “determinacy” and “precision” as interchangeable terms. “Precision”, in turn, is often thought to be about accuracy: a measurement is more precise the more it approaches the actual value of the magnitude measured. Accordingly, comparing the precision of two color percepts requires comparing how closely each matches the real color of the perceived surface. It is not clear how this could be done, if there are no color facts. Thus, it might seem that arguing for differences in determinacy between attentive and inattentive perception requires a commitment to some form of color realism. Objective colors would be needed to provide a normative standard for the precision of perception.

However, on a closer look the question about precision might well turn out to be orthogonal to the question about color realism. True, describing the passing car as cobalt blue is at once a more precise and a more accurate description than describing it as blue. But the facts that make this description precise are dissociable from the facts that make it accurate. The description is accurate because it captures a property of the car, but it is precise because of how closely it captures the property. Figure 7  illustrates this conceptual distinction.³⁷ Thus, to compare the precision of two percepts, one need not look at what wordly properties they match; rather, one must look at, say, how many distinct properties they attribute – i.e., the range of attributed values. It follows that, even if there are no color facts, we can still meaningfully talk about the precision of our color percepts.

Fig. 7

Precision and accuracy as distinct concepts. We can provide both precise, but inaccurate description of a color, and imprecise, but accurate description of the color