From Towards a Science of Consciousness 3         Section 4: Vision and Consciousness -- Introduction       CogNet Proceedings

Selective Peripheral Fading: How Attention Leads to Loss of Visual Consciousness

Lianggang Lou

Consciousness and attention are closely related concepts. In this chapter, I will follow the common practice to use the term consciousness to refer to what we can report having been aware of, and the term attention to refer to the selection and maintenance of a portion of information for privileged access, whatever its consequences on consciousness. I will provide a case that seems to illuminate, in an unexpected way, the neural substrate for the distinction and relation between attention and consciousness in vision.

It is well known that if one attends to a stationary point in peripheral vision while maintaining fixation in central vision, the peripheral point will fade from awareness within a few seconds. This effect, known as Troxler fading (Troxler, 1804), has been considered to reflect local adaptations in lower visual pathways (Clarke and Belcher 1962, Millodisk 1967). Local adaptation here means that the receptors or neurons responsible for detecting stimulus-background edges by luminance or chromatic contrast cease to respond to steady images. Because in Troxler fading the stimulus is perceptually replaced by whatever in the surrounding area rather than a black patch, it has been likened to the phenomena of blindspot filling-in (Ramachandran 1992). Thus, the conscious percept corresponding to the missing sensory input has been thought to be somehow "fillled in" at higher visual centers by interpolating from the surrounding visual areas. However, it is not certain whether local adaptation is an accurate account of what leads to the missing sensory input in the case of Troxler fading. The following observation suggests that, unlike the blind spots, the voluntary attention required to observe the fading may actually be responsible, at least partly, for the fading.

[Insert Figure 17.1; figures not yet available]

A circular array of disks, three green and three orange in alternate positions, was presented against a gray background (figure 17.1). Observers (this author and several postgraduate students and colleagues) fixated on the display center and voluntarily attend to three disks of one color (e.g., the green disks) and tried to ignore disks of the other color (e.g., the orange disks). In a few seconds, one or more of the disks faded from awareness. The disks that faded first tended to be those selected for attention. The phenomenal experience was that the background color quite abruptly filled the disk areas. The disappearance of the disks was sometimes quite fleeting and sometimes seemed to last a few seconds.

A similar informal observation was reported by Barbington-Smith (1961), but was never followed up in the thirty-seven years since its publication. A potential problem regarding these observations (henceforth, selective fading of attended stimuli or selective peripheral fading) is how certain one can be that the fading selectively occurs to the attended disks. Suppose lower-level sensory adaptation occurs equally to both the attended and unattended disks; it may be impossible or difficult to consciously perceive the fading of the unattended disks. At least two kinds of evidence may be adduced in support of this possibility. First, recent studies showed that people can fail to detect very salient changes in a visual scene if the change is not relevant to their current goal or intention (Grime 1996). Similarly, subjects could be so occupied with the stimuli selected for attention that they failed to access the changes occurring to the unattended stimuli. Therefore, while they truly perceived the fading of the attended disks, it is unclear whether they perceived, or simply believed in, the constancy of the unattended disks.1 Second, attention could induce a bias of "prior entry" (Titchener 1908), which leads the sensory adaptation of the attended stimuli to enter the consciousness sooner than that of the unattended stimuli.

To address these concerns, a formal experiment was conducted, in which various parameters of the fading, notably the duration of faded awareness and the proportion of attended disks involved, were measured to see how plausible the alternative explanation of the selective fading can be. On some trials (control trials), a pair of disks, one of which selected for attention and the other ignored, was physically extinguished for a while following a period long enough for attention to be engaged to the selected disks. If attention leads one to be selectively aware of the disappearance of the attended disks, the bias would also be expected in perceiving the fading on other trials, where no disks were physically extinguished from the display. In addition, the extent of the fading was also measured at various eccentricities in order to confirm that, like Troxler fading, it occurs primarily in peripheral vision.

Methods


Eighteen subjects participated in the experiment. They were University of Hong Kong undergraduate or postgraduate students. None of them had previous experience in experiments of visual perception and they were na´ve as to the purpose of the study. In a pretest with the above display (figure 17.1), two subjects failed to observe any fading after repeated trying and were subsequently dismissed. Each of the remaining 16 subjects was tested in 6 blocks of trials. The first two blocks consisted of seven trials each, including five test trials and two control trials. The same display (figure 17.1) was presented, rotated randomly from trial to trial. On the control trials, one attended disk and one unattended disk were simultaneously extinguished from the display three to four seconds following the display onset and recovered one second later. Subjects sat 35 cm from the screen and were instructed to fixate the display center, attend to the disks of one color, and ignore the disks of the other color. To facilitate this task, they were told to perceive the to-be-attended disks as forming an equilateral triangle and the to-be-ignored disks as part of the background. Their chin and forehead were supported to minimize fixation shift. Half of them were instructed to attend to the green disks in the first block and the orange disks in the second block, and the other half of them were told to do the reverse. The same displays, but of smaller sizes, hence smaller stimulus eccentricities, were used in blocks three to six, in which attention was directed to the green disks throughout.

Both the perceptual fading and the physical disappearance of the peripheral stimuli were indicated by manual response on a two-key response box. Pressing the left key indicated the onset of stimulus fading or disappearance, and releasing it indicated the recovery of the faded stimulus in awareness. The right key was to be pressed if more disks faded subsequently. The whole display ended one second after the release of the left key or the press of the right key. The maximum viewing time was 40 seconds. After each trial, subjects were prompted to recall the number of faded disks of each color when they pressed the left key.

Results


Table 17.1 presents four different aspects of the fading from each of the 16 subjects: the percentage of attended disks in those that faded first, the latency, the duration, and the number of disks faded simultaneously. Although there were quite a lot of individual differences, it is clear that fading occurred mostly to the attended disks. On average, in about 10 (mean = 11.35) seconds of viewing time, some of the six disks started to fade from awareness. Of the disks that faded, most (mean = 81.3%) were those selected for attention. This was true regardless of attention being directed to the green disks (mean = 94.6%) or to the orange disks (mean = 68.0%). Except for one subject (#13), disks of a particular color faded more frequently when the color was selected for attention than when it was ignored, t(15) = 7.67, p < .001. The faded disks remained out of awareness for up to a few (mean = 1.55, 95% confidence Interval = 0.49) seconds during which the other disks were clearly visible. On control trials, in contrast, all subjects on all trials perceived the simultaneous physical disappearance of the two disks, one of them selected for attention. The mean duration of the perceived disappearance was 1. 12 (95% confidence interval = 0.34) seconds, very close to the actual duration of disk disappearance (one second).

With decreasing eccentricity, the fading occurred less frequently. At the smallest stimulus eccentricity manipulated (0.8), the fading occurred on less than 30 percent of the trials, and with longer latency, shorter duration, and fewer number of disks involved. These characteristics suggest the fading to be primarily a phenomenon of peripheral vision.

Discussion


Some obviously implausible explanations can be ruled out in accounting for the selective fading of the attended stimuli. First, while small fixation drifts or short saccades may occur in the present experiment, they are unlikely to account for the effect. The eye movements would either follow a random trajectory or, given the close functional association between attention and eye movements, be biased toward the direction of the attended disks. In both cases, the eye movements would not bring the images of the attended disks to the more peripheral retinal regions, thereby set in their earlier fading.

Second, the disks that faded more frequently were unlikely to have been the less attended. To be sure, attention may be spread or inadvertently directed to the to-be-ignored disks. Indeed, the less than perfect selectivity of the fading (94.6% when the green disks were to be attended and 68.0% when the orange disks were to be attended) may be accounted for, at least partially, by the less-than-perfect selectivity of attention. However, it is extremely implausible that attention was directed primarily to disks other than those specified in the instruction. To suggest the latter would imply either that the brain systematically misinterprets the instruction for the allocation of attention, or that subjects were playing pranks on the experimenter. Both scenarios were implausible.

Therefore, voluntary attention must have facilitated the fading, or inhibited the awareness of the attended stimuli. The problem is how to explain it. Two possibilities were alluded to earlier. First, sensory adaptation occurs evenly at all disk locations of the same eccentricity, but only that at the attended locations is translated into the fading (the hypothesis of attentional bias). This hypothesis does not seem tenable. In particular, it does not seem likely that any of the unattended disks would remain visible for 1.5 seconds or longer while its sensory basis has been removed because of adaptation. In addition, because attention was not found to bias the perception of the physical stimulus disappearance, it was unlikely to significantly bias the perception of the fading caused by low-level adaptation either. Therefore, we are left with the second possibility, that attention actually triggers or precipitates the fading of the attended disks (the hypothesis of inhibitory effect of attention).

Visual selective attention is known to involve many different cortical areas, notably the V4, IT, posterior parietal and prefrontal areas (Desimone and Duncan 1995). As some of these cortical areas are likely where visual consciousness arises (Crick and Koch 1995), they could underlie both selective attention and its inhibitory consequence on visual consciousness. However, it is equally important to consider the fading as a special case of Troxler fading because of the following defining characteristics: the static stimuli, the low stimulus-background contrast, the specificity to peripheral stimuli and the time (about 10 seconds) that it took for the fading to occur. Thus, what has been known of the neural basis of Troxler fading should also hold true for the selective fading of the attended stimuli. Clarke and Belcher (1962) showed that the saccadic displacement needed for a faded image to recover is too large to suggest a ganglion or preganglion origin of the fading. On the other hand, they failed to find binocular interactions of the luminance threshold for the fading, suggesting a precortical origin of the fading. They concluded, therefore, that Troxler fading most likely arises from the adaptation at LGN, the relay station between the retina and the cortical visual areas. Putting these threads together, the selective fading of the attended disks is likely owing to adaptation at the LGN being triggered, or accelerated, by the activation from cortical areas associated with selective attention.

The above proposal is consistent with the models positing that consciousness relies on the reverberating neural activities in the cortico-thalamic loop (Newmann 1995, LaBerge 1995). To explain the selective fading, it can be assumed that the fading results from a breakdown of the reverberation, and that the breakdown occurs more easily at locations under additional influence from the cortical areas associated with voluntary selective attention.

Implications


The neural mechanism proposed for the selective fading of the attended stimuli allows an evaluation of some common assumptions about the functional relations between visual attention and visual consciousness. The function of attention is commonly assumed to be to select a small set of information from what is potentially available. Indeed, that has been the working definition of attention in this chapter. There has been a lot of debate as to where in the information-processing streams the selection occurs. Early selection theories posit that selection occurs relative early, and leads elementary features of the attended stimuli to undergo further processing and those of the unattended stimuli to be "filtered" (e.g., Broadbent 1958). Late selection proponents (e.g., Deutch and Deutch 1963) argue that selection occurs late in the information processing stream, after much of the semantic processing has been completed. To be compatible with these theories of attention, consciousness has to be conceived as emerging from the further processing after the locus of selection. It is almost certain that neural activities at LGN would not by themselves give rise to visual consciousness; not even those at V1, according to Crick and Koch (1995). Rather, the activities at LGN would provide more or less preprocessed sensory information to be further processed and integrated at the cortical areas, which gives rise to visual consciousness. If, however, attention intensifies and then inhibits the activities at LGN, as has been suggested, then the above explanatory framework has to be modified. It would not be appropriate to assume that visual consciousness is merely due to the additional processing after the locus of selection. For a stimulus to be visually conscious, its retinotopic sensory input (in this case, at the LGN) may need to be amplified by the covert attention directed toward it. Clearly, this proposal is incompatible with the late selection theories. On the other hand, it may be consistent with the early selection theories if the term selection is amended to mean "pick and boost."

As the selective fading of attended stimuli has been found only in peripheral vision and only after seconds of voluntary attention toward the stimuli, it is certainly a special case. Nevertheless, like many visual illusions, it may have revealed a general brain mechanism of visual attention and visual consciousness through its breakdown under special circumstances. Thus, voluntary visual attention may entail a centrifugal influence from extra-striate cortical areas, including the prefrontal areas, to lower visual centers, for example, V1, V2, and LGN. Indeed, evidence from electrophysiological studies of selective attention has suggested sensory modulation by selective attention, but perhaps only at the early extrastriate level (Clark and Hillyard 1996). With more sensitive experimental designs and neuroimaging techniques, direct evidence supporting the notion of selective modulation of cortico-LGN reverberation may be obtainable.

Note

1. This is similar to Dennett's idea about what is involved in the filling-in phenomena that occur at the blindspots (Dennett 1992). Instead of perceptual filling-in (Ramachandran 1992), Dennett believed that the brain simply ignores the blindspot and when forced to report what is in the blindspot relies on conceptual interpolation from what is perceived from the area surrounding the blindspot.

Table 17.1

The means of four dependent measures of fading from the first and second blocks, catch trials excluded. G: Green disks were selected for attention; O: Orange disks were selected for attention.

Subject

%Attended disks

Onset latency (sec.)

Duration (sec.)

Number of disks

 

G

O

G

O

G

O

G

O

1

83.33

43.33

6.12

5.23

2.13

0.72

2.17

2.40

2

100.00

100.00

10.90

8.47

1.30

1.26

1.20

1.40

3

100.00

50.00

9.20

8.71

1.20

1.14

1.60

1.20

4

60.00

90.00

10.00

8.71

1.45

0.89

1.60

1.60

5

100.00

100.00

5.03

9.26

1.03

0.47

1.40

1.20

6

100.00

53.33

8.27

9.90

1.44

1.11

1.40

1.40

7

90.00

50.00

8.38

12.07

1.18

1.47

1.40

1.20

8

100.00

100.00

27.91

17.19

4.31

2.80

1.80

1.60

9

100.00

40.00

14.87

18.58

5.24

2.88

1.80

1.80

10

100.00

50.00

8.99

6.60

1.09

1.04

1.00

1.20

11

80.00

46.67

9.10

6.73

1.90

0.87

2.00

1.80

12

100.00

100.00

20.48

25.71

1.91

1.18

1.00

1.00

13

100.00

0.00

6.90

4.69

2.05

1.09

1.00

1.00

14

100.00

90.00

12.30

12.24

0.98

0.80

1.20

1.40

15

100.00

100.00

12.45

9.45

1.47

1.18

2.20

1.60

16

100.00

75.00

12.18

16.58

1.35

0.81

1.20

1.00

Mean

94.58

68.02

11.44

11.26

1.88

1.23

1.50

1.43

References


Barbington-Smith, B. 1961. An unexpected effect of attention in peripheral vision. In Nature 189:776.

Broadbent, D. E. 1958. Perception and Communication. London: Pergamon Press.

Clark, V. P., and S. A. Hillyard. 1996. Spatial selective attention affects early extrastriate but not striate components of the visual evoked potential. In Journal of Cognitive Neuroscience 8(5):387-402

Clarke, F. J. J. 1962. On the localization of Troxler's effect in the visual pathway. In Vision Research 2:53-68.

Crick, F. and Koch, C. 1995. Are we aware of neural activity in primary visual cortex? In Nature 375:121-123.

Dennett, D. C. 1991. Consciousness explained. Boston: Little, Brown.

Desimone, R., and J. Duncan. 1995. Neural mechanisms of selective visual attention. In Annual Review of Neuroscience 18:193-222.

Deutch J. A., and D. Deutch. 1963. Attention: Some theoretical considerations. In Psychological Review 70:80-90.

Grimes, J. 1996. On the failure to detect changes in scenes across saccades. In Perception. Edited by A. Kathleen. Oxford: Oxford University Press, 89-110.

LaBerge, D. 1995. Attentional processing: The brain's art of mindfulness. Cambridge, Mass.: Harvard University Press.

Millodisk, M. 1967. Extra foveal variations of the phenomenon of Troxler. In Psychologie Francaise 12(3):190-196.

Newman, J. 1995. Review: Thalamic contributions to attention and consciousness. In Consciousness and Cognition 4(2):172-193.

Ramachandran, V. 1992. Blind spots. In Scientific American 266(May):44-49.

Titchener, E. B. 1908. Lecture on the Elememtary Psychology of Feeling and Attention. New York: Macmillan.

Troxler D, 1804. Uber das Verschwindern gegebener Gegenstande innerhalb unsers Gesichtskrcises. In Ophthalmologisches Bibliothek, Vol. II. Edited by K. Himley and J. A. Schmidt. Jena: Fromann, 51-53.