We found that the time course of attention in the amblyopic fovea differed significantly from that obtained when the same observers viewed the stimuli with their preferred eyes. Specifically, when viewing with their amblyopic eyes, observers had a shallower ‘attentional blink,’ with less Lag 1 sparing and greater accuracy of T2 report 200 ms after T1, as a result of making fewer neighbor errors and more errors consisting of letters that were not presented in the sequence. The question is whether these differences arise from changes in the quality or function of attention, when viewing through amblyopic eyes, or changes in the quality of the image, perhaps as a result of masking. Intriguingly, responses to T1 presented to the amblyopic eye were significantly less accurate than responses from the preferred eye, and this difference resulted entirely from misreporting preceding letters in place of T1. This latter finding implicates changes in attentional processing and specifically temporal binding in amblyopia. Nevertheless, in the discussion that follows, we first consider the possibility that differences in T2 report resulted from impaired image quality in amblyopic eyes.
Vision in the amblyopic fovea is degraded, as compared with the normal fovea, both in terms of single-letter acuity and vulnerability to crowding by nearby stimuli. However, even normal observers experience considerable crowding (and reduced acuity) when stimuli are presented outside the fovea, and therefore the normal periphery has frequently been considered a model for amblyopic vision (e.g., Levi & Klein,
1985). What is the attentional blink like in normal peripheral vision? Kristjánssonn and Nakayama (
2002) studied the effects of spatial separation on the attentional blink, and therefore had to place their stimuli outside the fovea. They found no Lag 1 sparing, for stimuli presented at 6 deg eccentricity and with a 140 ms SOA. It is possible that this characteristic of amblyopic performance is the result of degraded vision, similar in the amblyopic fovea to the normal periphery. However, this speculation must be treated with caution, because it is not at all clear that, even in the normal fovea, Lag 1 sparing would be as pronounced at such a long SOA (Popple & Levi,
2007) and see solid circles in
Figure A1,
1. Additionally, although targets in Kristjánsson and Nakayama's (
2002) study were marked by a luminance cue, they were primarily distinguished from non-targets by virtue of being digits among letters, and results in this task may differ from the task in the present study. Further experiments need to be carried out to directly compare the amblyopic fovea and the normal periphery in identical attentional blink tasks.
Bowman and Wyble (
2007) recently proposed a new and comprehensive theory of encoding RSVP targets in visual working memory. Their theory brings together several previous models and ideas, particularly those relating to the attentional blink. The theory has two stages:
-
a perceptual stage, in which many ‘types’ or object representations become activated in a parallel fashion, depending on their presence in the environment, and the task-dependent operations of a ‘salience filter;’
-
an encoding stage, when a series of time-labeled ‘tokens’ are bound to the activation of the salience-filtered ‘types.’
This ‘tokenization’ requires the action of a ‘temporary attentional enhancement’ (TAE) mechanism that selectively enhances the activation of salient ‘types,’ so that they can be bound to ‘tokens’ in the ‘binding pool.’ According to this theory, in the attentional blink the TAE is still suppressed by the processing of the first target when the second one arrives.
We have argued elsewhere (Popple & Levi,
2007) that Bowman and Wyble's (
2007) theory is better than other popular theories of the attentional blink (e.g., Chun & Potter,
1995; Di Lollo, Kawahara, Shahab Ghorashi, & Enns,
2005; Nieuwenhuis, Gilzenrat, Holmes, & Cohen,
2005; Shapiro, Arnell, & Raymond,
1997) at explaining the pattern of intrusion errors in this task. How might the patterns of errors found in amblyopic vision occur, according to their model? Perhaps there is a delay in the processing of the letters (‘types’) presented to the amblyopic eye, which results in mis-‘tokenization’ such that a previously presented letter becomes erroneously bound to the T1 ‘token’ in the ‘binding pool.’ This might explain the prevalence of pre-T1 errors in amblyopia but not the changes in T2 performance, which result from suppression of the TAE mechanism. One possibility is that the timing of all the letters is slightly erratic in amblyopia. Bowman and Wyble (
2007) explain Lag 1 sparing as the unordered capture of both T1 and T2 in a single sweep of the TAE, and perhaps if their ‘types’ arrive erratically over time this may result in a blurring between Lag 1 and Lag 2 performance in amblyopia. However, the pattern of T2 errors in amblyopia suggests that something else may be going on.
Pelli, Palomares, and Majaj (
2004) argued that the phenomenon of crowding involves excessive feature integration between spatially nearby stimuli. They distinguished this spatial crowding from ‘temporal crowding’ in which, they proposed, the illusory conjunction of sometimes spatially distant features arises from a failure of attentive binding. Our results show that in normal vision, color may erroneously bind with letter identity when stimuli are presented in close temporal proximity, as is the case in the attentional blink (Popple & Levi,
2007). However, we did not find evidence for the exchange of features between different letters in the normal fovea, as occurs in spatial crowding in normal peripheral vision and the amblyopic fovea (Levi, Hariharan, & Klein,
2002; Pelli et al.,
2004). In the present study, we did find suggestive evidence for this in the form of an increased prevalence of non-stimulus letters in the responses to T2 from amblyopic eyes. This cannot be explained by a bias toward a particular set of letters whenever the target is missed, because that would predict more non-stimulus responses where there were more errors, and instead more non-stimulus responses occurred with the amblyopic eye responding to T2 (although performance was the same as the preferred eye) and not when responding to T1 (although performance was worse with the amblyopic eye). These non-stimulus responses may have been formed by the integration between features of successive letters, resulting in the perception of letters that were not present. Perhaps certain stimulus representations—in the normal periphery and the amblyopic fovea—are simply more vulnerable to crowding and over-integration, whether in space or in time. In space, this vulnerability is inevitable, but in time, it may be contingent on the suppression of the TAE by the tokenization of a prior target. As with the degraded image hypothesis, this idea needs to be tested by investigating the pattern of errors produced by the attentional blink in both the fovea and the periphery, with identical targets, distractors, and time course.
Bonneh et al. (
2007) studied temporal crowding in amblyopia by looking at the size threshold for identifying a single RSVP target digit and found that larger sizes were required for target identification in amblyopia (specifically in strabismic amblyopes). We too found a reduction in amblyopic performance on T1 (although not T2). In our study, this effect was due to misreporting preceding targets in place of T1. Bonneh et al. did not report the distribution of errors, as compared with distractor identity. Perhaps the target digit identity was unavailable for binding with the size ‘token’ in Bonneh's experiment, unless the target was made larger. There is evidence that larger, or more salient, stimuli are processed more rapidly and can be identified with smaller exposure times (Rimmer, Iragui, Klauber, & Katz,
1989). However, the different sizes of the letters used in amblyopic and non-amblyopic eyes in our study probably had little effect on performance, at least by inference from their effects on a normal observer (see
1).
Asper et al. (
2003) found only small effects of amblyopia on T2 detection and did not report on T1 performance. However, they may have confounded their results with differences in practice between the amblyopic and preferred eye sessions, since they also showed better performance in the non-dominant eyes of normal observers as compared with their dominant eyes, tested in an earlier session (Asper, 2007; personal communication). We found no such difference in a control test of two experienced observers (
1). We also used a larger sample than they did (12 in place of 10 observers) and a more sensitive task (identifying T2 in place of detecting an ‘X’), thereby improving our chances of detecting a small effect. The small effect we found on the temporal properties of the attentional blink in amblyopic eyes, compared with preferred eyes, suggests that attention itself may be disrupted in amblyopia.
What is attention? Some argue that attention is the preparation for making a saccadic eye movement to an interesting target, in order to better inspect it using high-resolution foveal vision (Findlay & Gilchrist,
2003). If so, it would make sense for the process of attention to be disrupted in amblyopia, because the amblyopic fovea is not necessarily an improvement over peripheral vision in the amblyopic eye, and it would be counterproductive for the individual with amblyopia to make saccadic eye movements to stimuli detected, and perhaps distorted, in this eye. Whatever the circuits in the brain that drive bottom-up pre-attentive vision and the capture of attention by salient stimuli, they may be muted in amblyopia. Evidence for this comes from the tendency in amblyopia to undercount pop-out stimuli (Sharma, Levi, & Klein,
2000) and from the poor localization of such stimuli in amblyopic eyes (Popple & Levi,
2005). Additional evidence comes from the finding that observers with amblyopia show small but significant abnormalities in attentionally tracking multiple suprathreshold objects (Ho et al.,
2006; Tripathy & Levi,
2008). Perhaps the altered time course of attention in amblyopia reflects genuine changes in the process of attention when viewing through amblyopic eyes. Alternatively, the temporal dynamics of masking might be different in amblyopic and preferred eyes. Another possibility is that accurate temporal binding relies on intracortical connectivity, which might be reduced in amblyopia. There is no evidence that selective attention to each eye is possible in the course of normal visual development, but it can be found in certain cases of strabismus, where there is alternating fixation.