Elsevier

Cognition

Volume 102, Issue 3, March 2007, Pages 321-340
Cognition

The role of perceptual load in inattentional blindness,☆☆

https://doi.org/10.1016/j.cognition.2006.01.002Get rights and content

Abstract

Perceptual load theory offers a resolution to the long-standing early vs. late selection debate over whether task-irrelevant stimuli are perceived, suggesting that irrelevant perception depends upon the perceptual load of task-relevant processing. However, previous evidence for this theory has relied on RTs and neuroimaging. Here we tested the effects of load on conscious perception using the “inattentional blindness” paradigm. As predicted by load theory, awareness of a task-irrelevant stimulus was significantly reduced by higher perceptual load (with increased numbers of search items, or a harder discrimination vs. detection task). These results demonstrate that conscious perception of task-irrelevant stimuli critically depends upon the level of task-relevant perceptual load rather than intentions or expectations, thus enhancing the resolution to the early vs. late selection debate offered by the perceptual load theory.

Introduction

Would focusing attention on a current task prevent the intrusion of task-irrelevant stimuli into awareness? This fundamental issue has intrigued psychologists for many years and has led to an enduring controversy between early selection views suggesting that attention can prevent irrelevant stimuli from reaching awareness (e.g., Broadbent, 1958, Neisser and Becklen, 1975) and late selection views proposing that attention does not affect perceptual awareness, but rather affects later processes such as response selection and memory (e.g., Deutsch and Deutsch, 1963, Tipper, 1985).

A possible resolution to this debate has been offered within a hybrid perceptual load model (Lavie, 1995, Lavie, 2000). According to this model, focusing attention on a current task can prevent the perception of task-irrelevant stimuli (i.e., early selection) when the task-relevant processing involves a high level of perceptual load which consumes all available capacity. By contrast, when the processing of task-relevant stimuli involves only low perceptual load, any spare capacity spills over involuntarily to the perception of irrelevant stimuli (i.e., late selection).

Although the perceptual load model would clearly predict that irrelevant stimuli should not reach awareness under conditions of high perceptual load, this hypothesis has not been directly tested as yet. Previous evidence for the load theory has so far relied upon either indirect behavioral measures of the effects of distractors on target reaction times (RTs) or functional imaging experiments assessing neural activity in sensory cortices related to distractor perception (see Lavie, 2005, for review). However, with one exception (Rees, Frith, & Lavie, 1997), previous experiments have not examined whether load influences the entry of distractors into visual awareness.

For example, in a series of behavioral experiments, Lavie, 1995, Lavie and Fox, 2000 assessed the effects of perceptual load on distractor processing within the response competition paradigm (Eriksen & Eriksen, 1974). Participants made speeded choice responses deciding whether a central target letter was one of two pre-specified letters while attempting to ignore a peripheral distractor letter. Distractor letters could be either congruent (e.g., distractor “X” for target “X”) or incongruent (distractor “X” for target “Z”) with the target. In some experiments, the perceptual load of this task was manipulated by increasing the number of letters among which the target appeared: In addition to the distractor, the target letter appeared either alone (low load) or among five other non-target letters (high load). In other experiments, perceptual load was manipulated by increasing the demands placed on attention for identical displays. For example, participants either detected presence/absence (low load) or (for the same stimuli) made a difficult size and position discrimination (high load). The results showed that distractor congruency with the target influenced target RTs in tasks of low perceptual load, but that such distractor effects were eliminated under high perceptual load. Lavie and Fox (2000) similarly found that distractors produced negative priming effects (i.e., they increased RTs when presented as targets in subsequent trials) in tasks of low perceptual load, but not in tasks of high perceptual load.

These studies clearly demonstrate that interference effects from distractors are reduced in tasks of high (compared to low) perceptual load. However, they cannot provide information about the effects of perceptual load on conscious perception of such distractors. Although the perceptual load model interprets the elimination of distractor effects on RTs by higher loads as reflecting an overall reduction in distractor perception (i.e., implying no conscious perception of distractors with high perceptual load), these effects are equally consistent with alternative interpretations proposing no role for perceptual load in determining conscious perception. For example, one could claim that perceptual load influences unconscious perceptual processes but has no effects on conscious perception. On such an interpretation, task-irrelevant distractors never enter awareness under either condition of load: distractor interference effects seen in conditions of low load merely reflect unconscious recognition of distractor-/target–response associations. Alternatively, one could also claim that distractors always enter awareness, regardless of load. With this interpretation, the reduction of distractor effects on RTs by higher perceptual loads reflects either an influence of load on post-perceptual processes such as response selection, or a dissipation of distractor effects during longer RTs for high load (but see Lavie and De Fockert, 2003, Lavie and Fox, 2000; for counter-evidence).

Functional imaging tests of the perceptual load theory demonstrate that activity in the visual cortex related to distractor presence is eliminated under high perceptual load. For example, Rees et al. (1997) showed that visual cortex activity related to moving (vs. stationary) distractors (e.g., in MT), was found in conditions of low task-relevant load (requiring detection of letter case in fixated target words) but was eliminated by high task-relevant load (involving a more complex word discrimination). Other studies found that visual cortex activity related to various other task-irrelevant stimuli (e.g., checkerboards, meaningful pictures) was significantly reduced, indeed typically eliminated, with increased perceptual load in a relevant task (Pessoa et al., 2002, Pinsk et al., 2003, Schwartz et al., 2005, Yi et al., 2004).

The convergence of results from behavioral and imaging experiments makes an appealing case for the hypothesized role of perceptual load in distractor processing. It also clearly demonstrates that the perceptual load modulation of distractor processing is evident at multiple different levels and does not merely reflect effects on RTs (e.g., as stipulated in dissipation accounts). However, assessment of neural activity cannot provide information about subjective conscious experience and, as discussed above, neither can indirect behavioral measures (RTs).

As mentioned above, one imaging study (Rees et al., 1997) included an assessment of the effects of perceptual load on awareness. Rees et al. (1997) measured the subjective duration of a motion after effect caused by the irrelevant distractor-motion in their imaging task. They found that the duration of a motion after effect was significantly reduced when participants performed the high load task compared with the low load task. As this test involved participants directly reporting their subjective motion experience, the results are encouraging for our suggestion that perceptual load determines awareness. However, without further corroboration and extension to other measures of awareness, Rees et al.’s (1997) results remain confined to the case of motion after effects.

The purpose of the present study was therefore to directly examine the role of perceptual load in visual awareness, using a more general measure of awareness within the “inattentional blindness” paradigm (Mack & Rock, 1998). In a typical inattentional blindness procedure, an unexpected, task-irrelevant object appears either in the final “critical trial” of a small set of experimental trials (e.g., a square presented in the periphery while participants perform a line-length judgment on a cross-target at fixation, Mack & Rock, 1998; a triangle moving across the screen while participants attend to one of two subsets of moving shapes, Most et al., 2001), or for some duration during a continuous task (e.g., a woman with an umbrella passing by people playing a ball game while participants count the number of ball-passes made, Neisser, 1979, Simons and Chabris, 1999). Following the usual task–response on the critical trial, participants are asked to report whether they were aware of any extra task-irrelevant stimulus, or anything unusual on the screen. Findings show that participants often fail to notice the unexpected task-irrelevant stimulus. By contrast, the same stimulus is often detected on a following control trial in which participants do not perform the task but instead pay attention to whether there is any extra stimulus on the screen. This contrast has been taken to reflect blindness due to inattention hence the term “inattentional blindness”.

Fully-attended control trials differ from experimental trials however, in several aspects that entail processes other than attention. First, the critical stimulus is expected on the control trials, and participants are likely to look for it intentionally (either due to explicit instruction to look for something extra in some studies, or due to the preceding awareness probe raising their expectations of something unusual). Thus, the comparison of control trials with experimental trials confounds effects of attention with effects of expectation and intention (see Braun, 2001). Second, awareness reports are made after a task–response and a surprise awareness question in critical trials, but can be made immediately following display presentation in control trials. Reduced rates of awareness in critical (vs. control) trials may therefore reflect greater rates of forgetting during the longer delay from display presentation until the awareness question in the critical (vs. control) trials. In other words, inattentional blindness may be conceptualized as “inattentional amnesia” (e.g., Wolfe, 1999).

Thus, the contrast of awareness between critical trials and control trials in previous studies cannot lead to clear conclusions about the pure role of inattention in the phenomenon of inattentional blindness and may, at least in part, reflect effects of expectation, intention and memory. The present study therefore also served to clarify the role of inattention in “inattentional blindness”. To avoid the expectation and memory confound in this study, we did not compare rates of inattentional blindness between critical trials and control trials. Instead we compared rates of inattentional blindness between critical trials with different levels of attention available, as determined by manipulations of perceptual load in the relevant task. Awareness reports in the control trial were used solely as an exclusion criterion: participants that could not report the critical stimulus in the fully-attended control trial were excluded from analysis (thus ensuring that any failures to report the critical stimulus in the critical trial could not be explained by an inability to see that stimulus). In this way, our comparisons were not confounded with varying levels of expectation: the additional task-irrelevant stimulus in the critical trial is equally unexpected at both levels of perceptual load. “Inattention” was manipulated through varying perceptual load. Determining the relationship between inattentional blindness and perceptual load in this way will not only establish the role of perceptual load in awareness but will also allow us to confirm that reported “blindness” within the inattentional blindness paradigm is indeed due to inattention.

A role for general task difficulty, and hence a possible role for perceptual load, in inattentional blindness has been hinted at in two previous studies. An early study (reported in Neisser, 1979) using the “selective looking” paradigm found greater rates of awareness for an irrelevant stimulus (e.g., a woman with an umbrella walking across the screen whilst participants attend to a ball-game) in the third repetition of the same video clip compared with the first viewing. The increase in awareness with practice may result from a reduction in perceptual load following greater practice in the relevant task. However, Neisser’s (1979) report does not establish that task performance became any easier with practice, since results regarding task performance were not reported. Moreover, although practice is expected to reduce perceptual load, it is also expected to reduce load on other task-processes, including memory. In addition, practice is also expected to speed-up task–responses, and hence may have reduced the delay between stimulus presentation and the questioning of awareness (as awareness questioning always followed task–responses). Thus, the increase in rates of awareness reports with practice may simply reflect a lower likelihood of forgetting due to effects of practice on processes other than perceptual load.

Simons and Chabris (1999) varied task difficulty more directly. They asked participants either to count the number of ball-passes made between one of two teams of basketball players in a video-clip (“easy task” condition), or to maintain two separate counts for the number of bounce passes and aerial passes (“hard task” condition). Awareness for an unexpected “Gorilla-man” walking through the playing space was reported more often by participants in the easy task condition than participants in the hard task condition, in line with our load hypothesis. The particular difficulty manipulation used in this study however, was likely to have involved a greater tendency for eye movements in the hard task condition than the easy task condition, as the discrimination between aerial and bounce passes would benefit from looking up (for aerial throws) and down (for bounce passes) whereas monitoring all ball-passes can be made without this discrimination. Thus, since eye movements cause blur on the retina, the critical stimulus may simply have been less visible in the hard task (separate-count) condition. Moreover, maintaining two separate counts (as in the hard task condition) places a greater load on working memory than maintaining just one count (as in the easy task condition). Since encoding into long-term memory is known to be determined by the availability of working memory (Baddeley, 1986), lower awareness in the hard task condition may have been caused by a reduction in the encoding of critical stimuli into memory (where it had to be retained until the awareness question following the rest of the video clip and the report of the count). As such, the role of attentional load per se (e.g., without the potential effects of eye movements and working memory load) in determining awareness within this task remains unclear.

In the present study, we systematically varied the level of perceptual load in the relevant task while assessing the rates of awareness reports for a task-irrelevant stimulus presented on a final trial. Static displays were presented for only a brief duration (200 ms, or less in some experiments) in order to preclude alternative accounts for effects of perceptual load in terms of eye movements. As described earlier, increased perceptual load means either that the number of relevant items with different identities is increased (e.g., a search task with many items is harder than searching amongst relatively few) or that a more demanding perceptual task is carried out for the same number of items (for review see Lavie, 2005). Accordingly, in the following experiments we manipulated perceptual load both by increasing the number of letters (with different identities) in a visual search task, and by varying the demands of perceptual judgments; comparing a subtle length discrimination (high load) either with a more obvious length discrimination (low load, Experiment 3) or with a simple color detection on identical stimuli (low load, Experiment 1).

Section snippets

Experiment 1

In Experiment 1, we modified the conventional inattentional blindness cross-task procedure (Mack & Rock, 1998) to incorporate a manipulation of perceptual load. Participants in each condition of load were given identical series of central cross-targets with two arms of clearly different color (blue and green) and slightly different length. Participants in the low load condition performed a simple color discrimination task (indicating which cross-arm was blue), typically thought to impose low

Experiment 2

In Experiment 2, we sought to generalize the effects of perceptual load on inattentional blindness across another manipulation of perceptual load, which varied the number of different identity items in the relevant task. Thus, the typical inattentional blindness cross-task was replaced by a visual search task. Participants were asked to search a circular array for a target letter amongst either five non-target letters (high load) or five place-holders (low load). A critical stimulus identical

Experiment 3

The small number of trials presented to participants in Experiments 1 and 2 precluded the assessment of effects of perceptual load on task RTs. Although it is well-established that the visual search set size manipulation of perceptual load used in Experiment 2 produces slower RTs in high load compared to low load (e.g., Lavie, 1995, Lavie and Cox, 1997, Maylor and Lavie, 1998, Lavie and Fox, 2000, Lavie, 2005, for review), the effects of the cross-task load manipulation on RTs are yet to be

Experiment 4

Experiment 3 clearly replicated the typical effects of perceptual load on target RTs, with slower RTs in conditions of high perceptual load compared with conditions of low perceptual load. Although the measure of awareness we used in Experiment 3 (i.e., via Yes/No reports) was not based on RTs, it might have been affected by slowing of task responses under higher perceptual load. Such slowing of responses would introduce a longer delay from presentation of the critical display until the

General discussion

The present research shows that the level of perceptual load in a current task determines whether a task-irrelevant stimulus will enter visual awareness or not. When load is increased in the relevant task (either through a greater number of items among which the target has to be found in search tasks as in Experiments 2 and 4; or through the cross-task typically used in inattentional blindness studies requiring a more subtle perceptual discrimination as in Experiments 1 and 3) more participants

References (32)

  • B.A. Eriksen et al.

    Effects of noise letters upon the identification of a target letter in a non search task

    Perception & Psychophysics

    (1974)
  • C.L. Folk et al.

    Involuntary covert orienting is contingent on attentional control settings

    Journal of Experimental Psychology: Human Perception and Performance

    (1992)
  • N. Lavie

    Perceptual load as a necessary condition for selective attention

    Journal of Experimental Psychology: Human Perception and Performance

    (1995)
  • N. Lavie

    Selective attention and cognitive control: dissociating attentional functions through different types of load

  • N. Lavie et al.

    The role of perceptual load in negative priming

    Journal of Experimental Psychology: Human Perception and Performance

    (2000)
  • N. Lavie et al.

    Contrasting effects of sensory limits and capacity limits in visual selective attention

    Perception & Psychophysics

    (2003)
  • Cited by (214)

    • Inattentional blindness in dogs

      2023, Applied Animal Behaviour Science
    View all citing articles on Scopus

    This research was supported by a Biotechnology and Biological Sciences Research Council (UK) studentship (UCF). We thank Live Science at the Science Museum, London UK for allowing data collection.

    ☆☆

    This manuscript was accepted under the editorship of Jacques Mehler.

    View full text