It's a Jungle in There: How Competition and Cooperation in the Brain Shape the Mind (12 page)

Schneider and Shiffrin pursued this idea with the more easily managed stimuli of letters in the alphabet. They sought to find out how quickly people could find a letter depending on the size of the target set. To vary the size of the target set, they varied the potential targets in successive blocks of trials. In a given block, their participants were given instructions such as, “The possible target in this block of trials is
H
,” or “The possible targets in this block of trials are
H
or
J
.”
³²

Behind Schneider and Shiffrin’s interest in the size of the potential target set was the idea, already mentioned, that if detecting a target is automatic, it shouldn’t matter how many potential targets there are. In other words, if you learn to automatically detect Bernie, Bill, and Buster (your three siblings) in a crowd, it shouldn’t take you longer to search for any one of them when you know all three of them are possibly there than when you know that only two of them are possibly there or when you know that only one of them is possibly there.
This predicted result should hold, however, only if the detection of targets is automatic. If the detection of targets is
not
automatic, the number of potential targets
should
impact the detection time: The more potential targets there are, the longer the detection time should be.

Schneider and Shiffrin found that as people looked for the same targets over and over again, the number of potential targets mattered less and less. With practice, the time to detect
H
, for example, depended to a decreasing degree on the possibility that other targets could appear. This outcome showed that participants developed the capacity to automatically detect the targets. In terms of the jungle principle, the demons responsible for finding
H
or
J
or
K
became sufficiently strong that they could yell equally loudly, or be heard equally well, no matter how many of their clique joined the claque.

What was the evidence that stimulus-response associations became automatic? The critical finding was that participants could detect targets automatically only when the targets were always associated with positive responses, and never or rarely when the targets were also associated with negative responses. If targets sometimes served as distracters, and so summoned negative responses, participants did not reach the level of automaticity they could if the same target was always a target. What was automatically triggered, therefore, were stimulus-response mappings, not stimuli alone or responses alone. Where automaticity took hold was at the point of response selection, where stimuli were used to determine which responses would be made. This is just what you’d expect if you thought attention aids action selection.

I’ve said that where the action is when it comes to attention is in action selection. My argument for this view echoes the ideas of Harold Pashler and his colleagues, as I said before. My position also accords with the so-called late-selection view of attention, a view that ascribes attention to a “late” process in stimulus-response translation—namely, at the response-selection side of the ledger, not at the stimulus processing side.
³³
The antithesis to the late-selection view is the early-selection view, advocated most famously by Donald Broadbent in his sensory gate hypothesis (illustrated with “opening an ear” or “closing an ear”). I favor the late-selection view on the grounds that action selection is where, ultimately, decisions must be made.

Despite this perspective, I want to emphasize that it would be a mistake to say that attention cannot be directed to stimuli
per se
or to the features comprising stimuli. Clearly, you can listen for high-pitched tones rather than
low-pitched tones, you can look for dark hues rather than light hues, and so on. There must be a way of accounting for the fact that low-level stimulus features can be objects of attention. Similarly, there must be a way of accounting for the fact that when people participate in studies where they attend to one type of stimulus, they invariably do worse at recalling or recognizing the stimuli to which they did not attend. For example, in dichotic listening experiments (experiments where two messages come to the two ears simultaneously), though participants may perk up when they hear their names, they cannot recall or recognize most other unattended stimuli.
³⁴
This shows that there are objective grounds for saying that unattended stimuli receive less-favored status than do attended stimuli, which raises the question of whether the brain responds differently to stimuli depending on how much attention is directed to them. If it does but if no overt action is made, is this a problem for the view that attention is action-limiting?

With respect to the first question—does the brain actually respond differently to stimuli depending on how much attention is directed to them?—it turns out that neural activity in the auditory cortex is enhanced if sounds come in at or near the expected pitch.
³⁵
Analogous effects are obtained for visual stimuli when recordings are made in the primary visual cortex.
³⁶
In one particularly dramatic illustration of selective visual effects, researchers recorded from two regions of the brain known to respond to different aspects of visual inputs. One region, within the fusiform gyrus, is activated by seen
faces
. The other region, the parahippocampal place area, is activated by seen
places
(for example, seen
houses
). Using fMRI techniques, the scientists who did this study found that the face region of the brain became more activated when attention was directed to a seen face than to a seen house, but the house region of the brain became more activated when attention was directed to a seen house than to a seen face.
³⁷
An especially striking feature of this demonstration was that the face and the house were overlaid. The fact that the two brain regions were activated to different degrees based on what the participants were attending to shows that different brain areas can be “goosed up” or “goosed down” depending on what information is sought. Another implication is that attention can be directed to
objects
as well as, or instead of, to
locations
.

Other studies have supported the hypothesis that attention can be object-based rather than location-based.
³⁸
In all these cases, participants were more primed to direct their actions to the items being attended to than to the items not being attended to, thereby addressing the second question posed above: Is it problematic for the view that attention is action-limiting that participants don’t have to make overt actions following the direction of
attention? The answer is no. The actions can be inner actions or precursors of actions.
Chapter 7
is all about what happens before actions are made.
³⁹

Can the phenomena I’ve summarized be accommodated with a single theory of attention? If so, is that theory of attention effectively embodied in the jungle principle? Toward answering that question, it’s useful to say first what theory of attention
won’t
work. Agreeing on what won’t work can help narrow the field of what may.

One theory that won’t work likens attention to a spotlight. Think of a prison escape movie. Imagine a scene where the convicts are trying to break out and find themselves in the prison yard at night. Periodically, a spotlight roves over the yard and the cons dash into the darkness to evade the roving beam. The guard guiding the spotlight is supposed to focus his or her attention on every place being illuminated. This, at least, is the guard’s job description. If you think of the guard as the homunculus and the moving spotlight as the focus of the guard’s attention, you have what amounts to the spotlight theory of attention.

I’ve already said that the spotlight theory of attention is wanting. What’s wrong with it? One problem is that someone or something needs to move the spotlight. That someone or something is the homunculus, the irksome bloke we want to evict from our theory of the mind.

A second problem is that the enhancement allowed by the spotlight does…what exactly? The light makes it possible to see more and so, in theory, to attend to more detail than was possible otherwise. Attending more is a bit like casting a brighter light on an area of interest. But how do you cast a brighter light on a sound you’re listening for, on a smell you’re savoring, or on a feature of a mathematics problem you’re trying to solve? The point is that casting more light is a nice metaphor that, ironically, may shed little light on the deeper problem to be solved.

A third difficulty is that spotlights can run in parallel. Theaters and circuses have multiple spotlights. Think of the proverbial three-ring circus. Nothing in the spotlight metaphor says there can be only one spotlight, but if there can be more than one, how many more can there be? 10
¹
? 10
²
? 10
¹²
(the number of neurons in the brain)? 10
¹⁴
(the number of synapses in the brain)?

A fourth problem is that the spotlight idea is fuzzy in scope. How wide can a spotlight spread? If it does spread, does that mean everything under it can be attended to equally well? If what’s in the center of the spotlight can be
attended to more effectively than what’s far from the center of the spotlight (but still within it), you’ve got spotlights within spotlights—not the sort of theory you want (Russian dolls within Russian dolls all over again).

Fifth, an important property of attention that you’d expect to hold if the spotlight metaphor were true turns out to be violated. Because a spotlight moves continuously between adjacent places of interest, any out-of-place object that’s suddenly interposed at one of those places should disrupt processing. Yet adding such an object doesn’t have this effect.
⁴⁰
The implication is that if attention moves, it can move discontinuously, which is something a spotlight can’t do unless it’s granted so many
ad hoc
properties that its theoretical value dims.

Considering all the things I’ve said in this chapter, is there a general theory of attention in reach? The theory I want to endorse is, of course, the one I’ve been endorsing all along, the jungle theory. Happily, a theory of attention that has achieved considerable notoriety has as its core claim one that aligns with the jungle perspective. The theory, advanced by Robert Desimone, now of MIT, and John Duncan, of the University of Cambridge,
⁴¹
was summarized as follows in a leading textbook on cognitive psychology: “…attention is seen as a form of competition among different inputs that can take place between different representations at all stages of processing.”
⁴²

This
biased competition
theory, as it is known, departs from its predecessors in that it is remarkably unconstrained. It doesn’t say attention works like a spotlight, with all the limitations that spotlights have. It doesn’t say that attention works like a sensory gate, with the restrictions that sensory gates impose. It doesn’t say that attention can operate only at the level of greatest meaning and personal significance, as you might suppose if you hark back to the cocktail party phenomenon. It doesn’t even say that attention operates at the level of response selection, though I happen to believe that that’s the case.

Desimone and Duncan didn’t develop their theory in consultation with me or, apparently, in consultation with Harold Pashler, the advocate of the response selection model of attention (and a leading scholar in the field of attention). Desimone and Duncan based their conclusion on the basis of a wide range of considerations, including some data not reviewed here. All of these facts suggested to Desimone and Duncan that it’s possible to see attentional bias effects virtually everywhere in the brain. For example, cells in one
part of the brain concerned with vision (area V4 of the extra-striate visual cortex) that fire in response to either of two visual stimuli show greatly reduced firing to one of those stimuli if it’s treated as a distracter.
⁴³

Saying that attention may be more important in some domains than others, as in saying that attention may be more important in response selection than in sensory intake, doesn’t imply that attention must operate only in the supposedly most-important areas. Attention can operate anywhere, as Desimone and Duncan allowed. Their key idea, in fact, is that attention is, at heart, a competitive process that gets resolved by creating biases on the fly.
⁴⁴
Those biases favor some outcomes over others wherever and whenever conflicts arise. Attention is also a cooperative process in that bias reflects preference. Giving preference to some events means granting cooperative relations between the relevant functions in the brain.

Because of this appeal to bias and competition, Desimone and Duncan called their theory the
biased competition
model, as I’ve already mentioned.
⁴⁵
Their model is so similar in spirit to the jungle principle that I’d be less than candid if I didn’t say that when I learned about their model I felt even more encouraged than I already was to embrace the jungle principle.
⁴⁶

You’ve got to be quick to survive in a jungle, but speed alone isn’t enough. You’ve also got to wait, and wait wisely. If you’re a cat on the prowl who has spotted a munchy-looking mouse, you need to wait for the right moment to pounce on that unsuspecting prey. The mouse, meanwhile, has its own agenda. It wants to fulfill its feline-free fate for as long as possible, striving for a quick getaway no matter how pugnacious the puss.

The two characters in this little drama are in different states of preparedness. One is primed for pouncing. The other is braced for bolting. Each needs to act as quickly as possible. The situation they’re in—having to respond as quickly as they can while maintaining preparedness in different ways—has been studied in detail by cognitive psychologists. Their favorite measure of preparedness is reaction time, or RT for short. In this chapter I’ll consider how data from RT tasks bear on the jungle principle.