Source : Memory and the brain
Color coding : Violet - Question Red - Claim Blue r- Evidence Green - Assumption Yellow - Constraints
To become aware of sensory experience apparently requires more than the activity of the different sense modalities. It requires a turning toward the source of such experience: it requires attention.
According to William James, everyone knows what attention is: "It is the taking possession by the mind, in clear and vivid form, of one out of . . . several simultaneously possible objects or trains of thought." ( 1890, p. 403.) But how do we pay attention, what is this "taking possession by the mind"? Is it a cognitive process? Most psychologists would unhesitatingly answer that it is. However, our experience speaks against this interpretation.
Take, for example, seeing, hearing, or touching: these are cognitive processes. They mediate knowledge and do so directly, immediately. We see as soon as we open our eyes, provided there is enough light; we hear whenever there is a sound, provided we are awake and our attention is not otherwise engaged. In contrast, attention is directed here or there, and can be withdrawn from something in front of our eyes and directed toward something imagined or remembered. Attention can increase the intensity of a sense experience: if we expect a sound, attention can lower the threshold of hearing. When directed elsewhere, attention may altogether prevent perception. Attention is a condition of cognition just as light is a condition of sight. A man engrossed in a book may not be aware that somebody is asking him a question. That he hears is shown by the fact that calling his name will attract his attention, although the questioner may not have raised his voice. Cognitive functions do not act in this manner. Sight, hearing, touch, smell, taste, simply mediate experience, they do not change it. They neither increase nor decrease the intensity of experience. Sensory receptors receive changes in cortical sensory areas even during sleep or light anesthesia.
Attention makes reflective knowledge or awareness possible; but is attention itself a cognitive function? This problem is usually not talked about. Recent experiments and theories of attention ( Norman, 1968, 1969; Kahneman, 1973) treat the effect of attention on sensory experience at great length, but never raise the question whether attention itself is cognitive in nature. As a result, theorists often conceive of the sensory system as a single communication channel with a selective filter (attention) that is part of the system and filters out irrelevant sensory impressions ( Broadbent, 1958; Bower, 1967). This filter, they postulate, can be tuned to any one of several inputs. Only information passed by the filter can be further processed and remembered.
Such a model seems to fit experiments on selective listening in which the subject is asked to repeat aloud what he hears in one ear while a different (irrelevant) message is fed to the other ear. In these experiments, the subject can switch attention at will from one message to the other, even when the voices in the two channels are identical. In fact, Spieth et al. ( 1954) found that such selective listening was possible when the voices were heard over loudspeakers placed only ten to twenty degrees apart. The subjects could also follow a voice coming through a low-pass filter, rather than one sent through a high-pass filter. Finally, Egan et al. ( 1954) found that the subjects can attend either to intensity or to frequency of sounds. In short, it is possible to attend to one of two simultaneous sequences of sounds different in quality, location, intensity, and frequency.
However, when the irrelevant (unattended) message contains the subject's name, he begins to attend to it ( Moray, 1959); also, he switches from one ear to the other when the message in that ear becomes relevant, for instance, when the message in the non-attending ear continues the passage read to the attending ear ( Treisman, 1960, 1964a,b). This switching in the middle of the message is difficult to explain on the basis of a filter theory. How can relevant stimuli on unattended channels capture attention? This problem is discussed by Craik and Jacoby ( 1979). For instance, several subjects recognized the rejected message if it was identical with the message they attended to. In fact, some bilingual subjects recognized such identity although the unattended message was in another language ( Treisman, 1964a,b). Treisman tried to explain this phenomenon by postulating several levels of processing, and insisted that the level reached depends not only on the physical characteristic of the stimulus but on current expectations in the analyzing system. Broadbent ( 1971) later accepted this modification of his filter theory. But neither the original theory nor Treisman's modification can explain how attention can be directed toward meaningful material when the filter is only equipped to deal with the physical features of the stimulus. It can hardly have biases and expectations that go beyond these characteristics. Norman ( 1968, 1969) has faced the difficulty of explaining a switch of attention on the basis of a filter model. He claims that a model that can account for such changed filters needs a more complicated system. According to him,
------------------------
Notes : Attention according ..... filter theory is equiped to deal with physicalfeatures of the stimulus but it does not explain how attention can be directed toward meaningful material.
------------------------
"all signals arriving at sensory receptors pass through a stage of analysis performed by the early physiological processes. Then, the parameters extracted from the processes are used to determine where the representation of the sensory signal is stored. Thus . . . all sensory signals excite their stored representation in memory. Now, at the same time, we assume that an analysis of previous signals is going on. This establishes a class of events deemed to be pertinent to the ongoing analysis. The set of pertinent items also excite their representation in memory. The item most highly excited by the combination of sensory and pertinent inputs is selected for further analysis." ( Norman, 1969, p. 33)
The difficulty with this description is that we are not certain what Norman is speaking about. Is he referring to a neural network, a computer model, a flow diagram, or a set of human activities? Often, Norman speaks explicitly of human mental processes, particularly in his later articles ( Norman, 1979). But he also describes processes that operate on knowledge structures in an almost mechanical way. When he speaks of "the item most highly excited by the combination of sensory and pertinent inputs," which is selected for further analysis (and presumably attended to), we begin to wonder whether he is perhaps alluding to the neural network. In such a network, the most intense excitation wins out. But the human being may pay attention not to the lecturer's loud voice and erudite discourse but to a fleeting distraction -- and who is to decide whether the distracting thought is more intense? Norman's notion of attention as the algebraic sum of excitation from perception and memory analyses does not explain why a subject should recognize a message read to the non-attending ear as identical with the message to which he has been paying attention all along, particularly when the two messages are in different languages. Nor does it explain why a memory analysis should establish a pertinent item in the first place -- unless it is the human being who decides what is relevant and directs his recall accordingly.
Both Norman's and Broadbent's models depend on the presence of memory traces, some of which are excited by the sensory input. But can we assume that there will always be stored memories to select an item for attention? Supposing a subject is instructed to press a button every time he sees a shade of red in the tachistoscope. No matter how many shades are shown, the subject has no difficulty in judging them correctly as red and pressing the button. The notion that there must be a memory trace for every type and intensity of sensory experience leads, as Norman says in another connection, "to a staggering amount of complexity in the neural network required to do the task. Moreover, it is not clear that any finite network could ever suffice, for there are truly an infinite number of ways we can see well-known objects" ( 1969, p. 40). Norman's solution is to suggest that stimulus-analyzing mechanisms recognize patterns, and recognize them by rules: "a set of rather general purpose operations examine the set of features extracted by the basic physiological analyzers and classify patterns, more by synthesizing them anew than by any other procedure" (p. 40). But if attention is merely the effect of various operations, as Norman says, the model does not explain how the subject decides that a given color is a shade of red. If he is instructed to look for "red," so that the instructions make this color pertinent, how does the expectation so aroused excite the trace for that particular shade? The concept "red" has no particular shade, hence there can be no memory trace for any given shade. Even if we assume, as some theorists do, that a concept is the generalized image of a class of objects, that is, a generalized shade of red, how can a particular shade excite the trace of such a generalized memory image? Again, it is the human being and not a stimulus-analyzing mechanism that forms concepts and recognizes a given shade of red as belonging to the class of "red."
Neisser ( 1967), a cognitive psychologist, also insists that attention "is simply an allotment of analyzing mechanisms to a limited region of the field. To pay attention to a figure is to make certain analyses of, or certain constructions in, the corresponding part of the icon." (p. 89.) For Neisser, focal attention can function only after "preattentive" global processes have already segregated the figural units involved. He insists that even a mechanical recognition system would have to have such a feature for separating certain units and selecting one after another for further processing.
Neisser's "preattentive processes" have an obvious similarity with the integrative sense discussed in chapter 1; so do Norman's stimulus-analyzing mechanisms. Without the ability to see a well-articulated scene, or to distinguish patterns of sound, it would be impossible to direct attention to any part of the visual field, or to begin listening to discover the drift of a conversation or the theme of a musical composition.
Like Norman, Neisser seems to consider attention simply as the result of certain preconscious operations. When he says that attention is "the allotment of analyzing mechanisms," who or what makes the allotment? Neisser disagrees with psychoanalytic writers who regard attention as the manifestation of a single process, an allocation of energy; but his own formulation shows little improvement. Whether attention is a process of allocating energy, or the allotment of analyzing mechanisms, we have no clue how this allotment is done, just as we do not know how the allocation of energy can result in attention.
It is more than doubtful that the same preattentive or preconscious process that segregates the manifold of sensory impressions into distinguishable units could also select a given unit for further processing. The preattentive global process that results in crude percepts cannot direct what is to be done next. Another function is needed that will allot mechanisms, allocate energy, and direct the memory search. And when we ask with Neisser: Who does the searching, the allotting? Is there a little man in the head, a homunculus? Neisser answers: As recently as a generation ago, processes of control had to be thought of as homunculi, because man was the only known model of the executive agent. Today, the stored-program computer has provided us with an alternative possibility, in the form of the executive routine. This is a concept which may be of considerable use to psychology. ( 1967, p. 295.)
Of course, Neisser admits that the executive routines must be programmed into the computer from the beginning and would not allow a change of direction in midstream, yet this is a common phenomenon in human experience. In man, he thinks, such executive functions are acquired through experience -- a case of learning like any other.
But learning is not the explanation. We do not learn attention, we merely learn to what we should attend, just as we do not learn to see, we merely learn to interpret what we see. It is odd that psychologists should be willing to accept "executive routines" that somehow (nobody knows how) direct man's behavior, when they are unwilling to agree that the human being directs his own behavior, yet that is a fact of everybody's direct experience. The very notion of a "little man in the head" could gain credence only because psychologists were convinced that psychological activities, like the movements of physical bodies, must be explained objectively, without appealing to any human agency. Yet a human being sees, pays attention to what he sees, appraises it, thinks about it, chooses to react to it in some way. He sometimes throws a ball or digs a ditch. We do not assume that a little man in the head pulls strings to make us see, make us dig ditches or throw a ball -- why should we need a homunculus to attend to something, to remember, to think?
To insist that human behavior is the result of learned executive routines is to indulge in a double analogy, from man to machine and from machine to man. Executive routines are instructions directing computer searches, on the analogy of a human being directing himself or others to investigate a particular area of interest. Instead of assuming that the human being uses executive routines, it is much simpler and closer to human experience to assume that we use various functions to deal with the world around us. When we do that, the only problem left is to identify the functions needed in a particular activity.
There has never been any difficulty about obvious sensory functions. Seeing, hearing, smelling, touching, and tasting have always been accepted as something we do -- except for early behaviorists who admitted responses to stimuli, but not the perception of stimuli. But the notion of an integrative sense that enables us to see objects rather than colored patches has found few adherents until it surfaced comparatively recently in the guise of global preattentive processes ( Neisser, 1967). Yet it is obvious that the visual sense itself can only mediate color. However, special cell columns in the visual cortex have been found to respond to vertical, horizontal, and oblique lines, and to movement ( Hubel & Wiesel, 1962). Hence it is the visual cortex with its six layers that seems to mediate vis- ual form and movement. It is a reasonable inference that the feltwork of the sensory cortex mediates sensory integration and articulates the visual field in such a way that we can focus on any part of it.
The primary function of attention is not that of analyzing or integrating sensory impressions, but of focusing other functions on the area of interest. We turn toward the object of attention, look, listen, and try to discover what we can about it. In other words, attention is a desire to see, hear, taste, smell, touch, think about, and so to know something as best we may. Attention requires some kind of intuitive or deliberate, unconscious or conscious judgment that something is worth knowing or investigating. Attention can be voluntary or involuntary. Involuntary attention is based on an intuitive, unwitting appraisal: a sudden loud sound, a bright color in a dull landscape, a whiff of perfume, or an unpleasant smell, all of these can alert us and draw our attention. Voluntary attention, on the other hand, is based on a person's conscious, deliberate judgment that this particular thing is something worth knowing about. If an experimental subject is instructed to press a button every time he sees a shade of red, he attends to the visual array. He wants to know it so that he can respond to it. As long as this desire lasts, he will be "set" to notice and report the shade that fits the concept. He may see other colors but will not react to them and may not even remember them.
Posner ( 1973) allows for both preconscious and conscious perceptual processing in his model of attention. A single-channel central processing capacity represents conscious attention, and is followed by rehearsal and the choice of response. Conscious attention is made possible by two stages of perceptual processing: the encoding of the stimulus, and a comparison of the stimulus with the contents of long-term memory. In his experiments, Posner showed a target stimulus (a digit or letter) to the subject, and later an array of similar stimuli with the instruction to locate the target stimulus. According to Posner, this search and comparison is preconscious. He comments that "both the target and non-target items are subject to a memory search process but only the target item gives rise to the phenomenological experience of jumping out at the subject." (p. 41.) But there is considerable doubt that this experience of the stimulus jumping out at the subject is the result of a comparison of each array stimulus with the remembered target stimulus. Indeed, research findings seem to be against such an assumption. For instance, Sternberg ( 1975) had his subjects commit a set of stimuli (letters, digits, figures) to memory; he called this the "positive set." He then showed a single test stimulus and asked his subjects to press one switch if the test stimulus was a member of the positive set, another if it was not. Sternberg found that the reaction time was so short that the subjects must have made about thirty comparisons per second, a rate considerably faster than the rate of subvocal speech which might have to be used for such comparisons. Sternberg also reported that the subjects were not aware of any search or comparison. Apparently the test stimulus simply felt familiar.
This is a prime example of an assumption based on a computer analogy: because the computer can locate the test stimulus only by search and comparison, the human being must do so, too. But the human being can recognize the test stimulus if he has seen it before (or memorized it, in Sternberg's experiment), that is, he can find it familiar. Recognition has always been considered different from recall. It is inaccessible to analysis, accounts for the target stimulus "jumping out" at the subject and is far faster than recall and comparison could ever be. In chapter 5, I discuss the difference between recall and recognition at greater length. Here, I will merely point out that recognition attracts attention (the stimulus "jumps out").
Shevrin and Dickman ( 1980) reviewed several different theories of attention and perception, all of which assume preconscious processing. A few of them also assume that such processing is qualitatively different from conscious attention. These authors suggested, on the basis of the available body of experimental research, that an initial preconscious stage of cognition is psychological (not physiological), affects conscious experience and may differ from conscious cognition in its principles of operation; conscious cognition is a later stage.
I would like to suggest that this preconscious cognition is an intuitive appraisal of something as familiar, or as good to know better because unfamiliar; this arouses a desire to turn toward it, look, listen, and investigate it. In the case of familiar objects, their recognition brings about a desire to do something about it, whether to report the occurrence or respond to it as usual. Hence attention is an appetitive and not a cognitive function. We want to turn toward the object, look, and listen. This immediately solves the problem of selective attention: all appetition implies a choice of goal, whether we want a new hat, a meal, to travel to Europe, or to look at a new car. If we attend to something, we select out of many possible things one we want to know better, whether we want to listen to a recording, watch an eclipse, or report on shades of red, locate a target stimulus, or identify the test stimulus in the positive set.
Involuntary attention, then, is the same impulse to look and listen, but based on the completely unconscious intuitive appraisal that this is unfamiliar, and should be investigated. Voluntary attention is based on a deliberate, conscious appraisal. In both cases, we have an impulse to focus on the selected object. But in voluntary attention, we want to look, in involuntary attention, we feel an urge to look.
No comments:
Post a Comment