If we could assume that awareness is not a simple, all or none phenomenon, then the evaluation of self-report in blindsight would itself be clearer. Campion, Latto & Smith are right to worry about inconsistent measures of awareness in the study of blindsight, but they seem to assume that introspective knowledge is dichotomous and, worse yet, without use to us as investigators. They also confuse the terms “consciousness” and “verbal report”; hence the discussion here is concerned mainly with verbal reports of awareness in studies of visual perception.
Investigations of blindsight have in the past confounded two measures of awareness, a threshold determination by the ascending method of limits, and localisation determination by forced-choice guessing. The fact that there exists a difference in the conclusions suggested by these methods (which is itself the primary evidence of blindsight) is important in its own right, and Campion et al. are in danger of dismissing vital evidence concerning perceptual processing.
When threshold determinations give an indication of a scotoma, the conclusion is based (in humans) upon verbal self-report, and yet when pointing or guidance behavior is used as evidence of vision within the scotoma, the conclusion is based upon forced-choice guessing behavior. Response conservatism can account for reports of blindness in the case of partial sight, and using a forced-choice task eliminates the incidence of falsely negative reports. Different criteria may affect the patients' response decisions in the two tasks, and Campion et al. are therefore justified in questioning the conclusion that a response to a hemianopic stimulus is evidence of vision without awareness. We cannot conclude, on the basis of this inconsistency alone, that subcortical blindsight mediates guessing performance. Responsibility may reside in criterion differences in the preparedness to respond.
Campion et al. go further than to suggest alternatives to the notion of subcortical blindsight, and in particular, they argue against the usefulness of self-report. When they follow Corso (Reference Corso1967) in suggesting (a) that self-report is beyond verification, and (b) that self-report of states of awareness is not always possible, the issues are clear. The first point is valid but restrictive. If we are not to ask humans about their beliefs about their internal states, then a great deal of activity in psychology and medicine is called into question. These are questions of motives, and in what follows here, as in most research, the assumption is that subjects have no reason for behaving dishonestly. The second point is also appropriate. We cannot give veridical verbal reports of our perceptions of a nonverbal world, partly for reasons of linguistic poverty. When blindsight patients report being “aware of something” without being able to describe the stimulus, their remarks resemble Sperling’s (1960) reports of being able to report far less than he was aware of in tachistoscopic displays. This discrepancy should not be confused with a lack of willingness to respond. Patients give good reports of nonhemianopic stimuli, and so their reports on stimuli presented to the scotoma gain validity. The feelings of awareness associated with these stimuli contain vital perceptual evidence, and their importance as a dissociation of perception and awareness has been discussed by Humphrey (Reference Humphrey, Josephson and Ramachandran1980), Natsoulas (Reference Natsoulas and Underwood1982), and Searle (Reference Searle1979), among others. In the final section of their paper, Campion et al. acknowledge the importance of dissociation, but in their consideration of the evidence, they prefer to exclude verbal reports (now in the guise of “consciousness” as a whole).
Verbal reports from threshold determinations are inconsistent with pointing behavior in a forced-choice paradigm, but this is no reason to recommend that we take the behaviorist stance of ignoring the self-report of awareness of a stimulus. When the forced-choice guessing task employs a verbal report, an indication of vision is again obtained (Weiskrantz, Warrington, Sanders & Marshall Reference Weiskrantz, Warrington, Sanders and Marshall1974). If we are to ignore reports of states of awareness, as Campion et al. recommend, then should we ignore this vital verbal evidence? The problem is not in the use of verbal self-report, then, so much as in the preparedness to respond indicated by the measure of β in signal detection analyses. It is quite unnecessary to question the use of verbal self-report if similar response criteria are used in the two tasks.
It is well established that not only can stimuli not available for verbal report lead to feelings of awareness (Sperling Reference Sperling1960), but also that those that do not lead to such feelings can also influence behavior. Accordingly, we would not expect verbal reports to be necessarily available to blindsight patients in all circumstances in which behavior is affected. For the case of verbal stimuli, Underwood (Reference Underwood1977) and Fischler and Goodman (Reference Fischler and Goodman1978) have established that awareness of an effective stimulus changes its direction of effect upon an attended stimulus. Reducing the exposure duration of an effective stimulus below its threshold for awareness can lead to an effect not apparent with supraliminal presentations. A recent result from our laboratory (Underwood, Whitfield & Briggs, unpublished) illustrates this point. In this experiment normal adults were required to respond to a series of line drawings of common objects by naming them, and the naming latencies were recorded. Prior to the onset of each picture a word appeared on the projection screen, below the area reserved for the picture. The word was presented either subliminally (25-msec exposure, pre- and postexposure masked with a pattern of letter fragments), or supraliminally (250 msec), with 250-msec interval between the word and the picture. With a 25-msec exposure, our subjects were unable to report any words. Results indicated that when the word was a semantic associate of the picture the naming response was faster than when the word was unassociated, but this effect held for subliminal presentations only. Whereas a smaller semantic priming effect did appear with the supraliminal words, it was statistically unreliable.
The changing pattern of influence of subthreshold and suprathreshold stimuli is itself evidence for a correlation between cognitive processing and verbal reports of awareness. Not only can self-report be investigated within the domain of inquiry of cognition, it can also give valuable evidence of the progress of cognitive processing. The threshold does not have a step function however, and the ogive curve from threshold determinations (see Campion et al.'s Figure 2A) may be indicative of a continuous change of state from absence of awareness to full awareness. Blindsight patients give reports of partial awareness of the nature of hemianopic stimuli, perhaps as a function of the processing of degraded sensations. This may be an improvement in the use of sensations of degraded stimuli which is observed during blindsight training (Zihl Reference Zihl1980) or a changed criterion for response accompanying learned associations between sensations and stimuli. Blindsight patients typically express surprise over the accuracy of their guesses, suggesting, perhaps, that their inadequate verbal reports reflect impoverished sensations, or sensations qualitatively different from those of nonhemianopic stimuli. Reports of the quality of blind-sight perceptions led Searle (Reference Searle1979) to conclude that they have the intentionality of beliefs without phenomenal properties. This certainly describes the accuracy of performance when the criterion for belief becomes irrelevant. The intentionality that accompanies normal vision is absent with hemianopic stimuli, but performance is available. Near-threshold stimuli that give rise to feelings of awareness without full knowledge (or total intentionality) are stimuli that are not generally useful. These feelings of awareness, identified with the “non-conscious visual perceptions of blindsight” by Natsoulas (Reference Natsoulas and Underwood1982), are insufficient to give rise to beliefs about objective stimuli, at least for untrained observers. Rather than exclude these reports from any consideration, as Campion et al. inconsistently recommend – they use them in their own experiments – we must collect introspective reports as evidence. They are responses to sensory stimulation, and thus correspond to changes in the nervous system.
We cannot assume that awareness is all or none, and the paucity of verbal reports of hemianopic stimuli in combination with (a) good reports of other stimuli, and (b) feelings of “awareness of something,” must be taken together as evidence of processing without awareness of the products of processing.
If we could assume that awareness is not a simple, all or none phenomenon, then the evaluation of self-report in blindsight would itself be clearer. Campion, Latto & Smith are right to worry about inconsistent measures of awareness in the study of blindsight, but they seem to assume that introspective knowledge is dichotomous and, worse yet, without use to us as investigators. They also confuse the terms “consciousness” and “verbal report”; hence the discussion here is concerned mainly with verbal reports of awareness in studies of visual perception.
Investigations of blindsight have in the past confounded two measures of awareness, a threshold determination by the ascending method of limits, and localisation determination by forced-choice guessing. The fact that there exists a difference in the conclusions suggested by these methods (which is itself the primary evidence of blindsight) is important in its own right, and Campion et al. are in danger of dismissing vital evidence concerning perceptual processing.
When threshold determinations give an indication of a scotoma, the conclusion is based (in humans) upon verbal self-report, and yet when pointing or guidance behavior is used as evidence of vision within the scotoma, the conclusion is based upon forced-choice guessing behavior. Response conservatism can account for reports of blindness in the case of partial sight, and using a forced-choice task eliminates the incidence of falsely negative reports. Different criteria may affect the patients' response decisions in the two tasks, and Campion et al. are therefore justified in questioning the conclusion that a response to a hemianopic stimulus is evidence of vision without awareness. We cannot conclude, on the basis of this inconsistency alone, that subcortical blindsight mediates guessing performance. Responsibility may reside in criterion differences in the preparedness to respond.
Campion et al. go further than to suggest alternatives to the notion of subcortical blindsight, and in particular, they argue against the usefulness of self-report. When they follow Corso (Reference Corso1967) in suggesting (a) that self-report is beyond verification, and (b) that self-report of states of awareness is not always possible, the issues are clear. The first point is valid but restrictive. If we are not to ask humans about their beliefs about their internal states, then a great deal of activity in psychology and medicine is called into question. These are questions of motives, and in what follows here, as in most research, the assumption is that subjects have no reason for behaving dishonestly. The second point is also appropriate. We cannot give veridical verbal reports of our perceptions of a nonverbal world, partly for reasons of linguistic poverty. When blindsight patients report being “aware of something” without being able to describe the stimulus, their remarks resemble Sperling’s (1960) reports of being able to report far less than he was aware of in tachistoscopic displays. This discrepancy should not be confused with a lack of willingness to respond. Patients give good reports of nonhemianopic stimuli, and so their reports on stimuli presented to the scotoma gain validity. The feelings of awareness associated with these stimuli contain vital perceptual evidence, and their importance as a dissociation of perception and awareness has been discussed by Humphrey (Reference Humphrey, Josephson and Ramachandran1980), Natsoulas (Reference Natsoulas and Underwood1982), and Searle (Reference Searle1979), among others. In the final section of their paper, Campion et al. acknowledge the importance of dissociation, but in their consideration of the evidence, they prefer to exclude verbal reports (now in the guise of “consciousness” as a whole).
Verbal reports from threshold determinations are inconsistent with pointing behavior in a forced-choice paradigm, but this is no reason to recommend that we take the behaviorist stance of ignoring the self-report of awareness of a stimulus. When the forced-choice guessing task employs a verbal report, an indication of vision is again obtained (Weiskrantz, Warrington, Sanders & Marshall Reference Weiskrantz, Warrington, Sanders and Marshall1974). If we are to ignore reports of states of awareness, as Campion et al. recommend, then should we ignore this vital verbal evidence? The problem is not in the use of verbal self-report, then, so much as in the preparedness to respond indicated by the measure of β in signal detection analyses. It is quite unnecessary to question the use of verbal self-report if similar response criteria are used in the two tasks.
It is well established that not only can stimuli not available for verbal report lead to feelings of awareness (Sperling Reference Sperling1960), but also that those that do not lead to such feelings can also influence behavior. Accordingly, we would not expect verbal reports to be necessarily available to blindsight patients in all circumstances in which behavior is affected. For the case of verbal stimuli, Underwood (Reference Underwood1977) and Fischler and Goodman (Reference Fischler and Goodman1978) have established that awareness of an effective stimulus changes its direction of effect upon an attended stimulus. Reducing the exposure duration of an effective stimulus below its threshold for awareness can lead to an effect not apparent with supraliminal presentations. A recent result from our laboratory (Underwood, Whitfield & Briggs, unpublished) illustrates this point. In this experiment normal adults were required to respond to a series of line drawings of common objects by naming them, and the naming latencies were recorded. Prior to the onset of each picture a word appeared on the projection screen, below the area reserved for the picture. The word was presented either subliminally (25-msec exposure, pre- and postexposure masked with a pattern of letter fragments), or supraliminally (250 msec), with 250-msec interval between the word and the picture. With a 25-msec exposure, our subjects were unable to report any words. Results indicated that when the word was a semantic associate of the picture the naming response was faster than when the word was unassociated, but this effect held for subliminal presentations only. Whereas a smaller semantic priming effect did appear with the supraliminal words, it was statistically unreliable.
The changing pattern of influence of subthreshold and suprathreshold stimuli is itself evidence for a correlation between cognitive processing and verbal reports of awareness. Not only can self-report be investigated within the domain of inquiry of cognition, it can also give valuable evidence of the progress of cognitive processing. The threshold does not have a step function however, and the ogive curve from threshold determinations (see Campion et al.'s Figure 2A) may be indicative of a continuous change of state from absence of awareness to full awareness. Blindsight patients give reports of partial awareness of the nature of hemianopic stimuli, perhaps as a function of the processing of degraded sensations. This may be an improvement in the use of sensations of degraded stimuli which is observed during blindsight training (Zihl Reference Zihl1980) or a changed criterion for response accompanying learned associations between sensations and stimuli. Blindsight patients typically express surprise over the accuracy of their guesses, suggesting, perhaps, that their inadequate verbal reports reflect impoverished sensations, or sensations qualitatively different from those of nonhemianopic stimuli. Reports of the quality of blind-sight perceptions led Searle (Reference Searle1979) to conclude that they have the intentionality of beliefs without phenomenal properties. This certainly describes the accuracy of performance when the criterion for belief becomes irrelevant. The intentionality that accompanies normal vision is absent with hemianopic stimuli, but performance is available. Near-threshold stimuli that give rise to feelings of awareness without full knowledge (or total intentionality) are stimuli that are not generally useful. These feelings of awareness, identified with the “non-conscious visual perceptions of blindsight” by Natsoulas (Reference Natsoulas and Underwood1982), are insufficient to give rise to beliefs about objective stimuli, at least for untrained observers. Rather than exclude these reports from any consideration, as Campion et al. inconsistently recommend – they use them in their own experiments – we must collect introspective reports as evidence. They are responses to sensory stimulation, and thus correspond to changes in the nervous system.
We cannot assume that awareness is all or none, and the paucity of verbal reports of hemianopic stimuli in combination with (a) good reports of other stimuli, and (b) feelings of “awareness of something,” must be taken together as evidence of processing without awareness of the products of processing.
Acknowledgment
Preparation of this paper was supported by grant no. G/8127736N from the Medical Research Council.