Marcia Johnson’s research tells us that source monitoring is a skill, much like the skill of discriminating between rhinoceroses and elephants. Imagine that you are a game warden and, in order to conduct some kind of zoological research, you are required to shoot with an anaesthetic dart the former beast but not the latter. You are standing in a jungle clearing, when a large grey animal charges into view. Pulling the trigger, you discover to your dismay that you have knocked out an elephant. On asking yourself how you could have made such a bad mistake, you will discover that your judgement was undermined by the same factors that undermine the source-monitoring judgements of hallucinating patients.
Perhaps your beliefs and expectations overrode the evidence of your eyes. One of your companions may have told you that only rhinoceroses live at this particular location. You therefore assumed that you were seeing a rhinoceros and pulled the trigger without making sure. Similarly, hallucinators who mistake their inner speech or mental imagery for voices or visions of other people may be influenced by what they expect to hear or see. This is why culture has such a profound impact on hallucinations. Someone who grows up in a society that recognizes the existence of ghosts, or that values spiritual experiences, will be more likely to mistake the mental image of a deceased relative for an apparition than someone who has grown to maturity in a society that emphasizes the scientific world view.
Another possibility is that your judgement was impaired by degradation of the stimulus. You may have been stalking your prey in darkness, or your view may have been obscured by foliage, so that it was difficult to tell a rhinoceros from an elephant. Our ability to vary the level of evidence required before deciding that a stimulus is present, allows us to be flexible when making judgements about the external world. In situations in which our perception of our surroundings is degraded, it is often necessary to adopt a weak ‘criterion’ for deciding that an expected event has in fact occurred. (The rustle of the undergrowth, and the vague perception of a bulky object moving before us, may be enough to persuade us that we are in the presence of a rhinoceros.) For similar reasons, hallucinating patients are especially likely to mistake their thoughts for voices in conditions in which there is very little external stimulation (for example, when alone at night), or when external stimulation is chaotic and unpatterned (for example, in the middle of a noisy crowd, or when bombarded by white noise during a psychology experiment).
Finally, stress and a sense of urgency may have played a role. You may have feared that a failure to anaesthetize the rhinoceros quickly would leave you vulnerable to being charged and injured. Of course, stress undermines mental efficiency. The extent to which we process the information that is available to us is reduced, and our decisions tend to be hasty and less accurate as a consequence. Just as we are more likely to pull the trigger inappropriately when adrenalin is running through our veins so, too, patients under stress are especially likely to mistake their imaginings for real events.
The influence of these factors on the source-monitoring judgements of hallucinating patients is illustrated in Figure 14.3. I have laboured my analogy with hunting because the idea that source monitoring is a skill is not intuitively obvious to many people. After all, we rarely consciously think about the source of our perceptions, and source-monitoring judgements are nearly always automatic (but, then, so are many of our judgements about rhinoceroses and elephants). The analogy shows how hallucinations arise from an error of judgement rather than an error of perception. It also shows how hallucinating can be explained in terms of the same kinds of mental processes that affect normal perceptual judgements.
Testing the Source-Monitoring Hypothesis
The argument we have pursued so far states that people who hallucinate make faulty judgements about the sources of their experiences, and it is for this reason that they mistake their inner speech or visual imagery for stimuli external to themselves. However, we have not so far considered direct evidence that hallucinating patients make this kind of error more frequently than other people. The experiments that have addressed this issue have been complex, and so the following account will necessarily highlight only the essential elements of key studies.
Figure 14.3 An outline model of hallucinatory experiences. Adapted from R. P. Bentall (2000) ‘Hallucinatory experiences’, in E. Cardena, S. J. Lynn and S. Krippner (eds.) Varieties of Anomalous Experience: Examining the Scientific Evidence, Washington, D.C.: American Psychological Association.
Source monitoring for memories
The earliest source-monitoring experiment with hallucinating patients was reported by psychologist Arthur Heilbrun in 1980.62 Heilbrun, based at Emory University in Atlanta, USA, interviewed hallucinating and non-hallucinating patients and recorded their opinions about a number of topics. Their verbatim remarks were later included in a multiple-choice test alongside other statements which were similar but not identical. The hallucinators had particular difficulty when asked to recognize their own statements, and were less able to do this than the non-hallucinating patients.
Since this promising start, other source-monitoring experiments have been conducted with hallucinating patients by Heilbrun, by myself and by other investigators.63 Rather than describing each of these experiments in detail, I will focus on a recent study reported by Liverpool-based psychologists Peter Rankin and Pierce O’Carrol, which highlights some of the complexities involved in this kind of research.64 Rankin and O’Carrol decided not to study patients, but instead compared students who had high scores on a questionnaire measure of hallucinatory experiences with students who had low scores. Their experimental design was based on the work of Marcia Johnson. The students were asked to learn ‘paired associates’ (pairs of words, for example, ‘vehicle–car’). Sometimes the word-pairs were presented to the students and at other times their memory was tested by presenting them with the first word in each pair and asking them to recall the second (for example, ‘Which word goes with “vehicle”?’). The number of times that the words were presented and tested was cunningly varied so that, for example, some of the words were presented many times but tested on only a few occasions, whereas others were presented on only a few occasions but tested on many. At the end of this sequence of presentation and testing, the participants were asked to say how many times they thought they had been presented with each item. As Marcia Johnson had previously found, participants gave inflated estimates for those words which had been most tested, indicating that they were mistaking occasions on which they had recalled the words for occasions on which the words had been presented. In Rankin and O’Carrol’s study, this effect was most evident in those students who scored high on the hallucination measure, indicating that these people were most prone to misattributing their mental events (the times they recalled the words) for events in the world (presentations).
Despite these positive results, experimental studies of source monitoring for memories with psychiatric patients, using similar methods, have produced mixed results. In one investigation, Gus Baker, Sue Havers and I found that patients who experienced hallucinations were especially likely to mistake words that they had generated themselves for words that they had heard,65 but Mark Seal and his colleagues in Melbourne, Australia, failed to find the same result with an almost identical procedure.66 More recently, a similar study was conducted by Gildas Brebion at the New York State Psychiatric Institute and his colleagues, who did find evidence that hallucinators were more likely than non-hallucinating patients to misattribute to an external source words that they had produced themselves.67
Monitoring current perceptions
One reason why the experiments we have considered so far may have been inconclusive is that they have focused on source monitoring for memories. However, hallucinations presumably involve errors in the source monitoring of current perceptions. The measurement of these kind of errors is particularly challenging, and was the focus of my first experimental study of hallucinations, which utilized a technique known as signal detec
tion theory (SDT).
SDT is a mathematical model of perceptual judgement. When I was a psychology student, it was one of my least favourite topics in the undergraduate curriculum, partly because I found its mathematical realization hard going. It was with some dismay that I later concluded that SDT was exactly what I needed to test the source-monitoring theory of hallucinations. In order to avoid generating the same level of distress in my readers, I will try to outline the basic idea behind the theory in a non-mathematical form (see Figure 14.4 for a diagrammatic explanation).
SDT attempts to explain how we decide that we really are perceiving something. Although this is precisely the problem confronted by the person suffering from hallucinations, the theory was initially designed to account for our ability to detect weak stimuli (or ‘signals’ in the jargon of the theory) when working with communication devices such as telephones or radios. Imagine that you are in a darkened room. You think that you have seen a chink of light but you cannot be certain. Before the 1960s, psychologists would explain what happens in these circumstances by ‘threshold theory’, according to which the light is perceived only if it is above a certain magnitude. However, attempts to measure the thresholds for particular kinds of stimuli (lights or noises) failed to yield consistent results, so a different kind of approach was required.
SDT proposes that our ability to detect a signal in these kinds of circumstances depends on two factors. The first factor is perceptual sensitivity, which might be thought of as the efficiency of our perceptual system. The greater our perceptual sensitivity, the easier it is to detect signals. The second factor is known as perceptual bias, which can be thought of as our willingness to assume that the signal is present.
Figure 14.4 Relationship between choices made in signal-detection experiments and source-monitoring judgements. High perceptual sensitivity results in a high proportion of hits and correct rejections and few signal-detection errors (misses and false alarms). Increasing bias towards detecting signals results in more hits and fewer misses, but at the expense of an increase in false alarms and a decrease in correct rejections (from R. P. Bentall (2000) ‘Hallucinatory experiences’, in E. Cardena, S. J. Lynn and S. Krippner (eds.), Varieties of Anomolous Experience: Examining the Scientific Evidence. Washington, DC: American Psychological Association).
Unlike perceptual sensitivity, which can only be affected by fairly gross physiological changes, perceptual bias can change from moment to moment, and is particularly affected by our beliefs and expectations.
Despite the dangers of analogy-overload, I will resort to one more in the hope of making these ideas clear. In a typical signal-detection experiment, participants listen to bursts of white noise, some of which contain a signal (for example, a brief voice). After each burst, the participant is asked to say whether the signal was present or not. In these circumstances, the participant is behaving like a Cold War radar operator whose job is to detect a possible incoming Soviet air strike. Four things can happen. The person can detect the signal (the Soviets attack, the radar operator spots them and they are intercepted), which is known as a ‘hit’. The person can also correctly fail to report a signal (a ‘correct rejection’). However, there are also two kinds of possible error: a ‘miss’ (the Soviets strike without being spotted, so that civilization as we know and love it is laid to waste) and a ‘false alarm’ (the radar operator incorrectly believes that the Soviet air force is on its way, and a lot of aviation fuel is wasted before the error is discovered). Pursuing this analogy further, an improvement in perceptual sensitivity is equivalent to an upgrade of the radar system, resulting in an increase in hits and correct rejections and a decrease in misses and false alarms. However, a change in perceptual bias involves a greater willingness to believe that the signal is present without any improvement in sensitivity, as if the radar operator’s haste to sound the alarm is influenced by the prevailing political climate. An increase in bias (for example, during periods of political tension such as the Cuban missile crisis) will produce an increased probability of a hit (if the Soviets come they will be detected) but also an increase in false alarms (the nervous radar operator will be more likely to scramble the intercepting aircraft unnecessarily). A decrease in bias (perhaps following a period of détente) will reduce the risk of false alarms (aviation fuel will be conserved) but increase the risk of misses (if the Soviets do attack unexpectedly, they have a better chance of getting through).
By now, it should be obvious that false alarms are the hallucinations of signal detection theory. The theory therefore provides us with a method for determining whether hallucinations are the product of perceptual problems (which would be evident as impaired perceptual sensitivity) or biased source monitoring. Peter Slade and I therefore conducted two signal-detection experiments, one comparing hallucinating and non-hallucinating patients and one comparing students who had high scores on a hallucination questionnaire with students who had low scores. Both involved participants listening to many brief episodes of white noise in order to detect occasions when a voice was also present in the background. Applying a bit of mathematics to the participants’ judgements allowed us obtain separate measures for sensitivity and bias. The results, which are shown in Figure 14.5, show clearly that the hallucinating participants differed from their respective control participants in bias but not in sensitivity.68
This finding was later replicated by Peter Rankin and Pierce O’Carrol in their study of Liverpool students,69 and also, using a slightly different methodology, in an experiment conducted with patients by Gildas Brebion and his colleagues.70 Overall, these observations are consistent with our hypothesis that hallucinations arise from peculiar judgements about what is real, rather than from deficits in patients’ perceptual systems.
Other researchers have devised simpler methods for measuring source monitoring for perceptions. My colleague Tony Morrison has carried out a series of experiments in which participants were asked to provide associations for words which were either emotionally positive (for example, ‘courage’), emotionally negative (for example, ‘crazy’) or neutral (for example, ‘bookcase’). Immediately after responding to each cue word, the participants were asked to rate their responses for ‘internality’ (‘How much was the word that came to mind your own?’). The hallucinating patients gave much lower ratings (stating that the words that came to mind did not seem to be their own) than the other participants, especially when the cue words were either positively or negatively emotionally salient.71 In a recent study that followed up these findings, Morrison found that these effects were more marked when patients were first encouraged to focus their thoughts on themselves (by first being required to make up a short story about themselves) and were less evident when the patients’ attention was first directed elsewhere (by asking them to make up a short story about someone else).72
Figure 14.5 Bias (B) and sensitivity (P(A)) scores from signal-detection experiments conducted by Bentall and Slade (1985). The left panels show comparisons between students scoring high and low on the Launay-Slade Hallucination Scale (LSHS-A), and the right panels show comparisons between hallucinating and non-hallucinating schizophrenia patients. Note that low B scores indicate a strong bias towards detecting signals.
An even more direct measure of immediate source monitoring has been developed by Louise Johns and Phil McGuire at the Institute of Psychiatry in London. Johns and McGuire had participants speak into a microphone. Their speech was immediately played back to them through earphones after being distorted electronically. They were then asked to say whether the speech they were hearing was theirs or someone else’s. In this experiment, hallucinating patients, in comparison with non-hallucinating patients and ordinary people, were much more likely to say that their voice belonged to someone else, especially if the content of what was being said was derogatory.73
This progress in establishing that hallucinating patients experience difficulty with source monitoring has been matched by progress in identifying the ne
ural structures involved. Of course, we should not expect these to be the same structures as those involved in generating the inner speech which hallucinating people mistake for a voice. In a study carried out in Toronto, Canada, Henry Szechtman and his colleagues PET scanned ordinary people, some of whom were highly hypnotizable and had been selected for their ability to experience hypnotic hallucinations. Blood flow to different regions of the brain was measured as the participants listened to speech, imagined speech and, in the case of the hypnotizable participants, hallucinated a voice. A region in the area on the right side of the cerebral cortex known as the right anterior cingulate (also known as Brodmann area 32)* was found to be active in the hypnotizable participants when they heard or hallucinated a voice, but not when they imagined hearing it. The same conditions did not cause similar activations in the non-hypnotizable participants. Szechtman and his colleagues have concluded that the right anterior cingulate may contain neural circuits that are responsible for deciding whether experiences originate from the external world.74
Madness Explained Page 43