Professor Chris Roe at the Anomalistic Psychology Research Unit on 25 November 2014
Approximately two thirds of all reported spontaneous cases of extrasensory perception (ESP) have occurred while the experient was in an altered state of consciousness, particularly while dreaming (Rhine, 1962). Early experimental attempts at the Maimonides sleep laboratory to elicit ESP by monitoring participants and waking them during REM sleep were remarkably successful, with an overall hit rate after 450 trials of 63% (where MCE = 50%), that has odds against chance of 75 million to one (Radin, 1997). Attempts to replicate this promising finding have been limited by the prohibitive costs of maintaining a sleep laboratory and difficulties in recruiting participants for studies that require them to stay overnight. However, some researchers have continued to investigate dream ESP using cheaper and less labour-intensive methods.
In this presentation I will outline some of the methods adopted by teams working post-Maimonides and consider recent reviews of this database (Roe & Sherwood, 2009; Storm, Tressoldi, & Di Risio, 2010) to draw conclusions as to whether an effect has been demonstrated. I will pay particular attention to conceptual and methodological weakness in the approaches taken (cf. Roe, 2009a, 2009b) and make recommendations for future work.
The strongest evidence, it turns out, falls way short of demonstrating that there is even a phenomenon of dream telepathy to research (and there is no physical model for how such communication might work). That said, Roe is certainly at the sensible end of parapsychology, and, as well as being a good speaker, he gave a lesson in how to apply sceptical thinking in scientific research, asking the sorts of questions sceptics ask, working hard to provide naturalistic explanations and not jumping to the woo conclusion at the first hint of an effect.
He began by noting that personal experience is the primary driver for belief in the paranormal, which also makes it difficult to investigate using scientific methods. However, he believes there is no such thing as an unscientific claim. Before coming on to his own work, he looked at the history of studying spontaneous cases of ESP and dreams, which tend to involve those who are emotionally close to us and which reflect serious rather than trivial events.
Louisa Rhine documented these cases but knew that they were not evidential, since they were experiences in the real world and not in controlled situations. Their usefulness lay in that they might lead to experiments that were designed to detect an effect if there was something there to detect.
I had just been reading a similar example of how to approach difficult-to-interpret experiences in philosopher Daniel Dennett’s essay Intentional Systems in Cognitive Ethology. He became interested in how primatologists study vervet monkeys in the wild, and in what was going on in the minds of these monkeys as they gave different alarm calls to different predators (Dennett 1989:239):
How much of a language, one wants to know, do the vervets really have? Do they really communicate? Do they mean what they say?
At a conference on “Animal Mind—Human Mind” Dennett was pleased to discover that his “impromptu exercises in applying the intentional stance to their research problems did in fact generate some novel testable hypotheses, designs for further experiments, and methods of developing interpretations” (269–70).
The following is most relevant for understanding the difference between simply collecting stories of unusual experiences and conducting a rigorous scientific investigation into what may be going on (271):
A few, I gather, have mistaken my advocacy of the Sherlock Holmes method of creating (and controlling) “anecdotes” for a wholesale defense of casually obtained advocates as evidence! So I should reiterate and emphasize the point I was trying to make: a unique or one-off bit of behavior is useless as evidence for an attribution of an intentional state (however valuable to the researcher as a hint for further experiments) unless it can be shown to be an otherwise unlikely behavior, provoked just by the conditions that would provoke, in a rational agent, beliefs and desires which would render the unlikely behavior rational. Showing this always requires running controlled experiments. The method I was extolling was not a substitute for experimentation, but a way of seeing which experiments needed to be done.
Back to Chris Roe, who gave several examples of the common experience of a dream giving a clear sense of déjà vu, and seeming to confirm a subsequent event. The circumstances of occurrence include minimally demanding physical and mental activity, and the subject is not really thinking about anything in particular (that would cover most sleep states!).
The difficulty is controlling for counter explanations: any sceptical scientist will have many alternative, naturalistic explanations. For example, there are lots of people in the world, who have lots of dreams every night of the year, and so it is not surprising that some of these seem to correspond to events that later occur in the real world.
Roe listed five problems:
- Finding a chance baseline for coincidences.
- Self-selection of cases (only those with something interesting come forward, so we’re cherry picking the best examples).
- Shared antecedents.
- Sensory leakage (while listening to the radio you hear an old song that, unconsciously, reminds you of an old friend, who then rings you up, and who had been listening to same song, which reminded him of you).
- Convergence during recall (we selectively remember events that fit).
We need to “hoik the phenomenon” out of the open, noisy system that is the real world and into the laboratory, where we can impose some controls and rule out these explanations. Roe described several ways in which more objective outcomes could be arrived, by pre-specifying and accurately reporting all trials, by using random selection of targets, by ensuring that barriers are in place to prevent “normal” means of communication from being possible, and by producing independent records.
The best-known dream ESP research was carried out at the Maimonides Medical Center in New York. A person tries to communicate a target image to the sleeping participant. Researchers have no idea what the target is, and record what the dreamer experiences. The problem is that we can be very creative in looking for correspondences in a metaphorical way, and this itself creates enormous difficulties when interpreting the results of these kinds of experiment.
How do we evaluate dream correspondences? Consider the following dream:
I can hear water all around me — I have a sense of being in the open air — I can see clouds above me — there’s green…
Roe showed us several possible target images, including an image of the sinking Titanic and a spider in web. These couldn’t be much further apart when we use our visual perception, but the water that is cold and vast around the Titanic could also be taken as droplets of dew on the spider’s web, and so on. We have to be able to take account of this flexibility of interpretation, by showing decoys at the same time as the true target, and then quantifying the correspondences using rank ordering.
Later studies were less successful than Maimonides studies at finding effects that were better than chance.
Noise reduction model: the idea is to damp down ordinary perception and reduce the information entering the system via the usual sensory modalities, to draw attention away from ordinary perception. Such “ganzfeld stimulation” encourages a hypnagogic state in which the mind creates internal imagery, and the subject then reports their ongoing experience.
Ganzfeld has a chequered history of success. Ray Hyman disagreed with Chuck Honorton, attributing the above-chance performance to methodological flaws. For example, an effect is not robust if the statistics change depending on the inclusion or exclusion of a single experiment. Better-designed experiments were needed and Hyman and Honorton collaborated.
There are still serious methodological challenges, such as the one-size-fits-all assumption (some people fall sleep, others don’t) and the basic fact that our consciousness is constantly in a state of flux as our attention varies. We need to show a difference between ganzfeld and ordinary consciousness, and Roe believes that focusing on altered states of consciousness is a useful strategy for eliciting above-chance scoring.
Roe gave an example of a striking miss, which illustrates the danger of reading much more meaning than is there: Dr David Luke reported experiencing the distinctive opening credits of Trainspotting, which would have been a striking correspondence, except that this wasn’t the target.
After 150 years of failure to detect any paranormality, there is a strong tendency to assume anyone involved in the area must be a crank. While we may think there are better questions for science to be asking, Chris Roe keeps at the sensible end of parapsychology. He recognizes that this human capacity — “if it exists” — would have its own evolutionary history, and he acknowledges that test–retest reliability of subjects is atrocious.