Thursday, May 19, 2011

Free Will, Neurobiology, and Presentiment - response to a posted comment


Since my recent post occasioned the first comment being posted on The Mystical Positivist blog site since its inception last December, I want to respond to the points it raises. Also, it presented me with a considerable challenge in that I needed to familiarize myself with the arguments pro and con on the use of Bayesian statistics in the analysis of Psi data. 

The commenter posted a link to a paper entitled Why Psychologists Must Change the Way They Analyze Their Data: The Case of Psi. The authors, who are well known in their development of Bayesian statistical analysis, apply their techniques to a paper published in 2011 by Dr. D.J. Bem called Feeling the future: Experimental evidence for anomalous retroactive influences on cognition and affect. This is a fairly controversial paper in which Bem reports on nine experiments that he conducted in "retro-causation" that show findings not unlike the paper I cited in my original post by Dr. Dean Radin. Bem used a frequentist analysis to claim that eight of his nine experiments showed strong evidence (as measured by p< 0.05) of the validity of his alternative hypothesis (i.e. the affirmation that a retro-causal effect was present and the rejection of the null hypothesis which says that there is not effect). In the paper my commenter cites, Wagenmakers et al. claim that by using a Bayesian approach which takes into account assumptions about the prior probabilities of both the null hypothesis and the alternative hypothesis, they can reanalyze the data and show that Bem's claims of significance are wildly overstated.

It turns out that Bem and two co-authors (who have previously authored a number of Bayesian analyses of Psi data) published a response to Wagenmakers et al. called Must Psychologists Change the Way They Analyze Their Data? A Response to Wagenmakers, Wetzels, Borsboom, & Van der Maas (2011). They argue that Wagenmakers et al. used the subjectivity inherent in the selection of the Bayesian distribution of the prior probability for the alternative hypothesis to inappropriately bias their analysis toward the null hypothesis. In very simple terms (but terms that do not really do justice to the subtlety of the argument here), they state that Wagenmakers et al. are demanding such a large effect size from the prior probability distribution of the alternative hypothesis that their resulting analysis of Bem's data drives the Bayes factors (the ratio of the marginal probability for the null hypothesis given the data to the marginal probability of the alternative hypothesis given the data) to relative insignificance (technically they use a diffuse Cauchy prior distribution). Bem et al. utilize a different probability distribution for the alternative hypothesis based on prior knowledge of similar psychological studies of priming effects on reaction time, and they demonstrate that a Bayesian analysis of Bem's original data yields strong affirmative results for the alternative hypothesis in five of the nine experiments. They go on to do a Bayesian meta-analysis of the aggregate dataset and show a Bayes factor of 13.6K (extremely strong confirmation of the alternative hypothesis).

Lest we leave it there, Wagenmakers et al. offer a response to the response here. In looking into this question, I found a useful background paper on the subject by authors with no particular philosophic axe to grind that explains the issues behind conducting a Bayesian t-test. So based on all of this material, I have drawn a number of conclusions - some of which are actually different from where I started from:
  • The most important lesson here is that the ordinary criterion for statistical significance (p<.05) in a frequentist analysis is way too generous and does not give sufficient weight to the null hypothesis. Stated differently, I will no longer lazily think that a p<0.05 result in a Psi test is necessarily significant. I will however continue to think that a p<0.001 is significant (such as in Radin's paper). Fortunately for my previously held opinions, there are plenty of solid studies out there that show results of this magnitude.
  • The critiques in Wagenmakers et al. against exploratory versus confirmatory testing of data are excellent points and important ones to watch for in the Psi debate. Effectively they say that if you use a data set to identify potential hypotheses and then run a statistical analysis testing for those hypotheses against the same data set, then you will be biasing your results away from the null hypothesis. Unfortunately, they miss the point that four or five of the nine Bem experiments in the original paper were strictly replications and not exploratory in nature. And as far as the paper I cited in my original blog post, Radin reports on an initial study and three strict replication studies that do not attempt to extract new hypothesis out of the data.
  • To cite Wagenmakers et al. as an indictment of all data analysis in the field of Psi research and to use that conclusion as a justification of not examining more rigorously the large body of published data on the subject smacks of confirmation bias. The appropriate application of Bayesian statistics to hypothesis testing is far from settled even outside of the field of Psi. The virtue of the Bayesian approach is that it forces one to be explicit about the subjective choices one makes in determining the prior probabilities to use in an analysis. Reasonable people can disagree on this and that disagreement can make the difference between a significant and not so significant result, but at least one's biases are out in the open. The more common frequentist approach has a built in bias against the null hypothesis that is not so obvious. In just this, Wagenmakers et al.'s larger critique of how statistical analysis is used in psychology and social sciences is a valid and strong one.
To put a bow on this particular discussion, I found a video on Dean Radin's blog of a recent debate at Harvard that featured Daryl Bem, Jonathan Schooler, and Sam Moulton. Bem is the author of the retro-causation survey discussed in Wagenmakers et al. Jonathan Schooler is the researcher featured in the articles recently in the New Yorker on the Decline Effect. Sam Mouton is a Harvard Psychologist who started out as a parapsychologist but later rejected the field when he consistently got null results in a number of studies he conducted. Bem claims to have started out as a skeptic but was led by his results over a number of studies to believe there was something there in his tests. It is a good discussion that raises some of the challenges in this kind of research. Schooler comes out in this discussion as a Psi researcher and discusses the impact of the Decline Effect on replication studies in Psi (and other fields). 

Sunday, May 15, 2011

Free Will, Neurobiology, and Presentiment - a Defense of the Soul

Yesterday we had an interesting discussion on the Mystical Positivist radio show with Dr. David Weisman, a neurologist and author who has written about how current findings in neurobiology falsify the idea of a unified soul governing the human organism (podcast here). Likewise I have discussed this issue in the context of Sam Harris' book, The Moral Landscape - How Science Can Determine Human Values in an early entry on this blog site (post here). The basic idea is that fMRI studies definitely show that activity in the brain's motor regions can be detected as much as 350 milliseconds to 10 seconds before the conscious decision to move has entered a test subject's field of awareness. The general conclusion that neuobiologists such as Harris or our guest, neurologist Dr. Weisman, tend to draw is that these results demonstrate that however much we think we are the authors of our own actions, physical processes below our level of awareness are determining how our physical system manifests. The inescapable conclusion is that what we subjectively experience as our "self" is in reality a fiction conjured up post hoc by the brain to maintain the illusion of an independent being having agency over the functioning of the body.

As I have written in the past, it is hard to escape this conclusion regardless of your spiritual persuasion if one is being intellectually honest. But I made a connection this morning that I had not quite registered before that provides an interesting and challenging response to this neurobiological argument against free will. It involves the fairly large and independently reproduced data sets of what consciousness researcher, Dean Radin, calls "presentiment."

Presentiment refers to measurable physiological behavior that anticipates a future stimulus regardless of a subject's conscious awareness of the physiological shift. Radin uses this term to distinguish presentiment from precognition which refers to a subject's conscious awareness of a future physical event. From 1997 through 2000, Radin conducted a rigorously designed experimental test of presentiment with three follow-on replications, all of which demonstrate to a large level of statistical significance (p=.001) that the body's nervous system can accurately anticipate a randomly generated future stimulus several seconds before the stimulus is presented in time (full PDF paper here).

The basic idea of Radin's presentiment studies is as follows. Electrodermal activity (EDA) of test subjects is measured before and after the subjects are presented with an image that depicts either a calm image (e.g. nature scenes) or an emotionally charged image (e.g. violence, accidents), and the EDA data are analyzed for differential effects between the image sets. In the experiments, test a subject sits two feet in front of a computer monitor that displays a blank screen. The subject then presses a button and 5 seconds later a random number generator selects an image to display from a repository of several hundred images classified across a spectrum of calm to highly emotionally charged. The random image is presented for 3 seconds and then the screen goes blank again for 10 seconds.

Depending on the experiment, subjects would engage in 30 to 40 replications of pressing the button to gather data for a single session. In these individual sessions, the system was designed not to show the same image twice to eliminate effects due to recognition or expectation. The EDA data measured in these specific experiments were skin conductance level (SCL). A continuous physiological record would be taken in each test extending from the point the subject presses a button, through the 5 seconds before the system randomly selects an image, through the 3 seconds during which the image is displayed, to the 10 seconds after the  image has been replaced by a blank screen. These data were then analyzed to test whether a statistically significant differential response could be observed between the SCL prior to the presentation of the image for calm images versus emotionally charged images. Typical results appear as follows:


In the graph above, -5 seconds is the point at which the subject presses the button to initiate the test, and 0 seconds is the point at which a random number generator determines which image (from a set of calm or emotional images) is presented to the subject. After the image is presented for 3 seconds, we see a strongly differential change in SCL when the subject sees an emotional versus a calm image. That is expected. But the radical thing in these data is that there is a statistically significant differential effect between the calm and emotional images 5 seconds prior to the random selection of the image yet to be viewed. This bears repeating: the SCL measurements were different 5 seconds before a randomly selected image was presented only when in fact it turned out that the image was emotionally charged. The subject's body was accurately responding in advance to a future stimulus all at a level below that of the subject's conscious awareness.

If you are sufficiently startled by these results, I encourage you to read Radin's paper in full. He details how his experimental setup anticipates the possible objections to the methodology, he cites an independent review by skeptics of his protocol, and he cites independently conducted experiments that replicate his findings. I am personally convinced of the integrity of the results and the researcher here to move on to the question of what these results might say about my earlier discussion on neurobiology and the brain.

These presentiment results change the character of the argument about free will and the brain considerably. The results cited by Harris show, for instance, that 350 milliseconds prior to a conscious willing to action of a physical movement, the motor centers in the brain are activating and preparing for the movement. The results cited by Radin show that as much as 5 seconds prior to presentation of a randomly selected stimulus, a subject's autonomic activity is responding differentially according to the character of the yet to be presented stimulus. Taking these results together, it is not a far stretch to suggest that the moment of the arising in my field of awareness of my conscious intention to move my arm might function similarly to the random presentation of an image in Radin's presentiment experiments. The fact that we can measure activation of my brain's motor center seconds before my intention to move my arm registers in my field of awareness can be understood as an example of presentiment. Yes my nervous system is responding in advance of my intention to move my arm, but this does not necessarily mean that my conscious experience of intentionality is a fiction. The results of Radin's presentiment experiments support as plausible the interpretation that a "soul's" coupling to the body's nervous system models like the random presentation of emotionally charged images to a test subject. However counter-intuitive to our notions of causality, the nervous and autonomic systems of the body seem to demonstrate the ability to couple across time and anticipate as yet undetermined stimuli. Conscious intentionality can be modeled as one such future stimulus.

Lest I be accused of overreaching here, I am not claiming that Radin's presentiment results prove the existence of a unified soul. But I am claiming that Radin's data negate the argument described above that the results from neurobiology necessarily falsify the idea that we have free will of some sort. The retrocausal anticipation in our nervous systems to a yet to be determined future event (e.g. the formation in consciousness of an intention to move) can be seen as a more general and strange property of our nervous systems. It does not prove that what we construe as our conscious intention is a fiction asserted post hoc by the brain system.

The Mystical Positivist - Radio Show #22 - 14MAY11

The Mystical Positivist is now a weekly radio show on KOWS-LP 107.3 FM, Occidental, CA. Listen live on Saturday evenings from 4:00 - 6:00pm, PST, via the web at KOWS Live Stream.

This week's podcast features:
  • A discussion with David Weisman M.D. Dr. David Weisman is a neurologist in PA, and author of "Buddhism and the Brain" and "From Divided Minds, A Specious Soul" in Seed Magazine. He discusses data that demonstrate that although humans have a strong idea their minds have unity and self control, neurology and neuroscience indicate that reality is nothing like that.

More information about Dr. David Weisman's work can be found at:

Monday, May 9, 2011

The Mystical Positivist - Radio Show #21 - 07MAY11

The Mystical Positivist is now a weekly radio show on KOWS-LP 107.3 FM, Occidental, CA. Listen live on Saturday evenings from 4:00 - 6:00pm, PST, via the web at KOWS Live Stream.
This week's podcast features:
  • Discussion of Landmark Forum (a descendant of EST)
  • Work of the Week: Separating out the events of one's life from the interpretations
  • A discussion with Mirka Knaster, Ph.D. Mirka Knaster is the author of Living This Life Fully: Stories and Teachings of Munindra, a book about the Bengali meditation master who was a grandfather of the vipassana/​mindfulness movement in the West and who taught many of today's most prominent Western dharma teachers. She interviewed nearly 200 people around the world for their down-to-earth yet inspiring poignant and humorous remembrances of someone who embodied the qualities of awakening and who believed it was possible for all of us to cultivate them. The book also draws on discussions with Munindra before his death in 2003, early talks Munindra gave in the U.S., and includes rare photographs. Shambhala is the publisher (October 2010). Mirka collaborated with Robert Pryor on this project.
More information about Mirka Knaster's work and interests can be found at:

Wednesday, May 4, 2011

The Mystical Positivist - Radio Show #20 - 30APR11

The Mystical Positivist is now a weekly radio show on KOWS-LP 107.3 FM, Occidental, CA. Listen live on Saturday evenings from 4:00 - 6:00pm, PST, via the web at KOWS Live Stream.
This week's podcast features:
  • The Creation of music through the instrument of the body
  • Work of the Week: The Tayu New Day Exercise
  • A discussion with Jisho Warner Roshi. Jisho Warner Roshi, head of Stone Creek Zen Center of Sebastopol, is a Buddhist priest and Zen teacher in the Soto Zen School. Warner Roshi trained in the United States under Tozen Akiyama and Dainin Katagiri Roshis, and in Japan under Shundo Aoyama Roshi. She received authorization to teach and dharma transmission from Tozen Akiyama Roshi.
More information about Jisho Warner Roshi's work and interests can be found at: