Edited by: Rufin VanRullen, Centre de Recherche Cerveau et Cognition, France
Reviewed by: Rufin VanRullen, Centre de Recherche Cerveau et Cognition, France; Arnaud Delorme, Centre de Recherche Cerveau et Cognition, France; Sofia Gameiro, Cardiff University, UK
*Correspondence: Julia Mossbridge, Department of Psychology, Northwestern University, Evanston, IL 60208, USA. e-mail:
This article was submitted to Frontiers in Perception Science, a specialty of Frontiers in Psychology.
This is an open-access article distributed under the terms of the
This meta-analysis of 26 reports published between 1978 and 2010 tests an unusual hypothesis: for stimuli of two or more types that are presented in an order designed to be unpredictable and that produce different post-stimulus physiological activity, the direction of pre-stimulus physiological activity reflects the direction of post-stimulus physiological activity, resulting in an unexplained anticipatory effect. The reports we examined used one of two paradigms: (1) randomly ordered presentations of arousing vs. neutral stimuli, or (2) guessing tasks with feedback (correct vs. incorrect). Dependent variables included: electrodermal activity, heart rate, blood volume, pupil dilation, electroencephalographic activity, and blood oxygenation level dependent (BOLD) activity. To avoid including data hand-picked from multiple different analyses, no
Predicting the future is an essential function of the nervous system. If we see dark clouds and smell a certain scent in the air, we predict that rain is likely to fall. If we hear a dog bark, we predict that we will see a dog nearby. These everyday predictions are based on experience (e.g., memory) and perceptual cues. If even without experience and perceptual cues we could somehow prepare for important imminent events by activating the sympathetic nervous system prior to such events, this skill would of course be highly adaptive. More than forty experiments published over the past 32 years examine the claim that human physiology predicts future important or arousing events, even though we do not currently understand how such a thing could be accomplished. This meta-analysis examines a subset of these experiments allowing us to test the hypothesis that seemingly without experience and perceptual cues, human physiological measures anticipate what seem to be unpredictable future events by deviating from a baseline before an event occurs, in the same direction that they will continue to deviate after that event occurs. This is a controversial but important hypothesis. Thus, although there is no known mechanism for the effect reported in such studies, the implications of such an effect are far-reaching enough to justify a careful meta-analysis.
The studies we include in this meta-analysis make direct comparisons between pre-stimulus physiological activity measures using paradigms that produce a contrast in post-stimulus physiological activity between responses to stimuli from different categories. Two paradigms are used: (1) randomly ordered presentations of arousing vs. neutral stimuli, or (2) guessing tasks for which the stimulus is the feedback about the participant’s guess (correct vs. incorrect). In arousing vs. neutral stimulus paradigms, participants are shown, for example, a randomly intermixed series of violent and emotionally neutral photographs on each trial, and there is no
It has been known for some time that arousing and neutral stimuli produce somewhat different
The primary value of this meta-analysis is that it tests a hypothesis that is different from those examined in most of the studies included in it. For the included studies, the hypotheses were, for the most part, bidirectional – namely, that the data would reveal a significant difference between physiological activity preceding two (or more) seemingly unpredictable stimulus types, regardless of the direction of that difference. A meta-analysis of these data would certainly be significant, as any deviation between the two physiological activity measures would produce a positive effect size (ES), in favor of a hypothesis that there is any difference between the measures. In contrast, adopting a more conservative approach, ours is a directional hypothesis: for paradigms producing post-stimulus effects differing between two or more stimulus types, and with randomized and theoretically unpredictable stimulus orders, the pre-stimulus difference between those same stimulus categories will have the same sign as the post-stimulus difference. In other words, we use meta-analytic techniques to test the hypothesis that the direction of pre-stimulus activity is predictive of the direction of post-stimulus activity, even when the stimulus category itself seems to be unpredictable
We took a relatively inclusive approach to ensure that all studies with negative and null results were included along with those supporting the hypothesis. A study was defined as a unique (not previously reported) examination of physiological responses to stimuli or events in one group of human participants; a report could include more than one study. Included studies were required to provide quantitative data or descriptive statistics reflecting physiological measures recorded during a period of time preceding stimulus presentation. This requirement excluded examinations of post-stimulus emotional responses that did not also report pre-stimulus activity. Further, only studies that marked stimulus event times using automatic (software) methods were included. Prospective (not
Several potential moderators of apparent pre-stimulus differences were examined, including: study quality and whether an expectation bias analysis was performed. Because several authors have suggested that unexplained anticipatory activity is stronger among women than men (McCraty et al.,
All three authors were familiar with the unexplained anticipatory physiology literature, but to ensure consideration of studies about which we were not familiar, we performed broad web searches for studies reported between 1978 and 2010. We conducted the searches using PubMed, PsycInfo, Google Scholar, and the OAIster database from OCLC, which is useful for such gray literature searches. We also searched the archives of the
The first two authors independently coded each of the studies before analyzing the results of the meta-analysis. The first author coded the studies before seeing the individual ES calculated for each study by the second author (see
The first author coded the sign of the ES in all studies with a second pass (quality check) by the second author. The sign of the ES is one of the most critical parameters to be coded in any meta-analysis that tests a hypothesis that differs from some or all of the included studies. The sign of the ES could not be taken as the original sign given to the
We coded study quality weighted by three factors: peer review, use of hardware RNGs, and expectation bias analysis. Peer review is subjective and can be unreliable, and therefore is at best a guess about study quality (Casati et al.,
Initially, the first and second authors subjectively rated study quality, then chose a numeric ranking for each study. Intercoder agreement was fair, achieving a significant correlation between quality rankings from the first and second authors:
Each study reported physiological dependent variables for each of the stimulus categories the authors were comparing (Table
As is common in the analysis of psychophysiological data, in all included studies except one (Tressoldi et al.,
It is important to note that when determining the sign of the ES, the same measure(s) used to calculate the pre-stimulus effect was (were) used to determine the direction of the post-stimulus effect. In most cases, the direction was obvious from group data presented in figures or tables. In other cases, direction was determined as described above (see
We calculated a unique ES for each study, based on
In terms of the original
Standard error (SE) was calculated for each ES derived from baselined data with the formula
In terms of models, the fixed-effect model is based on the assumption that the true ES is the same for all studies, while the random-effects model is based on the assumption that the true ESs differ across studies, and are sampled from a distribution comprising multiple different ESs. Both models are plausible here because we are not sure about the underlying distribution. Our heterogeneity analysis (see
To test for “filedrawer effects” resulting from possible publication bias and/or selective reporting, we used two standard methods [classical fail-safe (Rosenthal,
The statistical power of this meta-analysis is 0.90 assuming the true ES = 0.01 and variance = 0.002 (the observed variance in the random effects model; Hedges and Pigott,
One explanation for the anticipatory activity reported in the included studies is that researchers performed multiple analyses to find the dependent variable that produced the effect. This approach is more likely when the dependent variable arises from fMRI or EEG data, because multiple spatial and temporal locations can be used to define the dependent variables. However, electrodermal activity is a physiological endpoint that provides fewer opportunities for multiple analyses, because: (1) it offers only one spatial position (the point at which the electrodes were attached), and (2) the response time course of skin conductance measures is very sluggish (2–3 s), so that manipulating temporal parameters such as the duration of the pre-stimulus and baseline periods could influence the size of the result but it would only alter its direction if the two conditions produced phasic physiological responses that differed in phase during the pre-stimulus period. Most of the ESs of the studies included in this meta-analysis are based at least partially on electrodermal data (21 out of 26 studies, see Table
Our search strategies retrieved 49 published and unpublished studies that initially seemed to fit our constraints (see
The overall ES for all included studies is small, while the overall statistical significance is high [fixed effect: overall ES = 0.21, 95% CI = 0.15–0.27,
Study ESs ranged from −0.138 to 0.67. Tests of homogeneity reflected relatively low heterogeneity (as defined in Huedo-Medina et al.,
Although heterogeneity is low, we chose to examine several potential moderators of the effect. First, we examined study quality. Studies were scored for quality using a scoring procedure that encompassed level of peer review, type of RNG used, and whether or not an expectation bias analysis was performed (and if it was, whether expectation bias could have explained the results; see
As pointed out previously (see
Finally, neither the ratio of male-to-female participants nor the number of trials in each study were related to the ES. The correlation between male-to-female participant ratio and study ES was not significant (Pearson
Given such a surprising result, it is critical to investigate the potential influence of reporting bias. Skeptical mainstream scientific researchers would be unlikely to under-report negative results, as the effect examined here is controversial enough that reports of supporting evidence are not likely to further a mainstream scientific career. In contrast, there may be a sub-community of paranormal researchers who could be tempted to file away null or negative results. We think this unlikely for two reasons. First, many paranormal researchers were not investigating the directionally dependent hypothesis examined by this meta-analysis (see
On the other hand, it is still possible that the highly significant overall effect reported here could be explained by a “filedrawer effect” if negative findings were under-reported for some reason. To examine this possibility, we performed a trim-and-fill analysis (Duval and Tweedie,
To further investigate the possibility that a persistent bias toward underreporting negative or null results could explain the significance of the overall effect, we used two methods to determine the number of contrary unpublished reports that would be necessary to reduce the level of significance to chance (
As discussed previously (see
The available data support the hypothesis tested by the current meta-analysis. Specifically, for paradigms producing post-stimulus physiological effects that differ among two or more intermixed and randomized stimulus classes, the group mean difference between physiological responses accompanying these stimulus classes seems to be in the same direction before and after stimulus presentation. For the 26 studies that fit our inclusion criteria (see Table
These results seem not to be an artifact of poor experimental design, as higher-quality experiments that addressed known methodological concerns (randomization and expectation bias analysis) produced a quantitatively if not significantly higher overall ES and level of significance than lower-quality studies. Further, the unexplained anticipatory effect examined here seems not to be due to expectation bias, as the overall effect was still highly significant when we included only those studies that reported expectation bias analyses and found that expectation bias could not explain the effects. Additional examination of other potential moderators of the effect revealed that the male-to-female ratio among study participants was not correlated with study ES; neither was the number of trials performed by each participant in a study correlated with ES.
Calculations to determine the number of contrary unpublished reports that would be necessary to reduce the level of significance to chance provided a fail-safe number of reports of 87 for the most conservative estimate. Seven laboratories contributed to the experiments included in this meta-analysis. Five more laboratories produced data that were related to our question, and many of them reported significant anticipatory effects, but they were excluded from this meta-analysis (see
The results of the overall analysis are surprising, especially because in order to be inclusive we have combined data from multiple experimental paradigms and physiological measures that fit our constraints (see
In summary, the overall effect is small but statistically significant, seems not to be due to expectation bias, and is unlikely to be due to publication bias. Thus there seems to be a small, predictive anticipatory physiological shift in the seconds preceding apparently unpredictable stimuli. What could explain this effect?
One trivial explanation for the effect is that the compared stimulus categories did not differentially affect participants’ physiology, so it follows that the same random differences between physiological traces preceding the presentation of the two different stimulus classes also occur after stimulus presentation. However, our inclusion/exclusion criteria specifically required that all studies included in the meta-analysis use tasks presenting intermixed and randomized stimulus classes that produce appreciable physiological post-stimulus effects (see
A more reasonable explanation for the predictive anticipatory effect could be sensory cueing. Sensory cueing occurs when an experimenter allows information about a future stimulus to be obtained by the participant. Experiments using intentional sensory cueing were not considered for this meta-analysis, as the stimulus order would then not be random (see
Another explanation includes the idea that the filtering of physiological data can produce artifacts, some of which can appear in the pre-stimulus period. A recent review of this phenomenon as demonstrated in EEG data showed that high-pass filters with low frequency cut-offs greater than 0.1 Hz can produce pre-stimulus effects that differ in direction from the post-stimulus response, assuming causal filtering is not used (Rousselet,
One might suspect that order effects could explain the predictive anticipatory effect described here. Order effects become more likely when fewer trials are performed, as order effects tied to a given stimulus order generally “wash out” when a greater number of randomly ordered trials are performed. Other order effects, specifically expectation bias, can occur when the two stimulus classes are not presented equiprobably, and a participant learns that one type of stimulus is more common among the potential stimuli. But we found that expectation bias could not explain the anticipatory effects in any of the studies in which these analyses were performed. However, different authors used different analyses, and it is critical to determine the most sensitive expectation bias analysis and to use that method in future studies of unexplained anticipatory activity. Other order effects, including forward priming, were not widely examined in these studies. Because experimenters randomized stimulus selection and order, because we assume that in most studies experimenters correctly initialized their RNGs and therefore presented a different stimulus order to each of their participants, and because most studies described tests of randomness passed by the RNGs, it is unlikely but not impossible that orders were consistent across most of the participant runs in one study. However, the chance of this occurring consistently in most of these 26 studies is vanishingly small, and even smaller in studies using hardware-number generators that do not require initialization. In spite of all these assurances, analyses of expectation bias and other order effects are critical to the clear understanding of the mechanisms underlying these predictive but seemingly anomalous anticipatory effects.
One possible way to address order effects is to determine whether a between-participants anomalous anticipatory effect exists when participants perform only one trial in which a single randomly selected stimulus is presented. In such a paradigm, statistical power should be weaker due to the between-participant design, but the ES might be large enough to detect a significant anticipatory difference – unless these unexplained anticipatory effects are by-products of mundane order effects. Interestingly, a
One unfortunate possibility we must examine is either participant or experimenter fraud. Participant fraud can be easily ruled out – it is not clear how participants would be able to change their own physiology, even if they knew the direction in which they should change it in order to produce an effect. Although we did not find studies showing evidence of participant or data selection, optional stopping, or data manipulation, it is still possible that an unscrupulous experimenter in any discipline who is willing to commit what amounts to this sort of scientific fraud would be careful enough not to provide evidence of their fraud for the reader. Thus, no scientific venture can completely rule out fraud. Based on the strong significance of the overall ES estimated from the pertinent studies available between 1978 and 2010, to explain the predictive anticipatory effect examined here, such fraud would have to be widespread and undetected. We find the possibility of such massive collusion highly unlikely.
Another seemingly tractable explanation for the currently unexplained anticipatory effect is that some of the experimenters performing these experiments are using many methods of analysis and reporting the results for the one method that produces the biggest effect. This is an understandable approach in the early stages of the discovery of any phenomenon, as the work is necessarily exploratory because none of the factors influencing the effect are known. However, after performing an exploratory analysis, researchers would ideally settle on both a single paradigm and a single analysis method, then attempt to replicate their work using exactly the same paradigm and analysis. All of the authors of the studies we have examined here are presumably careful researchers. However, for any researcher, it is tempting to tweak paradigms when attempting a replication in order to obtain more information about the phenomenon than is provided by an exact replication. Unfortunately, this temptation may have produced a situation in which a single, replicable unexplained anticipatory physiology experiment with a well-defined paradigm and analysis method is not yet available. Such an experiment is critical for the future understanding of this currently unexplained effect. Because of the potential importance of the phenomenon, we encourage multiple researchers to pursue this aim in parallel.
Critically, this multiple-analyses hypothesis cannot fully explain the results of the present meta-analysis, as the hypothesis tested by most of the studies we examined was different from the hypothesis tested by this meta-analysis. Presumably, researchers would be biased toward methods that supported their hypothesis (any pre-stimulus difference) rather than methods that supported ours (a pre-stimulus difference matched in direction to the eventual post-stimulus difference). Thus, even if all researchers used analyses that maximized the likelihood of supporting their hypothesis (which we personally know not to be the case at least in our own work), and even if there were no real unexplained anticipatory effects, roughly half of the studies should have positive ESs and half should have negative ESs (relative to our hypothesis), which is clearly not the case. However, it is possible that unstated assumptions about the directionality of the effect could bias researchers toward finding analyses for which the post-stimulus effect matched the pre-stimulus effect. This sort of explanation could potentially explain the results. However, if this explanation is correct, it is unclear why the meta-analysis constrained only to electrodermal data produced a highly significant effect. As described previously (see
The remarkably significant and homogeneous results of this meta-analysis suggest that the unexplained anticipatory effect is relatively consistent, if small in size. If so, the effect should be apparent in many experiments that present a series of emotional and calm stimuli. However, we agree with the scientists who design such experiments that both everyday experience and the second law of thermodynamics suggest a single direction for causality; causes normally precede effects. For these reasons, physiological effects preceding a subsequent cause are not generally assumed to exist, and are therefore not usually examined. In fact, one of the first analytical steps in most studies of physiological responses to distinct stimuli is to use the average of a time period preceding the stimulus as a baseline value. If this value is subtracted from all points in the physiological trace, such a baselining practice can effectively remove any evidence of a predictive anticipatory effect by zeroing out the anticipatory period (see below, Implications, for steps that can ameliorate this problem). Regardless of whether such a practice is followed, most researchers do not present much of the pre-stimulus period for comparison across conditions. For these reasons, predictive anticipatory effects may be both rampant yet invisible in mainstream psychophysiology results. Indeed, one study included in this meta-analysis that examined pre-stimulus data for three such experiments found anticipatory effects in all three mainstream studies investigated; one effect was significant (α = 0.05) and the other two were borderline (Bierman,
To determine whether other mainstream studies also contain evidence for similar anticipatory effects, we requested data from 14 researchers who published emotional physiology studies in non-parapsychology journals after 2000. Four offered to share data, but two of these four could not find the appropriate data files. Here we briefly report our analysis of the two data sets made available to us. For both data sets, multiple dependent variables were analyzed in the two published reports, which both focused on post-stimulus effects. Using the same methods we used to determine ESs for correlated data (see
As already briefly discussed, one possible explanation for the present results that may be made to fit the available data is that most researchers have an implicit assumption about the directionality of the effect and they used this assumption to select analysis methods that magnified the similarity between the pre- and post-stimulus effects as well as the ES. We consider this an unlikely but plausible explanation. Unlikely because we ourselves have analyzed our own data in multiple ways that produce larger pre-stimulus effects but feel constrained by scientific rigor to report only the results obtained with the originally selected analysis method. Further, we have had conversations with several of the other researchers whose studies we have examined here, and it is clear that their analysis methods were attempts at replications of previous analysis methods used by other researchers. However, the explanation is plausible because unexplained anticipatory activity is a phenomenon that is not well understood, and some researchers may feel justified in using multiple methods of analysis in order to better understand the effect. However, it is important to note that when researchers reported multiple statistical results from the same dataset we used the results leading to the smallest ES. Nevertheless, until this unexplained anticipatory effect is replicated multiple times using the same paradigm and method of analysis, we cannot completely rule out the multiple-analyses explanation. Further, there may be other explanations of which we are presently ignorant, but that will become clear over time. In the meantime, we speculate below about the implications of these results.
The most mundane implication of these results is that the existence of unexplained anticipatory effects could potentially either: (1) produce what seem to be null psychophysiological results due to baselining when in fact there is a significant pre-stimulus effect, or (2) produce significant psychophysiological results due to not baselining when there is a significant pre-stimulus effect accounting for the post-stimulus difference. Ideally, in future experiments the physiological variables preceding the stimuli or events of interest would be compared across stimulus classes first, before performing the usual baselining procedure. If there are significant baseline differences, then these differences should be reported in addition to any further post-stimulus effects observed after baselining.
More importantly, we feel that these predictive anticipatory effects constitute a fourth category in addition to three broad categories of anticipatory effects that have already been established in psychophysiology and neuroscience. The first category includes physiological anticipation of intentional motor activity, e.g., physiological anticipation of a willed movement begins at least 500 ms before the conscious report of the intention to move (Libet et al.,
Recently, a third category of anticipatory effect, dubbed “preplay,” was discovered when the pre-maze activity of mouse hippocampal neurons was shown to mimic the activity recorded during and after being in the maze, even in mice for whom a maze was novel (Dragoi and Tonegawa,
For all three categories of anticipatory effects described above, the usual cause-preceding-effect assumption is sufficient to construct reasonable explanations for the observed phenomena. The seemingly anomalous anticipatory effects investigated in this meta-analysis could have some influence on the each of these three types of phenomena, but these unexplained anticipatory effects are not necessary to explain these three types of established anticipatory effects. Conversely, the three types of established predictive effects cannot explain the unexplained anticipatory activity examined here. Thus we suggest that unexplained predictive anticipatory effects belong in a category independent from, but potentially overlapping with, the three other categories of anticipatory effects already described.
In sum, the results of this meta-analysis indicate a clear effect, but we are not at all clear about what explains it. We conclude that if this seemingly anomalous anticipatory activity is real, it should be possible to replicate it in multiple independent laboratories using agreed-upon protocols, dependent variables, and analysis methods. Once this occurs, the problem can be approached with greater confidence and rigor. The cause of this anticipatory activity, which undoubtedly lies within the realm of natural physical processes (as opposed to supernatural or paranormal ones), remains to be determined.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
When comparing control and experimental conditions using independent (or uncorrelated) samples, the usual effect size measure is
One possibility is to use the effect size used for the independent samples case, which can be estimated if the correlation is known or can be approximated:
where
Therefore,
and
The other possibility is to use
It is easy to see the relationship between these two effect sizes:
Therefore, the following relationships hold based on the correlation
When
When
In particular, when
Citation | Study component | Stimuli | Psychophysiological measure(s) | Anticipatory period (s) | Effect size |
SE | Expect. bias analysis | Quality score | SC effect size |
---|---|---|---|---|---|---|---|---|---|
Bierman ( |
Study 1 | Calm/violent images | Electrodermal activity | 7.5 | 0.68 | 0.27 | Yes | 4.750 | 0.68 |
Bierman ( |
Study 1 | IAPS calm/emotional images | Electrodermal activity | 7 | 0.17 | 0.0.17 | No | 3.250 | 0.17 |
Bierman ( |
Study 2 | IOWA gambling task | Electrodermal activity | not reported | 0.47 | 0.30 | No | 3.250 | 0.47 |
Bierman ( |
Study 3 | calm/erotic images | Electrodermal activity | 4 | 0.10 | 0.17 | No | 3.250 | 0.10 |
Bierman ( |
Pilot study | Pleasant/unpleasant sounds | Electrodermal activity | 3 | -0.18 | 0.14 | No | 2.250 | −0.18 |
Bierman and Radin ( |
Study 2 | Calm/violent images | Electrodermal activity | 7.5 | -0.08 | 0.17 | Yes | 4.250 | −0.08 |
Bierman and Radin ( |
Study 3 | “ | Electrodermal activity | 7.5 | 0.16 | 0.17 | Yes | 4.250 | 0.16 |
Bierman and Scholte ( |
Entire study | IAPS calm/violent/erotic images | BOLD in ROI: visual cortex | 8.4 | 0.33 | 0.32 | Yes | 4.500 | N/a |
Bierman and van Ditzhuyzen ( |
Entire study | Slot machine images (guessing task) | EEG/event-related potentials (ERP); entire pre-stimulus period pooled over Fz, Cz, and Pz | 1 | 0.42 | 0.18 | Yes | 4.000 | N/a |
Broughton ( |
Entire study | IAPS calm/emotional images | Electrodermal activity | 3 | 0.03 | 0.12 | Yes | 5.500 | 0.033 |
Gillin et al. ( |
Roulette | Roulette images (guessing task) | Electrodermal activity; heart rate | 12 | 0.29 | 0.36 | No | 4.250 | 0.046 |
Gillin et al. ( |
Business case | Business images (guessing task) | Electrodermal activity; heart rate | 6 | 0.38 | 0.36 | No | 4.250 | 0.74 |
May et al. ( |
Entire study | Alerting sounds | Electrodermal activity | 3.5 | 0.29 | 0.14 | Yes | 6.750 | 0.29 |
McCraty et al. ( |
Entire study | IAPS calm/emotional images | Electrodermal activity; heart rate | 6 | 0.29 | 0.20 | Yes | 4.375 | 0.13 |
Mossbridge et al. ( |
Exp 1 | Calm IAPS images (guessing task) | Electrodermal activity | 10 | −0.14 | 0.16 | Yes | 4.750 | −0.14 |
Mossbridge et al. ( |
Exp 3 | “ | Electrodermal activity | 10 | 0.06 | 0.18 | Yes | 4.750 | 0.06 |
Radin ( |
Entire study | Calm/emotional images | Electrodermal activity; blood volume | 5 | 0.30 | 0.18 | Yes | 4.750 | N/a |
Radin ( |
Entire study | IAPS calm/emotional images | Electrodermal activity; blood volume; heart rate | 5 | 0.09 | 0.14 | Yes | 4.000 | 0.09 |
Radin ( |
Study 1 | Calm/emotional images | Electrodermal activity | 5 | 0.60 | 0.22 | Yes | 5.000 | 0.60 |
Radin ( |
Study 2 | “ | Electrodermal activity | 5 | 0.16 | 0.13 | Yes | 5.750 | 0.16 |
Radin ( |
Study 3 | IAPS calm/emotional images | Electrodermal activity | 5 | 0.49 | 0.15 | Yes | 6.000 | 0.49 |
Radin ( |
Study 4 | “ | Electrodermal activity | 5 | 0.24 | 0.41 | Yes | 5.000 | 0.24 |
Radin and Lobach ( |
Entire study | Light flashes vs. no flashes | EEG/event-related potentials (ERP): Slow cortical potentials at Oz only | 1 | 0.12 | 0.22 | Yes | 6.375 | N/a |
Radin and Borges ( |
Exp 1 | IAPS calm/emotional images | Pupil dilation | 3 | 0.46 | 0.18 | Yes | 5.000 | N/a |
Spottiswoode and May ( |
Entire study | Startle sounds vs. no sounds | Electrodermal activity | 3 | 0.29 | 0.09 | Yes | 6.750 | 0.29 |
Tressoldi et al. ( |
Study 1 | Pleasant/alerting sounds | Heart rate | 5 | 0.29 | 0.07 | No | 4.625 | N/a |
We thank the Bial Foundation for awards supporting the work of the first and second authors. The first author was also funded under NIH training grant 5T32NS047987-05 during a portion of the period used to write this manuscript. We are grateful to Dean Radin, Rollin McCraty, Norman Don, Ed May, Ed Modestino, George Dragoi, and Barbara Spellman for feedback that greatly improved this manuscript.
1This hypothesis is likely too simplistic, in that some physiological measures may consistently reveal different pre-stimulus anticipatory directions depending on the valence of the upcoming stimulus (Rollin McCraty, personal communication) and/or participant characteristics such as gender. However, for a first meta-analysis of these phenomena, we thought it best to keep the hypothesis simple.
2We selected 1978 as the early cut-off because the first study we could find that was relevant to the meta-analysis was published in that year (Hartwell,
3While some peer review committees for conferences are very strict, it is our experience that some conferences have lax peer reviews or none at all. Therefore, to be conservative, we have given studies that appear in conference proceedings (and were not later published in a journal) a lesser score than those appearing in peer reviewed journals.
4It is important to note here that most authors using pseudo-RNGs reported tests of randomness passed by those generators. Thus, the sequences were clearly random. As a result, this ranking could be considered quite conservative, but we have chosen to take conservative steps wherever possible.
5It has been suggested that we use a fourth quality index reflecting the number of hypotheses tested by each study. However, for all studies included here, the main hypothesis was in regard to unexplained physiological anticipation of unpredictable future events. For studies testing other hypotheses, these hypotheses were either orthogonal to the main hypothesis (e.g., fMRI pattern differences in resting state activity for meditators and non-meditators), or were