BY J. E. KENNEDY

ABSTRACT: Many parapsychological writers have suggested that psi may be capricious or actively evasive. The evidence for this includes the unpredictable, significant reversal of direction for psi effects, the loss of intended psi effects while unintended secondary or internal effects occur, and the pervasive declines in effect for participants, experimenters, and lines of research. Also, attempts to apply psi typically result in a few very impressive cases among a much larger number of unsuccessful results. The term unsustainable is applicable because psi is sometimes impressive and reliable, but then becomes actively evasive. One of the most testable models for this property is that psi effects occur against a background of supporting and opposing motivation and psi influence due to the extreme polarization of attitudes toward psi in the population. These attitudes may have genetic and gender associated components. Another possible explanation is that the primary function of psi is to induce a sense of mystery and wonder. Other possible functions of psi also need to be investigated. For example, psi could contribute to evolution by briefly influencing random processes to enhance diversity, without specifically guiding evolution or having sustained effects. Some type of higher consciousness may influence or control psi effects.

In describing the evolution of ESP research at the Duke laboratory, Palmer (1981, p. 31) noted: “more typical of the new trend was the performance of a teenage boy identified as P.H., whose promising psi-hitting when tested informally reverted to significant psi-missing when better controls were applied (Russell & Rhine, 1942).”

Rao (1965) used terms like “bidirectionality” and “differential effect” to characterize the property that “shifts the mode of psi response from hitting to missing in a rather capricious manner” (p. 245). He summarized numerous examples and described this characteristic as preventing the useful application of psi.

Bierman (1981) pointed out that the shift between psi hitting and psi missing is not just low reliability, but negative reliability. He noted that  the direction of scoring  frequently reverses from pilot to confirmation which is negative reliability. He also commented on several studies he carried out and other studies in the literature that showed unexpected reversals of scoring when the data for one study were divided into two groups. This division of data is essentially the same as a split-half method to estimate reliability.

These unintended, undesired, significant reversals in direction are an unusual form of unreliability that seem evasive and almost defiant. The normal manifestation of unreliability is results that are sometimes significant and sometimes nonsignificant (i.e., reliability approaching zero rather than being negative).

Another seemingly capricious or defiant psi manifestation is when the overall intended effect becomes nonsignificant, but unintended secondary effects provide evidence for psi. The Princeton Engineering Anomalies Research (PEAR) laboratory provides a recent example. Studies with electronic random event generators (REGs) had small but significant effects for a decade. A recent large-scale replication effort obtained nonsignificant results overall, but Jahn et al. (2000) reported internal structural effects that appeared to indicate psi. The analyses for these effects were based on findings in the previous data, and the effects were reported as significant after adjusting for multiple analyses. However, the effects had different patterns than the earlier results and were not consistent across the three laboratories participating in the project.

Jahn and Dunne (2001) summarized the situation as follows: “At the end of the day, we are confronted with an archive of irregular, irrational, yet indismissable data that testifies, almost impishly, to our enduring lack of comprehension of the basic nature of these phenomena” (p. 300). They noted that these changes in psi manifestations are not consciously intended or desired by those conducting the studies, and suggested that unconscious processes may have a major role in psi effects and the associated inconsistencies.

The evolution of research at the Princeton laboratory is notably similar to the earlier experience at the Duke laboratory. At Duke the initial research was remarkably successful in demonstrating the intended effects. However, a decade later, unintended, internal effects were increasingly being reported as the primary finding. J. B. Rhine (1946a) commented:

[M]any of the experiments yielding only chance totals have proved to be fruitful in other respects. Analyses of different character than those initially applied have in some cases revealed hidden relationships that were first overlooked. Some of the most important discoveries concerning PK, like those concerning ESP, have emerged in this way. (p. 73)

In fact, J. B. Rhine (1974) argued that these internal effects were some of the best evidence for psi because the lack of motivation, intention, and expectation for their occurrence reduced the likelihood of fraud or errors. The position effects and other internal effects were thought to be a psychological reaction to the testing methods. However, there is also evidence that they may be associated with certain experimenters, and at least some of the internal effects appear to be “mind prints” or “signatures” of experimenter effects (Kennedy & Taddonio, 1976; McConnell, 1989). The Duke group also concluded that the elusive, inconsistencies of psi were due to the unconscious nature of psi (J. B. Rhine, 1946c).

Palmer (1981) noted that position effects received relatively little attention after the 1940s. We are left with the impression that these internal effects become elusive when they become an expected or intended outcome. Of course, skeptics would argue that the internal effects were (are) simply post hoc data selection in an effort to salvage nonsignificant results. However, the position effects were remarkably consistent in the early studies—at least when they were not intended or expected.
Batcheldor (1994) commented that his efforts to investigate macro (nonstatistical) PK phenomena had similar properties. He summarized his experience as follows:

The evidence exists, but it fluctuates. And a closer examination shows that . . . the evidence seems to “avoid” those positions in time or space where we are actively looking for it. . . . And then, later, new evidence is found elsewhere and under tight conditions. (p. 93)

In addition to the changes in psi manifestations noted above, the loss of psi effects may also indicate the evasiveness and unsustainability of psi. The loss of psi effects occurs for individual subjects, experimenters, and lines of research.

Pratt (1975) summarized the universal loss of psi effects with individual subjects: “we must recognize what has been the most serious limitation on psi research with outstanding subjects. This is the unexplained loss of ability that has always brought their successful performance in the test situation to an end” (p. 159).

Houtkooper (1994, 2002; Haraldsson & Houtkooper, 1995) proposed the term “meta-analysis demolition” to describe the loss of effect for an experimenter or experimenter group. His evaluation of seven different series of studies found that a summary evaluation was followed by an average of 90% reduction in effect size. Kennedy and Taddonio (1976) noted other examples of declining effects for experimenters. In early parapsychological research, Taves and Dale (1943) used the term experimenter “Midas touch in reverse” to describe the tendency for effects to decline within a study.

Bierman (2001) showed that declines in effect are typical for most lines of research in parapsychology. His evaluation used several meta-analysis databases. Because most popular lines of research were initiated by psi-conducive experimenters, these declines probably result from declines for individual researchers combined with failed replication efforts by experimenters who are not psi-conducive. Beloff (1994) and Pratt (1978) commented on the apparently universal decline in effect for psi research before the advent of meta-analyses.

Despite these various declines, the overall significance level for most of these participants, experimental series, and lines of research remains significant. There is evidence for psi, but the effects seem systematically unstable.

These declines are an anomalous form of unreliability that indicate an inhibitory process. Unreliability normally manifests in one of two patterns over time. The effect sizes and proportion of studies with statistically significant outcomes will tend to   increase for later studies if relevant variables are identified and controlled. Alternatively, the effects will remain relatively uniform across studies if there is no progress in understanding the phenomenon. The widespread declines found in psi research indicate an inhibition of psi as well as a general lack of progress in understanding psi.

Braud (1985) and Batcheldor (1994) have suggested that psi appears to occur more readily in situations with lower methodological quality. For example, Dunne and Jahn (2002) noted that their 25-year experience with remote viewing found a steady decline that occurred as they attempted to develop increasingly objective, quantitative evaluation methods. Braud (1985) also noted that more sophisticated, process-oriented studies seem less likely to obtain psi effects than simple studies intended to provide evidence for the existence of psi. Of course, skeptics interpret these trends as evidence that ostensible psi phenomena actually result from methodological problems.

However, there is a clear mechanism for a misleading, noncausal association between psi effects and methodological quality. The normal process for scientific research is to develop increasingly controlled and analytical research methods as research progresses. The declines over time in psi effects for a line of research will therefore be correlated with higher quality methodology even though they may be independent processes.

This possibility is supported by cases when declines occur even when the methodology remains constant or becomes less well controlled. For example, in a study investigating the use of ESP to find land mines, the initially significant results declined even though the methodology did not change (J. B. Rhine, 1971). The loss of effect with the attempted replication of the PEAR REG studies occurred even though the replication efforts used an REG that was lower quality (had less redundant failsafe measures) than the REG for many of the early studies (Jahn et al., 2000). Also, the REG used in the replication studies had been used with positive results is some earlier studies. Given the evidence for widespread declines in psi effects, a more sophisticated analysis is needed before a negative association between psi and methodological quality can be taken as evidence for the evasive nature of psi.

At the same time, the idea that psi effects are inversely related to experimental sophistication needs to be evaluated. The notable lack of comparative studies in some of the most successful lines of psi research appears to indicate that parapsychologists have tacit, working assumptions favoring simple, unrevealing experiments (perhaps as a result of operant conditioning by actively evasive psi). As a conspicuous example, research attempting to show that the ganzfeld procedure is psi conducive has rarely included a control group even though that is a fundamental departure from the basic principles of research design. The assumptions underlying this practice need to be acknowledged and made explicit.