Real Versus Decoy Chinese Characters: Vernon, Hitchman and Roe's Pre-Registered Test of Morphic Resonance

Most of what is written about Morphic Resonance is argument: Rupert Sheldrake's own books, and the replies of his critics. "An Implicit and Explicit Assessment of Morphic Resonance Theory Using Chinese Characters", by David Vernon, Glenn Hitchman and Chris Roe, is something different: a controlled experiment. The paper describes the outcomes of earlier tests with Chinese characters as equivocal. This one was pre-registered, two of its authors had worked on those earlier Chinese-character studies, and its outcome went against the theory: the paper reports that 154 people with no knowledge of Chinese showed no ability to tell genuine characters from invented ones, and that when simply asked which they liked, they leaned towards the fakes.

Why Chinese characters?

The experiment rests on a prediction the authors draw from Sheldrake's The Presence of the Past (cited in its 2011 second edition; see The Presence of the Past: Morphic Resonance and the Habits of Nature). If Formative Causation is right, something learned by very many individuals should leave a strong collective trace in a morphic field, and later learners should find it slightly easier to pick up. A written language used by an enormous population over a long history ought, on that reasoning, to be an unusually favourable case. A real character has been learned by generations of readers; a made-up character that merely looks Chinese has been learned by nobody. People who know no Chinese should therefore respond differently to the two.

The paper places this in a short line of earlier work. It describes Sheldrake's 1983 hidden-image test, in which one of two puzzle pictures was shown on television to approximately 2 million people in the UK and was reportedly then identified more readily abroad, and Mahlberg's 1987 finding that novices learned Morse Code faster than an equally difficult invented code. It then turns to two direct predecessors:

• Robbins and Roe (2010) showed participants genuine and decoy Mandarin characters and reported significantly better recognition of the genuine ones, a result read as supporting the theory.
• Roe and Hitchman (2011) repeated the test with a larger stimulus set in which real and decoy characters were matched more strictly on features such as complexity and stroke count. The advantage for real characters disappeared.

Two of the three authors of the new study were thus involved in the earlier, conflicting results.

What was new in this study

The authors suggest one reason the earlier studies might have disagreed: the tests used were, as the paper puts it, predominantly reliant on explicit memory, which involves deliberate, conscious recollection. They note proposals in the wider literature that anomalous effects may show up more clearly in implicit measures, where people respond without being asked to recall or judge anything. So the design compared two tasks in the same people:

• an implicit preference task, in which participants saw a real character and its decoy side by side and clicked on the one they "preferred";
• an explicit identification task, in which they saw a pair and clicked on the one they thought was real.

Three predictions followed: real characters would be preferred above chance, real characters would be identified above chance, and the effect would be larger on the implicit task. The authors also explored whether scores tracked belief in the paranormal.

How the experiment was run

The study was pre-registered with the Koestler Parapsychology Unit (registry reference 1046), with the raw data deposited there, and had ethical approval from Canterbury Christ Church University. A power calculation, using an assumed effect size of 0.3 (deliberately below the 0.41 reported by Robbins and Roe), gave a minimum sample of 144; the target was 160.

It ran online in Qualtrics and was advertised as research on the "aesthetics of language", with the real purpose disclosed only in the debrief. Of 176 people who began, 14 were removed for answering yes to any of five screening questions about exposure to Mandarin (having studied it, visited mainland China, knowing five or more characters, and so on) and eight for not fully attending to the task. The remaining 154 comprised 106 women, 47 men and one who preferred not to say, aged 18 to 76 with a mean of 33.5 years.

The stimuli were 24 real characters and 24 decoys, the decoys created by a native Chinese speaker, paired so that each real character and its decoy matched in stroke count and complexity, following Roe and Hitchman. These were split into two matched lists, each produced in two left/right arrangements, giving four lists. Every participant did both tasks, 12 pairs each, on different lists, with task order and list counterbalanced and the pairs shown in random order. Each task was preceded by a one-minute relaxation period with a star-field image and music. Belief was measured with the Experience, Belief and Ability scales of the Anomalous Experience Inventory.

What the data showed

With 12 pairs per task, guessing gives an expected score of 6. The paper reports:

All tests were two-tailed. On the preference task the departure from chance was statistically significant, but in the wrong direction: participants chose the real characters less often than guessing would predict. On the identification task there was no reliable difference from chance, though the small shortfall pointed the same way. The two tasks did not differ in effect size (reported means of -0.215 and -0.152; p = 0.59), so the idea that an implicit measure would be more sensitive was not borne out. None of the six correlations with the belief scales was significant; the largest in magnitude was r = -.131 (explicit identification against the Belief scale, p = .105), and all p values exceeded 0.1.

How the authors read it

The abstract states plainly that the findings "fail to support the idea of morphic resonance". For the unexpected preference for decoys, the authors favour an ordinary explanation: something about the look of the invented characters appealed to people. They point out that this fits the earlier failure to replicate, where Roe and Hitchman found decoys recognised slightly more often than real characters (5.40 against 5.22, not significant) and falsely "recognised" more often when never shown (2.64 against 2.14).

They then list reasons the study might have missed a real effect, all offered as possibilities rather than findings:

• Online testing. Unsupervised participants may have paid little attention or simply guessed. The authors note that one of them had found significant precognition effects with a retroactive-priming task in the laboratory but not online, citing Vernon's papers of 2015 to 2018, and they suggest that online testing may lower engagement or lose an interaction between experimenter and participant.
• Effort and personality. Mahlberg had suggested that an effort to learn, and an introverted disposition, matter; Robbins and Roe had linked success to transliminality. Neither trait was measured here.
• Distance. Participants were probably in or near the UK, and the authors speculate that any effect might weaken with distance from China.
• Task order. Half the participants did the identification task first and may have carried a guess-the-real-one mindset into the preference task.
• The stimuli. The real characters were chosen by intuition rather than by measured frequency of use, and all the decoys were created by one person, whose judgement and artistic ability may have shaped the outcome to a degree the authors say is hard to gauge.

What the study does and does not establish

Neither pre-specified prediction was supported, in a pre-registered study whose analysed sample of 154 exceeded the minimum of 144 given by the authors' own power calculation. Taken with Roe and Hitchman, it means that two of the three Chinese-character studies described in the paper, and both of those using tightly matched stimuli, found no advantage for real characters.

It does not test Rupert Sheldrake's hypothesis as a whole. It examines a single prediction, in one online sample, with a stimulus set of 24 real characters and 24 decoys. The authors' preferred and, in their words, parsimonious reading of the significant result is an aesthetic one: something visually distinctive about the decoys appealed to participants. They add that they cannot judge with any accuracy how far the judgement and skill of the single person who created the decoys affected the outcome. On that reading the result says something about these particular stimuli, not about a "negative" morphic effect, and the authors do not claim one.

The proposed limitations also deserve some caution. Suggestions that the effect may depend on distance from China, on personality, or on participants' engagement and their interaction with an experimenter are untested here; the authors offer them as directions for future research. The paper itself does not say so, but a reader may reasonably observe that conditions put forward after a null result, if they accumulate, leave a prediction harder to test. The reading guide at The Presence of the Past: Morphic Resonance and the Habits of Nature raises a related concern about Sheldrake's own treatment of intrinsic differences between old and newly invented stimuli. The paper's remark that it is unclear how far a morphic resonance effect would localise in space also sits awkwardly beside Sheldrake's description, as the authors summarise it, of a field operating beyond space and time. For the wider record of claims and evidence, see Morphic Resonance: Formative Causation, Behavioral Fields and the Proposed Tests.

References

• Vernon, D., Hitchman, G., and Roe, C. "An Implicit and Explicit Assessment of Morphic Resonance Theory Using Chinese Characters." Author copy in the University of Northampton research repository.
• Robbins, K., and Roe, C. A. (2010). "An empirical test of the theory of morphic resonance by using recognition for Chinese symbols." Explore, 6(4), 256–262.
• Roe, C. A., and Hitchman, G. A. (2011). "Testing the theory of morphic resonance using recognition for Chinese symbols: A failure to replicate." Journal of the Society for Psychical Research, 75(4), 211–224.
• Mahlberg, A. (1987). "Evidence of a collective memory: A test of Sheldrake's theory." Journal of Analytical Psychology, 32, 23–34.