Den telepatiske papegøyen

Test av dyrs intelligens og telepatiske evner har lang tradisjon. Den mest kjente er nok historien om hesten som anga riktig antall ganger med hoven på enkle regnestykker som 2 + 2.
Mekanismen viste seg å være enkel: Hesten var verken telepatisk eller intelligent, ihvertfall ikke matematisk. Den svarte kun riktig hvis den kunne se eierens ansikt. Ørsmå og ubevisste endringer i ansiktsuttrykket fortalte hesten når den skulle slutte å stampe i gulvet.
Dette er nok kanskje grunnen til at man nå bruker mer fargerike dyr. En av de siste forsøkene er ihvertfall med papegøyen N’Kisi som angivelig kan lese eierens tanker. Forsøket omtales slik i abstraktet i Journal of Scientific Explorations (som utgis av SSE – The Society for Scientific Exploration som er dannet for å forske på slike ting og dermed nok i utgangspunktet ikke er helt uhildete):

Aimee Morgana noticed that her language-using African Grey parrot, N’kisi, often seemed to respond to her thoughts and intentions in a seemingly telepathic manner. We set up a series of trials to test whether this apparent telepathic ability would be expressed in formal tests in which Aimee and the parrot were in different rooms, on different floors, under conditions in which the parrot could receive no sensory information from Aimee or from anyone else. During these trials, Aimee and the parrot were both videotaped continuously. At the beginning of each trial, Aimee opened a numbered sealed envelope containing a photograph, and then looked at it for two minutes. These photographs corresponded to a prespecified list of words in N’kisi’s vocabulary, and were selected and randomized in advance by a third party. We conducted a total of 147 two-minute trials. The recordings of N’kisi during these trials were transcribed blind by three independent transcribers. Their transcripts were generally in good agreement. Using a majority scoring method, in which at least two of the three transcribers were in agreement, N’kisi said one or more of the key words in 71 trials. He scored 23 hits: the key words he said corresponded to the target pictures. In a Randomized Permutation Analysis (RPA), there were as many or more hits than N’kisi actually scored in only 5 out of 20,000 random permutations, giving a p value of 5/20,000 or 0.00025. In a Bootstrap Resampling Analysis (BRA), only 4 out of 20,000 permutations equaled or exceeded N’kisi’s actual score (p = 0.0002). Both by the RPA and BRA, the mean number of hits expected by chance was 12, with a standard deviation of 3. N’kisi repeated key words more when they were hits than when they were misses. These findings are consistent with the hypothesis that N’kisi was reacting telepathically to Aimee’s mental activity.

Det hele er nærmere beskrevet i denne pdf-filen, som i god tradisjon ved forskning på feltet bruker rimelig normal vitenskapelig metodologi.

Forsøket og analysen virker ved første øyekast ganske tilforlatelig. Det anvendes kjente statistiske metoder, selv om en av reviewerne finner noen praksisk sett betydningsløse feil ved måten man tenker permuteringer. Resultatet omtales i kvantitative termer. Og det er signifikant.
Men som så ofte, viser det seg at slike forsøk handler om mer enn matematikk. Det er i det hele tatt en god øvelse å tenke gjennom hvor snubletrådene kan være utlagt når man ser på parapsykologiske studier.
Du kan godt bruke litt tid før du sjekker kommentarene mot slutten av artikkelen, selv om de ikke er gitt opp ned på bakerste side.
Vi skal her gjengi de to viktigste:

In reviewing this paper, I commented on the preponderance of flowers in two aspects of the experiment. The parrot’s vocabulary and the selection of images are not random. Both show enhanced frequencies of certain culturally favored concepts—‘‘flower’’ being only the most obvious one. The statistical analysis presented in the paper assumes randomness, and is therefore inappropriate.
The authors’ response (that it would be unfair, statistically speaking, to arbitrarily remove the most obvious evidence) is correct but misses the point. For their comment to be relevant, ‘‘flower’’ and other concepts, such as ‘‘doctor,’’ and ‘‘medicine,’’ would have to be the most frequent concepts in this experiment simply because of statistical fluctuations. But both samples (the parrot vocabulary and suite of pictures) contain them with higher than average frequency because of their cultural importance.
It is very difficult, if not impossible, to account for this kind of cultural selection effect with statistical analysis of the data. Hence, I do not believe that this experiment provides any evidence supporting the claim of telepathy. What is
needed is an experimental protocol insensitive to selection effects in the first place.

Denne type kommentarer understreker at det er nødvendig å gå dypere inn i slike studier enn abstraktet. Noe man kan mistenke mange media for sjelden å ha tid til og telepatitroende sjelden ha lyst til. La oss derfor avrunde med tidsskriftredaktørens oppsummering:

This article is another instance of your Editor’s difficulties where research protocols and statistical inference are questioned. The first two reviewers of this manuscript made opposing recommendations, and two more reviewers were consulted. Publication was a majority recommendation, though some reviewers felt that the protocol was flawed for the reason described above by Scargle.

My personal reaction is that, once again, we have suggestive results, a level of statistical significance that is less than compelling, and the devout wish that further work with refined protocols will ensue.

«Once again» er nok riktig uttrykt.
Vi ser ellers av hjemmesiden at SSE har hatt årsmøte i 2007. I Roros i det kjente landet Normway.