Escolar Documentos
Profissional Documentos
Cultura Documentos
Perception
Day 12
27 Feb 03
Parapsychology
mediumship magicians vs. physicists
psychokinesis Uri Geller & James Randi
precognition seers
telepathy 1975, J.B. Rhine
AAAS accepts in 1969
clairvoyance autoganzfeld experiments
case studies vs. scientific method
replicability
probability & statistics a priori chance
a posteriori base rate
Does use of the scientific method imply that the
research is science?
Prelim: Thurs, 4 Mar
during class time
11:301:05
comprehensive: lectures 112 (including today);
chapters 1, 2, 10, 11, 12, 13, Appendix
Question answering session:
Mon evening, 3 Mar
7:009:00 PM in Uris 202
Why is extrasensory perception (ESP) interesting
theoretically?
Why is extrasensory perception (ESP) interesting
theoretically?
1. the role of belief in science (and all of academia)
2. Johannes Müller’s Doctine:
peripheral nervous system (PNS) > CNS
ESP would bypass PNS
no known receptors
direct knowledge of the world without sensory information
3. relation of scientific methods to science
Scientific Methods > search for patterns and their meaning
null hypothesis H0
“no pattern”,
no special need for understanding cause
experimental hypothesis H1
there is a pattern of interest
and we need to understand the cause
Scientific Methods
null hypothesis H0
experimental hypothesis H1
Logic: 3 steps
1. devise situation for statistical test
2. test Ho; that is, assess its probability
3. if improbable, assume H1 is true
criterion > p < .05
if not improbable, assume Ho is true
criterion > p > .05
One can never prove the experimental hypothesis true.
Results corroborate theories;
they do not prove them true or false
Proof of truth is possible only in sufficiently closed systems,
such as math and logic.
Why?
Issues:
1. many H1s (or H1, H2, H3, H4…)
There are an indefinitely large number of
theories that can account for any set of data.
2. occurrence of logical error in step 3
….if improbable, assume H1 is true
if not improbable, assume H0 is true
Type 1 error reject H0 when it is true
Type 2 error accept H
0 when it is false
Example of multiple H1s:
Linear induction
1, 2, 3, 4, 5, .....
k = place in sequence
n = number in that place
Linear induction
1, 2, 3, 4, 5, .....
k = place in sequence
n = number in that place
Theory 1: n = k = 6
Theory 2: n = (k1)(k2)(k3)(k4)(k5) + k
= 126
How to assess hypotheses (theories)
data (generated in some systematic way)
statistics (N>>1, not case studies)
our rhetorical devices
know this
The sign test:
z = x 0.5 NP__
sqrt[NP (1P)]
z = measure of distribution on a normal (bell
shaped) curve (like a standard deviation, same
units as measured with d’)
x = number of occurrences of a particular pattern
of interest
0.5 = "correction for continuity" (magic)
N = number of observations
P = the a priori probability of that pattern
Examples: coin flips,
expectation = 0.5 Heads/0.5 Tails
Heads 6/10 > z = 0.31 p > .75
12/20 ?
With small numbers we often assume
that things should occur with exact
probabilities. They usually don’t.
Examples: coin flips, expectation = 0.5
Heads 6/10 > z = 0.31 p ~ .75
Heads 12/20 > z = 0.67 p ~ .50
Examples: coin flips, expectation = 0.5
Heads 6/10 > z = 0.31 p ~ .75
Heads 12/20 > z = 0.67 p ~ .50
Heads 60/100 > z = 1.90 p < .055
Examples: coin flips, expectation = 0.5
Heads 6/10 > z = 0.31 p ~ .75
Heads 12/20 > z = 0.67 p ~ .50
Heads 60/100 > z = 1.90 p < .055
Heads 600/1000 > z = 6.29 p < .0000001
Law of Large Numbers:
likely occurrences converge rapidly
towards expected probabilities
A Sample ESP Study
The sign test:
z = x 0.5 NP__
sqrt[NP (1P)]
z = x 0.5 N/10__
sqrt[N*.1 *.9]
z = x 0.5 N/10__
.9 sqrt[N]
Reject H0 : null hypothesis is (incredibly) unlikely,
assume null hypothesis is not true
Accept H1 : ESP
telepathy
or something else?
chance vs. base rate
a priori vs. a posteriori probabilities
example: live births, male or female
chance vs. base rate
a priori vs. a posteriori probabilities
example: live births, male or female
“chance”: 50%
base rate: ~52%
stacking effect
“Pick a number between 0 and 9”
28% pick 7
14% pick 3
The sign test:
z = x 0.5 NP__
sqrt[NP (1P)]
z = x 0.5 .28N__
sqrt[N*.28* .72]
z = x 0.5 .28N__
.45 * sqrt[N]
“Pick a number between 0 and 9, like 7”
15% pick 7
15% pick 3
R.D. Laing Knots (1974)
“They are playing a game. They are playing
at not playing a game. If I show them I see
they are, I shall break the rules … I must play
their game of not seeing I see the game.”
Paraphrase of Laing
“Professor Cutting is playing a game [of spontaneous
choices]. And he is also playing at not playing a game.
If I show him I understand his game [of spontaneity], I
shall break the rules … I must choose my spontaneous
pattern carefully.”
One must be wary of stacked situations in
methodologies of all experiments, not just in those
investigating the paranormal. Questionnaire studies
are particularly susceptible.
Another set of studies on clairvoyance
Autoganzfeld experiments
Bem and Honorton, 1994
results favor the existence of some form of psi
anomalous information transfer
Methodology: not flawed
Results: more than reasonable to
reject the null hypotheses
chance 25%
results ~3055%
does one have to accept the
experimental hypothesis of
anomalous information transfer?
A framework
Response
Yes No
Response
Yes No
Experimental results
positive negative
(Ho rejected) (Ho accepted)
Exists "progress" Type II error
Phenomenon
Does Not Exist Type I error not part of
science
Experimental results
positive negative
(Ho rejected) (Ho accepted)
Exists psi exists Type II error
Phenomenon
Does Not Exist Type I error
Type I error one says the phenomenon
exists, but it doesn’t [false alarm]
Type II error one says the phenomenon does
not exist, but it does [miss]
the practice of science deplores (has a bias against)
Type I errors when a result does not mesh with an
existing fabric of logic and research;
in such situations it would rather make Type II errors
Why?
what are the
relative
costs?
Theories should be:
1. accurate in predictions
(across replications)
2. relatively simple
3. broad in scope
4. internally consistent
5. able to generate new research and
new findings
Theories should be:
1. accurate in predictions
(across replications)
*2. relatively simple > linear induction
3. broad in scope
4. internally consistent
*5. able to generate new research and
new findings
We are on the verge of breakthroughs in psychical phenomena.
William James, 1890
We are on the verge of breakthroughs in psychical phenomena.
William James, 1890
We are on the verge of vast development in psychic research.
Lord Rayleigh, 1919
We are on the verge of breakthroughs in psychical phenomena.
William James, 1890
We are on the verge of vast development in psychic research.
Lord Rayleigh, 1919
Parapsychology appears ready to make startling advances.
Time Magazine, 1974
We are on the verge of breakthroughs in psychical phenomena.
William James, 1890
We are on the verge of vast development in psychic research.
Lord Rayleigh, 1919
Parapsychology appears ready to make startling advances.
Time Magazine, 1974
We believe that the replication rates and the effect sizes achieved
by one particular method, the autoganzfeld procedure, are now
sufficient to warrant bringing this body of data to the attention of
the wider psychological community.
Bem & Honorton, 1994
Parapsychology
mediumship magicians vs. physicists > not fashionable
psychokinesis Uri Geller & James Randi > not fashionable
precognition seers > not fashionable, outside of
supermarket checkout lines
telepathy 1975, J.B. Rhine > not fashionable
AAAS accepts in 1969
clairvoyance autoganzfeld experiments > fashionable
case studies vs. scientific method
replicability
probability & statistics a priori chance
a posteriori base rate
> role of belief in science
Does use of the scientific method imply that the research is science?
> no, it does not,
but that also doesn’t mean individuals shouldn’t pursue their interests