On Being Stoned
Charles T. Tart, Ph. D.
Chapter 5. Methods of Analysis
ALL OF THE CHAPTERS in Part II, Phenomenology of Marijuana Intoxication,
are organized along the same general plan, for the convenience
of the reader. I shall outline the basic plan, give definitions
of terms, and present descriptions of methods here.
BASIC PLAN
General Format
Each chapter consists of the results of potential effect descriptions
(questions, items) dealing with a single area, such as vision,
thought processes, etc. Within each chapter are subgroupings of
related questions.
For each question I have given: (1) the actual wording used in
the questionnaire; (2) the percentage[1]
of users responding in each of the frequency of occurrence and
minimal level of intoxication categories; and (3) differences
in the effect related to the background variables when such differences
were statistically significant.
When the wording of a question does not completely explain the
nature of the effect, I have added explanatory comments, based
on my interviews with pilot subjects and informants. Many effects
deal with areas of knowledge that are not generally well known
even among scientists, such as those concerning meditation or
ostensible paranormal phenomena, so I have given literature references
to guide the reader seeking more understanding. I have tried to
avoid speculation and interpretation as much as possible and to
stick to the basic findings.
Each chapter also contains a section on additional effects, a
ranking of effects according to increasing minimal levels of intoxication,
a summary of background factors modulating the effects, and a
general summary.
Terminology
It is impossible to write about these phenomena in a readable
style without using descriptive adjectives. To avoid the ambiguity
usually inherent in quantity adjectives, I have used a standard
set of them, which are defined in Table 5-1. Whenever other adjectives
than those defined are used, I am speaking generally rather than
describing the exact form of the data.
To illustrate: if an intoxication effect is described as "very
characteristic" and "primarily beginning to occur at
Moderate levels," this indicates that more than 50 percent
of the users rated this effect as occurring Very Often or Usually
when they have been intoxicated in the last six months, and my
judgment of the distribution of responses on minimal levels of
intoxication is that the Moderate ("Fairly Stoned")
level is the most representative[2]
level indicated.
TABLE 5-1
DEFINITION OF TERMS
TERM | DEFINITION |
|
Frequency of Occurrence Terms | |
"Rare" | >/=75% indicate Never, Rarely |
"Infrequent" | >/=50% indicate Never, Rarely |
"Fairly Frequent" | </=50% indicate Sometimes, Very Often, Usually[a] |
"Common" | >/=50% indicate Sometimes, Very Often, Usually |
"Very Common" | >/=75% indicate Sometimes, Very Often, Usually |
"Characteristic" | 50% indicate Very Often, Usually |
"Characteristic" | Bottom third of distribution |
"More Characteristic" | Middle third of distribution |
"Very Characteristic" |
"Most Characteristic" | Top third of distribution |
"Extremely Characteristic" |
Levels of Intoxication Terms | |
"Low" | Questionnaire term Just |
"Moderate" | Questionnaire term Fairly |
"Strong" | Questionnaire term Strongly |
"Very Strong" ("Very High") | Questionnaire term Very Strongly |
"Maximum" ("Very High") | Questionnaire term Maximum |
|
a. Infrequent and Fairly Frequent are not always
identical in practice
because of variable numbers of users skipping
particular questions.
Linking
Many pairs or sets of question called for statistical comparison
because of obvious similarity or because they described converse
effects. This was always done by a chi-square test of the distributions.
I have usually presented graphical results when they would be
illustrative, as well as the probability figures.
Many other links exist that I have not analyzed in the text. The
reader interested in particular comparisons may perform such analyses
himself from the percentage data presented for each item. Only
slight errors will result from using percentages rather than the
raw data I worked from.
Background Variables
The background information on the first page of the questionnaire
was used to divide the users into a number of groups, and every
question was subjected to a chi-square analysis for differences
in the distributions among the groups. Only significant (p <
.05) differences are presented in the text.
The groups compared were as follows:
Males versus females. Forty-nine percent of the
users were men, 27 percent women. The remainder were not used
in male-female comparisons because this question was inadvertently
left off some of the questionnaires.
Older and younger users were defined as those 25
years of age or older versus those from 16 to 24.
Educational Level was compared for the College-educated
(at least some college up to and including bachelor's degree or
equivalent) versus the Professionals (graduate training
or master's or doctor's degrees). The users with only a high school
education were too few (6 percent) to constitute a group for valid
analysis and so were omitted from the educational level comparison.
Frequency of use of marijuana in the last six months was
broken into three groups: the Occasional user ("occasional"
or "less than once/month" on the questionnaire), the
Weekly user ("once/week or more"), and the Daily
user ("almost every day or more"). With a three-way
classification, it was found that some of the frequency and intoxication
level categories had to be combined to avoid having too many cells
with low expected frequencies for the chi-square tests,[3]
so all analyses with three-way classifications were done against
frequencies of Never, Rarely/Sometimes, and Very Often/Usually.
Similarly, levels were uniformly condensed into Just, Fairly/Strongly,
and Very Strongly/Maximum.
Because a given degree of marijuana use in the last six months
might mean different things for one user who had followed that
pattern for ten years and for another who had used it for just
one year, a three-way analysis was also made for total marijuana
use. Categories were Heavy Total users, Moderate
Total users, and Light Total users. These categories
were obtained in the following way. Using the number of uses per
month as a basic unit, the self-rated frequency of use over the
user's whole use-history was assigned the value of 20/month ("almost
every day or more"), 8/month ("once/week or more")
or 2/month ("once/month or more" plus "occasionally").
Total length of time in years that the users had used marijuana
was weighted as I for one year or less, 2.25 for three years or
less, and 6 for more than three years.
The combinations of these weightings are shown in Table 5-2. They
fell into three natural groupings, which were designated the Heavy
(21 percent of the users), Moderate (44 percent), and Light (32
percent) Total users. A few users did not provide enough information
to be classified.
Users and Non-users of Psychedelics were classified
on the basis of whether they had ever used LSD, mescaline,
peyote, psilocybin, dimethyltryptamine (DMT), diethyltryptamine
(DET), STP (2, 5dimethoxy-4-methylamphetamine), MDA (3, 4-methylene
dioxy-amphetamine) or PEACE (a street drug supposed to contain
phencyclidines, such as we legitimately market under the trade
name Ketamine or Sernyl). Seventy-two percent of the users had
tried at least one of these powerful psychedelic drugs at least
once.
TABLE 5-2
DIVISION FOR TOTAL MARIJUANA USE:
WEIGHTING FACTORS
FREQUENCY OF USING MARIJUANA IN TOTAL USE PERIOD | LENGTH OF TIME MARIJUANA HAS BEEN USED
|
ONE YEAR OR LESS | THREE YEARS OR LESS | FOUR OR MORE YEARS |
|
Almost every day | 20 | 45 | 120 |
Once a week or more | 8 | 18 | 48 |
Once a month or more or occasionally | 2 | 4.5 | 12 |
|
Light Total Use: figures in italics
Heavy Total Use: figures in boldface
The final background analysis, dealing with commitment to personal
growth, divided the users into Meditators, the Therapy
and Growth Group, and Ordinary Users. Meditators were
so classified if they indicated that they regularly practiced
some form of meditation. They comprised 16 percent of the users.
The Therapy and Growth group were those who indicated they had
been in regular psychotherapy (2 percent) or the new growth-oriented
therapies (5 percent), such as Gestalt therapy (Perls, Hefferline,
& Goodman, 1951) or encounter groups (Schutz, 1967). Ordinary
users may have tried meditation exercises or the like occasionally,
but did not indicate any regular, systematic approach to personal
growth as the other two groups did.
Additional Effects
This section includes any further phenomena, volunteered by the
users at the end of the questionnaire, that were not already covered
in one of the regular questions. These have not been included
in any tabulations or analyses, and are added in each chapter
to further indicate the range of effects.
Levels of Intoxication
Except when there are too few effects of a given type to warrant
it, each chapter has all the effects discussed ordered by the
representative minimal level of intoxication. Categories are the
five divisions of level of the questionnaire (Just, Fairly, Strongly,
Very Strongly, Maximum) and levels halfway between these. Relevant
effects from other chapters also appear in the graphs.
Within each level, effects are ordered in terms of the arithmetic
mean of the intoxication levels reported, from lowest at the bottom
to highest at the top. Within a level, chi-square tests
of the distributions practically never reach significance. Overall
differences in levels for the phenomena of a particular chapter
were tested by a chi-square test using the lowest level (by arithmetic
mean) effect within each level category as the entry for that
level. They were usually extremely significant.
Variations in type style are also used in these graphs to indicate
the frequency of occurrence of an effect. Characteristic phenomena
are in bold capital letters, common are in bold lower case, infrequent
(fairly frequent is combined with infrequent here) in small capitals,
and rare phenomena are set in capitals with lower case letters.
Thus if one wants to know what is very likely to happen at various
levels for a given category of phenomena, one can look only at
the characteristic or common effects (in boldface). If one wants
to flesh this out with what may also happen if psychological factors
assume the correct values, all the phenomena may be looked at.
I have occasionally inserted question marks after particular phenomena
on the graphs, indicating that comments of informants raise some
doubts as to its fitting into the minimal level model, i.e., it
may cease to be available after some higher level.
Modulating Factors
Each chapter contains a table summarizing the effects of all significant
background factors. I have combined the categories of frequency
of use of marijuana in the last six months, total marijuana use,
and psychedelic drug use into a single category of more drug
experience for convenience here. The reader who needs these
separated can go back to the original item descriptions in the
text.
Almost all background variables had relatively linear effects.
Where they did not, the text in this section mentions the fact,
and they are not included in the table.
Statistical Notes
In addition to the various statistical considerations mentioned
above, it should be realized that about 5 percent of the significant
differences reported herein are due only to chance, i.e., are
not really reflecting a genuine effect. In the many thousands
of comparisons made in this large mass of data, 5 percent will
come out at the .05 level of probability by chance alone. I debated
on whether to try to eliminate these false positives, but the
only way would be by the criterion of whether the differences
"made sense" to me. Rather than impose my judgment on
the data, I have let it stand. As the main purpose of this study
is to stimulate research rather than provide final answers on
the nature of marijuana intoxication, these occasional false positives
will be weeded out by lack of confirmation in future studies.
Footnotes
1. I have generally used percentages rather
than actual numbers for clarity of presentation, All statistical
tests, however, were performed on the raw data to avoid the slight
rounding errors involved in using percentages. (back)
2. While it would have been possible to assign
the intoxication levels the values 0, 1, 2, 3, and 4 and use the
arithmetic mean as the average value, I did not want to make the
questionable assumption of equal intervals between categories.
Also, many of the distributions were highly skewed, so I would
judge the most representative intoxication level as half-way between
two of the defined levels. In practice, a correlation between
my judgments and arithmetic means would be extremely high. (back)
3. The technical question of how many cells
in a chi-square table can have expected frequencies below a certain
value is still hotly debated in the psychological literature.
Rather than arbitrarily combine the data on every question in
ways to eliminate low expected values, I have used the uniform
rules above, plus the rule, used only rarely, that in any chi-square
table with more than four cells having expected frequencies of
less than five I would combine whichever end category eliminated
the largest number of low cells with the adjacent category, i.e.,
Never or Just with Rarely or Fairly, etc. If this was not sufficient,
the analysis was thrown out. Allowing as many as four cells to
have low expected values is a fairly liberal position, but seemed
appropriate in an initial exploration of an important area. (back)