In the preceding chapters we have seen that the characteristics of students—background, activities, interests, viewpoints and the like—are linked in some consistent ways to differential use (experience with) drugs. These trends show not only that students who use one drug tend to differ from those who do not use it but, more generally, that one set of characteristics—for example, family wealth or mobility or student school dissatisfaction—are linked to the use of all or nearly all drugs which constitute our spectrum of interest. There are exceptions. It is also important to realize that students with diverse characteristics are included in user and non-user groups; the larger the group, the more diverse the characteristics therein within the range of characteristics in the sample. Nevertheless, what emerges from these chapters is the impression that the use of drugs as such is a general phenomenon, that particular drugs, alike in either the style of use (that is, social, medical, illicit) or in their presumed probable pharmacological effects, "go together" in the sense that people who use one also use another and, conversely, that people who reject one also reject another. What these observations imply is that certain drugs are correlated or, expressed differently, that a regular association exists between a person's use of one drug and his use of another. Secondly, since it appears that there are subgroups of drugs in which the use of one is linked to another but not to drugs in a different subgroup, one may posit that intercorrelations exist among subgroups. Put another way, this means that clusters of people use certain groups of drugs. The search for intercorrelations specific to one subset and not to another is the search for factors and can be accomplished through factor-analytic procedures.

In this chapter we present results of a factor analysis showing the intercorrelations which stand out and demonstrate subsets or clusters of drugs that show communality in their use. Our factor-analytic procedures are based on normalized scores which correct for age— that is, on the z scores derived from the drug-profile scores of each person's lifetime drug history. Note that the significance of correlations rests, in some cases, upon the high number of cases, many of which may be zero cases. For example, a correlation of .07 can be significant at the .05 level because there are 1,312 degrees of freedom.

We see that each drug or drug class is significantly correlated with most, if not all, other drugs and that the order of correlation ranges from a low .07 (alcohol and sedatives, alcohol and tranquilizers) to highs of .55 (marijuana and hallucinogens) and .46 (illicit opiates and special substances). Note that the correlation with other drug use is least and lowest for sedatives and tranquilizers and highest for marijuana. What this shows is that use of sedatives and tranquilizers, while related to the use of other drugs, is a more independent phenomenon than in the case of marijuana; here, it is evident that marijuana users are also likely to be users of all other drug classes.

These preliminary observations on intercorrelations now lead us to the examination of common factors-that is, clusters of drug-use habits that link one class of drugs to another.

Table 9 below presents the principal-components solution, using as a criterion that the second last latent root (eigenvalue) be equal to or less than 1. (The latent root is the square root of the sum of the squared loadings for a given factor.) These calculations are used to determine whether the maximum (residual) variance is being accounted for by a given factor. With this method of factoring, each factor extracts the maximum amount of variance (that is, the sum of squares of factor loadings is maximized on each factor) and gives the smallest possible residuals. Thus, the correlation matrix is condensed into the smallest number of orthogonal factors. This method, applied to our data by Switalski and Bonato at the George Washington University Biometrics Laboratory, has the advantage of giving a mathematically unique (least squares) solution.

Sum of latent roots = 6.015. Percentage variance explained = 66.83. Proportion of total variance explained by each root: 1 = .296; 2 = .149; 3= .126; 4 = .097.
Proportion of cumulative explained variance accounted for by each added root: 1 = 1.000; 2 = .335; 3 = .221; 4 = .145.

From Table 9 we see that Factor I is found to account for the greatest proportion of the variances-that is, it yields the highest intercorrelation, one' which accounts for about 10 per cent of the observed drug-use patterns. We see that tobacco, amphetamines, marijuana, hallucinogens, illicit opiates, and special substances all have high loadings on Factor I. Another cluster finds sedatives and tranquilizers with high loadings; these are the only two drugs sharing Factor II. Alcohol and tobacco share high loadings on Factor III to the exclusion of all other drugs. (Note the consistent negative loadings on Factor III for all illicit drugs, plus amphetamines.) There is no important fourth factor; the highest loadings are in the .40 to .44 (rounded) range and are shared by hallucinogens, illicit opiates, and special substances.

The next step, presented in Table 10, offers the four-factor principal-components solution with rotations for four-, three-, and two-factor coordinate systems. Rotation is a procedure in which coordinates (factors) are moved so as to transform the data to optimize
meaning-that is, interpretation-by virtue of getting greater precision.

Using a cut-off level obtained by calculating the mean of the means for the two highest loadings of each drug across all factors, a value of .476, one sees four clear factors emerge. Factor I has high loadings for marijuana and hallucinogens only. Factor II has high loadings for sedatives and tranquilizers; Factor III for tobacco and alcohol, and Factor IV for amphetamines, illicit opiates and special substances. These are reasonably precise factors and differ from the results of Table 9 by differentiating out of Factor I in Table 9 two different substructures of illicit-exotic drug use, one linking marijuana and hallucinogens, the other among amphetamines, opiates, and special substances. We already know, from Table 8, that significant correlations do link marijuana with the amphetamines (r = .33) and with special substances (r = .27), which is why Factor I in Table 10 must be interpreted to represent substructures in illicit use rather than a dichotomy. It remains of interest that, on the most precise factor analysis, the amphetamines are linked to hard narcotics, volatile intoxicants, and other unusual materials. We cannot by means of factor analysis identify the underlying variables which account for this link, whether these be drug action, settings for use, personality of users, or what-have-you.

Although we reject it as a less than optimal solution, a three-factor rotation was also conducted, the results of which are presented in Table 11.

Underlined are loadings near or above the cut-off point of .453. We see in Table 11 that amphetamines and special substances are grouped with marijuana and hallucinogens under Factor Ia. These, however, are split loadings which, when combined with the fact of the high correlation between special substances and opiates (r = .46), point to the four-factor solution of Table 10 as optimal. Nevertheless, the three-factor solution in Table 11 merits comment, especially the high loadings under Factor lia for sedatives, tranquilizers, and illicit opiates. We see that Factor IIIa with its high tobacco and alcohol loadings is identical to Factor III in Table 10.

From the foregoing analysis, we believe it reasonable to propose that there is, first, a general drug-taking disposition. This is inferred from the consistent positive correlations found in Table 8 and is compatible with the distributions set forth in preceding chapters. The notion of a general drug-taking disposition is imprecise, and by factor analysis a set of discrete subordinate clusters of patterns emerge. The unrefined Factor I of Table 9 and Ia of Table 11 allow the presumption of one subordinate pattern which is of illicit-exotic drug use as such. Upon more precise analysis, Table 10, this is found to consist of at least two separate substructures.

There is Factor I on which marijuana and the hallucinogens have high loadings and there is Factor IV on which amphetamines, illicit opiates, and special substances have high loadings. Sedatives and tranquilizers, two medically prescribed classes of psychoactive drugs, have high loadings on Factor II. In regard to Factor II, note that another prescribable class, the amphetamines, are not to be found. Thus, although we are dealing with drugs capable of being prescribed, Factor II cannot be considered a prescription factor per se. This is a limited interpretation for, as we shall see later, amphetamines are often obtained by students under nonprescription circumstances. A potential link between Factor II and the amphetamines, one found among persons heavily involved with drugs on the up-and-down cycle (sedatives as "downers" and methamphetamines as "uppers"), does not appear. The implication is that the cycle has not made its appearance among students.

In discussing Factor II, we should not overlook the correlation appearing in Table 9 which shows that among all drug classes correlated with sedatives and tranquilizers, the illicit opiates yield the highest, r = .25 in both cases. Looking at Table 11, we see that all three drug classes share high loadings on Factor lib. Given the fact that tranquilizers, sedatives, and opiates have much in common in terms of probable pharmacological effects (Irwin, 1968) in reducing arousal-excitability, fighting behavior, responsiveness to pain, activity, learning, and probably anxiety, we posit that Factor II has a behavioral component, the desire for these suppressive effects. Factor lib bridges legal- and illicit-drug use whereas the final Factor II does not. We shall consider either factor as the distress- and activity-diminishing factor.

Factor III is found to underlie only tobacco and alcohol. It is obviously not linked to illicit-drug use, to prescription drug taking, or—given our assumptions about Factor II—to depressing centralnervous-system effects. Were there to be a component related to pharmacological alerting or stimulation, one would expect some strong link between these drugs and the amphetamines. In the absence of these presumed components, we shall consider Factor III a conventional social-drug-use cluster.

Returning to Factors I and IV, we are uncertain as to their interpretation. Both embrace illicit-exotic drug use, although the amphetamine loadings in Factor IV may bridge into prescription and informal amphetamine use as well. One can only speculate, suggesting, for example, that the marijuana-hallucinogen Factor I may reflect the psychedelic enthusiast or drug "intellectuals" or the curious drug-experimenter, whereas Factor IV may be a drug-immersion factor. Consider, for example, that from the findings in Chapter Five, the greatest number and variety of self-ascribed drug motives occur among users of opiates, special substances, and amphetamines, in that order. Given the high intercorrelations and the motivational polyvalence, we prefer the immersion construct. If so many functions are linked to drugs or if so many gratifications are connected to them, one may presume that the user's life is centered about drugs as such, a phenomenon implying dependency or addiction. If in this light one infers from the data reported in Chapters Eleven and Nineteen the increasing likelihood that students will be using methamphetamine and heroin intravenously, it would follow that Factor IV will be of increasing interest to epidemiologists. Manner of drug administration should also be considered as a possible component in Factor IV, since methamphetamine and heroin are both injected while heroin and the volatile intoxicants can both be sniffed. Consideration of Factors I and IV should also mark well the psychology of the street—that is, the characterizations which drug users apply to persons fancying one drug over another. Thus, there are acid heads, potheads, speed freaks, junkies, drunks, and others. It is possible that psychological components in the two factors include typologies which can be constructed on the basis of drug preferences, these preferences in turn probably being related to personality, social settings, and biochemical constants. We offer the foregoing as speculations; what is required is further investigation into the variables accounting for the observed factors.

CONCLUSIONS

Intercorrelations and factor analysis, supported by data from the other chapters, reveal a number of relationships capable of interpretation as patterns, communalities, or substructures linking drug use. We propose that there is, first, a general disposition toward psychoactive drug use as such. Although of a low order statistically, such an orientation reflects the widespread willingness to use a variety of drugs as tools to alter states of consciousness, biological cycles, and social relations. Much more precise are the subsets of dispositions which link particular drugs and exclude others. One factor tentatively identified is style of drug use by source, which has as separate components conventional social-drug use, the employment of illicit-exotic substances, and, less well supported, reliance on prescription drugs. Given the data in Chapter Five on home-remedy use, we would posit self-medication as a behavior linked to the prescription-drug cluster. A second 'factor tentatively identified concerns seeking similar drug effects among several classes of substances. In particular there is a component interpreted as pharmacological effects which diminishes activity and anxiety.

With regard to the more uncertain Factors I and IV, we propose—as a basis for exploration rather than final explanation—alterna,- tive components: immersion in drug use, an interpretation based on the multiplicity of after-the-fact functions said to be met by drug use along with high intercorrelations with other drug classes; a typology of users, which we presume reflects personality and social circumstances and is based on street types such as acid head, speed freak, junkie, and the like; and also manner of drug administration. The special reference is to the intravenous injection and sniffing ("snorting") which link methamphetamine, heroin, and the volatile intoxicants. Among these constructs, immersion is probably most closely linked to what is ordinarily considered addiction. The clusters revealed by factor analysis show the importance of discriminating among a variety of components in drug use.

* Here, and in the following two tables, negative as well as positive signs appear. The negative covariance should be examined in contrast to positive covariance only within columns. What is expressed is the relationship between the drug use and each factor. It may be conceived as an arbitrary description with a factor which is a vector in an n-dimensional field, thus the sign of the loading implies drug use and factor compatibility or antagonism. It is the magnitude which occupies our attention.