NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Research Council (US) and Institute of Medicine (US) Committee on Drug Use in the Workplace; Normand J, Lempert RO, O'Brien CP, editors. Under the Influence? Drugs and the American Work Force. Washington (DC): National Academies Press (US); 1994.

Cover of Under the Influence?

Under the Influence? Drugs and the American Work Force.

Show details

7Impact of Drug-Testing Programs on Productivity

Many employers in the United States have attempted to address the problems they perceive to be related to alcohol and other drug use by establishing drug-testing programs, as described in Chapter 6. What effects do drug-testing policies and programs have on people's productivity at work? Unfortunately, there have been few systematic studies relating these drug-testing programs to workers' productivity, and those that have been done are often flawed in significant ways. This chapter critically reviews the literature that does exist and discusses how the effects of such programs may best be evaluated. It also reviews the literature on the attitudes of workers and job applicants toward such programs and the potential effects of these attitudes on productivity.

We reiterate here an important point that was mentioned in Chapters 1 and 6. It is related to the methods and measures used by the studies reviewed here: most drug-testing programs do not include alcohol among the drugs to be tested. Executive Order #12564 signed by President Reagan in 1986 mandated testing for illicit drugs; subsequently, in 1988 the HHS ''Mandatory Guidelines for Federal Workplace Drug Testing Programs" limited the number of drugs to be tested to the following commonly used illicit drug classes: (1) marijuana, (2) opiates (heroin, morphine), (3) cocaine, (4) amphetamine and methamphetamine, and (5) phencyclidine. These guidelines have served as a model for most drug testing programs that have been implemented to date. The only exception to this omission of alcohol is the recent Omnibus Transportation Act of 1991, which requires the Department of Transportation to include alcohol as a target drug in their testing program. To date no evaluation of adding alcohol to the list of drugs to be tested for has been carried out.

Types And Uses Of Drug-Testing Programs

Workplace drug-testing programs are of three distinct types: (1) preemployment testing of job applicants; (2) incident-driven or for-cause testing of employees (e.g., post-accident, fitness for duty); and (3) postemployment testing without specific cause, often selected at random from a pool of targeted (usually sensitive) positions (Walsh et al., 1992). In 1988, 3 percent of the employers surveyed by the U.S. Department of Labor (1989) had some type of drug-testing program; by 1990, this figure had risen to 4 percent (Employee Assistance Professional Association, 1991). The majority of this testing has been undertaken by major corporations. Among the largest employers (with more than 1,000 employees) surveyed by the Department of Labor, 43 percent had some type of drug-testing program, whereas only 2 percent of the smallest establishments (with fewer than 50 workers) had such programs. More recent surveys of workplace drug testing indicate that between 50 and 75 percent of medium and large organizations now test current or prospective employees for drugs (American Management Association, 1992; Axel, 1990; Hayghe, 1991).

Preemployment drug testing is the most prevalent form of drug testing: 85 percent of those companies the Department of Labor surveyed with a testing program tested job applicants. Findings from the Monitoring the Future project (Patrick O'Malley, personal communication, 1992) show that increasing proportions of the work force have undergone testing for drug use in recent years: in 1987, 15 percent of men and 6 percent of women surveyed by the follow-up study reported having had a drug test; the corresponding figures in 1991 were 33 and 16 percent. In other words, by 1991, one in three young adult men and one in six women report having been tested. Most of these tests were preemployment drug tests.

Although the surveys reveal that corporate executives and human resources managers believe that drug-testing programs are effective tools for improving workplace safety, health, and productivity, there is little empirical evidence pertaining to their efficacy. To date, most evaluation attempts have consisted of simply monitoring and interpreting trends in laboratory tests results.

The main impetus for the rapid diffusion of these programs appears to have evolved from the preeminence given by the Reagan and Bush administrations to their "war on drugs" policies, from which the concept of a "drug-free workplace" first emerged, the substantial amount of publicity given to recent tragic accidents, and numerous government regulations and directives. Favorable court rulings have also substantially contributed to making drug testing common in corporate America.

Over the past decade, incumbent presidents and Congress have not only authorized drug testing by public and private employers, but also have required or encouraged it in some workplaces. Their actions provide legal authority for drug testing that overrides all contrary authority except the U.S. Constitution. The scope of workplace drug testing is limited by certain state and federal constitutional restrictions, particularly in the public sector and in postemployment settings; these limits, however, are generous and allow a broad range of employees to be tested using a wide range of reliable methodologies. A detailed treatment of the legal issues surrounding drug testing is provided in Appendix B. The argument here is that the policies of previous administrations, the legislature, and the judicial systems have all contributed to establishing drug testing as a major component of the nation's "war on drugs."

Unfortunately, to date, little consideration if any, has been given to assessing the impact of such programs on public health. Although this chapter is concerned with the effectiveness of workplace drug-testing programs on productivity, the reviewed literature still provides valuable information to policy makers. The emphasis on workplace productivity rather than public health or social welfare affects the evaluation method and criteria used in this chapter. In evaluating the efficacy of drug-testing programs, the researcher's perspective will influence what are regarded as acceptable program objectives, which in turn will determine the proper criteria for evaluating program effectiveness. Walsh et al. (1992) point out that, from a public health perspective, the primary question is whether drug intervention programs prevent or postpone any death, disease, disability, or dysfunction associated with the use or abuse of drugs.

From business or human resource management perspectives, the prevailing question is different from that of the public health perspective. In a business context, the primary concern is whether such programs allow the corporation to function more efficiently in a competitive market. Acceptable evaluation criteria are whatever business decision makers consider relevant productivity indices (e.g., absenteeism, work output, safety and health).

Job-site drug-testing programs have some promise in this respect. About three-fourths of adult men (16 years and older) and over half of adult women are in the labor force (Bureau of the Census, 1989). A sizable proportion of a worker's waking hours are spent at work, and a common sense of identity often exists within the work setting, where coworkers share norms and values. Workers are also subject to powerful influences that encourage conformity that can be used constructively to tackle problem behaviors such as alcohol and drug abuse. Formal and informal channels for communicating messages about drug abuse and treatment abound. Workplace drug programs thus have the potential for extensive social benefits. At the same time, such programs may have social costs. Some have argued, for example, that business drug intervention programs such as preemployment testing may aggravate society's drug problem by making those in need of assistance (e.g., applicants testing positive who are denied a job) unemployable.

Impact Of Drug-Testing Programs

Preemployment Drug Testing

Preemployment testing is not only the most common form of drug-testing currently in place in organizations, but it has also received the largest amount of attention from the research community. One of the earliest evaluation studies attempting to assess the efficacy of preemployment drug testing that appeared in the published literature was carried out by Lewy (1983), who reported that 13 of 500 hospital job applicants tested positive for one or more of 5 drug classes (benzodiazapine, barbiturates, amphetamines, phencyclidine, and opiates). Based on this limited descriptive information and the cost of testing, the author concluded that preemployment drug testing is not a financially viable selection procedure. A significant limitation of the study is that the investigator did not test for the most prevalent illicit drugs (marijuana and cocaine) and made no attempt to assess the relationship between test results and performance.

Parish (1989) also attempted to evaluate the effectiveness of an applicant drug-testing program at a large hospital. This study focused on the important question, which had not yet been addressed in the literature, of whether preemployment drug test results were associated with job performance indicators. All employees hired over a 6-month period were tested for illicit drug use. The drug test results were kept confidential and, after 1 year of employment, job performance measures were extracted from personnel files. The 22 employees who had tested positive were found to have had a 28 percent higher turnover rate and a 64 percent higher rate of receiving disciplinary warnings. Twenty-five percent of the identified drug users received poor performance evaluations from their supervisors, compared with 5 percent of those who tested negative. Despite these observed disparities, no statistically significant relationship was detected between drug test results and job performance characteristics. The author commented that the null results were probably due, at least in part, to the small number of cases in the drug-positive group, an observation that has since been confirmed by other reviewers (Normand et al., 1990; Zwerling et al., 1990). These cautions are well taken. Given the low power of the statistical tests used, the brief tenure of the employees, and the complete absence of hired applicants testing positive for cocaine, the study's failure to reject the null hypothesis provides little reason to believe that preemployment drug use cannot predict some employment behavior.

Blank and Fenton (1989) carried out a similar predictive study with U.S. Navy recruits who were screened for illicit drugs before being sent to recruit training. Only those who tested positive for marijuana alone or negative for all drugs were retained. The 500 recruits who tested positive for marijuana were compared with a matched group of approximately 500 recruits who tested negative for all drugs. The results showed that 43 percent of the marijuana positives but only 19 percent of the negatives had been discharged from the Navy after 2.5 years. Despite this dramatic difference, the study should not be read as establishing the efficacy of drug testing as a selection procedure, even in the naval context. When evaluating the predictive value of an employee selection procedure, it is essential that the research sample include applicants who are typical of the population to which the results will be generalized and that the procedure be assessed in a way that is consistent with its operational use (Society for Industrial and Organizational Psychology, 1987). In the Blank and Fenton study, the use of a matched control group of 500 recruits resulted in a prevalence rate of 50 percent in the studied sample, which is substantially higher than what would be expected in the population of interest (i.e., Navy recruit applicants). The inflation of the prevalence rate upwardly biases the estimated predictive validity coefficient (i.e., the correlation between test results and turnover). Furthermore, the drug-positive group consisted only of identified marijuana users, whereas most organizations with a preemployment drug-testing program screen applicants on a wide variety of illicit substances. More importantly, recruits who tested positive for marijuana at selection, though retained by the Navy, were subject to follow-up and occasional random testing. Failing a subsequent drug test was grounds for dismissal. Thus, at least some of the turnover was probably a direct result of ongoing efforts to detect and discharge drug-using sailors and was not necessarily based on actual performance decrements or behavioral problems that might have resulted from the use of drugs. Moreover, since recruits who had tested positive for marijuana before induction were more likely to be tested subsequently then controls, a higher rate of dismissal for subsequent drug use would not show that they were more likely than the control group members to have used drugs after induction.

A large-scale attempt to assess the predictive efficacy of preemployment drug testing was conducted at the Boston general mail facility (Zwerling et al., 1990). The authors of this prospective study tested 4,964 job applicants in the Boston division of the U.S. Postal Service at the time of their preemployment medical examinations. Data from the urinalysis were collected for research purposes only and had no bearing (with the exception of positive opiates) on hiring or other employment decisions. Outcome measures were later obtained from personnel files (subjects were followed up, on average, for 406 calendar days). Of those tested, 2,537 applicants accepted the postal job offer. A proportional hazards model was used to analyze the data and to statistically control for age, sex, smoking, exercise status, race, and job classification. Identified marijuana users showed increased risks (relative risk ratio) of termination (1.56) (i.e., positive marijuana employees were 1.56 times more likely than negatives to turnover), accidents (1.55), injuries (1.85), disciplinary actions (1.55), and absenteeism (1.56) when compared with employees who tested negative. Cocaine positives showed an increased risk of absenteeism (2.37) and injuries (1.85).

One problem with this study is that the representativeness of the study sample (and therefore the generalizability of the results) was compromised by arbitrarily deleting whole applicant subgroups from the analyses (e.g., Hispanics, Native Americans, employees in professional or technical positions). If these individuals belong to the population of job applicants, then excluding them from the study sample limits the generalizability of the results and biases the parameter estimates. Furthermore, the authors elected to operationalize drug test results into discrete categories (e.g., marijuana, cocaine) rather than defining it as it is typically operationalized (positive on any drug versus negative on all drugs). Also, the use of statistical control techniques for potential confounding factors is questionable (Lord, 1967; Cochran and Rubin, 1973; Reichardt, 1979) when the predictive value of a selection device is being assessed.

Normand and Salyards (1989) reported the results of a large-scale longitudinal study that assessed the contribution of preemployment drug testing to the prediction of future job success. In this study, 5,465 U.S. Postal Service job applicants at 21 sites across the country were tested for use of illicit drugs as part of their preemployment medical examination. Test results were kept confidential and had no bearing on any subsequent personnel actions. Job outcome measures for 4,396 applicants who were hired were extracted from personnel records at four different time points (see Figures 7.1 and 7.2). After having been on the job for an average of 8.2 months, employees who tested positive were found to be absent at a rate 45 percent higher than those who tested negative (4 percent of scheduled work hours versus 3 percent). The drug-positive group also showed a 40 percent higher dismissal rate compared with the group testing negative for drugs (13 compared with 10 percent). A follow-up study indicated that, after an average of 1.3 years on the job, the observed absence and turnover differences between the two groups had increased substantially (Normand et al., 1990). The positive group showed a 59 percent higher rate of absenteeism (7 compared with 4 percent) and a 47 percent higher rate of firing (15 compared with 11 percent) compared with those testing negative. Despite a high level of statistical power in this study (the probability of detecting even weak effects exceeded 0.95 in this study), no significant associations were detected between drug test results and accidents and injuries.

FIGURE 7.1. Absence rates at four time points.


Absence rates at four time points. NOTE: employee tenure: 8.2 months, 1.3 years, 2.4 years, and 3.3 years.

FIGURE 7.2. Firing rates at four time points.


Firing rates at four time points. NOTE: employee tenure: 8.2 months, 1.3 years, 2.4 years, and 3.3 years.

In July 1990, after having been on the job for an average of 2.4 years, both the absenteeism and turnover disparities across the two groups had further increased: the positive group's absenteeism and firing rates were found to be 66 and 69 percent higher, respectively, than those of the negative group (Normand and Salyards, 1991). Finally, after 3.3 years on the job, the disparity in absenteeism rates plateaued at 66 percent, whereas the disparity in firing rates had increased to 77 percent (Normand and Salyards, 1992). As part of this latest follow-up, the investigator examined more specific indicators of job absence. A notable finding was that the observed disparities in absenteeism rates across the two groups were mirrored when annual absence frequencies were examined (30.61 compared with 18.75). Furthermore, more specific analyses by type of leave (sick leave, leave without pay, and absent without official leave—AWOL) showed that the disparities across groups were greatest for AWOLs, the most dysfunctional leave type. This latest update also extracted information pertaining to disciplinary infractions (Figure 7.3) and employee assistance program referrals (Figure 7.4). Results indicated that the employees who tested positive for drugs at preemployment were much more likely to be formally disciplined (a risk ratio of 2.44) and to experience problems requiring the intervention of an employee assistance program (EAP) (a risk ratio of 2.67). Further analyses revealed that the most common disciplinary action was due to attendance infractions and that the most commonly diagnosed EAP assessment problem at induction was alcohol-related. Employees who tested positive were 3.47 times more likely to be referred to an EAP for drinking problems and 5.69 times more likely to be referred for drug abuse problems as those who tested negative. A similar pattern was observed when medical claims data were examined (see Figure 7.5). The drug-positive employees were 3.42 times more likely as drug-negative employees to file medical claims having alcohol or drug-related diagnosis. Accident and injury rates continued to show no difference.

FIGURE 7.3. Disciplinary actions: drug-test results by infraction.


Disciplinary actions: drug-test results by infraction.

FIGURE 7.4. EAP referrals: drug-test results by problem identification.


EAP referrals: drug-test results by problem identification.

FIGURE 7.5. Drug-test results by medical claims information: number, dollar amount, and drug-related claims.


Drug-test results by medical claims information: number, dollar amount, and drug-related claims.

The generalizability of the results of this study may be limited by the peculiarities of the work setting. That is, the positive rates and outcome measures distributions may not necessarily be reflective of other organizational settings. Until more empirical studies of this kind are carried out, one will not be able to determine with a high degree of accuracy to what extent these results can be generalized to other organizations. One factor that further complicates the generalizability issue is the change in prevalence rates of illicit drug use over time. The proportion of people in the general population who use illicit drugs has been declining relatively quickly since 1980. Such fluctuations in the prevalence of drug use are bound to affect the rate of positive preemployment drug-test results and consequently the effectiveness of preemployment drug-testing programs.

This study also highlights the importance of the perspective in which statistics are presented. For example, a 77 percent higher rate of firing after 3 years may suggest great benefits to preemployment drug screening and the policy of not hiring positive testers. However, as depicted in Figure 7.2, 76 percent of those who tested positive at preemployment had not been fired after 3 years, compared with 86 percent of those testing negative. This represents a 10 percent absolute difference between groups in retention rates, which appears to be less drastic than the reported 77 percent higher rate of firing.

For-Cause Testing

Only two published studies are available that attempt to evaluate the effectiveness of for-cause drug-testing programs (Crouch et al., 1989; Sheridan and Winkler, 1989). Each of these studies appeared in a nonrefereed publication.

Crouch et al. (1989) compared employees (N = 12) who tested positive in a for-cause drug-testing program to a control group (N = 47) selected from the total employee pool who had similar demographic characteristics at the Utah Power and Light Company. Retrospective employment information was collected over a 2-year period. The mean number of sick hours accumulated by the drug-positive group was 75.3 compared with the control mean of 55.8. Employees testing positive used an average of 63.8 hours of unexcused leave compared with 18.7 hours for the control group. An analysis of vehicle accidents revealed that the identified drug users were significantly more likely to be involved in an accident compared with the matched control group. These differences were all found to be statistically significant. Surprisingly, a comparison of the medical expenditures used by each group indicated that the control group's use of medical benefits ($719) actually exceeded that of the drug-positive group ($504).

Findings from a similar study conducted at Georgia Power Company (Sheridan and Winkler, 1989; Winkler and Sheridan, 1989) compared employees who tested positive in a for-cause drug-testing program to a control group on the use of medical benefits, absenteeism, and accidents. Drugpositive and control groups were matched on sex, age, ethnicity, job category, and length of service in an attempt to control for nonrandom selection of subjects. When compared with their controls, individuals testing positive used significantly more medical benefits ($1,377 compared with $163), showed greater absenteeism (165 compared with 47 hours of yearly absenteeism), and had a higher rates of vehicular accidents (23 compared with 11 percent).

These two studies, however, are fraught with both conceptual and methodological difficulties. From a conceptual view point, it is not clear whether the objective of these studies was to evaluate the impact of a for-cause drug-testing program or the impact of individual drug use on productivity indices. Because both studies attempted to isolate, through matching techniques, the effect of testing positive from potential individual confounding factors, the intent appears to be to assess the impact of drug use on outcome measures, rather than to isolate the effect of the for-cause drug-testing programs on outcome measures.

Methodologically, however, the matching technique used is inadequate to this task. Matching has serious, well-recognized methodological weaknesses (Campbell and Erlebacher, 1975; Campbell and Stanley, 1966; Cochran and Rubin, 1973; Craig and Metze, 1979; Kerlinger, 1986; Reichardt, 1979) that are exacerbated in these studies for two reasons. First, the drug-tested group may have been selected in part by behaviors related to the dependent variables and, second, drug use, which was a basis for separating the groups, is no doubt correlated with a number of variables on which the subjects are not matched and that are plausibly related to the dependent variables. Thus, these two studies cannot tell us whether for-cause drug testing deters drug use or affects an organization's productivity, and they provide no reliable information on the effects of drug use on job-related performance measures.

Random Drug Testing

The committee was not able to locate any published studies that examine the effects of random drug testing on the productivity of the work force. The published literature on the effects of random drug testing consists mostly of descriptive reports of trends. Deterrence effects of random testing programs are often inferred from observed decreases in positive drug test rates, which frequently follow the implementation of such programs (Osborn and Sokolov, 1989; Taggart, 1989). However, the evaluation of random drug-testing programs is complicated by the fact that very few, if any, corporations have implemented such programs in isolation. Most corporations with random drug-testing programs have also implemented applicant and/or for-cause drug-testing programs. This makes it difficult to isolate the effect of random drug testing on productivity. The trend studies done to date have not overcome these difficulties, for they lack the kinds of controls that would allow one to confidently attribute observed changes in drug-use patterns to the implementation of random drug-testing programs. Furthermore, even if a reduction in drug use could be attributed to the implementation of a random drug-testing program, this would not in itself be evidence that the program has affected productivity. More research and more sophisticated research is clearly needed. Information on the relative effectiveness of the various components of a universal drug-testing program (i.e., applicant, forcause, and random drug-testing programs) would be useful to business and could have substantial policy relevance.

Workplace Drug Testing and Productivity

What can we conclude from the extant research on the efficacy of drugtesting programs? Contrary to what certain popular publications or newsletters may indicate, there are few empirically based conclusions that may be reached concerning the effectiveness of drug-testing programs in improving workplace productivity. The two for-cause drug-testing evaluation studies published to date suffer from serious methodological problems that preclude any scientific assessment of the impact of for-cause testing on work force productivity, and no evidence evaluating the effects of random drug testing on worker productivity has yet appeared in the published literature. Enough studies of preemployment drug-testing programs have been published with sufficiently consistent results that we can conclude that preemployment drug-test results are, in at least some job settings, valid predictors of some job-related behaviors. Those who test positive for drugs before employment are, as a group, likely to have higher rates of absenteeism, turnover, and disciplinary actions than those who test negative (Blank and Fenton, 1989; Normand et al., 1990; Zwerling et al., 1990).

In addition, it appears that preemployment drug-test results are also predictive of subsequent alcohol and drug problems. Salyards (1993) reported that job applicants who tested positive were 3.47 times more likely to be referred to an EAP for drinking problems and 5.69 times more likely to be referred for drug abuse problems as those testing negative at preemployment. Furthermore, drug-positive individuals were 3.42 times more likely as other employees to file medical claims involving an alcohol-or drugrelated diagnosis. Blank and Fenton (1989) found that 14 percent of marijuana-positive Navy recruits were discharged for drug-or alcohol-related problems compared with 1 percent of those who tested negative.

These figures do not necessarily mean, however, that preemployment drug testing is a cost-effective selection program. Finding that job applicants' drug-test results are predictive of critical job behaviors means that employers can select, on average, more productive workers if they attend to drug test information than if that information is not used to make hiring decisions. However, as with any other type of selection program, the predictive efficacy of drug test results depends on a few critical selection parameters. In particular, the prevalence of drug use among potential job applicants has to be sufficiently large to yield meaningful measures of association between test results and the outcome measure and to justify the cost of testing (see Chapter 6 for discussion on low base rate). Thus a test that is cost-effective and predictively valid at one point in time may cease to be or, conversely, may become even more useful, if the pattern of drug use in the larger population changes.

Another important characteristic of most of the preemployment drug studies we have reviewed is that they have used analytical drug-testing procedures that conform with the guidelines of the National Institute on Drug Abuse. Many private organizations (especially smaller employers) that test today use less stringent testing procedures and methods that introduce sources of error that can be expected to make their programs less efficacious than the programs that have been studied to date (Murphy and Thornton, 1992a).

It is also possible for employers to overweight preemployment drug-test results. In the best studies done to date, employers chose applicants using the cues they ordinarily apply without reference to evidence of drug use. The research shows that, among this group, those who tested positive performed worse by certain job-related criteria than those who showed no evidence of recent drug use. The studies do not show that positive testers performed worse than those who would have replaced them had they been rejected on the basis of drug-test information. In particular, there is a danger that, if drug-test results trump other signs that warn of job difficulties, negative drug testers will be preferred to positive testers who exceed them on job-related criteria. A major gap in the extant research is its failure to examine the interaction of recent drug usage and other job-related applicant characteristics. Although the studies show preemployment drug testing to be predictively valid, they also indicate that many applicants who test positive could be hired without producing any job-related difficulties.

Conceptual Incongruities In Research Methods

Conceptual Issues

As stated at the beginning of this chapter, considerable confusion surrounds the basic evaluation research issue of how to assess the efficacy of drug-testing intervention programs. Most studies we reviewed claim to be investigating the same basic issue (i.e., the efficacy of drug-testing programs on productivity), and many employ similar designs. In many instances, however, the methods used to compare the work outcomes of drugpositive employees with those found to be drug-negative (or a matched drug-negative control group) are conceptually distinct from one another and reflect divergent research purposes. This state of affairs can be attributed to a lack of clarity in defining the research purpose.

Central to any scientific investigation is an analysis of the problem situation and a clear-cut formulation of the research purpose. Broadly speaking, a research study may be designed primarily for the purpose of explaining or predicting phenomena (e.g., see Cook and Campbell, 1979). In explanatory research, the emphasis is on identifying variables and understanding the processes by which they influence the phenomenon of interest. Establishing that covariation among variables exists is not sufficient. What is sought is an explanation as to why variables covary (Blalock, 1968). The hope is that by understanding the conditions that influence these phenomena, researchers and policy makers will be in a better position to recommend courses of action that will help to alleviate individual, organizational, and social woes. The task is difficult to accomplish. Most problems in the social and behavioral sciences have multiple and entangled causes that are often difficult to identify, isolate, and measure. Because the number of possible causal patterns among variables is virtually infinite, it is necessary to rely on theory to provide guidance and direction (James et al., 1982) in attempts to explain why observed differences exist. Theory confirmed by data can provide powerful explanations.

In predictive research, the emphasis is on studying the degree of covariation among variables and using this information to make predictions about subjects' future behavior or performance. Predictive research does not require a theoretical framework to proceed, even though theory may play a role in the initial development or selection of predictor variables and may be useful in generating finer predictions. Prediction is particularly important in practical applications, for example, the selection of applicants for employment, college entrance, and training programs. Prediction studies are descriptive by nature—they describe differences and examine their potential impact on aggregate outcomes.

The appropriateness of the methods used to assess the efficacy of an intervention program depends on the goals or purpose of the programs and the evaluation. If the primary objective for implementing drug-testing programs is to enhance work force productivity, what is crucial to the evaluation of preemployment testing is the predictive validity of drug test results (Society for Industrial and Organizational Psychology, 1987). If the drug test results are found to be valid predictors of critical job outcome measures, the average quality of new employees is likely to improve by implementing a selection program that takes drug test results into account. Evidence that such a program is predictively valid (i.e., that it enhances the average quality of the selected workers) means that the program has an effect on aggregate measures (e.g., mean value) of the outcome variables—not that drug use has an effect on the outcome variables. This is equally true of assessments of postemployment, for-cause, and random drug-testing programs. To evaluate their efficacy is to assess the impact of implementing the programs on productivity measures and not the impact of individual drug use on productivity measures. Within this framework, controlling for potential individual confounding variables (e.g., Crouch et al., 1989; Sheridan and Winkler, 1989; Winkler and Sheridan, 1989; Zwerling et al., 1990, 1992) is irrelevant and potentially erroneous.

There seems to be an implicit notion in the evaluation literature on drug intervention that negative employment behaviors or outcomes must be shown to be due to the effects of drug use, separate from other ''extraneous" influences, in order to establish the efficiency of a drug intervention program. This might possibly explain why so many drug intervention evaluation studies employ matching or statistical control procedures. The notion that drug use must be causally linked to critical employment criteria (otherwise why control for the effects of other influences?) goes well beyond the requirements for establishing the efficacy of drug intervention programs on productivity. As we note elsewhere in this report (see Chapter 5), it may be that drug use is more of a symptom of problematic job behavior than a cause, at least for the early stages of drug involvement.

These comments are not intended to detract from the importance of carrying out studies that attempt to untangle the complex causal relationship between drug use and problem behavior at work or elsewhere. Indeed, even though program evaluation and understanding human behaviors are different research issues, there is an important practical intersection. The better we understand the relationship between drug use and problem behavior, the more likely we are to be able to design an efficacious intervention program. In doing so, we can limit the unfairness that results from false negative predictions that have negative consequences for individuals. Moreover, in certain situations, it may be wrong or even illegal to employ valid selection programs. Thus there are reasons for research on drug-testing programs to control for individual variables. The reasons, however, are not to ascertain whether drug-testing programs have predictive validity but rather to further specify the conditions under which individual drug use influences individual performance. Because of conceptual confusion, statistical and matching techniques tend to be used for the former purpose.

The type or nature of drug-testing programs (preemployment, for-cause, or random) is another factor that is critical to the selection of an adequate evaluation strategy, because different types of drug-testing programs have different objectives. Program objectives determine the research purpose, which in turn dictate the choice of the study design and methods (e.g., variables to be measured, their operationalization, and statistical models) to be used. Unfortunately, most organizations that implement such testing programs do not articulate their program objectives clearly. This complicates matters for a researcher charged with evaluating the effectiveness of these programs. The lack of congruency between research purpose and method is an underlying source of much of the current conceptual and interpretational confusion regarding acceptable strategies for assessing the efficacy of drug intervention programs. To deal with this problem, researchers must secure clear statements of the program goals before designing their studies, and they must specify the goals that have motivated their evaluations in writing their research reports.

Cost-Effectiveness of Drug-Testing Programs

Two distinct audiences use the results of cost-effectiveness evaluations of intervention programs: policy makers and program administrators (or organizational decision makers). These groups need different types of information in order to make informed decisions. Policy makers, for example, are particularly concerned with the costs and benefits associated with various drug intervention programs (e.g., treatment modalities, testing) for society as a whole, while organizational decision makers need program-specific information pertaining to the costs and benefits of programs in their organizations or organizations like theirs.

Business decision makers want to know whether drug intervention programs enhance productivity, and if so, whether the consequent financial benefits justify program expenditures (relative to the costs of not intervening or to implementing another intervention). Unlike the cost of illness studies that are reviewed in Chapter 5, cost factors such as criminal activities, social welfare, incarceration, and early death were not typically included in utility models implemented by business. It is not that these factors are unimportant, but, from a business perspective, the interest is on relatively short-term return on investment at the organizational level. This is in contrast to policy decision makers, who are more interested in the long-term economic impact of broad policy change on society's welfare.

Only two peer-reviewed studies (Normand et al., 1990; Zwerling et al., 1992) have reported empirical estimates of the costs and benefits associated with implementing a drug-testing program. Both of these were carried out within the U.S. Postal Service and dealt with preemployment drug testing. In both instances the findings revealed that preemployment drug testing was a cost-efficient selection program. However, both studies neglected to include some important parameters into their models (e.g., variable cost, corporate taxes). Since there have been almost no empirical studies of the cost-effectiveness of drug-testing programs, it appears that decisions by organizations to adopt such programs have often been made without a well-grounded consideration of the likely benefits associated with their implementation. To the extent that governmental pressure has induced such programs, this is particularly likely to be the case.

There is a substantial body of literature that deals with the difficult issue of translating the statistical impact of organizational intervention programs into financial productivity indicators (i.e., dollars). Researchers in human resource management and related fields have applied the methods of utility analysis to estimate the probable benefits of programs ranging from ability testing to improved performance appraisal and feedback (see Boudreau, 1991; Jones and Wright, 1992; and Judiesch et al., 1992 for recent reviews of utility estimation techniques).

Both the utility and fairness of preemployment drug testing as a selection device depend in part on the type of employment errors that result from implementing such programs. Two types of errors are of concern. First, as discussed in Chapter 6, false positive classification error (mislabeling an individual as a drug user when in reality he or she is not a user) not only is disconcerting but also reduces the efficacy of the selection process. In the case of drug testing, it is well established that, when performed by a competent (NIDA-certified) laboratory, the rate of false-positive classification is virtually zero. But the superficial screening tests used by some businesses without confirmation increase the probability of false-positive classification error substantially. Second, and numerically more important, are what can be thought of as false-negative prediction errors. There will always be a number of applicants who test positive who would have been successful in their jobs had they been hired; to judge by the findings reported in the studies reviewed in this chapter, these falsely negative prediction errors will constitute the majority of drug-positive job applicants. False negative prediction errors also affect the efficacy of drug-testing programs, and minimizing them will increase a program's benefit-cost ratio.

This discussion of false-positive and false-negative errors should not obscure a central fact: the use of a valid selection program increases the proportion of successful new hires over what would be obtained had the selection program not been implemented. The prevalence of errors does affect the magnitude of the gains from a cost-benefit standpoint, so totally apart from fairness there are reasons to work to minimize them.

Effects Of Drug Testing On Attitudes And Morale

Another issue that merits discussion is the impact of drug-testing programs on the attitudes and morale of applicants and employees. Despite its increasing frequency, employee drug testing is still controversial. Several authors (e.g., Crant and Bateman, 1989; Murphy et al., 1990, 1991; Stone and Kotch, 1988) have suggested that workers and job applicants may react negatively to drug testing. Although different studies yield different estimates of the frequency of negative reactions, it is common for 40 to 50 percent of those surveyed to express reservations about employee drug testing (Hanson, 1990; Konovsky and Cropanzano, 1993; Murphy et al., 1990).

Attitudes toward drug testing may affect the behavior of applicants and incumbents, especially if drug testing involves policies or procedures that are objectionable to large numbers of individuals (Chadwick-Jones et al., 1982; Goodman and Friedman, 1971). In particular, attitudes toward practices such as drug testing may affect an individual's job search and job choice and his or her subsequent satisfaction with the job and the organization (Murphy and Thornton, 1992b; Schwab et al., 1987).

Murphy and Thornton (1992b) present evidence that individuals with negative attitudes toward drug testing are less likely to apply to, and may be less likely to accept, jobs in organizations that test for drugs. Their study suggests that attitudes toward drug testing and the probability that attitudes would affect subsequent job search behavior were largely unrelated to grades and academic qualifications, which implies that highly qualified applicants are as likely as less qualified applicants to be influenced in their job choices by their attitudes toward testing. Given the frequency of negative reactions to drug testing and the possible consequences of those attitudes, it is important to examine the ways in which specific characteristics of drug-testing programs affect attitudes and to explore ways in which negative reactions to drug testing might be minimized. The apparent efficacy of preemployment drug testing could be illusory if testing programs bias applicant pools so as to overrepresent those with few job options.

Influence of Job Characteristics

Reactions to employee drug testing vary substantially, depending on the job in question (Murphy and Thornton, 1992b; Murphy et al., 1991). In general, the higher the likelihood that impaired job performance could pose a danger to an individual, his or her coworkers, and the public, the higher the level of approval for drug testing. Thus, drug testing is likely to be seen as more acceptable for airline pilots than for accountants. Murphy et al. suggest that drug testing will also meet with higher levels of approval if the job involves activities or functions that are believed to be substantially impaired by drug use. Finally, there is evidence that drug testing is more common and more likely to be accepted in lower-level jobs than in managerial and executive jobs (Murphy and Thornton, 1992a).

Influence of Program Characteristics

There is considerable evidence that the policies and practices that define employee drug-testing programs can substantially affect reactions to drug testing. First, employee drug testing may be seen as an invasion of privacy, which is likely to lead to negative reactions (Stone and Stone, 1990). Urine testing is an especially sensitive procedure; the need to provide urine samples strikes many people as offensive, and the need to do so in front of witnesses or under tightly monitored conditions may seem particularly offensive. Privacy-related concerns might also be more salient in situations in which drug tests provide information about behavior outside the workplace. For example, it is well known that an individual who is a chronic user of marijuana can test positive for days or even weeks after having stopped using the drug, and that marijuana use that is completely divorced from the work setting can nevertheless lead to a positive drug test.

Research on attitudes toward drug testing and on perceptions of justice and equity in organizations (e.g., Konovsky and Cropanzano, 1993; Murphy et al., 1990) suggests that there are three factors that substantially affect the likelihood of negative reactions. First, drug-testing programs vary in the extent to which they are seen as reasonable. Testing programs that are restricted to high-risk occupations, or that are clearly and convincingly justified by management, are not likely to be a significant source of controversy. Second, drug-testing programs vary in the extent to which their overall orientation is seen as punitive. Testing programs that result in severe or irrevocable sanctions (e.g., dismissal) are more likely to be seen in a negative light than programs that are designed to help people deal constructively with substance abuse (e.g., by recommending counseling). Finally, drug-testing programs that involve consultation between labor and management are less likely to be seen in a negative light than those that are unilaterally imposed by management.

Limitations of Research on Reactions to Drug Testing

In evaluating research on attitudes toward drug testing, there are two reasons for caution in making broad generalizations. First, the majority of the studies in this area have employed convenience samples, usually college students. The attitudes of college students might differ in a number of ways from those of a general work force population (Murphy et al., 1991). Second, these studies often employ simulation methods (e.g., asking students to go through a simulated job interview) whose external validity is unclear (for an exception, see Murphy and Thornton, 1992b). The fact that an individual is willing to turn down a job offer in an experimental simulation does not necessarily predict behavior in real job interviews.

One of the few studies that has investigated the reactions of work force members is the High School Senior follow-up surveys (Patrick O'Malley, personal communication, 1992). These follow-up surveys show that, in general, most young adults are supportive of both preemployment and postemployment drug testing. In 1991, 65 percent "approved" or "strongly approved" of urine tests as a condition for getting a job like their own job, and 60 percent approved of urine tests as a condition for keeping a job such as their own. These figures reflect increases of about 15 percent in approval since 1987; for getting a job, the 1987 figure was 49 percent, and for keeping a job, the 1987 figure was 46 percent.

Approval rates have consistently been slightly higher among those who had actually been required to take a urine test. In 1991, among those who had been tested, either pre-or postemployment, 78 percent approved of preemployment testing compared with 61 percent of those who had not been tested. Postemployment testing was approved by 70 percent of those who had been tested, compared with 57 percent of those who had not been tested. These differences may reflect the types of people who get tested, positive experience with drug testing, or cognitive dissonance resulting from consent to a previously disapproved procedure.

On the basis of the research literature reviewed in this chapter, the committee provides the following conclusions and recommendations.

Conclusions And Recommendations

  • The empirical evidence pertaining to the efficacy of preemployment drug testing indicates that such programs may be useful to employers in choosing wisely among job applicants. However, regardless of the magnitude of the correlations between drug use and dysfunctional job behavior measures, the practical effectiveness of any drug-testing program depends on other parameters such as the prevalence of drug use in the population tested. The presence of significant relationships between drug use and workplace performance measures does not necessarily mean that an effective drug-testing program will substantially improve work force performance, and a program that substantially improves performance with some employees or in some job settings may do little to improve performance with other employees or in other job settings.
  • Despite beliefs to the contrary, the preventive effects of drug-testing programs have never been adequately demonstrated. Although, there are some suggestive data (e.g., see the military data in Chapter 3) that allude to the deterrent effect of employment drug-testing programs, there is as yet no conclusive scientific evidence from properly controlled studies that employment drug-testing programs widely discourage drug use or encourage rehabilitation.

Recommendation: Longitudinal research should be conducted to determine whether drug-testing programs have deterrent effects.

  • Many studies of alcohol and other drug use by the work force have been flawed in their design and implementation. Organizations that conduct their own drug studies can, by encouraging their researchers to publish in professional journals, enhance quality control and contribute to a knowledge base that will enable them to deal more effectively with future alcohol and other drug problems.
  • Different objectives have been suggested for work site drug-testing and diverse alcohol and other drug intervention programs. These include improving workers' performance, preventing accidents, saving on health costs, and working toward a drug-free society by deterring drug use. The effectiveness of alcohol and other drug intervention programs cannot be adequately evaluated unless the goals of such programs are clear.

Recommendation: Organizations should clearly articulate their objectives prior to initiating alcohol and other drug intervention programs and should regularly evaluate their programs in light of these objectives.

Among job applicants and workers, testing for drugs other than alcohol is already common and generally accepted. Of young men in a 1991 general population survey of high school graduates, 33 percent reported that they had been tested, 61 percent reported that they approved of preemployment testing, and 60 percent reported that they approved of postemployment testing. Approval rates were even higher among those who had been tested.

• Very little is known about what happens to job applicants who are not hired or to employees who are fired as a consequence of a positive drug test.

Recommendation: Research should be conducted on the impact of drug-testing programs with attention not only to those who remain within the organization as well as to those who are not hired or are dismissed. In particular, more information is needed about the impact of drug-testing programs on the health and productivity of the work force.

Recommendation: In light of the relatively low rates of alcohol and other drug abuse among the work force (see Chapter 3), the moderate predictive validity of testing programs (see Chapter 7), and the fact that many factors other than drug use may cause performance deficiencies seen in drug users (see Chapter 5), drug-testing programs should not be viewed as a panacea for curing workplace performance problems. Nonetheless, drug-testing for safety-sensitive positions may still be justified in the interest of public safety.


  • American Management Association 1992. 1992. AMA Survey on Workplace Drug Testing and Drug Abuse Policies. New York: American Management Association.
  • Axel, H. 1990. Corporate Experiences with Drug Testing Programs. New York: The Conference Board.
  • Blalock, H.M. 1968. Theory building and causal inferences. In H.M. Blalock, editor; and A.B. Blalock, editor. , eds., Methodology in Social Research. New York: McGraw-Hill.
  • Blank, D.L., and J.W. Fenton 1989. Early employment testing for marijuana: demographic and employee retention patterns. Pp. 139-150 in S.W. Gust, editor; and J.M. Walsh, editor. , eds., Drugs in the Workplace: Research and Evaluation Data. Rockville, Md.: National Institute on Drug Abuse. [PubMed: 2554140]
  • Boudreau, J.W. 1991. Utility analysis for decisions in human resource management. Pp. 621-746 in M. Dunnette, editor; and L. Hough, editor. , eds., Handbook of Industrial and Organizational Psychology, Vol. 2. Palo Alto, Calif.: Consulting Psychologists Press.
  • Bureau of the Census 1989. Statistical Abstract of the United States, 1988. Washington, D.C.: U.S. Department of Commerce.
  • Campbell, D.T., and A. Erlebacher 1975. How regression artifacts in quasi-experimental evaluations can mistakenly make compensatory education look harmful. Pp. 597-617 in E.L. Struening, editor; and M. Guttentag, editor. , eds., Handbook of Evaluation Research. Beverly Hills, Calif.: Sage Publications.
  • Campbell, D.T., and J.C. Stanley 1966. Experimental and Quasi-Experimental Designs for Research. Chicago: Rand McNally.
  • Chadwick-Jones, J.K., N. Nicholson, and C. Brown 1982. The Social Psychology of Absenteeism. New York: Praeger.
  • Cochran, W.G., and D.B. Rubin 1973. Controlling bias in observational studies: a review. Sankhya, The Indian Journal of Statistics 35:417-446.
  • Cook, T.D., and D.T. Campbell 1979. Quasi-Experimentation: Design and Analysis Issues for Field Settings. Chicago: Rand McNally.
  • Craig, J.R., and L.P. Metze 1979. Methods of Psychological Research. Philadelphia, Pa.: W.B. Saunders.
  • Crant, J.M., and T.S. Bateman 1989. A model of employee responses to drug-testing programs. Employee Responsibilities and Rights Journal 2:173-190.
  • Crouch, D.J., D.O. Webb, L.V. Peterson, P.F. Buller, and D.E. Rollins 1989. A critical evaluation of the Utah Power and Light Company's substance abuse program: absenteeism, accidents, and costs. Pp. 169-193 in S.W. Gust, editor; and J.M. Walsh, editor. , eds., Drugs in the Workplace: Research and Evaluation Data. Rockville, Md.: National Institute on Drug Abuse. [PubMed: 2509923]
  • Employee Assistance Professional Association 1991. EAPA Exchange October:23-25. Employee Assistance Professional Association, Alexandria, Va.
  • Goodman, P., and A. Friedman 1971. An examination of Adams' theory of inequity. Administrative Science Quarterly 16:271-288.
  • Hanson, A. 1990. What employees say about drug testing. Personnel July:32-36.
  • Hayghe, H.V. 1991. Anti-drug programs in the workplace: are they here to stay? Monthly Labor Review 114:26-29.
  • James, L.R., S.A. Mulaik, and J.M. Brett 1982. Casual Analysis: Assumptions, Models, and Data. Beverly Hills, Calif.: Sage Publications.
  • Jones, G.R., and Wright, P.M. 1992. An economic approach to conceptualizing the utility of human resource management practices. Pp. 271-300 in G. Ferris, editor; and K. Rowland, editor. , eds., Research in Personnel and Human Resources Management, Vol. 10. Greenwich, Conn.: JAI Press.
  • Judiesch, M.K., F.L. Schmidt, and M.K. Mount 1992. Estimates of the dollar value of employee output in utility analysis: an empirical test of two theories. Journal of Applied Psychology 77:234-250.
  • Kerlinger, F.N. 1986. Foundations of Behavioral Research, 3rd ed. New York: Holt, Rinehart, and Winston.
  • Konovsky, M.A., and R. Cropanzano 1993. Justice considerations in employee drug testing. In R. Cropanzano, editor. , ed., Justice in the Workplace: Approaching Fairness in Human Resource Management. Hillsdale, N.J.: Erlbaum. (In press.)
  • Lewy, R. 1983. Preemployment qualitative urine toxicology screening. Journal of Occupational Medicine 25:579-580. [PubMed: 6886865]
  • Lord, F.M. 1967. A paradox in the interpretation of group comparisons. Psychological Bulletin 68:304-305. [PubMed: 6062585]
  • Murphy, K.R., and G.C. Thornton, III 1992. a Characteristics of employee drug testing programs. Journal of Business and Psychology 6:295-309. 1992. b Development and validation of a measure of attitudes toward employee drug testing. Educational and Psychological Measurement 52: 189-201.
  • Murphy, K.R., G.C. Thornton, III, and D.H. Reynolds 1990. College students' attitudes toward employee drug testing programs. Personnel Psychology 43:615-631.
  • Murphy, K.R., G.C. Thornton, III, and K. Prue 1991. The influence of job characteristics on the acceptability of employee drug testing. Journal of Applied Psychology 76:447-453. [PubMed: 1864814]
  • Normand, J., and S.D. Salyards 1989. An empirical evaluation of preemployment drug testing in the United States Postal Service: interim report of findings. Pp. 111-138 in S.W. Gust, editor; and J.M. Walsh, editor. , eds., Drugs in the Workplace: Research and Evaluation Data. Rockville, Md.: National Institute on Drug Abuse. 1991. Applicant Drug Testing: Update of a Longitudinal Study. Paper presented at the 1991 Annual Convention of the IPMA Assessment Council, Chicago, Illinois. 1992Applicant Drug Testing: An Update of the Empirical Validation Evidence. Paper presented at the 7th annual conference of the Society for Industrial and Organizational Psychology, Montreal.
  • Normand, J., S.D. Salyards, and J.J. Mahoney 1990. An evaluation of preemployment drug testing. Journal of Applied Psychology 75:629-639. [PubMed: 2286599]
  • Osborn, C.E., and J.J. Sokolov 1989. Drug use trends in a nuclear power company: cumulative data from an ongoing testing program. In S.W. Gust, editor; and J.M. Walsh, editor. , eds., Drugs in the Workplace: Research and Evaluation Data. NIDA Research Monograph 91. Rockville, Md.: National Institute on Drug Abuse. [PubMed: 2509935]
  • Parish, D.C. 1989. Relation of the pre-employment drug testing result to employment status: a one-year follow-up. Journal of General Internal Medicine 4:44-47. [PubMed: 2915272]
  • Reichardt, C.S. 1979. The statistical analysis of data from nonequivalent group designs. Pp. 147-205 in T.D. Cook, editor; and D.T. Campbell, editor. , eds., Quasi-Experimentation: Design and Analysis Issues for Field Settings. Chicago: Rand NcNally.
  • Salyards, S.D. 1993. Preemployment drug testing: associations with EAP, disciplinary, and medical claims information. PharmChem Newsletter 21(1):1-4.
  • Schwab, D.P., S.L. Rynes, and R.J. Aldag 1987. Theories and research on job choice. In K. Rowland, editor; and G. Ferris, editor. , eds., Research in Personnel and Human Resource Management, Vol. 5. Greenwich, Conn.: JAI Press.
  • Sheridan, J.R., and H. Winkler 1989. An evaluation of drug testing in the workplace. In S.W. Gust, editor; and J.M. Walsh, editor. , eds., Drugs in the Workplace: Research and Evaluation Data. NIDA Research Monograph 91. Rockville, Md.: National Institute on Drug Abuse.
  • Society for Industrial and Organizational Psychology 1987. Principles for the Validation and Use of Personnel Selection Procedures, 3rd ed. College Park, Md.: Division of Industrial-Organizational Psychology, American Psychological Association.
  • Stone, D.L., and D.A. Kotch 1988. Individuals' attitudes toward organizational drug testing policies and practices. Journal of Applied Psychology 74:518-521. [PubMed: 2737995]
  • Stone, E.F, and D.L. Stone 1990. Privacy in organizations: theoretical issues, research findings, and protection mechanisms. In G. Ferris, editor; and K. Rowland, editor. , eds., Research in Personnel and Human Resource Management, Vol. 8. Greenwich, Conn.: JAI Press.
  • Taggart, R.W. 1989. Results of the drug testing program at Southern Pacific Railroad. In S.W. Gust, editor; and J.M. Walsh, editor. , eds., Drugs in the Workplace: Research and Evaluation Data. NIDA Research Monograph 91. Rockville, Md.: National Institute on Drug Abuse.
  • U.S. Department of Labor 1989. Survey of Employer Anti-Drug Programs. Washington, D.C.: U.S. Department of Labor.
  • Walsh, D.C., L. Elinson, and L. Gostin 1992. Worksite drug testing. Annual Review of Public Health 13:197-221. [PubMed: 1599585]
  • Winkler, H., and J. Sheridan 1989. An Analysis of Workplace Behaviors of Substance Abusers. Paper presented at the National Institute on Drug Abuse conference on Drugs in the Workplace: Research and Evaluation Data, Bethesda, Md.
  • Zwerling, C., J. Ryan, and E.J. Orav 1990. The efficacy of preemployment drug screening for marijuana and cocaine in predicting employment outcome. Journal of the American Medical Association 264:2639-2643. 1992 Costs and benefits of preemployment drug screening. Journal of the American Medical Association 267: 91-93. [PubMed: 2232039]
Copyright 1994 by the National Academy of Sciences. All rights reserved.
Bookshelf ID: NBK236244


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (2.6M)

Related information

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...