Exploring Internet addiction: Demographic characteristics and stereotypes of heavy Internet users

INTRODUCTION

With the increased role of the Internet in society and commerce, some of its users have become addicted; these addicts exhibit a variety of symptoms that are accompanied by a variety of negative outcomes. Expert estimates of the extent of this addiction range from 5-10% of Internet users (34, 41). Since the number of U.S. users is estimated to be over 100 million and rising, several million Americans may be addicted. Shapira and fellow researchers believe that Internet addiction will get worse as the size and speed of the Internet continue to increase (36). In this current study, the researchers use the term Internet addiction as an umbrella term for the various technological disorders described by psychologists.

The most common symptom of Internet dependence is the amount of time spent online. The researchers analyzed a well-known and publicly available dataset to determine which independent variables explain the amount of time spent online. This report reviews general addiction types as well as symptoms and outcomes of Internet addiction. If the researchers' models point to a "profile" of behaviors and attributes that lead to addiction, then counselors, human resource managers, IT managers, and IT trainers will have additional tools with which to identify and to interdict problems. The analysis is limited to Internet users in the U.S. and uses GVU survey data. The data were collected from online respondents who were self-selected. The data analysis covers only variables and attributes related to time spent online and does not cover cross-addiction or the criteria of compulsion. The researchers will pay special attention to stereotypes of who is prone to Internet addiction and, where permitted by the data, test the validity of such stereotypes.

PREVIOUS RESEARCH

The researchers have reviewed the prior research and have organized it to answer the following questions.

* What is addiction?

* Does Internet addiction exist?

* What are the types of technological addiction?

* What are the symptoms of Internet addiction?

* What are the outcomes of Internet addiction?

* How extensive is the problem of Internet addiction?

* What groups are vulnerable to Internet addiction?

Addiction Definitions

Mosby's Medical, Nursing & Allied Health Dictionary (31) defines addiction as "compulsive, uncontrollable dependence on a substance, habit, or practice to such a degree that cessation causes severe emotional, mental, or physiologic reactions." According to the Gale Encyclopedia of Medicine (13), "Addiction is a dependence, on a behavior or substance that a person is powerless to stop." Merriam-Webster's Collegiate Dictionary (29) defines compulsion as "an irresistible impulse to perform an irrational act," and addiction as a "compulsive need for and use of a habit-forming substance." The term addiction has been to some extent replaced by the word dependence for substance abuse. Researchers recognize substance addiction - alcoholism, drug abuse, and smoking - and process addiction - gambling, spending, shopping, eating, and sexual activity. In addition, researchers are beginning to acknowledge that many addicts are addicted to more than one substance or process (13, 33, 41, 42, 44).

Internet Addiction

Excessive human-machine interaction can be labeled as behavioral or process addiction; as noted by Griffiths, such technological addictions may be either passive or active (19). Their characteristics include the core components of addiction, including salience, mood modification, tolerance, withdrawal, conflict, and relapse. Researchers investigating the addictive potential of the Internet have noted that correlations exist between time spent online and the negative consequences reported by users.

Several psychological professionals have studied pathological Internet use. Young uses the phrase "Internet Addiction" when talking with the general public, but as early as 1996 and 1997, she used the term "Pathological Internet Use" when presenting a paper at the American Psychological Association's (APA) meeting. The addiction is modeled after pathological gambling in the Diagnostic and Statistical Manual of Mental Disorders 4th ed. (DSM-IV). Modifying criteria for pathological gambling, Young developed a screening test for addictive Internet use (43). Other psychologists such as Orzack also identified symptoms lists for Internet addiction, encompassing both psychological and physical symptoms (33).

Young (41) noted anonymity, convenience, and escape help make the Internet so alluring. She also stated that most addicts are not of the "computer nerd" type; instead, they are successful and outgoing. However, Internet "dependents" gradually spend less time with real people in exchange for solitary time in front of a computer. McCormick quotes John Sealy, M.D., medical director of a full-time inpatient recovery center for sexual addiction in a California hospital, who noted that because addiction is based on isolation and intense loneliness, the very isolation of the Internet appeals to and entraps its addicts (28).

Psychologist John Grohol wrote that it is not the technology that is addictive but the behavior, noting that it would appear that socialization is what makes the Internet so addicting (20). He also suggested that some newcomers to the technology get caught in Stage I (enchantment (obsession)) in his model of pathological Internet use, never reading Stage III (balance). Young described online content as immediate, constant, uncensored, and unregulated (44). A 1999 Newsweek article noted that the repetitive nature of online tasks has a soothing, ritualistic quality, somewhat like preparing and using drugs (7).

Types of Technological or Internet Addiction

Internet Addiction Disorder (IAD) is a broad term covering a wide variety of behavioral and impulse-control problems which include:

1. Cybersexual addiction - addiction to adult chat room or cyberporn.

2. Cyberrelationship addiction - online friendships made in chat rooms or cyberporn.

3. Net compulsion - compulsive gambling, day trading, or auction shopping.

4. Information overload - compulsive Web or database surfing.

5. Computer addiction - compulsive game playing or programming (4, 42, 44).

Cybersex and Cyber-Relationship Addiction. A Stanford University study found gay men to be among the highest at risk for becoming "cybersex compulsives" (10). McCormick (28) estimated that in 2000 over two million Internet addicts were Internet sex addicts. Schneider (35) noted that often the addicts begin to prefer their Internet encounters to their real-life partners. Reporting on a qualitative study of cybersex participants, she noted progression of cybersex addition was rapid, and study results supported previous studies on cybersex addiction.

Conlin (9) wrote in 2000 that the most commonly abused Internet web sites in the workplace were sexually oriented. In some cases, 70% of the activity levels at such sites occurred during working hours. Conlin estimated that one in five white-collar male workers accessed pornography sites at work. Griffiths (19) noted that growing empirical evidence (both direct and indirect studies) indicates Internet sex addiction exists. Young also reported heavy Internet users viewed cybersex to be completely anonymous and unable to be traced (44).

Chat Room, E-mail, Messaging, or Newsgroup Addiction. Writers cite anecdotal evidence of the addictive nature of chat and e-mail. One self-professed addict states," I can feel the pixels on the screen hit my face. Every time I'm not logged on, I feel empty, alone, wishing only to be there jumping from site to site, chatting, sending E-mail. I probably check my E-mail a couple dozen times a day" (39). "Opening E-mail is analogous to pulling the handle on a slot machine," (27). Some older addicts and women are drawn to chat rooms, engaging in conversations for many hours, often on sexual themes (6).

Gambling Addiction. Using the Internet as a means of gambling may be a more serious problem than when people gamble in other ways. Kevin O'Neill, deputy director of the Council on Compulsive Gambling in New Jersey, noted that the Internet is not monitored or regulated. Furthermore, when online betters are along in front of their screen, they can beg and get out of control quickly (32). The APA warned that Internet gambling could be more hazardous than other forms of gambling due to a lack of regulations and to the solitary nature of the activity (1).

Additionally, the explosive growth of the Internet can lead to more online gambling opportunities, resulting in more of the health and emotional difficulties that accompany gambling disorders. These difficulties include substance abuse, depression, and risky sexual behavior. Researchers (38) found that the majority of those with Internet gambling experience had the most serious levels of gambling behaviors, known as level 2 (problematic) and level 3 (pathological). Additional problems associated with gambling include circulatory disease and anxiety (11).

Information Overload. Some observers view the Internet as the primary cause of information overload due to its heterogeneous, unsorted, unfiltered information, and its ability to breed more information and greater speed (12). One article noted information overload has become almost synonymous with the Internet, and today's users are finding it increasingly difficult to efficiently locate precisely relevant content among its growth repository of 1+ billion pages (37). Four sources of information overload are identified in the literature: aggressive information, passive-aggressive information, friendly information, and customized information (12).

Gaming, Interactive Gaming, or Similar Addictions. Bower quotes Harvard Medical School clinical psychologist Orzack, who stated that younger users and men are drawn to games and pornographic websites (6). Multi-user dungeons (MUDs) allow users from around the world to access the host program, assume alternate identities, and develop an alternate self with differing personalities and characteristics. "Some of the tens of thousands of MUD users, who tend to be young males, spend as much as 80 hours a week playing their cyberspace persona" (8). As an example, one user's anecdotal reference to the game Doom said, "Time just flies - you don't notice it.... Somehow you think to yourself, 'Well, maybe I'll stop,' and then before you know where you are you're playing another round of the game" (8).

Cross-Addictions

Therapists note that the incidence of cross-addiction seems very common among Internet addicts. Most experts agree that the tools used for diagnosing and treating Internet-related addictions are similar to those used for other compulsions (28).

Internet use can weaken an individual's psychological well-being (25). Young reports that 54% of Internet addicts report a prior history of depression, 34% suffer from anxiety, and others suffer from low self-esteem (42, 44). Young has also seen individuals go from one addiction to another, has seen a relationship between women, depression, and the Internet, and seen men enjoy the struggle between dominance and power through sexual behaviors, interactive games, day trading, or sexually explicit chat rooms and cyberporn. The Mayo Clinic women's HealthSource showed a strong connection between compulsive computer use and mood, personality, and substance abuse disorders (3).

Of her patients addicted to the Internet, Orzack noted that all have at least one other problem - a history of another addiction, substance abuse, and so forth - and tend to be lonely, bored, depressed, or lacking in self-esteem (33). Although she noted no single pattern, the common thread was an emotional deregulation of some type. Shapira et al. (36) also found cross addiction in their study, concluding all participants' Internet use met established diagnostic criteria for the family of psychiatric illnesses known as impulse control disorders.

Addiction Symptoms

Internet addicts exhibit several conditions or symptoms. Long hours of use, upwards of six to nine hours a day, are displayed (17). Some addicts are online up to 40 hours per week spending time surfing or chatting (2, 42). These users can accumulate large phone bills and may neglect their work and their family (24). Other symptoms include lying about the level of use, preoccupation with the Internet, and using the Internet to escape other problems (18). These symptoms follow the necessary components for any addiction: increased tolerance, loss of control, and withdrawal (16). Some compulsive users have acquired the name "Internet vampires" because they emerge at dawn after a night on the Internet. Also emerging from excessive Internet use are "Internet affairs" (4).

Other authors (26) echo these symptoms. They cite numerous anecdotes of students sleeping in class, or employees missing work and missing deadlines, or parents/homemakers in financial difficulty and of individuals reducing their participation in social activities, all because of long hours spent online. Other physical symptoms cited in the same article include failure to attend to personal hygiene and changes in sleep patterns causing sleep disturbance.

White and Dorman (40) write that the phrase "Information Fatigue Syndrome" suggests a distinct set of physical and mental symptoms caused by unrelenting exposure to excess information. Increases in blood pressure, cardiovascular stress, memory difficulties, lack of concentration, headaches, stomach and muscle pain, and weakened vision are some of the physical symptoms of information overload; mental symptoms include lethargy, listlessness, sleeplessness, panic, irritability, and anger. Persons suffering from IAD often skip meals and suffer from repetitive stress injuries, dry eyes, headaches, backaches, and lack of sleep (15).

Besides physical and mental symptoms, emotional and social effects of information overload exist. "Our dependence on computers make [sic] them appear to be extensions of our bodies.... Computers comprise an integral aspect of their [the users'] self-image. This kind of technological symbiosis makes some people feel incomplete or undressed without a laptop or palm pilot close at hand" (40). Increased tension and competition in the workplace, lowered productivity, less time for family and friends, longer working days, and less leisure time are some of the social effects of information overload (40).

Consequences of Addiction

Various consequences follow from the symptoms identified in the above discussion. People who are Internet addicts cut back on the time they spend talking to family and friends. They spend less time being real human beings (6). Divorce, unemployment, financial and legal difficulties, and child neglect are some of the issues being faced by people who become addicted to online activity (42). Young has found that Internet addicts can suffer five categories of consequences: academic, relationship, financial, occupational, and physical (41). Orzack reports that addicts can lose their jobs as they become unable to limit their time spent online, either because they fail to turn up for work or because they misuse their office computer facilities (33). Schneider, documenting adverse effects of compulsive cybersex participants, noted adverse effects on the person's emotions, social life, work finances, and at times, legal status, as well as on the user's significant other and children (35).

Vulnerable Groups

Griffiths (19) discusses Young's and others' views on people who are more at risk developing cybersexual addictions. These people suffer from low self-esteem, a severely distorted body image, untreated sexual dysfunction, or a prior sexual addiction. Young and Rogers' study of the relationship between depression and Internet addiction suggests that increased levels' of depression are associated with those who become addicted to the Internet and increased levels of personal Internet use (45).

Mitka (30) and O'Neill (32) are both concerned with the possibility of children placing bets at Internet gambling sites and developing an addiction. Mitka's article quotes the APA, "...young people are at special risk for problem gambling and should be aware of the hazards of this activity, especially the danger of Internet gambling, which may pose an increased risk to high school- and college-aged populations."

A St. John's University study found that of the 138 undergraduates surveyed, 15% had seven or more of the 16 symptoms paralleling those of substance abuse and dependency found in DSM-IV (5). Young also identifies students as vulnerable groups (41, 42, 44).

In a study of Internet gambling (38), researchers found Internet gamblers were more likely to be unmarried or younger than those who never used the Internet for gambling, and they tended to have lower education and income levels than non-Internet gamblers did. This was surprising since Internet use is typically associated with people who have higher levels of education and higher income.

Number of Internet Addicts

The number of compulsive Internet users varies among researchers. Greenfield (17) estimated about 6% of people online in 2000 were compulsive users. His findings were from a survey conducted in conjunction with ABC News. Young in 1998 estimated 5 to 10% of online users were addicted (42). A New York-based research firm, Jupiter Communications, Inc., agreed with Young's 1998 estimates (16).

Summary

The phrases "Internet addiction," "pathological Internet use," and "Internet Addiction Disorder," are used interchangeable. Similarly, "online junkie" and "compulsive Internet user" are used interchangeably. Persons addicted to the Internet have symptoms similar to those associated with other addictions and may suffer from multiple disorders. Prior research suggests that several groups are vulnerable to Internet addiction. The research suggests that the most objective measure of Internet addiction is hours spent online. As suggested by Greenfield (17) in the literature, spending more than 40 hours per week on the Internet is a symptom of addiction.

DATA AND METHODOLOGY

Data for this study came from two data sets within The Graphic, Visualizations, and Usability Center (GVU) 10th survey conducted in the fall of 1998 (23). The data were used in accordance with policies published by GVU (21). GVU pioneered the first publicly accessible Web-based survey in 1994 and repeated the survey every six months until the fall of 1998. The GVU website makes public no data after 1998. Thus the researchers used the newest data of suitable quality available. GVU's website describes the survey as an independent, objective view of developing web demographics, culture, user attitudes, and usage patterns (22). Participants were solicited through announcements on Internet related newsgroups, banners randomly rotated through high-exposure sites, banners rotated through advertising networks (DoubleClick), announcements made to the www-surveying mailing list maintained by GVU, and announcements made in popular media. Thus respondents to the 10th survey were self-selected and self-reported. Five hundred records from the Web and Internet Usage dataset were merged with matching records in the General Use dataset. Gender, age, race, marital status, family household income, and hours of use were analyzed to determine whether correlations or causal relationships exist among the variables.

Prior research has suggested that the most effective measure of Internet addiction is hours spent online. On that basis, the researchers constructed a model in which the dependent variables were Total Hours Online and Recreational Hours Online. The demographic, economic, and behavioral variables were independent variables. Descriptive statistics and variable frequencies were prepared for all variables; correlation and cross-tabulation were used in preparation of the descriptive statistics. Fisher F (one-way ANOVA and regression) was used to test for causation where appropriate. Where ANOVA could not be used, the hypotheses were simplified and independent samples t-tests were used. The literature does not specify cause and effect between variables, only their associations. In the current researchers' model, hours spent online can be an independent variable or a dependent variable. Both versions were tested.

DESCRIPTIVE STATISTICS

This section first summarizes the essential facts and interrelationships among the variables. Cross-tabulations and Chi-square tests further explore variable relationships. In the final subsections, ANOVA, t-tests, and regression models are used to analyze the behavior of the heaviest users of the Internet and to analyze the usage rates of those respondents identified as members of groups vulnerable to addiction.

Independent Variables

Six percent of the 500 respondents were 20 years or younger. Thirty-one percent were ages 21 to 30, 23% were ages 31 to 40, 20% were ages 41 to 50, 13% were ages 51 to 60, and 7% were ages 61 and over. Males outnumbered females more than two to one (338 to 162). Nine percent had a high school education or less, almost 34% had vocational-tech/some college, 33% had a college degree, 18% had a master's degree, and 6% had a doctoral/professional degree. Fifty percent were married; 30% of the respondents were single. The category divorced/ separated/widowed was 10% of those responding, and the remaining 8% marked "other" as their marital status.

Nearly 4% of the respondents reported an annual household income of under $10,000; 7% reported $10,000 to $19,000. At the other extreme, 11% reported from $75,000 to $100,000, and 14% had income over $100,000. The modal income was $50,000 to $74,000 (28%). Over 91% of respondents were white; African-American users were almost 2% of users. Asian users comprised 3%. The ethnicity of the remaining 4% was classified as "other."

A little over 1% of respondents used the Internet under one hour per week (Total Hours Online). Nine percent spent two to four Total Hours Online. Thirteen percent spent five to six hours. Almost 14% reported seven to nine hours per week. The largest cell, over 34%, had a usage time of ten to 20 hours. Almost 18% had 21 to 40 hours of usage. Ten percent (51 of 500) spent over 40 hours per week online. Recreational Hours Online used the same categories with these results: 63 of the 500 (13%) spent less than one hour online per week; 36% spent between two and four hours per week; 19% spent between five and six hours per week; 19% spent between seven and nine hours per week; 9% spent between ten and 20 hours per week; and 4% spent over 20 hours per week.

Correlations in the Independent Variables

The researchers used correlation tools to look for relationships between pairs of independent variables and between dependent and independent variables. The following are worthy of note. Education Level and Gender are positively correlated with Household Income (0.309 and 0.122). Age is positively associated with Household Income (0.171) and Educational Level (0.133). Total Hours Online is not strongly correlated with any independent variable.

Describing Those Who Use the Internet the Most

The heaviest users of the Internet - persons online 40 or more hours per week - represented 51 of the 500 respondents. This group included 35 males and 16 females, about the same proportions as in the full study group of 500. Among these 51 cases, the income distribution is almost uniform in the five higher income classes, with smaller frequencies in the three lowest income classes. Twenty-three of the 51 indicated less than a college degree; 26 indicated a college degree or higher. About 10% of these heaviest users are under 20 years old. About 30% each are in the following categories: between 21 and 30 years old, between 31 and 40 years old, and over 40. Those who spend 40 or more hours online per week are younger than other classes of users.

STATISTICAL ANALYSIS

The review of previous research identified several groups vulnerable to Internet addiction: singles, young males, college students, gays, middle-aged females, and those less educated, among others. That review also identified symptoms of addiction, of which hours spent online was the most important indicator. The descriptive statistics in the previous section displayed patterns among the variables and characteristics of the respondents to the survey. In this section the researchers will connect the variables of the literature with the variables of the dataset. Shortcomings in the dataset - for example, no data were collected on any addiction symptoms other than hours online cause some claims in the literature to go untested.

The researchers formulated three broad types of investigative questions:

1. Noncausal: Is time spent online independent of each respondent characteristic (age, gender, etc.)? (Tested by Chi square and correlation)

2. Causal: Does time spent online vary significantly with each respondent characteristic? (Tested by Z, t and ANOVA)

3. Multivariate: Does the set of respondent characteristics explain time spent online? (Tested by regression)

Tests of Independence

The researchers conducted cross-tabulations and Chi-square tests on selected pairs of variables. The variables were Total Hours Online, Recreational Hours Online, Gender, Age, Household Income, Educational Level, Marital Status, and Race. The researchers constructed 12 hypotheses about the link between Total Hours Online and Recreational Hours Online, with each of the other six named variables. These hypotheses appear in Table 1.

Hours Used Versus Age. Chi-square was used to test whether hours online was independent of age. This is a pair of hypothesis tests formulated at H-1 and H-2. The significance level for H-1, Total Hours Online, (sig.=0.185) was greater than [alpha]=0.05; this led to failure to reject the null hypothesis. The Recreational Hours Online test, H-2, also failed to reject the null hypothesis (sig.=0.686). Thus, the data shows no dependency between age and hours online.

Hours Used Versus Gender. Chi-square was used to determine whether Total Hours Online and Recreational Hours Online were independent of Gender (H-3 and H-4). For Recreational Hours Online, it was 0.122. In both cases the researchers failed to reject the null hypothesis.

Marital Status. The researchers tested whether Total Hours Online and Recreational Hours Online were independent of Marital Status. The researchers could not reject hypothesis H-6 that Marital Status is independent of Recreational Hours Online (sig=0.083). However, Marital Status and Total Hours Online (H-5) are not independent (sig=0.015). As hours online rises, the percent of "Married" respondents falls - from 69% at the lowest usage rates to 35% at the highest. The lowest usage category was 10.6% of all respondents, but "Marrieds" were over-represented - 14.6% of this category. In the 21-40 hours category, "Divorced-Separated-Widowed" were 28% of the category, compared to 18% of all cases.

The Remaining Variables. The researchers were unable to reject a null hypothesis of independence for the pairs of variables Household Income versus Total Hours (H-7), Household Income versus Recreational Hours (H-8), Race versus Total Hours (H-11), and Race versus Recreational Hours (H-12). For hypotheses H-9 and H-10, the Chi-square results contained too many small cells, even after collapsing some rows and columns. The researchers thus report these tests "not conducted."

The Heaviest Users of the Internet

Prior research established 40 or more hours online per week as a sign of addiction. The researchers divided respondents into two groups, those using the Internet under 40 hours per week and those at or over 40 hours per week. This grouping leads to formulation of four hypotheses (H-13 through H-16). See Table 2.

Do those Internet users who spend the most time online differ from other users? Using independent samples t-test, the researchers found only one distinction. The two groups differed with respect to Age (H-13) - the heaviest users are a younger collection of respondents than other user groups. They did not differ on Gender (H-14), Education (H-15), or Income (H-16). Equal variances were assumed for H-14 through H-16; equal variances were not assumed for H-13.

The researchers also formulated hypotheses, again to be tested by student-t, about differences in mean hours online by Gender, by Age, etc. These hypotheses appear in Table 3 as hypotheses H-17 through H-22.

Gender and Age Differences

Do males spend the same Total Hours Online as females (H-17)? Males spent 19.02 hours online; females, 18.7. There was no evidence that Total Hours online by males and females (H-18) were different (sig.=0.838). For Recreational Hours Online, mean hours for males was 6.29 hours; females 6.28. The researchers found no difference between male and female Recreational Hours Online (sig.=.950).

Do younger users spend the same hours online as older users? Specifically, are recreational hours equal between those 20 and younger versus those over 20 (H-19)? Recreational Hours by users 20 and younger averaged 6.08. Recreational Hours by users over 20 averaged 8.79. Equal variances was assumed (sig.=.0236); and the hypothesis of equal hours of use (H-20) was rejected (sig.=0.024).

New Users of the Internet

One stereotype in the literature suggested that "newbies" spent more time online than more experienced Internet users. The researchers established hypotheses to test whether Total Hours Online is equal across all levels of Internet experience (H-21), and whether Recreational Hours Online is equal across all levels of Internet experience (H-22). For both hypotheses, equal variances could not be assumed, and no ANOVA was conducted.

The researchers then compared hours online between those with less than one year of Internet experience and those with a year or more. This simplified test used independent samples t-test. For Total Hours Online (H-21), those with less than one year's experience spent 14.6 hours online weekly; other users spent 19.4. This difference was significant (sig.=0.006), but the finding contradicts the stereotype - experienced users were heavier users of the Internet, not the "newbies." For Recreational Hours online (H-22), there was no difference.

ANOVA

The researchers established eight ANOVA tests (H-23 through H-30), where Total Hours Online and Recreational Hours Online were the factors, and Age, Gender, Household Income, and Education were the variables (Table 4). Of the eight hypotheses, equal variances could be assumed only for H-27, "Mean Age is the same across all categories of Recreational Hours Online," and H-30, "Mean Educational Level is the same across all categories of Recreational Hours Online." For Hypothesis 27, the ANOVA revealed no differences in the respondents' Age mix as Internet usage varied for Hypothesis 30, a significant variation was found. The mean Educational Level of respondents is significantly higher among those using the Internet four hours weekly or less, than for those using the Internet five to nine hours or 10-20 hours weekly.

The reader may have noticed two tests that seem similar. The two sets of hypotheses may seem similar but are approaching the data in different ways. Here is an example based on the variable Age. Based on the variable Age, the two related hypotheses are H-13 and H-23. In H-13, the researchers asked what is the difference in Age of those using the Internet most versus the Age of using the Internet at other levels. In H-23, the researchers asked whether older users spent different hours on line than younger users. The two hypotheses use the same variables, the same categories, but reverse the causation.

Usage Patterns of Female Respondents

Based on the literature concerning "middle-aged female" compulsive users, the researchers tested with one-way ANOVA for differences in Internet usage rates among female respondents only. The researchers established hypotheses to test whether Total Hours online in female users is equal among all age categories (H-31) and whether Recreational Hours Online in female users is equal among all age categories (H-32). For the Total Hours test, equal variances could not be assumed, and no ANOVA was performed. For the Recreational Hours test, equal variances was assumed, but H^sub o^ could not be rejected (sig.=0.366). There was no evidence that middle-aged women were more likely to be among the heaviest users of the Internet.

Linear Regression

The researchers modeled Total Hours Online and Recreational Hours Online with linear regressions. The dependent variable in each case was usage; the independent variables were Age, Gender, Educational Level, and Household Income (Table 5). The researchers' a priori expectation was that the sign of the Age variable was negative, the sign of the Education variable was negative, the sign of the Gender variable was positive, and the sign of the Income variable was negative.

For Recreational Hours Online, the regression equation was Recreational Hours = 8.582 + .029 Age - 1.058 Education - .141 Gender - .005 Income

Two signs differed from expectations, but neither was significant. For Total Hours (H-33), the regression was not significant. For H-34, r-square is small (sig.=0.031), but the regression is significant (sig.=0.002). Of the independent variables, only Education Level is significant. The regression models largely support the previously reported univariate tests, especially that Gender is not a significant variable and that Education is. The evidence was mixed for Age and Income.

SUMMARY, CONCLUSIONS, AND FURTHER RESEARCH

Internet dependence, or Internet Addiction Disorder, is a pathological condition that may affect several million Americans. The negative consequences of this behavior reach into their personal lives and workplaces. Early research on this phenomenon suggested young, computer-savvy males were the most vulnerable; later research suggested that middle-aged women, college students, gays, and children were also at risk. This study attempted to determine which stereotypes, if any, were substantiated by data in a publicly available database of Internet users.

The researchers concluded that the respondents' attributes - including race, gender, and age - were consistent with data in related studies. This study found that 10% of respondents reported 40 or more hours of online activity weekly - an amount labeled an important signal of addictive behavior. That percentage is consistent with results from related studies.

Those who spend the most hours online are younger than other users but do not differ on other attributes. There is no link between gender and level of use; this contradicts older stereotypes. Internet participation by married users falls as the quantity of hours online increases. Those users with more Internet experience spent more time online than less experienced users; this also contradicts the stereotype. There is no evidence that middle-aged women are more likely to be among the heaviest users of the Internet. The level of education is higher among those Internet users who use it the least. Regression models were of limited value; only education level stood out as a significant explanatory variable.

The researchers are concerned first that the phenomenon of Internet dependence is growing and second, that no reliable set of attributes and variables has been found that would identify an Internet addict. Those concerns prompt the researchers to call for additional surveys patterned after the GVU survey which provided the data used here. The researchers ask for two improvements in such surveys. First, the variables need smaller, more finely-tuned categories or need to be true interval data. Second, the surveys need to include questions specifically oriented to the symptoms and behaviors of Internet addiction.