Basic descriptive characteristics of the sample are shown in Table . Substance use was common. A third (33%) of the young adults were classed as 'heavy drinkers' and half (47%) as 'binge drinkers'. Over half (56%) reported ever use of cannabis, but many fewer (13%) reported ever use of one or more of the 'hard' drugs listed. Almost all (93%) reported that half or more of their friends drank alcohol (equivalent figure for ever use of cannabis, 21%).
Descriptive data: frequency of last week heavy and binge drinking, ever cannabis and 'hard' drugs, and of potential confounders.
Respondents had seen a mean of 19.0 (SD = 7.3, range 1-44) of the 50 films presented to them; mean film alcohol and drug exposures were 726 minutes (12.1 hours) and 45 minutes respectively. The mean number of films was higher for males (20.8) than females (17.3, F = 60.5, p = .000) and males' film alcohol and drug exposures were higher (770 vs. 682 minutes, F = 16.5, p = .000 and 51 vs. 38 minutes, F = 23.1, p = .000 respectively). There were no social class differences for films seen or film alcohol exposure, but film drug exposure was higher in those from higher social class backgrounds (non-manual = 51, skilled manual = 40, semi/unskilled manual = 41 minutes, F = 6.6, p = .001). There was a positive correlation between the film alcohol and drug exposure measures (r = .510).
Table reports the percentage of heavy and binge drinkers by quartile of film alcohol exposure, and the percentage of ever users of cannabis and ever users of 'hard' drugs by film drug exposure. The p values reported in the table relate to heterogeneity within the groups, but we also tested for linear trends. In the cross-tabulations, the tests for linear trends in the percentage of heavy drinkers (p = .018) and binge drinkers (p = 0.012) by film alcohol exposure quartiles (see Table ) were statistically significant. Similarly, there was an increase in the percent who had ever used cannabis with each quartile of film drugs exposure (linear trend p = .000). The percentage who had used 'hard' drugs was also highest in the highest quartile of film drug exposure (16%), but with less evidence of a stepwise increase (linear trend p = .033).
Alcohol and drug use by predictor variables - percentages (significance of chi-square).
Male gender and perceiving oneself as a risk-taker and rule-breaker were associated with all four substance use measures (p < 0.001 in all cases), and having no 'Highers' at 19 with all (p = 0.004 for heavy drinking, and p < 0.001 for ever use of cannabis, and ever use of hard drugs) except binge drinking (p = 0.65). Respondents from manual class backgrounds were more likely to have used 'hard' drugs (p = 0.037), and those reporting lower parental care were more likely to have ever used both cannabis (p = 0.003) and 'hard' drugs (p = 0.012). Friends' drinking and cannabis use were strongly associated with own drinking and drug status respectively. There were no associations between the substance use measures and parental structure, parental control, or hours per week watching television, videos or dvds.
Table shows the results of the logistic regression models for each outcome, both before and after adjusting for potential confounding or mediating variables. We consider the alcohol outcomes first. In the unadjusted model, those in the highest quartile of film alcohol exposure were more likely to be classed as both heavy and binge drinkers ((OR = 1.56, 95% CI 1.06-2.29) and 1.59 (1.10-2.30) respectively, compared with the lowest quartile) on the basis of their reported alcohol consumption the previous week. Adjustment for gender reduced the associations, but further adjustment for background characteristics returned the odds ratios for the drinking measures to the unadjusted levels. Adjusting for risk-taking, rule-breaking and qualifications, and particularly for friends' drinking status, reduced the odds ratios, but further adjustment for hours watching television, videos or dvds made no difference to the associations. In this final model only gender (OR for females 0.59 (95% CI 0.43-0.80) for heavy drinking and 0.64 (95% CI 0.48-0.85) for binge drinking) and friends' drinking status (OR for half or more friends drinking 1.44 (95% CI 1.27-1.64) for heavy drinking and 1.54 (95% CI 1.36-1.73) for binge drinking) had odds ratios which did not include unity (1.00), i.e. film alcohol exposure was no longer significantly associated with heavy or binge drinking. (Full tables showing OR and 95% CI for all variables included in all models available on request.)
Results of logistic regression models including film exposure quartiles, (a) unadjusted and (b-f) adjusted for potential confounders - ORs (95% CIs).
We turn now to consider the relationship between exposure to film images of illicit drug use and ever-use of cannabis and 'hard' drugs. In the unadjusted model, ever use of cannabis showed a stepped association with film drug exposure (OR for third and highest, compared with the lowest quartile of film drug exposure = 1.46 (95% CI 1.01-2.10) and 1.80 (95% CI 1.24-2.62)). Adjustment for gender attenuated the association, whereas adjusting additionally for family background made little difference. The OR was further attenuated (with 95% confidence which included unity) after adjusting for personal characteristics, friends' reported cannabis use, and then tv/dvd/video watching (see table ). In the final model having any 'Highers' (OR 0.56, 95% CI 0.39-0.80), seeing oneself as a rule-breaker (OR 10.70, 95% CI 3.43-3.38, comparing those saying 'very true' as compared with those saying 'very untrue'), and reporting that half or more of one's friends used cannabis (OR 2.14, 95% CI 1.87-2.45) were the only ORs in the model with 95% confidence intervals that did not include unity.
For 'hard' drug use the confidence intervals for the unadjusted ORs in third (OR 1.30, 95% CI 0.75-2.26) and highest (OR 1.57, 95% CI 0.91-2.69) quartiles overlapped with unity. Although the odds were greatest in the highest film drug exposure quartile in each of the models for ever use of 'hard' drugs, none of the associations reached conventional levels of significance, even in the unadjusted model. In the final model hard drug use was significantly inversely associated with having any 'Highers' (OR 0.26, 95% CI 0.15-0.45), and positively associated with seeing oneself as a rule-breaker (OR 3.02, 95% CI 1.05-8.69, comparing those saying 'very true' as compared with those saying 'very untrue') and reporting that half or more of one's friends used cannabis (OR 2.10, 95% CI 1.76-2.51).