PROCEEDINGS of the HUMAN FACTORS SOCIETY 34th ANNUAL MEETING-1990 INJURY SEVERITY AND LIKELIHOOD IN WARNINGS Michael S. Wogalter and Todd Barlow Department of Psychology Rensselaer Polytechnic Institute Troy, New York 12180 ABSTRACT Two experiments examined the influence of injury likelihood and severity in warnings on product hazard perceptions (Experiment 1)and behavioral compliance (Experiment 2). In Experiment 1,participants were given a set of front panel labels for 10 household consumer products. Warnings on the labels were constructed by manipulating the likelihood (low vs. high) and severity (low vs. high) of injury. Labels lacking a warning served as controls. Participants rated the product labels under the guise of a marketing study in which most of the questions concerned product familiarity, cost, and label attractiveness. Only one question was of interest which probed the level of hazard posed by the products. The results showed that (1) the presence of a warning increased the products' judged level of hazard, (2) products with high severity warnings were viewed to be more hazardous than products with low severity warnings, and (3) likelihood of injury in the warnings had no influence on hazard perceptions. Experiment 2 used a chemistry laboratory demonstration task to test the effects of injury likelihood and severity in a warning on compliance behavior (i.e., wearing gloves as directed by the warning). Greater compliance was shown when warned of a more severe injury, but only when the injury was of lower likelihood. In general, both experiments showed that injury severity influences warning effectiveness to a greater extent than injury likelihood. Th~ results suggest that to inform people of a hazard and to motivate them to comply with a directed behavior, product warnings should communicate the severity of consequences. INfRODUCflON What information do people use to determine the level of hazard posed by consumer products? This question is important because research suggests that hazard perception is closely related to people's willingness to read product warnings (Godfrey, Allender, Laughery, & Smith, 1983; Wogalter, Desaulniers, & Brelsford, 1986; Young, Brelsford, & Wogalter, 1990). Research also indicates that people's hazard perceptions are largely determined by the extent or severity of injury that might occur with virtually no contribution of how likely the injury might occur (Wogalter, Desaulniers, & Brelsford, 1987; Young et al., 1990). However, research on a related concept "risk" indicates that people's perceptions are determined by combining both severity and likelihood information (e.g., Lowrence, 1980; Slovic, Fischhoff, & Lichtenstein, 1979, 1980). The methodology employed in previous work examined the relative effectiveness of likelihood and severity information by asking people to judge the hazard of generic names of ?roducts (e.g., Wogalter et al. 1986, 1987a; Young et aI., 1990) or estimate the frequency or relative likelihood of accident events (e.g., Slovic et al. 1979, 1980). However, no study to date has investigated this issue using a more direct approach in which likelihood and severity information is manipulated in warnings. The present studies examine the effect of injury likelihood (low versus high) and severity (low versus high) information in warnings on hazard perceptions for commercially-available consumer products (Experiment 1) and measures their effect on behavioral compliance (Experiment 2). EXPERIMENT 1 Experiment 1investigated whether injury likelihood and severity information in warnings influences perceptions of 580 hazard. The study was conducted under the guise of "marketing research" concerned with factors affecting people's decisions to purchase certain consumer products. Participants answered a variety of questions for each of a set of products. One question, which asked how hazardous they perceived the product to be, was of primary interest. Method Participants. Forty-six undergraduates from Rensselaer Polytechnic Institute participated for course credit in introductory psychology courses. Prior to this experiment, a different set of 30 students from the same population participated in a preliminary word rating study. Stimuli and materials. Ten consumer products were chosen to represent a range of potentially hazardous household items (Alcon Optizyme Enzymatic contact lens cleaner, Dow bathroom cleaner, Excedrin extra-strength aspirin, Fresh Start laundry detergent, Klean-Strip paint thinner, Krylon fixative spray coating, Raid roach and flea fogger, Red Devil lye drain opener, Textra hair mousse, and Trugarde fabric stain protector). Labels from the products' front panels were duplicated using an optical, scanner/digitizer (Thunderware Thunderscan), stored and manipulated using a computer and software (Apple Macintosh and Silicon Beach Superpaint), and reproduced using a 300 dot per inch printer (Apple Laserwriter). Warnings on the front labels of the products were manipulated to differ with respect to conveyed injury likelihood (low versus high) and severity (low versus high) of injury that they conveyed. These two independent variables were orthogonally crossed to form four warning labels for each product: (1) Low likelihood, Low severity, (2) Low likelihood, High severity, (3) High likelihood, Low severity, (4) High likelihood, High severity. This