Elsevier

Drug and Alcohol Dependence

Volume 72, Issue 3, 11 December 2003, Pages 279-286
Drug and Alcohol Dependence

Brief measures of sensation seeking for screening and large-scale surveys

https://doi.org/10.1016/j.drugalcdep.2003.08.003Get rights and content

Abstract

Sensation seeking is central to research on the prevention of risky health behaviors, but current measures of sensation seeking are fairly long, thereby reducing their chances of inclusion in some research projects. Hence, we developed and evaluated two brief indices of sensation seeking, a four-item measure that retains the framework of the Sensation Seeking Scale-Form V (SSS-V) and a shorter two-item measure focusing on the risk-taking elements of sensation seeking. We compared the performance of the new indices with that of two well documented but longer measures of sensation seeking. The evaluation was based on data provided by more than 5000 teens and pre-teens in grades 7 through 11. Psychometric analyses revealed that the internal consistency of the two new measures was very good overall and across grade and sex categories. Additionally, the new indices correlated as expected with a series of risk and protective factors as well as tobacco, alcohol, and marijuana use. Both indices performed in ways remarkably similar to the established measures of sensation seeking and should prove useful for future research involving risky health behaviors.

Introduction

Sensation seeking is a personality trait believed to have a biological basis that expresses as a need for physiological arousal, novel experience, and a willingness to take social, physical, and financial risks to obtain such arousal (Bardo et al., 1996, Zuckerman, 1979, Zuckerman, 1994). Sensation seeking is associated with a variety of illegal and/or risky behaviors such as the use of illicit drugs (Newcomb and McGee, 1991, Palmgreen et al., 2001, Stephenson et al., 2002, Zuckerman et al., 1993), sexual risk-taking (Donohew et al., 2000, Hoyle et al., 2000), reckless driving (Heino et al., 1996), smoking (Zuckerman et al., 1990), and alcohol use (Stacy et al., 1993).

Until recently, virtually all research on sensation seeking assessed the construct using Form V of the Sensation Seeking Scale (SSS-V; Zuckerman et al., 1978), which consists of 40 items in forced-choice format. The SSS-V comprises four subscales that correspond to key features of the construct: experience seeking, disinhibition, thrill and adventure seeking, and boredom susceptibility. Some researchers have developed and tested predictions at the subscale level (Orlebeke et al., 1990), or used subscale scores as indicators of a latent variable (Newcomb and McGee, 1991). However, the preponderance of research on sensation seeking that has made use of the SSS-V has involved the computation of a single score based on responses to the complete set of 40 items.

The present research was motivated by the need for more economical measures of sensation seeking. Of specific concern is the need to include measures of risk factors, such as sensation seeking, in large-scale survey studies of problem behavior that influence prevention policy. These surveys typically tap many constructs using only one or two items. For instance, in a recent publication based on data from the Monitoring the Future National Survey, school bonding was measured with three items, school misbehavior was measured by four items, and academic achievement was assessed with a single item (Bryant et al., 2000). A recent National Household Survey on Drug Abuse allocates a single item to measuring perceived risks associated with use of specific substances (SAMHSA, 2000). Moreover, such surveys typically are dominated by questions that tap demographic or behavioral factors, while overlooking potent psychological risk factors such as sensation seeking. Because heritable risk factors such as sensation seeking likely give rise to or moderate the influence of more proximal factors such as perceived risks, it is important that influential, large-sample surveys tap into these factors.

There are, of course, risks involved with the move from a 40-item measure to the two–four-item composites characteristic of large-scale surveys. Sensation seeking is manifest in a broad range of behaviors and preferences and, therefore, content validity concerns would suggest the need for a large, heterogeneous pool of items that exhaust the content domain (Epstein, 1983). Moreover, each item on a scale such as the SSS-V is a fallible indicator of the construct. Some portion of the variability in responses to each item corresponds to the sensation seeking construct; the remaining portion of variability is either random or systematic measurement error (Nunally, 1978). By aggregating over multiple items, the influence of random error is minimized (Cronbach, 1951). In short, all else being equal, longer measures are better.

Although longer measures are advantageous for addressing content validity and measurement error concerns, they are not without their drawbacks. One cost of high internal consistency in a set of items is the perception by respondents that they are repeatedly being asked the same thing. For lengthy measures, this validity-threatening problem is very likely (Bollen and Lennox, 1991). Another potential drawback, the one that motivated our research, is that the constraints present in many testing situations preclude the use of lengthy measures of constructs. As noted earlier, the typical large-sample survey study targets a large number of constructs within the context of a relatively short survey instrument. Short instruments also characterize most “screener” surveys used to determine eligibility for large surveys, experiments, or focus groups. If the only means of measuring a construct is with a lengthy scale, then that construct will most certainly not be represented in such studies. Thus, there are clear advantages to brief measures of constructs, and, in some testing situations, those advantages outweigh the benefits of longer measures.

Recently, a number of alternative measures of sensation seeking have been proposed. Although, in some cases, the measures differ in substantive ways from the SSS-V (see Arnett, 1994, Zuckerman et al., 1993), a feature they all share is that they are shorter than the SSS-V. For instance, the impulsive sensation seeking item set from the Zuckerman–Kuhlman Personality Questionnaire comprises 19 items. The Arnett Inventory of Sensation Seeking (Arnett, 1994) includes 20 items. In addition to these measures, which reflect a reconceptualization of the sensation seeking construct, several shorter measures comprise a subset of the SSS-V item set. The Sensation Seeking Scale for Children (Russo et al., 1993) includes 26 items, of which all but a few are edited versions of items from the SSS-V. Huba et al. (1981) described a measure in which each of the four aspects of sensation seeking is measured using four items chosen from the 40-item SSS-V. An abbreviated and revised eight-item form of this measure, the Brief Sensation Seeking Scale (BSSS; Hoyle et al., 2002), includes two items representing each aspect of sensation seeking. An important feature of all of these measures is that their associations with variables such as drug and alcohol use and various risk and protective factors are at levels consistent with those based on measurement of sensation seeking using the SSS-V.

Although the BSSS comprises only eight items and has been used in a number of large-scale survey studies (Palmgreen et al., 2001), it is not brief enough to warrant inclusion generally in nationwide studies such as the Monitoring the Future National Survey and the National Household Survey on Drug Abuse. As noted earlier, most constructs in such surveys are represented by a single item, and a ceiling for number of items devoted to a single construct is about four. Hence, we endeavored to develop a reliable and valid measure of sensation seeking consisting of no more than four items.

A logical starting point for the development of such a measure is the Brief Sensation Seeking Scale, which traces its roots to the SSS-V and, like the SSS-V, includes equal numbers of items from the four content areas identified in early descriptions of the construct (see Zuckerman, 1979, Zuckerman et al., 1978). An alternative strategy is to build on work coupling sensation seeking and impulsivity in a comprehensive psychobiological model of personality (Zuckerman et al., 1993). This work has given rise to a measure of impulsive sensation seeking (ImpSS) that does not make explicit reference to the four content areas apparent in the SSS-V and its derivatives. Rather, items on this measure refer more broadly to the preferences of sensation seekers (e.g., “I like doing things just for the thrill of it”) and impulsive individuals (e.g., “I tend to begin a new job without much advance planning on how I will do it”). Rather than choosing one approach over the other, we set out to develop and evaluate two brief indices of sensation seeking, a four-item measure that preserves the basic framework of the SSS-V and a two-item measure that focuses more closely on the risk-taking elements of sensation seeking.

Section snippets

Participants

The data on which our evaluation is based were generated in a large-scale investigation of the Office of National Drug Control Policy’s (ONDCP) National Youth Anti-Drug Media Campaign (Crano and Burgoon, 2002). In this study, systematic random sampling with geographic and grade stratification were employed to draw monthly samples of about 200 teens and pre-teens from the Knoxville, Tennessee and Lexington, Kentucky public schools. For this analysis, we used 29 months of data collected from

Results

Descriptive statistics for the four measures of sensation seeking are reported in Table 1. Statistics for the BSSS-4 are comparable to those for the established measures, ImpSS and BSSS. The means for SS2 were slightly lower. In terms of mean scores, there were no problems with floor or ceiling effects for either of the new indices. Standard deviations for the BSSS-4 and SS2 were consistent across subsamples, although slightly higher for SS2.

We compared means on each measure in Gender×Grade

Discussion

Sensation seeking in central to much of the substance use research, yet it is generally absent from large-scale surveys. Hence, our goal in this investigation was to establish the predictive and convergent validity of two very brief measures of sensation seeking suitable for screening and inclusion in lengthy surveys. Results of our evaluation established the empirical utility of two new sensation seeking indices, a four- and a two-item measure (BSSS-4 and SS2). Both new measures displayed

Acknowledgements

This research was partially supported by grants DA-06892 and DA-12360 from the National Institute on Drug Abuse.

References (34)

  • J.M Cortina

    What is coefficient alpha? An examination of theory and applications

    J. Appl. Psychol.

    (1993)
  • Crano, W.D., Burgoon, M., 2002. Mass Media and Drug Prevention. Lawrence Erlbaum, Mahwah,...
  • L.J Cronbach

    Coefficient alpha and the internal structure of tests

    Psychometrika

    (1951)
  • L Donohew et al.

    Applications of a theoretic model of information exposure to health interventions

    Hum. Commun. Res.

    (1998)
  • S Epstein

    Aggregation and beyond: some basic issues on the prediction of behavior

    J. Pers.

    (1983)
  • Hornik, R., Maklan, D., Judkins, D., Cadell, D., Yanovitzky, I., Zador, P., Southwell, B., Mak, K., Das, B., Prado, A.,...
  • R.H Hoyle et al.

    Personality and sexual risk-taking: a quantitative review

    J. Pers.

    (2000)
  • Cited by (429)

    View all citing articles on Scopus
    View full text