Our primary collection of literature was gathered by searching online electronic databases for articles relevant to our topic of interest through December 2008. The four databases used were PsycINFO, Medline, Web of Science, and EBSCO MegaFile. In each of these databases, we conducted a search using a combination of the following terms: sleep deprivation or sleep restriction and cognition, attention, memory, performance, vigilance, and executive function (12 combinations in all). This search yielded 4,276 hits in total. We next scanned the abstracts of these articles to determine their suitability for inclusion in the analysis. In total, 176 of the articles were empirical studies that employed SD as a primary independent variable and used at least one cognitive measure as a dependent outcome. We subsequently obtained the full text of these articles to determine whether they met full inclusion criteria. These criteria were as follows:
Because of the restrictions imposed by Criterion 3, a number of subareas within the realm of SD research necessarily had to be omitted from this analysis. A survey of the literature on SD and decision making revealed that outcome variables on these tests did not form a cluster that was homogeneous enough to warrant a quantitative synthesis. This was because many of these experiments employed complex, real-world scenarios, opting for ecologically valid designs over more controlled neuropsychological tests (for a review, see Harrison & Horne, 2000 (link)). Moreover, it is unclear how outcome variables from standardized decision-making tests (e.g., the Iowa Gambling Test) should be compared with the accuracy measures obtained from other cognitive domains. Finally, experiments on implicit and procedural memory were also excluded, as these form a separate body of literature pertaining to sleep and memory consolidation (Stickgold & Walker, 2005 (link)), the analysis of which is beyond the scope of this article.
In addition to this online literature search, we obtained data from several other sources. We conducted hand searches of the journal Sleep and the Journal of Sleep Research from 1988 to 2008. We also reviewed the reference lists of the major review articles on SD and cognitive performance that have been published over the last several years. Finally, to mitigate the “file drawer” problem (Strube & Hartmann, 1983 ), we contacted eight major sleep laboratories conducting research in this field to request unpublished data from experiments, as well as master’s and doctoral theses. We received additional data from one of these laboratories, as well as replies from all but one of the remaining investigators informing us that they did not have suitable data for inclusion. In total, 70 articles and 147 data sets met inclusion criteria and were included in the meta-analysis (seeTable 1 ). Among these, numerous data sets contained more than one cognitive outcome; these were coded separately, according to the recommendations of Hunter and Schmidt (2004) . Altogether, 209 aggregated effect sizes and 5,104 individual effect sizes were calculated from these data sets.
Because of the restrictions imposed by Criterion 3, a number of subareas within the realm of SD research necessarily had to be omitted from this analysis. A survey of the literature on SD and decision making revealed that outcome variables on these tests did not form a cluster that was homogeneous enough to warrant a quantitative synthesis. This was because many of these experiments employed complex, real-world scenarios, opting for ecologically valid designs over more controlled neuropsychological tests (for a review, see Harrison & Horne, 2000 (link)). Moreover, it is unclear how outcome variables from standardized decision-making tests (e.g., the Iowa Gambling Test) should be compared with the accuracy measures obtained from other cognitive domains. Finally, experiments on implicit and procedural memory were also excluded, as these form a separate body of literature pertaining to sleep and memory consolidation (Stickgold & Walker, 2005 (link)), the analysis of which is beyond the scope of this article.
In addition to this online literature search, we obtained data from several other sources. We conducted hand searches of the journal Sleep and the Journal of Sleep Research from 1988 to 2008. We also reviewed the reference lists of the major review articles on SD and cognitive performance that have been published over the last several years. Finally, to mitigate the “file drawer” problem (Strube & Hartmann, 1983 ), we contacted eight major sleep laboratories conducting research in this field to request unpublished data from experiments, as well as master’s and doctoral theses. We received additional data from one of these laboratories, as well as replies from all but one of the remaining investigators informing us that they did not have suitable data for inclusion. In total, 70 articles and 147 data sets met inclusion criteria and were included in the meta-analysis (see
