The latent state-trait structure of resting EEG asymmetry: Replication and extension DIRK HAGEMANN, a JOHANNES HEWIG, b JAN SEIFERT, a EWALD NAUMANN, a and DIETER BARTUSSEK a a Fachbereich I–Psychologie, Universita¨ t Trier, Trier, Germany b Department of Psychology, Friedrich-Schiller-Universita¨ t Jena, Jena, Germany Abstract Recent research on brain asymmetry suggested that resting electroencephalographic (EEG) asymmetry represents a superimposition of a trait-like asymmetry with state-dependent fluctuations. The present study tested this hypothesis and additionally examined individual differences in state changes. A 61-channel EEG was collected from 59 par- ticipants in a resting state on three occasions of measurement. An analysis of latent state-trait models suggested that between 40% and 50% of the variance of anterior asymmetry measures was due to individual differences on a latent trait and approximately the same portion of the variance was due to occasion-specific fluctuations. A further analysis of true intraindividual change models indicated that there were large individual differences in intraindividual change over time. These data replicate previous findings and substantiate that resting asymmetry has trait and state properties. Descriptors: Brain asymmetry, Electroencephalography, Individual differences, State-trait distinction Since the first published report on a relation between frontal EEG asymmetry and emotional experience (Davidson, Schwartz, Saron, Bennett, & Goleman, 1979), about 100 em- pirical papers have studied the association between frontal EEG asymmetry and affective states, motivational tendencies, affec- tive dispositions, and psychopathology (for an overview, see Coan & Allen, 2004). In the majority of this work, the EEG was recorded in a resting state, and the resulting measure of frontal EEG asymmetry was considered as an individual differences variable, that is, a biological trait. However, obtaining stable trait measures from EEG recordings in resting situations may pose a delicate methodological problem (Davidson, 1998). Several studies could consistently demonstrate that resting asymmetry measures of healthy subjects show retest correlations in the .50s or .60s for time intervals between 2 and 6 weeks for a variety of EEG reference schemes (Debener et al., 2000; Hage- mann, Naumann, Thayer, & Bartussek, 2002; Tomarken, Dav- idson, Wheeler, & Kinney, 1992). Because the asymmetry measures of these studies typically showed excellent reliability, this finding suggests that resting EEG asymmetry is not (only) capturing a stable characteristic of the brain. Rather, as Tomar- ken et al. (1992) have suggested, these data might indicate that resting EEG asymmetry reflects the joint contribution of a trait that is superimposed on state-like factors (see also Davidson, 1992). This proposal on the dual nature of resting asymmetry was recently put to test in a study of Hagemann et al. (2002), who used the latent state-trait of theory of Steyer, Ferring, and Schmitt (1992) to decompose resting EEG asymmetry measures into latent (not observed) state, trait, and error components. Models of Latent State-Trait Theory The latent state-trait theory is an extension of the classical test theory that accounts for the fact that no measurement takes place in a situational vacuumFrather, it takes account of the effect of the person and the effects of the situation (and their interaction) on any measured variable. To identify the variances of latent variables in empirical applications of this theory, the manifest variables have to be measured with at least two indicators i on at least two occasions of measurement k. Figure 1A and 1B illus- trate two basic models for the case of two indicators in three occasions. In a latent trait model (Figure 1A), all six manifest variables Y ik are decomposed into a common latent trait T and into meas- urement errors e ik that are specific for each variable Y ik . This model assumes that there are no substantial effects of the situ- ation or the person–situation interaction on the manifest vari- ables. The latent state-trait model (Figure 1B), however, is based on a two-step decomposition. In a first step, the two manifest The authors are grateful to Rolf Steyer for useful suggestions and also gratefully acknowledge Renate Freudenreich and Helmut Peifer for technical support, and Katja Bartz, Patrick Britz, Sven Haarscheidt, Melanie Hahn, Henning Holle, Alexander Hug, Pascal Klingmann, Astrid Kronbergs, Sonja Ro¨ mer, Mirjam Rupp, and Olaf Schweisthal for data acquisition and processing. The authors are also grateful to two anonymous reviewers for valuable suggestions. This research was sup- ported by the Deutsche Forschungsgemeinschaft through grant Ha 3044/ 2-1 to the first author. Address reprint requests to: Dirk Hagemann, Fachbereich I–Psy- chologie, Universita¨ t Trier, Universita¨ tsring 15, 54286 Trier, Germany. E-mail: hagemann@uni-trier.de. Psychophysiology, 42 (2005), 740–752. Blackwell Publishing Inc. Printed in the USA. Copyright r 2005 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2005.00367.x 740