Human judgment is essentially relative in nature. Abundant empirical evidence demonstrates that targets of all levels of complexity – ranging from simple objects (e.g., Brown 1953; Helson 1964) to complex social stimuli (Festinger, 1954) – are evaluated in a relative or comparative manner (Kahneman and Miller, 1986). When humans judge or evaluate a particular target, it seems, they do so by implicitly comparing it to a relevant context, norm or standard (Mussweiler, 2003). This tendency towards comparative information processing is characterized by a striking ubiquity. Social cognition research has provided numerous demonstrations of how strongly comparisons pervade our thinking. Comparisons play a core role in areas as diverse as stereotyping (Biernat, 2003), attitudes (Sherif and Hovland, 1961), person perception (Herr 1986; Higgins and Lurie 1983), decision making (Kahneman and Miller 1986; Tversky and Kahneman 1974), affect (Higgins, 1987), and the self (Festinger 1954; Higgins et al. 1986).#

Why are comparisons so ubiquitous? We suggest that the answer to this question may be tied to one of the most fundamental characteristics of humans as social information processors, namely their need to process information in an efficient manner (Taylor, 1981). People are constantly overwhelmed by the abundance of cognitive tasks they have to master. In light of their limited information processing capacities, they typically have to rely on those processing strategies that save scarce cognitive resources. That is, people often behave as cognitive misers (Taylor, 1981). Consistent with this hallmark principle of social cognition research, ample research attests that people have a strong tendency to rely on efficient modes of information processing such as the use of heuristics (Tversky and Kahneman, 1974) or stereotypes (Bodenhausen 1990; Bodenhausen and Lichtenstein 1987; Macrae and Bodenhausen 2000; Macrae et al. 1994). In much the same way, people as cognitive misers (Taylor, 1981) may rely on comparisons to save scarce cognitive capacities.#

In fact, these potential efficiency advantages become apparent upon closer inspection of the psychological mechanisms that underlie comparisons. Recent research demonstrates that comparisons allow the social information processor to focus on a subset of potentially judgment-relevant information (Mussweiler, 2003). Rather than engaging in an exhaustive search of all information that is potentially relevant for the judgment at hand, comparative processing allows judges to selectively retrieve a specific subset of this information (e.g., Mussweiler and Strack, 2000). More specifically, comparisons appear to involve a hypothesis-testing process (Mussweiler, 2003) during which judges primarily focus on hypothesis-consistent information (Snyder and Swann, 1978). Judges who test the hypothesis that the target and the standard of a comparison are similar, for example, selectively search for information which indicates similarity and essentially ignore information which indicates dissimilarity. In this respect, judges selectively focus on a hypothesis-consistent subset of judgment-relevant information. This informational focus in comparative information processing has been demonstrated in a number of recent studies (for an overview, see Mussweiler, 2003). In one of them, for example, participants were induced to compare a target quantity – the average price of a German car – to a numeric standard by testing the hypothesis that target and standard are similar (Mussweiler and Strack, 2000). Subsequent to this comparison, they worked on a lexical decision task which assessed the accessibility of information associated with high versus low car prices. Results indicated that information which is associated with high car prices was more accessible after a comparison with a high rather than a low standard, whereas the reverse was true for information associated with low car prices. This suggests that while carrying out the comparison, participants had selectively focused on information that was consistent with the hypothesis that the target quantity is similar to the numeric standard. Participants who compared the average price of a German car to the high standard, on the one hand, selectively focused on information indicating high prices and basically ignored information indicating low prices. Participants who compared average car prices to the low standard, on the other hand, selectively focused on information indicating low prices and ignored information indicating high prices. When making comparisons, judges thus appear to essentially ignore information that is inconsistent with the hypothesis that guides the comparison process. In this way, processing information in a comparative manner allows judges to significantly simplify the task at hand. Rather than engaging in an exhaustive and representative search of their knowledge base, judges limit themselves to a specific subset of information.#

In this respect, comparisons may save scarce cognitive resources. Recent social cognition research has provided initial support for this notion (Mussweiler and Epstude, 2006). Using standard social cognition measures of cognitive efficiency, these studies demonstrate that judges, who were induced to rely more heavily on comparisons during a series of target judgments, were able to make these judgments in a more efficient manner. This research applied a procedural priming logic (Smith, 1994) to manipulate the extent to which judges relied on comparisons during judgment. Participants were asked to report numeric estimates about a series of characteristics (e.g., the number of clubs) of an unknown City X. Participants received a brief description of City X in which some general information (e.g., university town) that was relevant for the critical estimates (e.g., number of clubs) was given. City X was described in terms that made it comparable to participants' hometown. Prior to the critical numeric estimates, participants worked on an ostensibly unrelated task in which they were provided with two pictures. Experimental participants were asked to compare the two pictures in writing (comparative processing), whereas control participants were asked to describe them (absolute processing). Previous research suggests that these different processing modes will carry over to the subsequent target judgment (Mussweiler, 2001; for a survey see, Smith, 1994). As a consequence, participants who were procedurally primed on more comparative processing should rely more heavily on comparisons in judging City X than those who were primed on more absolute processing. In fact, past research indicates that this is the case (Mussweiler and Epstude, 2006). This is, for example, apparent in the fact that after judging City X the pertinent comparison standard (i.e., participants' hometown) is more accessible for participants who were primed on comparative processing rather than absolute processing. In light of the fact that heightened accessibility of a concept typically reflects that this concept was recently used (Higgins, 1996), this suggests that these participants activated information about the standard and compared this information to the City X.#

If comparative processing does indeed hold an efficiency advantage, then participants primed on comparative processing should be more efficient in making the numeric estimates. Using two distinct measures of processing efficiency, results suggest that comparative information processing does indeed hold efficiency advantages. For instance, participants were able to give a critical estimate faster, if they relied more heavily on comparisons (Mussweiler and Epstude, 2006). Furthermore, in a dual task paradigm (e.g., Macrae et al., 1994), participants who processed information in a more comparative manner were able to allocate more residual processing capacity to a secondary task, which was apparent in superior performance in that task (Mussweiler and Epstude, 2006). This research provides initial behavioral evidence in support of the notion that comparisons save scarce cognitive resources. Comparative information processing thus appears to reduce mental effort.#

The present research was designed to provide more direct evidence for the hypothesized efficiency advantages of comparative information processing by directly examining electrocortical changes that are closely associated with reduced mental effort. Combining methods from social cognition research and the neurosciences can help to develop a broader understanding of the role comparisons play in social information processing. Such a multi-level approach is in line with recent pleas that psychological research and theorizing should build on an integration of diverse methods and procedures (e.g., Cacioppo and Berntson 1992; Ochsner and Lieberman 2001). Following such suggestions, we will examine the neural correlates of comparison processes, one of the most fundamental processes in social cognition. Similar attempts have previously been made to examine the neural correlates of social cognitive functions, including stereotyping (e.g., Hart et al., 2000), categorization (e.g., Ito and Cacioppo 2000; Ito and Urland 2003) evaluation (e.g., Cacioppo et al. 1996 1994), self-related knowledge (e.g., Heatherton et al., 2004; Kelley et al., 2002), and interpersonal behavior (e.g., Gardner et al., 2000). The neural correlates of comparative processing, however, remain to be specified.#

In the present study, we assessed changes in alpha amplitude in task-related Electroencephalogram (EEG) as a measure of effortful processing. Classic findings in EEG research have suggested that changes in alpha amplitude are a direct and reliable indicator of mental effort in that greater alpha power reduction is typically associated with more effort, arousal, or resource allocation, depending on the experimental task (e.g., Berger 1938; Ray and Cole 1985; Fink et al. 2005). Alpha band (i.e., 8 to 12 Hz) spectral changes have thus been used as an inverse measure of cognitive processing since the days of Hans Berger (Berger, 1938). Although recent work has provided evidence for a role of phase-locked alpha oscillations in stimulus processing (Schürmann et al., 1997) and more complex cognitive tasks (Basar et al., 2001), most EEG work points to a parametric reduction of alpha amplitude as a function of task difficulty or attentional demands (Ray and Cole, 1985). Temporal dynamics of alpha reduction have been used to examine memory processes (Klimesch et al., 1999) as well as motor behavior (Pfurtscheller and Aranibar, 1979) and emotional states (Sobotka et al. 1992; Fink 2005). Taken together, this work has suggested that stimulus- or task-related suppression of spectral power in the alpha range reliably indexes effort associated with well-defined behavioral tasks (see Neuper et al., 2005, for a recent discussion). While the literature on the cortical/sub-cortical origin of alpha modulation is inconclusive, several studies have converged, suggesting modality and task-dependent foci of alpha suppression (Pfurtscheller et al., 1994). For instance, visual processing appears to be associated with pronounced posterior-parietal alpha reduction, which can be related to changes in the activity at a variety of cortical regions (Hari et al., 1997). In particular, power reduction of alpha oscillations which are not phase-locked to the onset of a visual stimulus (i.e., so-called induced alpha responses, see methods) has been reliably observed at parieto-occipital EEG leads during effortful or attentive processing (Klimesch et al., 2000). Recently, the literature of alpha power modulations has been integrated in a comprehensive review (Klimesch et al., 2006), which concludes that alpha enhancement may be related to inhibitory top-down control of executive functioning. The present task, however, did not tap this kind of process, but rather involved cognitive operations of the kind that are related to widespread activity in attention and memory networks. This task type has been related to alpha decrease both empirically (Ray and Cole, 1985) and conceptually (Klimesch et al., 2006). Thus, if comparative information processing does indeed reduce mental effort, then this should be apparent in differential power changes of induced alpha.#

We examined modulation of alpha amplitude as evidence of electrocortical processes that are closely associated with mental effort. More specifically, using the same methods that were previously employed in the behavioral studies (Mussweiler and Epstude, 2006), we induced participants to engage in more or less comparative processing while making a series of target judgments. Employing a procedural priming method (Mussweiler 2001; Smith 1994), experimental participants were induced to rely more heavily on comparative processing when generating a series of numeric estimates about an unknown City X than control participants (see Fig. 1).#

We recorded electrocortical changes in alpha amplitude while the estimates about City X were generated. To increase the topographical specificity of the data, we used the minimum norm estimate (MNE). As a reference-independent estimate of voltage currents, this approach also aimed to avoid problems associated with interpretation of spectral parameters derived from scalp voltage data: For instance, bipolar generators are often associated with two distinct topographical maxima in voltage space (e.g., over anterior and posterior regions), but may reflect spectral changes at either of the two brain regions or a third region projecting to the electrodes under consideration (see Hauk et al., 2002, for simulation analyses and an extensive discussion of this issue). The MNE is a mathematical procedure for estimating the primary electric current that underlies the scalp-measured EEG (see Experimental procedures). Applying MNE to the EEG data results in values representing estimated strength at source locations located on a spherical source space, which was used as a model for the brain volume. Continuous MNE data can be submitted to frequency-domain transformation, allowing analysis of alpha changes on the level of modeled sources rather than on the level of voltage maps. We interpreted greater reduction in alpha amplitude as indicating a greater degree of mental effort. Thus, if comparative information processing reduces mental effort, smaller changes in alpha amplitude should result for experimental rather than control participants.#

The present findings suggest that comparative information processing holds efficiency advantages. On the electrocortical level, a more comparative mode of information processing was associated with smaller changes of alpha amplitude. In light of the abundant research pointing to a close link between changes in alpha amplitude and mental effort (e.g., Ray and Cole, 1985), our data suggest that comparative information processing reduces mental effort. Compared to control participants who processed the critical judgment tasks in a more absolute manner, the experimental group required less mental effort to solve the very same task. Notably, this held true irrespective of participants' social comparison orientation. Thus, comparative information processing was associated with less mental effort for participants who do versus do not rely on comparisons.#

These findings help to understand why comparison processes play such a pivotal role in human judgment and decision making. Research across a variety of domains has demonstrated that human information processing is essentially comparative in nature. Clearly, logical and conversational inferences are important factors that contribute to this ubiquity of comparative processing. Often, human judgments only make sense if a comparison with a pertinent standard is made. The statement “Bob is a hostile person”, for example, can only be successfully processed and interpreted, if the pertinent standard is taken into account (Huttenlocher and Higgins, 1971). In addition to such inferences, the efficiency of comparative information processing appears to be a second factor that contributes to the ubiquity of comparisons. People as cognitive misers (Taylor, 1981) tend to rely on efficient processes. The present study suggests that comparative information processing fulfills this efficiency criterion and helps to simplify human judgment. This may be the case, because comparisons reduce the breadth of information that has to be considered before a judgment can be made (Mussweiler 2003; Mussweiler and Strack 2000). Rather than trying to obtain a representative set of information about the target, judges who engage in comparative processing only consider the information that is needed to make the critical comparison with the pertinent standard. Focusing on such a subset of information helps people as cognitive misers (Taylor, 1981) to save scarce cognitive resources.#

The current study extends previous work in neuroscience and social cognition in a number of ways. It demonstrates that one of the standard electrocortical measures of psychological processes, namely changes in alpha amplitude, can be influenced by subtle procedural priming manipulations (Smith, 1994). To date, changes in alpha amplitude have been observed in a large number of experimental paradigms. For instance, phasic changes in alpha amplitude have been reported as correlates of attentive stimulus processing (Ray and Cole, 1985), affective stimulus processing (Keil et al., 2001), or motor performance (Pfurtscheller et al., 1994), among others. The present research demonstrates that differential patterns of this parameter may also be related to procedural priming (Smith, 1994). Although the two participant groups worked on the same judgment task, they were subtly induced to rely on different processing strategies in doing so. Applying these different strategies was sufficient to yield reliable differences in the amount of alpha reduction. Interestingly, this was only evident for the response-related alpha reduction. While the limited number of trials present in this work does not allow for a high-resolution time-frequency analysis, the response-related result suggests a locus of differential modulations that occurs late in the judgment process. In particular, aspects of response selection, motor preparation, and task execution may benefit from operating in a comparison mode as induced by procedural priming. Thus, the response-locked analysis may be more sensitive to these late, higher-order processes. Alternatively, cognitive processes related to forming an estimate may temporally extend into the later segment. This notion is supported by the relatively long response times. Furthermore, the fact that the baseline segments used here (pre-stimulus versus post-response) did not statistically differ in terms of alpha amplitude suggests that changes in overall arousal over the duration of a trial cannot account for the specific finding in the response-locked data. Source space projection together with permutation testing of topographical differences showed differential alpha reduction in central parietal cortex, but also in a widespread network encompassing left temporal and bi-lateral frontal cortex. This pattern of differential activation is consistent with reduced effort in areas mediating visual attention and higher cognitive processes, including verbal encoding of the response. Thus, a plethora of higher cognitive processes may benefit from processing in a comparative manner. It should be kept in mind, however, that changes with respect to the target stimulus did not show such modulation.#

In terms of limitations relevant to the interpretation of the present study, one may consider the long durations of Fast Fourier Transform windows (see Experimental procedures) as well as the low number of trials available for analysis. In particular, the estimation of temporal dynamics being achieved by means of complex demodulation or wavelet-transform (e.g., Keil et al., 2001) would be desirable in the context of models of social cognition. While the present experimental paradigm did allow for a high degree of similarity between designs in the social cognition and the EEG laboratories, future work may attempt to enhance the number of trials and modify designs to optimize the measured signal. Another limitation lies in the fact that long segments were used to obtain reliable estimates of amplitude with a low number of experimental trials. Certainly peak frequencies of the bands used here are also sensitive to spectral changes in the high theta or low beta range. Our data do indeed suggest that the selected alpha bands differ in terms of their sensitivity to comparative processing. As discussed in the introduction, such task-dependent modulation of sub-bands within the alpha range has been observed in a variety of other studies (e.g., Klimesch et al., 1998) and has strongly suggested that the functional role of oscillations within the traditional alpha range should be subject to systematic research (e.g., Fink et al., 2005). However, the general pattern of a stronger reduction for the control group was present throughout bands in the present study, although reaching significance for the lower band only. This supports our view that lower frequency activity was suppressed in preparation of the response, which is assumed to reflect ‘activation’ (e.g., Sobotka et al., 1992) or task-related system changes in brain states (Pfurtscheller and Aranibar, 1979). The topographical distribution of this difference as shown in Fig. 5 suggests that a parieto-frontal network differentially correlated with task performance in the two groups.#

To summarize, the present findings highlight one of the most central psychological processes in social cognition. Comparison processes play a crucial role in some of the most central phenomena of social cognition research such as stereotyping (Biernat, 2003), attitudes (Sherif and Hovland, 1961), person perception (Herr 1986; Higgins and Lurie 1983), affect (Higgins, 1987), and the self (Festinger 1954; Higgins et al. 1986). Despite widespread theoretical and empirical agreement, little was known until recently about the psychological mechanisms that underlie comparisons. Recent work has started to shed light on these comparison mechanisms by applying behavioral methods (for a review, see Mussweiler, 2003). The present study extends this work by examining the neural correlates of comparison processes with the help of standard neuroscience methods. In combination, both approaches may help to shed light on one of the most central characteristics of human information processing, namely the tendency to process information in a comparative manner.#