Cognition and emotion are two, intrinsically connected, major elements of brain sciences, although often regarded as distinct. Many aspects of real life involve applying some kind of emotional criteria to make a decision or deal with a situation. Earlier studies have shown a high degree of connectivity between brain regions, such as frontal areas and limbic structures, as being central to cognitive–emotional interactions, critical for integration of information from different brain regions (e.g., Young et al., 1994). This relationship between cognition and emotion is performed via many different forms of interaction linked to the functional organization of the brain through many cognitive processes (Phelps, 2006; Pessoa, 2008) and also to body responses through the autonomic nervous system (Scherer, 1984; Lang, 1985; Umeno et al., 2002).

Working memory (WM) is a complex cognitive system designed to retain information until a certain problem has been solved (Baddeley, 1986). The proposal linked to this process is that the resolution of a problem involves the suitability of an overall decision, which requires the evaluation of experienced facts. Moreover, WM alludes to the need to hold a stimulus in a priority state while information is manipulated to reach a certain goal, this being an essential part of the attention process. This focus of attention results in the explicit representation of the information in WM, and the simultaneous inhibition of irrelevant information (Hester and Garavan, 2005; Kim et al., 2005; Korsten et al., 2006; Pratt et al., 2011). The central executive is considered an attention system that controls the explicit manipulation of information during problem solving. It is comprised of the machinery necessary for manipulation of information in collaboration with specific components. One of its servers is the so-called episodic memory buffer, which binds information from multiple systems and recalls memories of personal experiences (Baddeley, 2000).

Electrophysiological measures, such those in electroencephalography (EEG), galvanic skin response (GSR), and heart rate (HR) variability, have been used to evaluate the mental workload and emotion recognition during dual tasks. Some of these studies assess decision-making and perception by means of association between the arousal of emotional content and increases in GSR (Laparra-Hernández et al., 2009). Others use emotional content to induce changes in autonomic nervous system, especially in HR and in GSR, in an attempt to define standards of autonomic responses for different types of emotion (Lee et al., 2005). Earlier studies indicate that cortical damage, especially to frontal lobes, may selectively attenuate GSR to psychologically relevant stimuli (Zahn et al., 1999). Additionally, cognitive functions such as attention and emotional memory encoding have been related to changes in parasympathetic activity (McCraty and Tomasino, 2006; Critchley, 2009). Some studies found a relationship between frontal and temporal cortice activity and cardiovascular responses associated with increasing emotional intensity, due to high arousal emotional content (Foster and Harrison, 2004).

The WM neural network has been the subject of behavioral, psychophysiological, and functional neuroimaging studies. Neuroimaging and EEG research has revealed the brain connections that support cognition, emotion, and the integration of information. The mainstays of this circuitry are the prefrontal cortex, including the central area, the occipito-parietal, and medial temporal regions (Curtis and D’Esposito, 2003; Osaka et al., 2003; Wager and Smith, 2003; Mizuhara et al., 2004; Erk et al., 2007; Gläscher et al., 2007; Pyka et al., 2009; Michels et al., 2010). Some studies were critical to sustain that the hippocampus modulates the WM system through temporal oscillations (Axmacher et al., 2008; van Vugt et al., 2010). As well as the functional association effects on WM performance, there is a general need to appraise the temporal organization of participant mechanisms using electrophysiological recordings, to demonstrate task-related brain activity (Jensen and Tesche, 2002; Schack and Sauseng, 2005; Kawasaki et al., 2010; Khader et al., 2010; Moran et al., 2010).

Behavioral procedures are important to elicit cognitive processes. The delayed matching to sample (DMTS) and delayed non-matching to sample (DNMTS) paradigms are widely used to assess WM (Tavares and Tomaz, 2002; Winocur and Hasher, 2004; Leiberg et al., 2006) and have been associated with emotional factors (Gasbarri et al., 2008; Satler and Tomaz, 2011). Both tasks involve memory and combining stimuli with a correct response. The DNMTS also requires increased attention to the subject chooses the novel stimulus (Elliot and Dolan, 1999). The difference in performance between stimuli categories, including emotional factors, represents the interaction between emotion and cognition. There is evidence for engagement of frontal and temporal areas of the brain, as well as premotor and occipital cortices, during DMTS and DNMTS performance (e.g., Elliot and Dolan, 1999; Lamar et al., 2004).

Several studies, designed for different purposes, have revealed relationships between cognition/emotion with EEG, HR, and GSR signals. For emotional WM, there are no studies including EEG brain mapping associated with autonomic responses. Therefore, the aim of the present study was to investigate electrophysiological responses during WM performance associated with emotional content processing. Delayed matching and non-matching to sample tasks were used with simultaneous EEG, GSR, and HR recording evaluations.

This approach will serve to elucidate: (1) central and peripheral correlates of WM performance in each of the tasks, (2) central and peripheral correlates of emotional content processing while performing a cognitive tests, and (3) central and peripheral correlates of the interactions between these two factors.

Several hypotheses will be tested: (1) Parasympathetic activity will be higher during DMTS task performance compared to DNMTS; (2) WM performance in both tasks will be facilitated by stimuli with emotionally arousing content; (3) Emotionally arousing content will engage sympathetic activity; and (4) Task and content effects over autonomic activity will interact in conditions of high cognitive demand and emotion arousal (e.g., DNMTS with emotional stimuli).

This study investigated EEG power concomitantly with autonomic reactions (GSR and HR) during WM performance associated with emotional content processing.

Behavioral results showed a high performance in both conditions, but statistically significant different scores between them (DMTS > DNMTS). The task comprises the ability to maintain the sample information online and recall this information after a delay according to a specific rule. Participants recalled emotional stimuli, negative and positive, with performance comparable to the neutral and geometric stimuli. The high correct performance found in both tasks limits us to emphasize a possible memory enhancing effect of emotional content, unlike that reported by earlier studies (Kensinger and Corkin, 2003), but consistent with others where negative emotional content were effective over the positive contents (Satler and Tomaz, 2011), mainly in DNMTS condition where more attention is required. However, we found a difference between tasks independent of the stimuli category. This can be explained by the fact that DNMTS requires the choice of the unfamiliar response, inhibiting the instinctively preferred familiar response. This aspect makes DNMTS performance lower than DMTS, supposedly due to attention demand (Elliot and Dolan, 1999; Lamar et al., 2004).

Electroencephalographic findings showed diversified rhythms which were present mainly in frontal and midline brain regions. Studies have shown that WM cannot be viewed as a single or dedicated system. A network of brain regions is essential for mediating goal-directed behavior, characteristic of this process (Collette and Van der Linden, 2002; D’Esposito, 2007). Theta activity was mainly observed in the prefrontal cortex and posterior association areas. These areas are strongly associated with central executive actions responsible for manipulation of information and planning (Baddeley, 2002; Jensen et al., 2007). Theta activity may have been induced by the task conditions performed. These results suggest that theta oscillations may play an important role in organizing the information network provided by WM, and this network operates as an integrated unit by means of synchronization in the theta band, consistent with earlier studies (Sarnthein et al., 1998; Raghavachari et al., 2001).

Alpha activity extended from frontal to posterior areas. Earlier findings suggest that alpha, associated with theta, oscillations reflect the transition of information from manipulation to maintenance state of WM tasks (Klimesch, 1999; Schack and Sauseng, 2005; Kawasaki et al., 2010; Moran et al., 2010). Likewise, DMTS tasks reproduces successful memory encoding and loading for processing WM. Increased alpha and theta power are regularly related to WM encoding, considering similar memory encoding processes to increasing WM load. This is consistent with the suggestion that alpha and theta oscillations modulate successful memory encoding (Khader et al., 2010).

The present study shows an increase of beta oscillations in frontal and temporal regions and includes task-related activity on the left hemisphere of the brain. This may be associated with attention demand, as representation of stimuli for DNMTS condition. Frontal beta activation has been thought to be a memory representation of task stimuli (Dörfler et al., 2001; Vernon, 2005; Spitzer and Blankenburg, 2011). This band often appears with theta oscillations, WM load, and encoding processes (Babiloni et al., 2004).

Gamma band activity was distributed over frontal and temporal regions, and higher for DNMTS than DMTS condition. High frequency gamma oscillations have been considered an evidence of involvement in attention process (Mizuhara et al., 2004). Alternatively, gamma activity can reflect memory maintenance according to the short-term memory model, where each item of the memory set is encoded by cortical neurons at the gamma band frequency (Lisman and Idiart, 1995). Previous reports on gamma load-dependence in WM connected with theta oscillations also enhanced the idea that gamma oscillation increases as an expression of maintenance of the to-be-remembered stimuli (Howard et al., 2003; Jensen et al., 2007; Klimesch et al., 2007).

These findings also add some information on WM brain dynamics regarding temporal activity. In previous studies, beta rhythm has been observed at the left temporal and frontal electrodes during WM encoding phase (Onton et al., 2005; Pesonen et al., 2006; Altamura et al., 2010). In a general WM analysis, temporal activity was more prominent in the beta band than the gamma. These results attempt to others that indicate the gamma band leading WM load, with no reference to beta band in any manifestation (Axmacher et al., 2008; van Vugt et al., 2010).

Our results, therefore, show fluctuations in theta, alpha, beta, and gamma bands. Beta band was relevant in right frontopolar site as indicative of consistent task-related changes. Previous studies indicated oscillatory neuronal activity occurring across these frequency bands during WM load (Michels et al., 2010), and these fluctuations were also observed in tasks requiring organization and temporal segmentation of information (Howard et al., 2003).

Some studies have indicated that emotions impact on the memory encoding process (Abe, 2001; Richter-Levin and Akirav, 2003), while underscoring the importance of amygdala–hippocampal connection in emotional memory load (Frank and Tomaz, 2003). Feasible hippocampal projection related to left temporal activity could be favored by subcortical structures responsible for emotion, such as amygdala, so as to better respond to emotional stimuli (McGaugh, 2000). Prior studies with WM tasks showed that bilateral prefrontal cortex activity reflected equally the emotional and WM task components (Gray et al., 2002). Earlier experiments showed a theta band manifestation for emotional content and gamma band for emotion arousal, both in frontal regions (Balconi and Pozzoli, 2009). These studies, along with our results, emphasize a significant electrophysiological interaction between cognition and emotion.

Autonomic responses showed an increased sympathetic activity (high LFHF and low pNN50) for the DNMTS condition. Results in HR index demonstrate a parasympathetic dominance in the DMTS condition, when a lower demand for cognitive functions is expected. Parasympathetic mediation of HRV has been associated with efficient attentional regulation and greater ability to inhibit inappropriate responses. Parasympathetic nervous system control of cardiovascular function, as well as activity of the prefrontal cortex, is associated with these inhibitory processes (Hansen et al., 2003). HR is a physiological variable controlled by sympathetic and parasympathetic activity. However GSR is not influenced by the parasympathetic nervous system, thereby making it a reliable index of sympathetic activity level (Venables, 1991), and is frequently used as an indirect measure of attention, or emotional arousal. The level of skin conductance was consistent with the cognitive effort considering the low performance for geometric sample stimuli and the arousal of negative emotional pictures. Equivalent emotional charge was applied in both conditions, thus no significant differences in GSR were found between conditions, but the expression of the sympathetic nervous system through a cardiovascular response was seen through the increase of GSR in the DNMTS condition.

The role of emotion was analyzed by evaluating human body response by means of HR and GSR measurements. Emotional content somehow modulates WM performance (Kensinger and Corkin, 2003; Levens and Phelps, 2008). Cognitive effort can also exert influence over emotion-related brain activation (Erk et al., 2007). Extrapolating, WM development has some benefit in the comprehension of emotions (Morra et al., 2011). This cross relation suggests that WM is susceptible to emotional content, in the same way that minimizing emotional reactions favors goal maintenance. Thus, the relationship between WM and emotion could be evaluated taking into account the arrangement of electrophysiological responses found while performing a WM task with emotional content. The correlations found between physiological variables were higher in the DNMTS condition, when attention demands are higher.

These results suggest an electrophysiological framework with interactions between brain dynamics and autonomic responses elicited by emotional engagement in a WM task. This adds elements to the relationship between cognitive processes and emotion. Future studies should evaluate the distinction over each emotion valence in task-relevant effects, assessing the specific contribution of positive, negative, and neutral, including abstract, representations.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.