Please be logical, I am in a bad mood: An electrophysiological study of mood effects on reasoning

Highlights

• An ERP study was conducted on the influence of mood on reasoning.

• Compared to illogical, logical conclusions showed reduced N400 ERP responses.

• Under an induced negative mood, this reduction was rather extended.

• Participants anticipated logical conclusions when reading categorical syllogisms.

• A negative mood increased the expectation of logical conclusions.

Abstract

Several behavioral studies have reported a detrimental effect of emotion on reasoning tasks, either when the content of the reasoning and/or the mood state of the individual are emotionally loaded. However, the neural mechanisms involved in this phenomena remain largely unexplored. In an event-related potentials (ERPs) study, we examined the consequences of an induced mood over the electrophysiological signals obtained while processing logical and illogical categorical conclusions. Prior to performing a syllogism reading task, we aimed to induce, by using short film clips, high arousal negative and positive moods and neutral affective states to participants in three separate recording sessions. Our mood induction procedure was only successful at inducing a highly arousing negative state. Behaviorally, participants committed more errors overall while judging the invalidity versus the validity of illogical and logical conclusions, respectively, but no influences from mood state emerged at this logical validity task. Electrophysiologically and overall a negative going N400 deflection was larger for illogical relative to logical conclusions in a parietal region between 300 and 420 ms. However, further analysis revealed that the logical conclusions were only more expected (smaller N400 amplitudes) in the negative relative to the neutral and the positive sessions, providing support to theoretical views that posit that a more analytic reasoning style might be implemented under a negative mood state. These results provide further electrophysiological evidence of the influence of mood on other cognitive processes, particularly on the anticipation and processing of logical conclusions during online reasoning tasks.

Introduction

Reasoning is the psychological process through which individuals organize, structure, and draw inferences from information (Blanchette, 2014). The dual process model of reasoning postulates a distinction between heuristic (implicit, automatic, associative, and intuitive) and analytic processes (more effortful, explicit, rule-based, and slower) (Evans, 2003). Just like for other cognitive processes such as attention, perception, memory, and problem solving, reasoning has been found to be influenced by emotional content and emotional state (see review by Blanchette and Richards, 2010). Working memory load (i.e., requirement of additional cognitive resources to process emotional information) has been proposed as one of the potential mechanisms mediating this interference of emotion on logical reasoning tasks (Tremoliere et al., 2016, Tremoliere et al., 2018).

Formal logical reasoning has been shown to be impaired (prompted to error) both when the content of the matter is emotional versus neutral (Blanchette and Richards, 2004, Lefford, 1946, Tremoliere et al., 2016) and also under negative and positive induced emotional states (e.g. Jung et al., 2014; Melton, 1995; Salovey, 1993).

Despite early studies on emotion and reasoning led to the simplified conclusion that “emotion leads to faulty reasoning”, recent studies suggest that the interplay between emotion and reasoning is more nuanced and complex (Blanchette, 2014). For example, emotion can lead to better, not worse, logical reasoning as it may facilitate access to relevant information during reasoning (Gangemi et al., 2014). Moreover, rather than postulating whether reasoning is either right or wrong under emotional conditions, some authors posit that mood influences an overall individual's processing style. In their own words, ‘being happy or sad influences the content and style of thought’ (Clore and Huntsinger, 2007). The mood-influenced cognitive styles hypothesis (also called the affect-as-information hypothesis) suggests that, generally speaking, positive mood is associated with a more global and flexible processing mode that relies on heuristics (Ruder and Bless, 2003), whereas negative mood is thought to promote a relatively analytical, careful and effortful processing style (Clore and Huntsinger, 2007). This view, has however been challenged by a recent review (Huntsinger and Clore, 2014) that posits that positive affect may also lead to detailed processing and a narrowed focus, and negative affect may lead to heuristic processing and broadened focus.

From an anatomical perspective, even in the absence of behavioral effects (a similar accuracy in logical decision tasks under different emotionally induced conditions) some functional Magnetic Resonance Imaging (fMRI) studies reveal the existence of mood-dependent differences with regard to the pattern of brain activations at the time of syllogistic reasoning (Smith et al., 2015, Smith et al., 2014). Other fMRI studies found a worse behavioral performance in syllogisms together with a lateral/dorsolateral prefrontal cortex (lat/dlPFc) increased activation under negative mood conditions (Brunetti et al., 2014). Thus, brain imaging studies point to a differential recruitment of brain areas for reasoning under the influence of a negative mood, but it is not clear yet when, if so, emotional state exerts its influence upon reasoning tasks.

In recent years, the Event-Related Potential (ERP) technique has been used to examine the online processing of both conditional (Blanchette and El-Deredy, 2014, Bonnefond and Henst, 2013, Bonnefond et al., 2014, Bonnefond and Van der Henst, 2009, Qiu et al., 2007) and categorical reasoning (Rodríguez-Gómez et al., 2018). Conditional reasoning uses arguments in the form: “If you water the plants, then they will grow; You water the plants; The plants grow.” (i.e., the “Modus Ponens” argument form), whereas categorical reasoning uses syllogisms such as: “All cats are mammals. All mammals have lungs. Therefore, all cats have lungs”. In both cases, reading the argument and following its logic allows the reader to most likely anticipate the last word of the conclusion (grow and lungs, in previous examples). The ERP technique, with a high temporal resolution, allows to examine the time course at which a variable such as emotion can exert its influence over an ongoing process, in our case, a reading for comprehension task.

The study by Blanchette and El-Deredy (2014) manipulated the emotionality of the verbal content upon which conditional reasoning was performed (e.g. If a country is at war, then people die; Britain is at war; British people have died). They found that emotional content had no influence on early ERP components, nor interaction with inference making ERP responses at middle time ranges (N400) and only marginal main effects at late latencies (800–1050 ms). The authors concluded that the effect of emotional content on conditional reasoning “might occur after actual inference making has taken place, maybe at the stage of conclusion maintenance, or response selection” (Blanchette and El-Deredy, 2014).

In contrast, the temporal dynamics of an individual's emotional state upon reasoning tasks with emotionally neutral materials remains unexplored. We carried out an adaptation of a previous ERP study on categorical reasoning (Rodríguez-Gómez et al., 2018). In the original study we were interested in the processing of syllogisms conclusions as a function of whether the major premises of the syllogism had previously been rated as true or false. In the condition in which the major premises had been rated as true, illogical conclusions lead to an increase of the N400 ERP component (between 380 and 512 ms) compared to logical ones. The N400 ERP component was first discovered as the response to words that render a sentence semantically incongruent (i.e. a nonsense word) (Kutas and Hillyard, 1980a, Kutas and Hillyard, 1980b). Nowadays, a reduction of amplitude of the N400 is best viewed as an index of a facilitatory process for word items that could have been pre-activated or anticipated based on prior contextual constraints (Federmeier, 2007, Federmeier and Kutas, 1999).

Thus, according to N400 functional significance (Kutas and Federmeier, 2011), our result demonstrated that in the context of a reading task participants were most likely to anticipate logical rather than illogical conclusions as far as the major premises held true. Considering prior literature on how an emotional state might alter an individual's cognitive load, attentional resources and reasoning style, we replicated this part of the study manipulating current mood in three separate sessions. A large number of emotion-elicitation techniques have been used so far to induce emotional states: exposure to emotional slides (Bradley and Lang, 2000, Schaefer et al., 2009), autobiographical recollection (Schaefer and Philippot, 2005), mental imagery (Schaefer et al., 2003, Vrana et al., 1986), Velten mood-induction technique (Velten, 1968), facial feedback (Matsumoto, 1987), respiratory feedback (Philippot et al., 2002), and real-life techniques (Landis, 1924, Stemmler et al., 2001), but viewing emotional clips was chosen as it is one of the most commonly used and effective procedures to induce mood in participants (Gerrards-Hesse et al., 1994, Gross and Levenson, 1995, Schaefer et al., 2010, Westermann et al., 1996, Zhang et al., 2014).

Behaviorally, our results could either match those studies in which emotional states prompted to more errors in logical decision tasks (Jung et al., 2014) or those who failed to find such mood-related effects on logicality error rates (Smith et al., 2014).

Regarding brainwave responses, the study by Blanchette and El-Deredy (2014) only found a late marginal effect of emotional content during conditional reasoning. We aim to determine whether emotional state, in contrast, influences reasoning (i.e. categorical) and the time-course at which it might exert its influence, if any. Prior ERP work has shown that an induced positive mood facilitates the integration (smaller N400) of more distantly related semantic information during sentence comprehension (Federmeier et al., 2001, Pinheiro et al., 2013). However, influences of a negative mood have also been described (Chwilla et al., 2011), with a reduction of the N400 cloze probability effect (i.e. a reduction of the commonly larger N400 to low vs. high cloze probability target words embedded in sentences). Using ERP measurements, other linguistic phenomena have also been described to be mood-dependent, such as a better referential anticipation under a happy versus a sad mood (Van Berkum et al., 2013) or a better sensitivity to semantic reversals under a happy than a sad mood (Vissers et al., 2013). The results of these studies indicate that indeed mood is able to influence processes of language comprehension. The question is whether these mood effects also arise during the anticipation of conclusions for categorical syllogisms (reasoning), and whether the influence is early (at the stage of semantic integration) or late (at the stage of reanalysis processes). As we saw, the latter is the finding for conditional reasoning with emotional vs. neutral content (Blanchette and El-Deredy, 2014).

In summary, the goal of the current ERP study is to provide a direct test of the effects of mood state on categorical reasoning, particularly on whether the logical conclusion might have better be anticipated under the influence of a particular induced mood or alternatively whether the response to illogical conclusion is enhanced for a particular mood state. To achieve this purpose, we tried to induce positive, negative and neutral moods using short-duration videos in three different recording sessions. Subsequently, participants were engaged in a silent reading task of categorical syllogisms, while they were asked to decide whether the logical conclusion followed from prior premises.

Based on the results of the studies reviewed here and on theoretical accounts that posit that reasoning style is more analytic under a negative mood (Clore and Huntsinger, 2007), we expect to obtain a smaller N400 to logical conclusions under a negative mood, indicating that it facilitates anticipation of logically valid analytical conclusions. Likewise, if positive mood is associated with a more global and flexible processing mode that relies on heuristics (Ruder and Bless, 2003), the ERP response to illogical conclusions under a positive mood might show a reduced N400 response indicating that they became more acceptable. Alternatively, the influence of emotional state in online reasoning tasks might only be manifested in later ERP components in line with the study on conditional reasoning with emotional content that was conducted by Blanchette and El-Deredy (2014).