3.3 Discussion
Our results showed modulation of the P2 in response to emotionally-laden visually presented words, but not an EPN response in contrast to earlier studies (cf Citron, 2012 for a review) which considered the EPN to be a robust EEG signature. Indeed, several recent studies failed to demonstrate a negative deflection at posterior sites during early time windows for emotionally valenced words (Pauligk et al, 2019; Ku et al. 2020; Imbir et al. 2021). In addition, unlike previous research showing larger P2 amplitude to either only positive words (Schapkin et al.,2000) or only negative words (Huang & Luo, 2006), our results demonstrated an enhancement of positive amplitude across the 220-300ms time window for both positively and negatively-valenced words, compared to neutral words. However, the effect was more widespread for negatively valenced than positively valenced words. The P2 component is assumed to reflect the early, automatic allocation of attentional regardless of task-relevance of emotional context (González-Villa et al., 2014). Our findings thus indicate that more attentional resources were allocated to emotion stimuli during the recognition task, undoubtedly due to the inherent importance of such stimuli. Our results add to literature on the P2 effect for positive as well as for negative valence words (Herbert et al.,2006). As noted above, this early effect was widespread for negative words, whereas positive words elicited a larger P2 amplitude only over frontal-central sites. According to the automatic vigilance model of emotion, humans have learned to avoid adverse situations for survival in threatening environments. In contrast, leaving a positive reward unattended would not incur a cost. The mechanism that quickly detects undesirable messages thus developed, leading to an attentional bias to negatively-valenced information (Estes & Adelman, 2008). It is thus possible that such “negativity bias” of emotion processing may have increased the P2 effect, extending to a boarder cortical response to negative than positive words. The concept of “negative bias” is in line with Rohr and Rahman’s (2015) results. Although these authors reported an EPN, which exhibits a distinct distribution and opposite polarity to our P2, they observed an EPN for negative words, but not for positive words, suggesting that the processing of negative words is prioritized over positive words at an early perceptual stage.
In addition, our results showed reduced N400 amplitude with a frontal-central distribution to words that carried negative connotations in relation to neutral words. Interestingly, in roughly the same time window, Grass et al. (2016) found that negatively connotated spoken words elicited a greater anterior positivity compared to neutral words. Recent, Leynes and Upadhyay (2022) suggested that distinct scalp distribution of N400 can be associated with either relative or absolute familiarity. They suggested that N400 detected at posterior sites reflects integrating process to accumulated semantic knowledge based on lifetime experience, while the frontally-distributed N400 (FN400) is hypothesized to be sensitive to the exposure of current content, with an attenuated negative wave to old relative to new items, indicating ease of processing for words that were recently presented, for example during a recognition task. Taken together, our results support the hypothesis that valenced words were processed more easily than neutral words due to the higher level of relative familiarity from previous auditory exposure, as indexed by a reduced FN400.
In contrast, our positive stimuli failed to elicit a reduced FN400, suggesting negligible facilitated processing during the recognition task. This result could be explained by the “negativity bias” of recognition memory in younger adults (Carstensen & DeLiema, 2018). That is, while aging brains are prone to positive content, younger adults tend to remember more negative than positive information. Taking this into account, less priming resources of positive words were maintained during initial encoding, leading to little or non-existent processing facilitation for positive words. This pattern of EEG results does not, however, match that for our behavioral data, which did not showed a “negative bias” where both negative and positive words were recognized better than neutral words. We can note that participants were not required to respond within a given time frame, such that the conscious retrieval of words may have overridden the facilitation often observed for negative words.
GENERAL DISCUSSION
The aim of the current study was twofold. First, we explored how the valence of auditory words, presented in an immediately prior experiment, influenced the subsequent recognition of printed words. Second, this design allowed us to directly compare the ERP response to the same stimuli across auditory and visual format. In relation to the first question, our results demonstrate a clear effect of valence/arousal of auditory stimuli on the subsequent recognition of these stimuli presented in written format. This was borne out by both behavioral and ERP data. Our behavioral results revealed increased recognition for both positive and negative valenced words compared to neutral ones, in agreement with the existing literature (Kensinger & Corkin, 2003). The neural response associated with the recognition of visually-presented valenced words was revealed by two different components, which varied according to polarity of words. We observed an increase in the P2 component, for both negative and positive words, although the effect was more widespread for negative words. In contrast, we observed a reduced N400 only for negative words in relation to neutral words. Overall, the results suggest that more attentional resources were allocated to negative words during recognition. It is possible that the relatively young age of our participants contributed to the “negative bias” that we observed in the ERP record (Carstensen & DeLiema, 2018).
Our ERP results align with those reported by Maratos and colleagues (2000) for valenced words. They demonstrated reduced N400 amplitudes to words with negative connotations compared to neutral words during a surprise memory task, with a maximum amplitude at frontal sites. However, the FN400 effect was found not only at the recognition phase but also during the prior encoding phase. As such their results do not support our claim that the effect arises from facilitated processing during recognition. Indeed, the FN400 effect has been reported in prior ERP studies including negative and positive words in diverse tasks. For instance, in a lexical decision task, Kanske and Kotz (2007) reported a reduction of N400 amplitude for concrete valenced words, relative to neutral words, and the effect was stronger over anterior than posterior sites. In an emotion-color Stroop task, valenced words elicited a reduced N400 compared to neutral words, with a frontal distribution (Sass et al., 2010). Together, these results indicate that the facilitated processing of valenced words per se, as indexed by reduced FN400, is likely due to higher-level inter-item association among valenced words. Otherwise stated, words with emotional connotations tend to prime one another due to their higher degree of semantic relatedness (Buchanan et al. 2006). However, our results diverge from the above-mentioned studies, in that we only found a reduced fN400 for negative items. If said effect was only due to semantic relatedness among stimuli, we would expect to observe a similar effect for positive words, which we did not. Therefore, it is less likely that the FN400 effect found for our negative words stems from inter-item association.
Kaestner and Polich (2011) reported a similar pattern of “negative bias” for facilitated processing during recognition in an ERP study using pictorial stimuli. Participants were instructed to memorize a series of pictures with high or low arousing, negative or positive valence, and then asked to indicate if the image had been presented previously. Behavioral results showed that d-prime (hit minus false alarm) was higher for negative than positive stimuli, suggesting that negative pictures were more salient and easier to detect than positive pictures. EEG results showed different patterns across the first, passive viewing task and the subsequent recognition task. During the recognition phase, compared to positive pictures, negative pictures showed larger effects of arousal, including enhanced positivity 200-400 msec and reduced negativity 450-650 msec post-stimulus. This effect, however, was not found during the passive viewing task. Instead, they reported a larger positivity 500-900 msec post-stimulus in response to negative pictures with high-arousal level, compared to low arousal level, reflecting a sustained allocation of attentional resources. Their results suggest that the effects of emotional valence is dependent on task demands and, critically, these findings support the idea that stimuli with negative content may have a privileged status in memory processing. Our results join this account and further demonstrate that the negativity bias in recognition memory extends to lexical items. In relation to the second question, we found clear differences in the neural response to valenced words as a function of presentation format. In line with the vast majority of studies, we did not find any evidence of early ERP effects for auditory presentation, in contrast to written words. As concerns later components, we again only found modulations, in the N400 time window, for written words, with no reliable effects for words presented auditorily. This question remains to be explored. Indeed, whereas early effects are quite rare, previous work has shown a modulation of the LPC (Hatzidaki et al., 2015) although it can be noted that their results were limited to processing in specific contexts. Overall, our results provide a compelling and complete picture of the effect of valence on the immediate processing and subsequent recognition of words as a function of presentation format. Our study provides valuable, cross-linguistic evidence to the existing body of literature and calls for further studies to examine how memory is affected by the emotional content of words.