Love and hate do not modulate the attentional blink but improve overall performance*

ABSTRACT

How may feelings of love and hate impact people’s attention? We used a modified Attentional Blink (AB) task in which 300 participants were asked to categorise a name representing a person towards whom they felt either hate, love, or neutral (first target) plus identify a number word (second target), both embedded in a rapidly presented stream of other words. The lag to the second target was systematically varied. Contrary to our hypothesis, results revealed that both hated and loved names resulted in higher accuracy for the second target than neutral names, which was largely independent of lag. Also, there we observed no sustained transfer effects of love and hate onto neutral name trials. The findings differ from prior research on attentional blink and transient, non-personal, stimulus-driven emotions, suggesting that interpersonal feelings activate different attention-relevant mechanisms. Relevant to future research, we speculate that love and hate are motivators of goal-directed behaviour that facilitate subsequent information processing.

KEYWORDS:

• Love
• hate
• attention
• attentional blink
• cognition & emotion

Love and hate can have enormous influences on a person’s life. Many of us can probably relate to one or both of these feelings and the impactful consequences they can have. Beyond direct experience, one only has to watch films, read novels, or listen to music to see that love and hate are exceptionally strong feelings. Indeed, past research has provided evidence for the intensity of such feelings, and though love and hate can be inspired by specific circumstances, they often occupy one’s mind over longer periods of time (Fischer et al., 2018; Fitness & Fletcher, 1993; Martinez et al., 2022). These feelings are therefore likely to have a lasting impact. Here we investigate how love and hate may affect basic information processing abilities. Specifically, do feelings of love and hate undermine our ability to direct attention to task-relevant stimuli?

There has been a substantial amount of research on how specific emotions (e.g. anger, fear, disgust) might affect attention. In general, people automatically prioritise emotional over neutral information (e.g. Compton, 2003; Phelps, 2006; Yiend, 2010). However, the effect of emotional stimuli on attention towards non-emotional stimuli is rather complex. For example, emotional stimuli have been found to facilitate attention to non-emotional stimuli appearing in the same location (Fox et al., 2001), but undermine attention to stimuli that are temporally close (McHugo, 2013; Most & Wang, 2011). While most past research has focused on momentary, non-personal emotional experiences triggered by stimuli that are relevant within the context of the experiment, the present research seeks to extend this research by examining personal and relatively enduring feelings toward another individual that we assumed to be of central relevance to the participant, namely love and hate.

Love and hate: intense, enduring, and interpersonal

Love and hate are broad concepts. People can love or hate a person, a group, or an object. In the current study, we focus on interpersonal love and hate, thereby excluding manifestations of love and hate towards groups or objects. The primary reason is that we wanted to enhance the comparability of feelings of love and hate, and we assumed that both interpersonal love and hate are primarily rooted in past social interactions. The conceptualizations of love and hate are difficult, and researchers find it hard to reach a consensus (see Berscheid, 2010; Fehr & Russell, 1991; Martinez et al., 2022; Sternberg & Sternberg, 2008). Here we adopt the definition offered by Sternberg (1986, 2003) who defined love and hate within the same theoretical system. Specifically, according to his triangular theory of love and hate, love contains feelings of intimacy, motivating elements of passion or sexual consummation, and/or cognitive aspects of responsibility and commitment (Sternberg, 1986). In contrast, hate includes feelings of distancing, behavioural tendencies of approach or avoidance, as well as a cognitive aspect of devaluation of another person (Sternberg, 2003).

Love and hate sometimes share the features of other emotions that accompany them (Fischer et al., 2018). For example, romantic love contains the component of eroticism (Zeki & Romaya, 2010); disgust is also closely linked to hate (Martinez et al., 2022; Pretus et al., 2018), as are anger and contempt (Fischer et al., 2018; Martinez et al., 2022). Furthermore, despite the fact that love is positive and hate is negative, both feelings involve high levels of arousal (e.g. Martinez et al., 2022; Sternberg, 1986). In light of this similarity, it is possible that high arousal underlying love- and hate-related stimuli may have a similar effect on attention as discrete high-arousal emotions like eroticism and disgust.

But from another perspective, both love and hate differ from discrete emotions in several respects. First, love and hate are often more intense. For instance, hate has been shown to be more intense and involve greater arousal than anger and contempt (albeit that hate has been found to be equally intense as disgust; Martinez et al., 2022; see also Rempel & Burris, 2005). Second, love and hate are more enduring compared to the (volatile) emotions that have been studied in past research on attention (Martinez et al., 2022; Rempel & Burris, 2005). Third, the discrete emotions used in the past research such as disgust, fear, or happiness can be elicited by stimuli that are not inherently social (e.g. rotten food, snakes, beautiful nature) or unrelated to social interaction experiences (e.g. Flaisch et al., 2016; Perone et al., 2021; Sklenar & Mienaltowski, 2019; Stein et al., 2009). In contrast, love and hate are inherently social feelings as they are rooted in typically cumulative social interaction experiences (Aumer-Ryan & Hatfield, 2007; Cacioppo et al., 2012; Ortigue et al., 2007) and as such guide our behaviours in future social interactions (e.g. Acevedo et al., 2012; Fischer, 2018; Hazan & Shaver, 1987; Martinez et al., 2022). For instance, feelings of love motivate people to care about the welfare of others and give rise to various prosocial behaviours, such as accommodation, gratitude, and sacrifice (Canevello & Crocker, 2020; Rusbult & Van Lange, 2003). In contrast, feelings of hate may lead people to devalue and distrust others, as well as give rise to the development of hostility in the long term (Aumer et al., 2015; Halperin, 2008; Martinez et al., 2022).

To capture the long-term love and hate that tend to be relatively stable over substantial periods of time, here in the current study, we define love and hate as feelings instead of emotions. With all those similarities and differences, whether love and hate can have a similar influence as the transient emotions accompanying them, or do their unique features make them have a special influence on attention?

How do emotions and feelings affect attention?

Several studies examined the relationship between emotions/feelings and attention, thereby using a variety of different tasks (see Compton, 2003; Phelps, 2006; Yiend, 2010, for review). Of these, the Attentional Blink (AB) task is one of the most used and validated paradigms, which can provide information on how the processing of emotional information affects the selection, encoding, as well as retrieval of stimuli (Dux & Marois, 2009; McHugo, 2013). In a typical AB task, each trial consists of a rapid serial visual presentation (RSVP) stream that involves two targets that participants need to memorise and respond to at the end of the trial. The second target can appear at different positions relative to the first target (i.e. lag). It has been demonstrated that attention is temporarily unavailable for new input (second target, T2) when earlier relevant visual information (first target, T1) is being processed successfully. This phenomenon, referred to as the attentional blink (Raymond et al., 1992), usually happens when the second target is temporarily close to the first target (typically at lags 2 and 3) and weakens or disappears when T2 appears at later positions (typically after about lag 5).

As far as we know, there has been no study investigating the effect of love and hate on the attentional processing of the following stimuli using the AB task. The only study involving love and attentional blink used romantic faces as T2 and showed that romantic faces were better recognised than faces of a friend and strangers (Nakamura et al., 2017). However, we can make some inferences based on the previous literature on discrete emotions accompanying love and hate. Researchers found a larger attentional blink when angry faces were displayed as T1 than when happy or neutral faces were displayed as T1 (de Jong et al., 2010). It does not mean that positive stimuli cannot influence performance in the AB task. Instead, there is evidence that both positively and negatively laden T1 stimuli can elicit a larger attentional blink as long as the stimuli used at T1 are associated with high arousal (e.g. erotic or aversive pictures, Flaisch et al., 2016; aversive or sexual/taboo words, Mathewson et al., 2008; Schwabe et al., 2011; Schwabe & Wolf, 2010).

Even when the emotional T1 is not a target that participants need to identify, it has a similar effect as in the AB paradigm. Specifically, a large attentional blink is evoked when task-irrelevant or to-be-ignored T1 are high-arousal images (Ciesielski et al., 2010; Flaisch et al., 2016; Keefe et al., 2019; Most et al., 2005; Most et al., 2007; Perone et al., 2021) and words (Arnell et al., 2007; Mathewson et al., 2008). Scholars have suggested that competition for representation in capacity-limited working memory is one of the key explanations for the findings that high arousal and emotion-eliciting stimuli undermine attention to other stimuli in the short term; that is, the emotionally more salient stimulus is represented at the expense of the second target (e.g. Schwabe et al., 2011).

Moreover, love and hate are enduring feelings that could trigger sustained influences on performance even beyond the current trial. The relationship between attention and positive or negative mood is widely studied. Generally, this research has uncovered that positive mood – including joy and love – can broaden attention to new stimuli (see Fredrickson, 2001; Vanlessen et al., 2016) and improve even higher-order cognitive processes, such as working memory (e.g. Carpenter et al., 2013; Yang et al., 2013). In contrast, a negative mood can either undermine the span of attention or enhance the span of attention to other negative, especially threatening, information; yet it is usually related to reduced goal-directed attentional processing (Vanlessen et al., 2016). It is again important to note that most of the studies mentioned above focused on discrete emotions (e.g. joy and contentment, fear and anger; Fredrickson, 2001), rather than sustained feelings. This leaves open the question of how relatively enduring states may affect attention. While love and hate have not been addressed, other relatively enduring states have been studied. For example, there is some evidence suggesting that negative states (e.g. depression) may undermine attention, whereas positive states (e.g. optimism) do not or even may enhance attention (Basso et al., 1996). As a case in point, there is some research suggesting that positive affect reduces the magnitude of attentional blink (Olivers & Nieuwenhuis, 2006), although this awaits replication. In that study, positive, negative, or neutral stimuli were used at the beginning of each trial to induce the mood and researchers found that the magnitude of AB was smallest when stimuli conveying happiness were used. Other research suggests that negative affect is associated with increases in the magnitude of AB (Vermeulen, 2010). Hence, from the perspective of positive and negative affect, we may hypothesise that hate as a negative feeling is likely to undermine attention, whereas love as a positive feeling may (even) enhance attention.

Aim and hypotheses

The aims of the present study are twofold. First, we want to test whether and how the names associated with a person people love and hate affect people’s attention to the subsequent stimuli in an attentional blink task. To this end, we adopt an AB task in which participants need to categorise a name (of whom participants felt either hate, love, or neutral towards; T1) and identify a number (T2) in an RSVP. The name was preceded by the number for 120, 240, or 840 ms (i.e. lag 1, 2, or 7, respectively). Based on the previous results of high arousal discrete emotions associated with love and hate, we predict that the name of a person people love and hate could induce a larger attentional blink (i.e. lower accuracy) in close temporal proximity, compared to the name of a person to whom people feel neutral (Hypothesis 1).

Second, we aim to investigate whether seeing the names associated with feelings of love and hate results in a sustained transfer effect onto neutral, non-feeling-inducing trials. Feelings of love and hate are long-lasting, so they might have a sustained effect beyond the feeling-inducing trial (i.e. trials with loved or hated names). This sustained effect may then transfer to the trials containing neutral names. Specifically, neutral names in love vs. neutral blocks (which only contain trials with loved and neutral names) might be influenced by a positive state induced by the loved names in other trials. Likewise, neutral names in the hate vs. neutral blocks (which only contain trials with hated and neutral names) might be influenced by the negative state induced by the hated name trails. Based on the idea that positive mood can broaden attention to new stimuli and negative mood can undermine the span of attention, we propose that relative to the neutral names in love vs. neutral blocks, a larger attentional blink (lower accuracy, especially in close temporal proximity), as well as worse overall performance, would be elicited by the neutral names in the hate vs. neutral blocks (Hypothesis 2).¹

Methods

Participants

We used the online platform Prolific to collect data. The online platform allowed us to recruit a relatively multifaceted sample of adults so that there is a relatively rich variety of love and hate experiences compared to the student sample. To minimise variance unrelated to the factors of interest, we included participants who had at least completed high school and were no older than 35 years of age. Participants with migraine, epilepsy, or attention deficit hyperactivity disorder were excluded because the task involved rapidly flashing visual stimuli. We excluded participants who spent less than 30 minutes or more than 90 minutes on the task (i.e. 22 participants). Participants who failed the embedded attention check question or made invalid responses (i.e. non-sense, not related to the topic, copy-and-paste from elsewhere) in the open-ended questions were also excluded (i.e. 2 participants). We terminated data collection when we reached the preregistered number of participants. This sample size was calculated based on the pilot study (which can be found in the Supplementary Material Pilot Study). Eventually, three hundred participants living in the United Kingdom (91 men, 204 women, and 5 chose “Other”. M_age = 22.69, SD_age = 4.48) took part in the study in exchange for 5 British pounds. The study was approved by the ethics committee of the Faculty of Behavioural and Movement Sciences of the Vrije Universiteit Amsterdam (VCWE-2020-108).

Procedure

After signing the informed consent form, participants first filled out the two scales assessing social value orientation (SVO), or preferences for particular patterns of outcomes for self and others (e.g. Van Lange, 1999) and dark personality, including narcissism, psychopathy, and Machiavellianism (Jones & Paulhus, 2014).² Then, they were asked to recall a person they hated or strongly disliked, a person they loved or strongly liked, and a person they felt neutral towards, in random order. We did not restrict participants to recognising a person they loved or hated but used an open standard – a person they strongly liked or disliked was also accepted. We did so by considering that hate is not a feeling that everyone has ever experienced (or that people do not use that label in reference to that feeling). Although there are differences between love and like (Berscheid, 2010), hate and dislike (Martinez et al., 2022), strong like or dislike is most likely to share some key features with love and hate, respectively. In the current sample, 252 participants (85%) reported they wrote someone they loved, and 151 participants (51%) reported they wrote someone they hated. Additionally, 132 participants (44%) recalled a person they love and a person they hate. We conducted further analyses using participants who wrote a person they loved and hated. The results were similar to the ones in the main paper and led to the same final conclusion (for detailed results and discussion, see Supplementary Material Section 6).

As love and hate are inherently social feelings, we included this recall section to ensure participants get a vivid view of the specific person and to enhance the association between the specific names and the feelings they should trigger (e.g. they might know other people with that name). During this recall section, the persons participants selected should be someone they had personal interactions with. Participants were asked in three steps to describe: (1) who that person is and what characteristics of that person caused these feelings; (2) experiences or events that led you to develop the feelings (describe an event where both participants and the person they feel neutral towards were present); (3) how you would like to behave towards that person.³ These instructions should prevent participants from bringing to mind people they do not know personally but still love or hate (e.g. politicians or celebrities). Then, participants performed the AB task, which was programmed using Opensesame (Mathôt et al., 2012) and executed online using a JATOS server. After completing the AB task, participants rated the intensity and duration of their feelings (for the person they hated, we measured: hate, dislike, anger, disgust, and contempt; for the person they loved, we measured: love and like). Finally, participants were thanked and debriefed after they provided demographic information, including age, sex, ethnicity, and income level.

Materials

Attentional blink (AB) task

Prior to the experiment, participants provided the names of a person they hated or disliked (refer to as “hated name” in the following sections for the sake of convenience), a person they loved or liked (referred to as “loved name”), and a person they felt neutral towards (referred to as “neutral name”). The names in all conditions were five letters long on average and did not differ among the three conditions (all ts < 0.74, all ps > .05, M_hate = 5.05, SD_hate = 1.24; M_love = 5.03, SD_love = 1.49; M_neutral = 4.98, SD_neutral = 1.38).

The task consisted of 240 trials which were evenly divided into eight blocks. Each trial of the task included 18 words presented for 100 ms each in a rapid serial visual presentation (RSVP) stream with an inter-stimuli interval of 20 ms (see Figure 1). In the words series, one was treated as the first target (T1, i.e. the names provided by participants), one was the second target (T2, i.e. single-word numbers, for instance: one, eleven, range from zero to one hundred), and 16 were irrelevant distractors (i.e. things or events occurring in natural environments, e.g. island, snowstorm, as in Barnard et al., 2004). Among the words, only the names were in uppercase to enhance their noticeability. The first target was randomly presented in positions 6–9 of the RSVP stream, and the second target was presented at lag one, two, or seven (i.e. 120, 240, or 840 ms after the first target). At the end of each trial, participants were asked to categorise which name they saw in the previous word stream by pressing “Z” for the name of the person they loved or hated (depending on the block) and “X” for the name of the person to whom they feel neutral (T1). They then indicated the number word they saw by entering it using Arabic numerals (T2). Participants were instructed to guess if they were not consciously aware of having seen a target. The accuracy of identifying the second target was the primary dependent variable in the current study. We only focused on the accuracy of the T2 under the condition that T1 is correctly categorised. All words were shown centrally on the screen in white on a black background.

Figure 1. Illustration of a typical trial in the experiment. There were 18 words in an RSVP which presented for 100 ms. A blank interval of 20 ms followed each word. The first target was the name participants provided and was displayed in uppercase. The second target was a single-word number. Participants needed to categorise the name, as well as recall the number, at the end of the word stream. The sample in this illustration shows that the second target displays at lag 2.

The study contained a block design such that of the eight blocks, four of them were love vs. neutral blocks, which contained either the loved name or neutral name in the words series, while the other four blocks were hate vs. neutral blocks, in which either the hated name or the neutral name was displayed on each trial. Participants randomly started with the four love vs. neutral blocks or the four hate vs. neutral blocks. In each block, there were five feeling-related (love or hate) trials and five neutral trials under each level of Lag, resulting in 30 trials per block in total. Before the formal experiment, there were fifteen practice trials in which the RSVP started slow and gradually built up speed.

Data analysis

We have three variables in the current study. The first one was Emotionality (love and hate vs. neutral), which referred to whether the name at T1 was a feeling-related name (loved or hated) or a neutral name. The second variable was the Block Type (love vs. hate), which indicated whether the specific trial was in the love vs. neutral blocks or the hate vs. neutral blocks. The last one was the Lag (1, 2, vs. 7) which represented the position where T2 was displayed.

The data cleaning method included the following steps: (1) we excluded the data of one block if the accuracy of the categorisation of the names in that block was lower than 60%; (2) we calculated the overall accuracy of the discrimination of the target numbers (T2) under the circumstance that participants correctly categorise the name (T1). Outliers were detected using Hoaglin and Iglewicz (1987) outlier labeling rule, in which we calculated the lower fourth (F_L) and the upper fourth (F_U) quartiles of the individual mean T2 accuracy scores.⁴ The participants with a mean accuracy out of the range of F_L– 2.4 × (F_L− F_U) and F_U+ 2.4 × (F_L− F_U) were excluded from further analysis (for the current study, the data of two participants was excluded).

We used binomial generalised linear mixed models (GLMMs) by the glmer function of the “lme4” package of R (Version 3.5.1; R Core Team, 2018) as many previous studies did (e.g. Eiserbeck & Abdel Rahman, 2020; Maier & Abdel Rahman, 2018; Perone et al., 2021) to investigate whether and how loved and hated names affect the identification of T2. In general, the binomial family was used, and the categorical variables were coded based on the effect coding method. Subject was added as the random intercept. We began by including all possible random slopes in the model and simplified the model by deleting the random slope from the highest interaction and the lowest estimation until we found the simplest model that was not statistically different from the most complex model.⁵ If the model still failed to be convergent, a bobyqa optimiser with a maxfun of 2e5 would be used to facilitate model convergence. We used χ² values of the type III ANOVA by the Anova function of the “car” package to run on the model. To show the simple effects of interactions, we reported the coefficient β (the odds ratio), standard error (SE), z-value, and p-value, the (estimated marginal) mean⁶ and standard error of the accuracy were reported if necessary. The confidence intervals were calculated by the confint function of the “stats” package using the Wald method.

The analysis plan was preregistered. However, as the results were unexpected, we opted for a different way of analysing and presenting the results to improve clarity and comprehensibility. The results that followed the preregistered analysis plan can be found in Supplementary Material Section 4. The analysis presented here started by examining the pattern of a typical attentional blink task. We built a model with Emotionality (love and hate vs. neutral), Block Type (love vs. hate), Lag (1, 2, vs. 7), as well as all interactions as the fixed factors and with subject as the random intercept. To test whether loved and hated names elicit a larger attentional blink (H1), Lag 1 was excluded as it was less essential to testing the hypothesis.⁷ We then built two separate models, one for each block type. The models included Emotionality (love and hate vs. neutral), Lag (2 vs. 7), and their interaction as fixed factors. To test whether love and hate feelings had a sustaining effect on neutral trials (H2), we assessed the accuracy of only neutral trials and built a model with the Block Type as the fixed factor.

Besides the analyses mentioned above, we also tested the models with the random effect of the target word (i.e. which number was used, since “one” is much easier to recognise than “twenty”), models with the control of the sequences of recall (i.e. the order of ruminations of the persons participants loved, hated, and felt neutral towards before the AB task) and blocks (i.e. the order of love vs. neutral blocks and hate vs. neutral blocks in the AB task), as well as the model controlled for the length of names participants entered. The results of these models and the correlations among all continuous variables are reported in the Supplementary Material Section 2 and Section 5.

Results

First target performance: name categorisation

Overall performance on categorising the names (T1) was close to the ceiling at >97%. This might be because the uppercase made T1 stand out and thus easier to detect. A full analysis of T1 performance as a function of condition and lag can be found in the Supplementary Material Section 3. All second target (T2) results are contingent upon T1 being correctly categorised.

Attentional blink

Before testing the main hypotheses, we first examined whether the names at T1 induced an attentional blink for T2. To this end, we conducted a binomial generalised linear mixed models with Block Type (love-neutral block vs. hate-neutral block), Emotionality (emotional [love or hate] vs. neutral), and Lag(1, 2, vs. 7), as well as all interaction effects as fixed factors. In this section, we focus on the effects of Lag across all combinations of Block Type and Emotionality (see Table 1). Indeed, and as is typical for AB tasks, participants showed the lowest accuracy at lag 2, followed by lag 1, while the accuracy was highest at lag 7. This means that the task produced the well-replicated attentional blink pattern.

Table 1. Results for the simple effect analysis of lag as the function of block type and emotionality.

Block Type	Emotionality ofthe names	Contrastof Lag	Statistic coefficients					Mean (SE)
			β	SE	z	p	95% CI	Lag 1	Lag 2	Lag 7
Love-Neutral	Loved names	1 vs. 2	1.56	0.09	7.89	<.001	[1.39, 1.74]	0.834 (0.009)	0.763 (0.012)	0.888 (0.007)
		1 vs. 7	0.63	0.04	−6.90	<.001	[0.56, 0.72]
		2 vs. 7	0.41	0.02	−14.84	<.001	[0.36, 0.46]
	Neutral names	1 vs. 2	1.59	0.09	8.49	<.001	[1.43, 1.78]	0.816 (0.010)	0.736 (0.012)	0.877 (0.007)
		1 vs. 7	0.62	0.04	−7.35	<.001	[0.55, 0.71]
		2 vs. 7	0.39	0.02	−15.91	<.001	[0.35, 0.44]
Hate-Neutral	Hated names	1 vs. 2	1.58	0.09	8.23	<.001	[1.41, 1.76]	0.818 (0.010)	0.740 (0.012)	0.884 (0.007)
		1 vs. 7	0.59	0.04	−8.15	<.001	[0.52, 0.67]
		2 vs. 7	0.37	0.02	−16.48	<.001	[0.33, 0.42]
	Neutral names	1 vs. 2	1.86	0.10	11.23	<.001	[1.67, 2.07]	0.828 (0.010)	0.721 (0.012)	0.866 (0.008)
		1 vs. 7	0.74	0.05	−4.66	<.001	[0.65, 0.84]
		2 vs. 7	0.40	0.02	−15.77	<.001	[0.36, 0.45]

Test of Hypothesis 1

To test whether there was a difference between loved/hated names and neutral names on the magnitude of attentional blink, we included two levels of Lag (i.e. lag 2 and lag 7) and conducted two separate generalised linear mixed models, one for each block type. Thus, the model contains the fixed factors of Emotionality and Lag, as well as their interaction.

The results for the love-neutral block showed significant main effects of Emotionality (χ² (1) = 12.19, p < .001) and Lag (χ² (1) = 337.68, p < .001). As the results of Lag have already been reported in the previous section, here we only present the results of the main effect of Emotionality. Specifically, participants had a higher accuracy on detecting T2 when T1 was the love-related names (M = 0.834, SE = 0.008) than when the name was neutral-related (M = 0.817, SE = 0.009, β = .88, SE = .03, z = −3.49, p < .001, 95% CI [.83, .95], see Figure 2). The interaction of Emotionality and Lag was not significant (χ²(1) = 0.38, p = .535).

Figure 2. Estimated marginal mean accuracy of the second target under the condition that the first target was correctly categorised. The error bars represented the standard errors across each condition (Block Type × Emotionality × Lag). *p < 0.05, **p < 0.01, ***p < 0.001.

Similarly, the analysis for the hate-neutral block showed a similar pattern in that the main effects of Emotionality (χ² (1) = 14.21, p < .001) and Lag (χ² (1) = 349.07, p < .001) were significant, while the interaction was not significant (χ²(1) = 0.99, p = .319). As can be seen in Figure 2, hated names (M = 0.825, SE = 0.008) produced a higher accuracy than neutral names (M = 0.805, SE = 0.009, β = .88, SE = .03, z = −3.77, p < .001, 95% CI [.82, .94]).

We conclude that relative to neutral names, the name of a person participants said to love or hate elicited greater T2 accuracy. As this was not specific to lag 2 but appeared to extend to lag 7, we conclude that the loved and hated names did not appear to modulate the attentional blink as such, but rather overall attention across lag 2 and lag 7.

Test of Hypothesis 2

To test the second hypothesis, we examined the overall accuracy of neutral trials as a function of block type. In this analysis, we only include neutral trials in love-neutral and hate-neutral blocks. A generalised linear mixed model was conducted with the Block Type as the only fixed factor. Results showed that the main effect of the Block Type was not significant (χ²(1) = 0.94, p = .333), which means that participants behaved similarly on neutral trials in love-neutral (M = 0.807, SE = 0.009) and hate-neutral blocks (M = 0.802, SE = 0.008). Therefore, there is no evidence to suggest that love-associated and hate-associated names induced a different sustained influence on neutral trials.

To conclude, we did not find support for a different sustained effect of the feelings of love and hate that then transferred to the trials containing neutral names.

Discussion

In the current study, we extend previous studies by including key aspects of social life – a person one loves or hates – into a widely used task that assesses attention and information processing. Specifically, we tested whether love and hate affected the accuracy of detecting stimuli following them and whether they had a sustained effect on neutral trials. The study is innovative as it is the first study, to our knowledge, to examine the effect of reminders of names associated with love and hate on information processing in an AB task.

Regarding the first aim of the study, we hypothesised that reminders of a loved or hated (versus neutral) person would be detrimental to T2 accuracy in the AB task (H1). However, the results showed a different pattern: After seeing a loved or hated name, participants demonstrated a better overall performance in detecting the second target instead of a reduced attentional blink, and this beneficial effect of loved and hated names was sustained to lag 7 at where participants’ attention often returned to the default level of accuracy in a typical AB task. This is a surprising finding, especially in light of past research, in which both task-relevant and task-irrelevant emotion-related arousing words or pictures have resulted in worse T2 performance, specifically at short lags (e.g. Ciesielski et al., 2010; Flaisch et al., 2016; Keefe et al., 2019; Mathewson et al., 2008; Schwabe & Wolf, 2010). Hence, these findings are largely inconsistent with the idea that reminders of names of people we love or hate consume more working memory resources or otherwise interfere with the primary task.

Clearly, the result was not in line with our hypothesis. We offer several explanations, which while post-hoc, are informed by previous research. One key difference between the current study and previous studies is that the stimuli in previous studies were associated with emotions that are momentary and non-personal, while the feelings of love and hate are inherently interpersonal and relevant to past experience and memory (as well as guiding behaviour in the future). In this case, the beneficial effects of love and hate may be twofold. First, the feeling-related names may have evoked certain autobiographical memories about the “self” (Ortigue et al., 2007; Tacikowski et al., 2011), involving recall of specific social events and general self-relevant evaluations of them (e.g. the meaning of these events for the self, or the future relationship with the other). There is evidence that the magnitude of AB following one’s own name is smaller than that following a close other’s name or a stranger’s name (Nijhof et al., 2020, Study 1; see Shapiro et al., 1997 Study 3 for a non-significant result, albeit with n = 8). In line with the effect of participants’ own names, the names of a person people love or hate might require fewer resources to process because they are close to the representation of self. Thus, the better performance after a loved and hated name might be (partially) explained by the idea that loved and hated names occupy less attentional capacity, thereby alleviating competition with other incoming information. However, because we did not measure the familiarity or closeness of the person participants described, it awaits future studies to test whether familiarity or closeness might help explain the superior performance caused by names of people we love and hate.

A complementary explanation related to autobiographical memory is that the names of loved and hated persons, compared to neutral names, may form stronger semantic relationships with other concepts, whether value-laden or not, simply because the target person “means more” to the individuals. According to the density hypothesis (Unkelbach et al., 2010), the more strongly connected a concept, the faster it will be processed. It is therefore possible that loved and hated names are accessed more rapidly, resulting in less interference for subsequent targets.

Second, the loved and hated names may have highlighted the social functions of love and hate, which is reflected by the relatively enduring process that affects attention even at lag 7. Perhaps, reminding of loved and hated persons calls for a task-related preparedness and alertness that extends to the task itself (i.e. phasic alertness, Sturm & Willmes, 2001) and brings a goal-directed state that facilitates the attentional process for T2. Support for this idea comes from a previous study that used an AB task but with a constant T1-T2 delay (Kever et al., 2015). The study found that in comparison to a low arousal T1 target, a high arousal T1 target led to enhanced detection of neutral T2 targets. Kever and colleagues did not provide a further explanation for this finding because it was not the main interest of their research, but we suggest that the high arousal T1 might have caused an enhanced preparedness to detect T2 (cf. Derryberry, 1988) and that the same may have happened in our study.

This preparedness explanation is also consistent with research using a lexical decision-making paradigm rather than an AB paradigm. Studies have shown that subliminal priming of a beloved’s name (compared to a friend’s name) leads to faster responses in detecting whether the followed-up stimulus is a word or non-word (Bianchi-Demicheli et al., 2006; Ortigue et al., 2007). Overall, the facilitation of love and hate in the current study and the lexical decision-making task used in the previous work might be because of the intensity of feeling (Finucane et al., 2010), perhaps in a combination of implications for thoughts or actions beyond the task – after all, loved ones (or hated ones) are “significant” and meaningful for social life.

These interpretations are also in line with evidence that angry faces facilitated the processing of T2 when they were task-irrelevant (de Jong et al., 2014) but impaired the detection of T2 when they were task-relevant (de Jong et al., 2010). In this case, when T1 is a social-related stimulus and does not occupy attentional resources, it can facilitate the first unlimited sensory processing stage for the upcoming information (cf. Oatley & Johnson-Laird, 1987). In summary, we regard two complementary explanations for the beneficial effect of love and hate on attention as plausible. Love and hate may enhance a goal-directed mindset, and love and hate are exceptionally strong social feelings. Both features can enhance attention, and future research is needed to examine the relative importance of these explanations.

We should note that although both positive and negative feelings about other people can facilitate the processing of T2, the mechanisms behind them may be different. For example, positive social feelings like love may broaden the scope of attention (Fredrickson, 2001; Fredrickson & Branigan, 2005) so that thinking of someone you love can facilitate attention to various new pieces of information. In contrast, negative social feelings such as hate are more likely to elicit an alert state (Petersen & Posner, 2012; Posner & Petersen, 1990), including tendencies to focal processing of information (Sakaki et al., 2011). In this case, feelings of hate might improve the performance at T2, especially because the environments of T1 and T2 were very similar (e.g. same location, same task). Future studies can use a task in which the loved and hated names (or other types of cues) can appear at different spatial locations than the following target (e.g. a visual search task or a dot-probe task) to dissociate these two different mechanisms.

The second aim of the current study is to test whether love or hate as enduring feelings had a sustained influence on neutral trials. We supposed that participants had a better performance on neutral trials in love vs. neutral blocks than in hate vs. neutral blocks (H2). In contrast, the results of the current study did not reveal a different effect of love and hate feelings on neutral trials. It could be because neither loved nor hated names affected the neutral trial because the high-demanding attentional blink task can attenuate the perception of an affective state, especially a negative mood (Askim & Knardahl, 2021; Van Dillen & Koole, 2007). Alternatively, it is also possible that loved and hated names had a similar influence on the neutral trials, especially when considering our finding that loved and hated names had a similar influence on non-neutral trials. As a limitation of our study, we did not measure participants’ mood, valence, and arousal before or after the task, and therefore cannot provide insight into their mediating roles or their relative importance. Thus, we recommend that future studies include a baseline condition containing neutral names only (versus no names or stranger’s names) and measure the participants’ arousal before and after the experiment, to test whether feelings of love and hate overall facilitate or impede the processing of targets on neutral trials. More generally, we recommend future research that seeks to understand how and why reminders of love and hate may facilitate versus undermine attention and information processing.

Other limitations include the fact that the effect sizes were overall small. Unlike other attentional blink studies using pictures or emotional words, T1 was always the same within a condition (i.e. the names provided by participants). This might weaken the effect of names because of the repetition and lack of novelty. This will allow future studies to test whether presenting different aspects of a loved or hated person will have a larger effect. Besides, the current study was conducted online and therefore ill-controlled. We tried to make up for this shortcoming by using a large sample size and a generalised linear regression model to control the device and screen size variations. Although the variability caused by different devices is relatively small (Anwyl-Irvine et al., 2021), it can still make the effect size smaller than it actually is.

Another limitation is that as this study represented a first exploration, we did not examine features that point at the type of love or hate – the target person can be anyone, and the reason for love versus hate can be quite multifaceted. However, the experiences of both love and hate can vary as a function of the type of relationship (e.g. friendship versus close relationship), and even within a relationship (e.g. romantic love at early-stage versus in long-term relationships in terms of intensity or arousal; Aumer-Ryan & Hatfield, 2007; Berscheid, 2010; Sternberg, 1986, 2003). However, we do not anticipate that various types of love and hate have different effects on attention in an AB task as we anticipate that the social nature of love and hate, perhaps along with the perspective of long-term orientation, are critical features of both love and hate (see Acevedo et al., 2012). We tested this idea by conducting additional analyses to explore different types of love (e.g. compassionate vs. filial) and found no significant difference (see Supplementary Material Section 8).

Conclusion

Other people may be an important source of happiness and joy, as well as anger and disappointment. It is, therefore, no surprise that these emotions have received enormous attention in the social and behavioural sciences. One broad conclusion from this research is that feelings are easily evoked and often serve as a specific guide to human behaviour. Past research has shown that these emotions often interfere with cognitive performance. Contrary to our hypotheses, the present findings point to the performance-enhancing effects of love and hate. Specifically, exposure to the names of people we love or hate enhanced performance in detecting subsequent information. We suggest that the self-relevant, intense, and enduring feelings of love and hate might yield enhanced alertness and goal-directed orientations that support rather than impede the processing of the subsequent information.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The design of the study, as well as the hypotheses and analysis plan, were pre-registered at the OSF (https://osf.io/twjb9/). All data, analysis code and research materials are available at the OSF (https://osf.io/em84y/).

Additional information

Funding

This work was supported by China Scholarship Council [grant number: 201806360274].

Notes

1 We modified the second hypothesis from the pre-registered one. Detailed reasons can be found in the Supplementary Material Section 1.

2 We include these measures regarding individual differences, as well as the intensity and duration of feelings and emotions that are stated later in this section for the purpose of exploratory analysis. A detailed explanation of the measurements, the analyses we did, and the exploratory results are presented in Supplementary Material Section 2 and Section 9.

3 The three questions were a bit different when asking participants to describe a person they feel neutral towards. But in general, the three questions cover the characteristics of the person, an event that the person and participants involved, and the action tendency participants had towards the described person.

4 The outliers were detected at the level of participants instead of conditions which is different from the preregistration. We did so to avoid the potential high false-positive rate caused by removing outliers by conditions (see André [2022] for detailed arguments).

5 We ultimately removed the interaction term from the random slopes.

6 The mean is conditional probability in that it reflects the accuracy of identifying T2 under the condition that the name (T1) was correctly categorized. We reported estimated marginal means instead of the observed means because the estimated marginal means are adjusted for other variables in the model.

7 We excluded Lag 1 trials because the performance at Lag 1 shows different functions compared to attentional blink. Lag 1 is different because there are no distractors between T1 and T2. Several studies have proved that there is (merely) no attentional blink at lag 1, and this phenomenon is called the lag 1 sparing effect. One possible explanation for this is that the two targets are combined into a single attentional episode in working memory due to their close temporal proximity (Martens & Wyble, 2010).