Introduction

COVID-19 is a significant health threat with rapid transmission and widespread infection, causing more than 626 million infections and more than 6.88 million deaths as of March 21, 2023 (WHO 2023). Vaccines are widely recognized by health authorities as an effective method for preventing infectious diseases and achieving public health success (Doherty et al. 2016; Osterholm et al. 2012). Despite the severity of influenza, vaccine hesitancy—a behavior involving delayed or refused vaccination despite available services—remains a global challenge in reducing the burden of pandemic influenza. Vaccine hesitancy is complex and varies by time, place, and vaccine type (Osterholm et al. 2012). The success of any vaccine depends on the percentage of the population that is vaccinated. Yet, global public acceptance of vaccines is relatively low. For example, over 30% of respondents in the United Kingdom and Ireland reported a low intention to receive the COVID-19 vaccine (Murphy et al. 2012), and about 20% of Americans expressed similar sentiments (Thunstrom et al. 2020). Only 30.5% of French people agreed to receive the COVID-19 vaccine during the first half of 2021 (Guillon and Kergall 2021).

Therefore, more effective strategies are needed to encourage people to get vaccinated against COVID-19. Numerous studies have explored factors like gender, health status, and knowledge level that influence individuals’ intentions to receive the COVID-19 vaccine through self-report questionnaires (Ruiz and Bell 2021; Khubchandani et al. 2021; Zheng et al. 2021). Although these studies provide valuable insights, they offer limited understanding of the underlying implicit attitudes towards the COVID-19 vaccine (Sallam 2021; Akarsu et al. 2020; Chen et al. 2021). Implicit attitudes, which are crucial for understanding behavioral decisions, have been extensively studied in various contexts as explanations for biases or stereotypes (Greenwald et al. 1998; Chen et al. 2018). Implicit attitudes refer to unconscious thought processes that influence individuals’ responses to the COVID-19 vaccine, providing deeper insight into people’s unexpressed opinions about the vaccine.

In addition, individual differences have been identified as an important factor influencing intentions (Bandura 2001; Ryu et al. 2023). Researchers have conceptualized individual differences based on two dimensions: need for affect (NFA; Maio and Esses, 2001) and need for cognition (NFC; Cacioppo and Petty 1982). While extensive research has explored how individuals perceive and respond to COVID-19 vaccines and persuasive vaccination measures (Lin and Wang 2020; Wismans et al. 2021; Karlsson et al. 2021), few studies have assessed the influence of personal traits (NFA vs. NFC) on implicit attitudes towards the COVID-19 vaccine, especially through neural cross-validation.

The present research is divided into two phases, each utilizing the same cohort of participants. Before delving into these phases, it is crucial to outline the central research questions guiding this study. These questions are:

RQ1: How do personality traits (NFA vs. NFC) influence the consistency between implicit and explicit attitudes towards the COVID-19 vaccine?

RQ2: How do personalized persuasion strategies based on personality traits affect implicit attitudes towards the COVID-19 vaccine?

Phase 1 seeks to explore the differences in implicit attitudes towards the COVID-19 vaccine among individuals with different personalities. Phase 2 investigates the effects of personalized persuasion strategies on participants with different personalities.

Explicit attitude and implicit attitude

According to implicit social cognitive theory, individuals may have inconsistent implicit and explicit attitudes toward the same object. Implicit attitude is a key concept in cognitive psychology, characterized as attitudes or judgments influenced by automatically activated evaluations that occur without conscious effort or cognitive attention (Greenwald et al. 1998). In contrast, explicit attitudes are conceptualized as self-reported evaluations that can be consciously controlled by the individual (Rydell and McConnell 2006). Recently, there has been a significant increase in empirical research on implicit social cognition (Manca et al. 2020; Chen et al. 2018; Hofmann et al. 2010).

To date, substantial evidence has shown that implicit attitudes function as a spontaneous process that can directly guide behavior (Bargh et al. 1996; Bargh et al. 2001; Chen and Bargh, 1999; Dovidio et al. 1997; Fazio and Dunton, 1997). They have become a more reliable predictor of behavior, especially for individuals inclined towards effortful cognitive activities (Conner et al. 2007). Furthermore, Fazio and Olson (2003) demonstrated that the correlation between implicit and explicit evaluations decreases as cognitive elaboration increases. These findings emphasize that implicit and explicit attitudes operate as two distinct systems, and implicit attitudes can change independently of conscious awareness, supporting the notion of a weak correlation between these systems (Dasgupta and Greenwald, 2001; Olson and Fazio, 2004). This also suggests that even if people are implicitly opposed to something, they may still express support for it in their explicit attitudes.

Given the widespread interest in implicit cognition processes, it’s not surprising that there are various methods to measure these constructs, such as the Implicit Association Test (IAT) (Greenwald and Banaji, 1995), the Affective Priming Task (Fazio and Williams 1986), and the Semantic Priming Task (Wittenbrink et al. 1997). Among these methods, the IAT is considered one of the most prominent measures for accessing implicit attitudes, with a growing number of contributions in implicit cognition research relying on it (Chen et al. 2018; Lane et al. 2007; Greenwald et al. 1998).

Therefore, in this study, we examine individuals’ implicit attitudes toward the COVID-19 vaccine using the Implicit Association Test (IAT). The IAT uses target stimuli representing the COVID-19 vaccine associated with both positive and negative attributes. When participants associate the target stimuli (i.e., COVID-19 vaccine) with positive or negative attributes in different tasks, they are expected to respond more quickly and accurately to congruent combinations (i.e., COVID-19 vaccine with positive attributes) than to incongruent combinations (i.e., COVID-19 vaccine with negative attributes). In this way, the IAT serves as an indicator of implicit bias. Specifically, we hypothesize that individuals will attempt to inhibit their implicit attitudes between the COVID-19 vaccine and negative attributes, which will result in slower reaction times on congruent trials.

Personality traits based on cognitive and affective processing

Understanding the connection between vaccine hesitancy and personality traits is crucial for developing effective immunization strategies (Reagu et al. 2023). Extensive literature has demonstrated that individuals vary in their desirability of emotions and their tendency to engage in thinking (Cacioppo and Petty 1982; Huskinson and Haddock 2004). Accordingly, researchers have denoted each characteristic as the need for affect (NFA; Maio and Esses 2001) and the need for cognition (NFC; Cacioppo and Petty, 1982). More specifically, the NFA has been defined as the general motivation of people to approach or avoid situations and activities that are emotion-inducing for themselves and others (Maio and Esses, 2001). The NFC has been defined as the motivation for an individual to engage in and enjoy the effortful cognitive activity (Cacioppo and Petty 1982). NFA and NFC are two important psychological constructs that have been studied extensively to understand its influences on people’s decision making (e.g., Haddock et al. 2008). However, to our knowledge, there hasn’t been studies compared the effectiveness of individuals information processing tendencies (e.g., NFA vs NFC) in the context of vaccine acceptance. Understanding the message preferences of individuals with different information processing tendencies helps researchers and healthcare professionals to optimize strategies for different groups and maximize persuasion outcomes.

Guided by different motivations, individuals with high NFA and high NFC may take different approaches during information processing. As proposed in the Elaboration likelihood model (ELM), individuals with higher NFC are more motivated to elaborate arguments and process information using a central route (Cacioppo and Petty 1982; Olson et al. 1984). Therefore, they are more likely to be influenced by the quality of the message and strength of the argument (Haddock et al. 2008) and produce more individual thoughts (Petty et al. 2002). Once they form the initial attitude, their existing attitudes are also more resistant to persuasion (Haugtvedt and Petty 1992). On the other hand, prior research indicated that individuals with higher NFA are more easily influenced by peripheral and emotional cues, such as liking, perceived expertise, and credibility (Maio and Esses 2001). Therefore, their attitude formation may be more contextualized and easier to change, depending on the peripheral cues they receive every time (Haugtvedt and Petty 1992).

Most existing research accesses individuals’ existing attitudes using self-report questionnaires (Wu et al. 2022; Yang et al. 2022), which may only reflect individuals’ explicit attitudes with less information on their implicit attitudes. However, initial evidence from the neurocognitive paradigm suggests that individuals with high NFC engage in bottom-up involuntary cognitive processing of contextually relevant information and controlled top-down attention allocation towards target stimuli (Enge et al. 2008; Strobel et al. 2015). This result indicates that individuals with high NFC not only involuntarily reflect on external message arguments and pre-process messages, but also voluntarily regulate their internal thoughts and form their implicit attitudes accordingly. Therefore, we expect to see more consistency between explicit and implicit attitudes among individuals with high NFC. Conversely, as the attitude formation of those with high NFA is more contextualized, we anticipate that their explicit attitudes will change based on peripheral cues they receive at a given time, resulting in larger discrepancies between their internal and external attitudes.

Implicit conflict and response monitoring using ERPs

The accuracy of measuring implicit attitudes as automatic cognition has been a controversial topic (Kurdi et al. 2021). Although many studies have shown that cognitive processes occur unconsciously and have a significant impact on people’s behavior (Dijksterhuis 2004; Zajonc 1980; Zaltman 2015), it is challenging to assess this unconscious information at the behavioral level. Therefore, investigating implicit attitudes from a neuroscience perspective is of great importance.

Event-related potentials (ERPs) refer to electrocortical activity that is time-locked to each stimulus and averaged across trials in the same experimental condition, which is measured using electroencephalography (EEG). ERPs are used to access ongoing electrophysiological changes resulting from the synchronous activation of several neural subpopulations in response to sensory, motor, or cognitive events (Bouaffre and Faïta-Ainseba 2007). They are considered to be highly reliable indicators that accurately represent how people process information at different points in time (Gaspar et al. 2011; Bouaffre and Faïta-Ainseba 2007). Unlike self-reported measures, ERPs are regarded as a “window” for examining psychological activity and have the potential to accurately reflect individuals’ cognitive processes (Shang et al. 2018). Existing research has shown that implicit attitudes can be monitored by evaluating changes in ERPs (van Nunspeet et al. 2014).

Based on the evidence presented above, in this study, we use Event-Related Potentials (ERPs) to assess individuals’ implicit attitude processes and conflict monitoring. Specifically, we will focus on N1, P2, and N400, which are related to attention and cognitive conflict. These components indicate the extent to which individuals’ attention and cognitive conflict are directed toward the target stimulus (Ito and Urland 2003; Wu and Zhang 2019).

The N1 component, the first negative-going response occurring roughly 80–150 milliseconds post-stimulus onset, is a critical indicator of the discrimination process. It is primarily associated with the processing of sensory information and the early stages of attention (Vogel and Luck 2000). The N1 is sensitive to the physical and perceptual features of stimuli and is modulated by attention. It often shows larger amplitudes when participants are focused on specific aspects of a stimulus or when a stimulus is novel or salient (Mangun and Hillyard 1991; Eichenlaub et al. 2012). Studies have also observed congruency effects on the N1 component across various modalities; congruent conditions typically result in reduced N1 amplitudes compared to incongruent conditions. This reflects a lower cognitive workload or decreased attentional demand when processing stimuli that align with expectations. For instance, in visual tasks where subjects anticipate and receive a matching stimulus, the facilitated sensory processing leads to a less pronounced N1, signaling efficient perceptual encoding and attention modulation (Hillyard et al. 1973; Luck and Hillyard 1994).

P2 is a positive peak occurring around 150–220 ms, associated with attention allocation (Yang et al. 2012; Key et al. 2005) and indicating the level of information complexity (Pernet et al. 2003). Its amplitude increases with the presence of extrinsic stimuli. Research also indicates that the P2 amplitude can be modulated by stimuli in congruent conditions, reflecting enhanced cognitive processing in response to expected or normative sensory inputs. For instance, studies have shown that congruent auditory and visual stimuli can lead to increased P2 amplitudes, suggesting that the brain allocates more attentional resources to stimuli that align with contextual expectations (Barry et al. 2007).

N400—a negative-going deflection that peaks around 400 ms (Kutas and Hillyard 1980)—is primarily considered an indicator of semantic processing difficulty, although its functional significance remains a subject of debate (Luck 2005). Studies have shown that the N400 amplitude is greater for stimuli presented as incongruent compared to congruent under conditions like the Implicit Association Test. This suggests that the greater the predictability of the stimulus, the smaller the N400 amplitude elicited (Ito and Urland 2003; Kubota and Ito 2007; Yang et al. 2023; Steinhauer et al. 2017). This finding aligns with observations that incongruent conditions generally trigger stronger cognitive conflict than congruent conditions, reflecting the N400’s role in both conflict detection and integration during semantic information processing, where its amplitude positively correlates with task difficulty (Hilgard et al. 2014).

Phase 1: implicit attitudes based on personality traits

Phase 1 of the study aims to investigate the consistency of explicit and implicit attitudes towards the COVID-19 vaccine among people with different personality traits (NFA vs. NFC) and to determine whether personality traits influence how individuals frame their implicit attitudes towards COVID-19 vaccination. By controlling participants’ pre-existing attitudes towards the COVID-19 vaccine, we measured their implicit attitudes using the Implicit Association Test (IAT) and recorded their event-related potentials (ERPs). Phase 1 aimed to explore whether personality-based implicit cognitive processes presented inconsistencies in attitudes toward the COVID-19 vaccine. This process is dependent on the interaction between personality groups (NFA vs. NFC) and congruency conditions (target stimuli, e.g., COVID-19 vaccine) vs. positive/negative attributes. Congruency condition was performed in IAT and varied by including congruent and incongruent pairs. The human brain was expected to process these pairs differently, which may be detected by the ERPs approach, and we hypothesized that there would be differences in the IAT effect and the amplitudes of N1, P2, and N400.

More specifically, we propose that:

H1: The NFC group will show more consistent implicit and explicit attitudes towards the COVID-19 vaccine than the NFA group, reacting more quickly under congruent conditions and more slowly under incongruent conditions.

H2: Incongruent conditions will induce larger N1 and N400 ERP components, while congruent conditions will elicit a larger P2 component, with differences between NFA and NFC groups.

Method

Participants

In accordance with previous research (Kissler and Koessler 2011; Zabielskamendyk 2013; Han et al. 2020), we employed simple random sampling to publicly recruited 82 healthy college students (41 males and 41 females) in China via social media, aged 18 to 33 (M = 22.82, SD = 2.49). The participants were eligible for the experiment if they were vaccinated and had a positive attitude toward vaccination. All participants were right-handed with normal or corrected-to-normal vision and were compensated with money after the experiment.

Measurement

Personality trait measure

Need for affect

This scale, which consists of 26 items, was adopted from Maio and Esses (2001) and demonstrated excellent test-retest reliability (α = .85). In this scale, 13 items assess individuals’ tendency to approach emotional situations, such as “strong emotions are generally beneficial”; 13 items assess individuals’ tendency to avoid emotional situations, such as “I wish to feel less emotions”, which is reverse scored. Participants were asked to evaluate the extent to which they agree with items from 1, ‘strongly disagree’, to 7, ‘strongly agree’.

Need for cognition

This scale was adopted from Cacioppo et al. (1984), which consists of 18 items and yielded excellent test–rest reliability (α = .81). In this scale, 9 items measure individuals’ motivation to engage in and enjoy thinking, such as “I would prefer complex to simple problems”; 9 items measure individuals’ motivation to avoid cognitive tasks, such as “I feel relief rather than satisfaction after completing a task that required a lot of mental effort”, which is reverse scored. Participants were asked to evaluate the extent to which they agree with items from 1, ‘strongly disagree’, to 7, ‘strongly agree’.

We use Z-score as the indicator to evaluate individuals’ personality preferences, which means that individuals who have a higher score in NFA and a lower score in NFC were conceptualized as having a preference for affect, whereas those high in NFC and low in affect were conceptualized as having a preference for cognition. In the formal experiment, 41 participants are in the NFA group (17 males, MZ-score=0.64, s.d.=0.92; Mage = 22.82 years, s.d = 2.49) and 41 are in the NFC group (22 males, MZ-score=0.35, s.d. = 0.87; Mage = 22.78 years, s.d = 2.21).

Explicit attitude measure

Participants were asked to fill out a scale of vaccine explicit attitudes before enrolling in the experiment. The 7-point scale, adapted from Chien (2011) and Nan and Madden (2012), demonstrated good reliability (α = 0.86) and consists of four questions. For example, “Encouraging my friends and family to get vaccinated is a very bad/very good decision”.

Implicit attitude measure

In this experiment, we set three types of stimuli. We invited 26 graduate and undergraduate students to assess the relevance of these stimuli to the target words or attributes before the formal experiment. Stimuli representing the target concepts consisted of 10 COVID-19 vaccine-related terms (M = 4.45, s.d = 0.56). These target terms were measured using participant agreement with a 5-point Likert scale ranging from “1 = strongly irrelevant” to “5 = strongly relevant”; Stimuli that represented positive attributes (M = 4.44, s.d = 0.59) and negative attributes (M = 2.05, s.d = 0.78) consisted of 10 positive words and 10 negative words, these attributes were measured using participant agreement with a 5-point Likert scale ranging from “1 = strongly negative” to “5 = strongly positive”.

The procedure of the Implicit Attitude Test is shown in Table 1 and Fig. 1. In the block of congruent trials, positive attributes shared the same key with target words; In the block of incongruent trials, the negative attributes shared the same key with target words. Practice blocks (IAT steps 1, 2, and 4) consisted of 160 trials, and test blocks (steps 3 and 5) consisted of 180 trials each. At the beginning of each trial, a red fixation point “×” was presented by E- Prime 3.0 in the center of the screen (lasting between 500 and 1000 ms), followed by a random presentation of the stimulus (1000 ms). During the behavior experiment task, the ERP data of the participants were recorded simultaneously.

Table 1 Procedure of the implicit attitude test (IAT).
Full size table
Fig. 1: Experimental procedure.
figure 1

The red fixation point “x” that presented lasting between 500 and 1000 ms.

Full size image

Procedure

The study measured explicit attitudes towards COVID-19 vaccines one week prior to the experiment, and only participants with a positive attitude were enrolled. In the formal experiment, participants came to the lab and completed measures of Need for Affect and Need for Cognition, categorizing them into two groups based on their personality preferences. Subsequently, participants completed five blocks of the Implicit Association Test (IAT), designed to assess individuals’ implicit attitudes by accessing their underlying automatic responses (Greenwald et al. 1998). While completing the IAT, participants’ ERP data were also recorded. The entire experiment lasted approximately 25 min.

ERP data acquisition and processing

ERP data were recorded by SynAmps2 64 channel EEG recording system produced by NeuroScan Company. During recording, the reference electrode was the left mastoid and converted to the average reference of bilateral mastoids during offline analysis. The vertical electrooculogram (VEOG) and horizontal electrooculogram (HEOG) were recorded by bipolar recording. The VEOG electrodes were placed above and below the middle of the left orbit, and the HEOG were placed outboard of the left and right lateral canthus. All electrode impedances stayed below 10 kΩ. The filter bandpass was 0.05 Hz–200 Hz. The sampling frequency was 1000 Hz.

The data pre-processing and analysis were performed using EEGLAB 14.1.1 in the toolbox of MATLAB2016a. Data were bandpass filtered between 0.5–30 Hz. Artifacts, including blinking and EMG, were corrected offline using Independent Component Analysis (ICA). Epochs were segmented for a duration (−100 ms to 500 ms relative to stimulus onset), and baseline correction was applied within the 100 ms pre-stimulus window. Epochs with amplitude values exceeding ±100 μV were manually rejected as artifacts, while the number of valid trials in each experimental condition was more than 30. Averages were calculated for both the congruent conditions and incongruent conditions. According to previous studies (Chen et al. 2018; van Nunspeet et al. 2014; Paller and Kutas 1992), Fz, Cz, and Pz were selected as electrodes of interest, and their averages were treated as the dependent variables in the statistical process.

Visual inspection of the data suggested that the N1, P2, and N400 components were most evident at the midline electrode sites Fz, Cz, and Pz (van Nunspeet et al. 2014). These three ERP components were quantified as the maximum peak amplitude within a time window: N1 = 70–140 ms; P2 = 140–240 ms; N400 = 240–450 ms. Statistical analysis was performed by IBM SPSS Statistics 22.0. The Greenhouse-Geisser correction was applied to the p values, and the post hoc tests were conducted using pairwise comparisons with Bonferroni correction.

Results

Behavioral results

Hypothesis 1 proposed that NFC participants will show a more consistent implicit and explicit attitude, such that they would react faster to congruent stimuli or slower to incongruent stimuli in the IAT task. The IAT effect, represented by the D-score, was calculated as the difference in the reaction times on congruent and incongruent trials divided by a pooled s.d. of all correct trials (Greenwald et al. 2003). We included all trials, compensated 400 ms for error latencies with ACC = 0 and RT > 350 ms, and excluded data with response times less than 350 ms or more than 1000 ms. An analysis of variance (ANOVA) with correct response times as the dependent variable, congruency (congruent/incongruent trials) as the within-subject factor, and group (personality: NFA vs. NFC) as the between-subject factor. Results revealed that the main effect of congruency is significant [F(1,80) = 136.951, p < 0.001, η²p = 0.631]. More specifically, participants in the incongruent group reacted more slowly than those in the congruent group. The main effect of group is nonsignificant [F (1,80) = 0.618, p < 0.434, η2P = 0.008], nor interaction effect of group × congruency [F(1,80) = 0.781, p < 0.378, η2P = 0.010]. H1 was not supported.

In addition, all participants demonstrated the standard IAT effect, with a D-score significantly higher than zero (t (81) = 10.141, p < 0.001). Moreover, this bias was significant in both groups [NFA: t (40) = 7.108, p < 0.001; NFC: t (40) = 7.150, p < 0.001], suggesting that both groups have implicit bias and inconsistent implicit and explicit attitudes.

Table 2 shows the correct rate and response time of the NFA and NFC groups under the two conditions.

Table 2 Phase 1. Accuracy (ACC; %) and reaction time (RT; ms) of new congruent and incongruent conditions in NFA group and NFC group (M ± SD).
Full size table

ERP results

We submitted the average-amplitude values to a 2(group: NFA vs.NFC) ×2(congruency: congruent vs. incongruent trials) ×3(electrode: Fz vs. Cz vs. Pz) mixed-model ANOVA. The congruency and electrode were the within-subject factors, and the group was the between-subject factor (MeanERP waveforms at Fz, Cz, and Pz between two groups: Fig. 2).

Fig. 2: Grand averaged ERP of N1, P2, and N400 elicited by the NFA group and NFC group at Fz, Cz, and Pz.
figure 2

NFA group: need for affect group, NFC group: need for cognition group.

Full size image

N1(70–140 ms). The main effect of the group is significant: the N1 was larger for the NFA group than for the NFC group [F(1,78) = 4.864, p < 0.03, η2P = 0.059]. The main effect of congruency is significant: the N1 was larger for incongruent trials than for congruent trials [F(1,78) = 0.131, p < 0.004, η2P = 0.104]. The interaction of group × congruency is nonsignificant[F(1,78) = 0.328, p < 0.718, η2P = 0.002].

P2(140–240 ms). The main effect of the group is significant: the P2 was larger for the NFC than for the NFA [F(1,78) = 8.990, p < 0.004, η2P = 0.103]. The main effect of congruency is significant[F(1,78) = 5.559, p < 0.021, η2P = 0.067]: the P2 was larger for congruent trials than for incongruent trials. The interaction of group × congruency is nonsignificant [F(1,78) = 2.347, p < 0.130, η2P = 0.029].

N400(240–450 ms). The main effect of the group is significant: the N400 was larger for NFA than for the NFC [F(1,78) = 4.551, p < 0.036, η2P = 0.055]. The main effect of congruency is not significant [F(1,78 = 2.070), p < 0.154, η2P = 0.026]. The interaction of group × congruent is not significant [F(1,78) = 1.318, p < 0.254, η2P = 0.017].

Discussion

Phase 1 found that the NFC and the NFA exhibited different levels of inconsistency in their implicit and explicit attitudes. Controlling their pre-existing positive explicit attitudes toward the COVID-19 vaccine, we measured their implicit attitudes using an Implicit Association Test (IAT) and recording their Event-related potentials (ERPs). Results show that, regarding the positive attributes of the COVID-19 vaccine, the NFA group exhibits larger N1 and N400 components. This might suggest that they are not only more engaged and responsive at the sensory and attentional levels but also more reactive to the semantic content of stimuli and experience increased cognitive conflict. These findings support our hypothesis that implicit attitudes are associated with personality traits and highlight differences in implicit cognitive patterns between NFA and NFC. Specifically, NFC individuals have a more consistent attitude system, while NFA individuals do not.

The behavioral results revealed that both the NFA and NFC groups exhibited implicit biases and inconsistencies between their implicit and explicit attitudes. However, there was no significant evidence to indicate which group was more biased. It might because that the tools and tasks used to measure implicit attitudes may not have been sensitive enough to detect subtle differences in reaction times influenced by personality traits. Consequently, we further analyzed the ERP results. The ERP findings showed differences in the ERPs components (N1, P2, and N400) elicited by participants during congruent and incongruent conditions in both NFA and NFC groups, supporting Hypothesis 2.

First, the amplitudes of N1 elicited by the NFA group were larger than the NFC group. The visual N1 component is assumed to reflect the orientation of attention toward target stimuli and represent the operation of discriminative processes within the focus of attention (Vogel and Luck 2000; Luck et al. 1990). Therefore, the greater N1 wave in the congruent condition of the NFA group reflects that the NFA group primed more early attentional resources than the NFC group to process COVID-19 vaccine-related words that appeared together with positive adjectives, indicating that the NFA group is more cognitively engaged in processing the positive attributes of the COVID-19 vaccine.

Second, the markedly more negative P2 amplitude was observed under congruent conditions and the NFC group. P2 is an indicator of attention and motivated perception (Amodio, 2010), which reflects the level of complexity of task-related information processing and the valence of the target stimuli (Pernet et al. 2003; Yang et al. 2012; Olofsson and Polich, 2007). Our findings reveal that the NFC group applied more attentional resources than the NFA group. This result indirectly validates the information processing model of NFC, which means that they invest more controlled top-down attention allocation and cognitive resources to the target stimuli (Enge et al. 2008).

Another finding that is of great importance is that personality traits can influence the way people process their implicit attitudes. This is revealed by the result that the amplitude of N400 in the NFA group is significantly greater than the NFC group under the congruent condition. N400 is an event-related brain potential component that is associated with meaning processing (Kutas and Federmeier 2011). The greater N400 is, the more cognitive resources are required for semantic integration (Ruz et al. 2003). In the IAT task, generally, the target stimuli represented by the COVID-19 vaccine with positive attributes carried less conflict content than those with negative attributes, but the NFA group spent more cognitive resources to process the information under the congruent condition, which also implies that the NFA group has conflicting perceptions of the positive attributes of COVID-19 vaccine. This supports our hypothesis that the NFA have more negative implicit attitudes toward the COVID-19 vaccine.

Phase 2: personalized matching persuasion

Results from Phase 1 reveal that personal traits significantly influence implicit attitudes, highlighting differences in implicit cognitive patterns between participants with high NFA and those with high NFC. Specifically, individuals with NFC demonstrated more consistent attitude systems compared to their NFA counterparts. Since people’s implicit and explicit attitudes do not always coincide, this means that sometimes people may say they are in favor of something while internally opposing it. Therefore, it is necessary to explore which types of persuasion can change their implicit attitudes. Previous studies have proved that implementing diverse vaccine promotion strategies could enhance COVID-19 vaccine uptake and refine health decision-making, consequently diminishing the pandemic’s impact (Fenta et al. 2023).

Building on these findings, Phase 2 explores potential persuasion strategies that may influence individual attitudes. More specifically, Phase 2 examined the role of message appeal, which refers to the emotional message (affect-based message) and rational message (cognition-based message), and tested whether the effect of personalized matching would be enhanced or diminished if matched or mismatched messages were applied to different personality traits of NFA or NFC. The investigation into these tailored persuasion strategies sets a foundation for determining how individual differences in trait-driven information processing can influence responsiveness to different persuasive techniques.

Previous studies have demonstrated that one of the most effective ways to strengthen a persuasive strategy is to match the message appeal to individuals’ personality traits and their processing style. (Teeny et al. 2021). This personalized matching model suggests that the NFA are more persuaded by an affect-based appeal (appeals that feature emotional-evoking information), while the NFC were more persuaded by a cognitive appeal (appeals that feature logical reasoning information) (Haddock and Maio 2019; Huskinson and Haddock, 2004; Fabrigar and Petty 1999; Haddock et al. 2008). Recent findings in neuroscience also indicate that messages that are well-matched to an individual’s preferences can activate their ventromedial prefrontal cortex, which could potentially lead to greater receptiveness to the matched appeals (Aquino et al. 2020).

However, there is evidence from some studies that suggest a potential for mismatching effects. According to Millar and Millar’s (1990) perspective, the utilization of a matching strategy would be ineffective due to the increased probability of message recipients generating counterarguments against it; they found that individuals who are high in NFC are more sensitive to emotional arguments. Some other studies also indicate that although matching can evoke positive meanings and increase elaboration, it does not guarantee a corresponding enhancement of persuasion (Wan and Rucker 2013; Updegraff et al. 2007).

This persuasion mismatching effect could stem from differences in individuals’ motivation and cognitive capacity to process the message (Petty et al. 1999). Based on the Elaboration Likelihood Model (ELM), when an individual lacks the capacity for detailed thinking (e.g., NFA), personalized matching acts as a simple cue or heuristic. To elaborate, if a match generates a positive meaning, it can enhance persuasion, regardless of the quality of the arguments. On the other hand, when an individual has a chronic inclination to process information under high elaboration conditions (e.g., NFC), they are prone to detecting and correcting any undesired influence that may be affecting their judgments, viewing such influence as inappropriate (McCaslin et al. 2010; Wilson and Brekke 1994). Thus, individual differences moderate the impact of personalized matching on their judgments.

Moreover, the abundance of COVID-19 reports can induce messaging fatigue, leading to feelings of exhaustion and disinterest resulting from repeated exposure to similar information among the target audience (Islam et al. 2020; Koh, et al. 2020). Consequently, we propose that message fatigue can also impose limitations on personalized matching. If an individual is continually presented with similar personalized messages, they might experience boredom or disinterest, which could lead to diminished engagement or even active avoidance.

To gain a comprehensive understanding of how personalized matching can influence persuasiveness, it is essential to understand the mechanisms underlying personalized matching, which is based on individual differences. Therefore, we employed a content-matching paradigm to investigate whether individuals with predominantly affective or cognitive attitudes would exhibit greater or lesser persuasiveness in response to an appeal that aligns with their attitude. More specifically, we propose the following hypotheses:

H3: The NFA will be more persuaded by matched messages (emotional) than mismatched ones (rational), reacting faster in congruent conditions.

H4: The NFC will show greater resistance to both matched (rational) and mismatched (emotional) messages, reacting slower in congruent conditions.

H5: ERP components (N1, P2, N400) will differ between NFA and NFC groups under matched and mismatched conditions.

Method

Participants

Participants in Phase 2 were the same as in Phase 1. After completing the Need for Affect and Need for Cognition scales in Phase 1, Phase 2 asked them to view the experimental material and complete the post-IAT test. The participants were randomly divided into four groups: (1) NFA with the rational appeal (male=8, female=12, Mage ± SD = 22.95 ± 3.24); (2) NFA with the emotional appeal (male=9, female=12, Mage ± SD = 22.81 ± 2.32); (3) NFC with the rational appeal (male=12, female=9, Mage ± SD = 23.09 ± 2.45); (4) NFC with the emotional appeal (male=10, female=10, Mage ± SD = 22.40 ± 2.01).

Materials

The persuasive strategy was manipulated by presenting either a rational (cognitive-based) or emotional (affective-based) appeal, both of which were selected and edited from articles in the People’s Daily and Xinhua News Agency. The same number and quality of arguments were used in both appeals. The content was an introduction to the COVID-19 vaccination. The experimental material we allocated to the rational appeal group was an article titled “COVID-19 Vaccination: A Crucial Tool for Social Epidemic Prevention and Control”, which is a full text containing 1379 Chinese characters and took approximately 3–5 min to read thoroughly; The experimental material assigned to the affective-based group was edited so that the general idea remained consistent, but the objective and neutral narrative was modified into emotional expression, with the title of “Embracing the Warmth of COVID-19 Vaccination: Why You Shouldn’t Hesitate to Get Your Shot”. The full text contained 1435 Chinese characters and took approximately 3–5 min to read. All the news material read by the participants was presented on the computer screen.

This study operationalized personalized matching persuasion as the utilization of message appeal that corresponds with the personality traits of individuals (e.g., presenting a rational appeal to those with high NFC). Conversely, mismatching persuasion was defined as the application of message appeal that deviates from individuals’ personality traits (presenting an emotional appeal for people with high NFC).

Procedure

The experiment was conducted in two stages: a learning stage and an IAT stage. Before beginning the learning stage, participants in all four groups rested for 10 min and were then asked to read the assigned text carefully. In the learning stage, participants were exposed to either a rational (cognitive-based) or emotional (affective-based) appeal material, which took approximately 3–5 min to read.

After the learning stage, participants performed five blocks of the IAT and recorded their ERPs. To enhance the internal validity of the experiment by minimizing the potential for measurement error and ensuring consistency in the results, we utilize the same IAT test employed in Phase 1, but with randomized presentation order. Using this measure also enables a direct comparison of the results from both studies, facilitating the evaluation of the impact of message appeal on implicit attitudes, while accounting for individual differences (Whitford and Emerson, 2019). The entire experiment lasted approximately 30 min.

Results

Behavioral results

The paired t test for D scores revealed a marginally significant condition difference [t (81) = 1.893, p < 0.062], with a higher D-score for pre-test trials (D-score=0.54, s.d.=0.479) than for post-test trials (D-score=0.455, s.d.=0.484). These results showed that after the persuasion stimuli, people had a more positive implicit attitude toward the COVID-19 vaccination. Furthermore, we performed an ANOVA with the personality group condition (NFA vs. NFC) and persuasive strategy (emotional vs. rational) as between-subject factors and (reaction time: pre-test/post-test) as a within-subject factor. Separate statistics were calculated for both congruent and incongruent conditions. There is no significant difference in the interaction effect in congruent condition [F(1,78) = 2.920, p < 0.091, η2P = 0.036], and no significant difference in the interaction effect in incongruent condition [F(1,77) = 0.319, p < 0.574, η2P = 0.004]. Table 3 shows the correct rate and response time of the NFA and NFC groups under the two conditions.

Table 3 Phase 2. Accuracy (ACC; %) and reaction time (RT; ms) of rational and emotional message appeals in the NFA group and NFC group (M ± SD).
Full size table

ERP results

We submitted the average-amplitude values to a 2 (group: NFA vs. NFC) × 2 (message appeal: rational vs. emotional) × 3 (electrode: Fz vs. Cz vs. Pz) mixed-model ANOVA. Separate statistics were calculated for both congruent and incongruent conditions. The electrode was the within-subject factor, and the group and message appeal were the between-subject factor (MeanERP waveforms at Fz, Cz, and Pz between four groups: see Fig. 3).

Fig. 3: Grand average ERP of P2 and N400, elicited by the NFA and NFC groups in response to rational and emotional message appeals at the Fz, Cz, and Pz.
figure 3

NFA group: need for affect group, NFC group: need for cognition group.

Full size image

N1(70–140 ms). In the congruent condition, the main effect of the personality trait group is nonsignificant [F(1,72) = 2.060, p < 0.155, η2P < 0.028]. The main effect of message appeal is nonsignificant [F(1,72) = 0.760, p < 0.386, η2P < 0.010]. The interaction of group × message appeal is nonsignificant [F(1,72) = 0.658, p < 0.420, η2P = 0.009]. Furthermore, in the incongruent condition, nonsignificant results were observed for the main effect of the personality trait group [F(1,72) = 0.001, p < 0.976, η2P < 0.000] and message appeal [F(1,72) = 0.039, p < 0.843, η2P < 0.001], as well as for the interaction effect between the two groups [F(1,72) = 2.546, p < 0.115, η2P < 0.034].

P2(140–240 ms). In the congruent condition, the interaction of group × message appeal is significant [F(1,72) = 5.394, p < 0.023, η2P = 0.070]. Specifically, in the rational appeal condition, the difference in P2 amplitude between the NFC and NFA groups was significant, with the NFA group showing larger P2 amplitudes than the NFC group [F(1,72) = 9.433, p < 0.003, η2P = 0.116]. However, in the emotional appeal condition, this difference was not significant [F(1,72) = 0.019, p < 0.892, η2P < 0.001]. The main effect of the personality trait group is significant: the P2 was larger for the NFA group than for the NFC group [F(1,72) = 4.556, p < 0.036, η2P = 0.060]. The main effect of message appeal is nonsignificant [F(1,72) = 0.010, p < 0.920, η2P < 0.01]. Furthermore, in the incongruent condition, there was no significant main effect of the personality trait group [F(1,72) = 0.153, p < 0.697, η2P = 0.002], nor interaction effect of group × message appeal [F(1,72) = 0.011, p < 0.917, η2P = 0.000], and the main effect of message appeal was also nonsignificant [F(1,72) = 1.687, p < 0.198, η2P = 0.023].

N400 (240–450). In the congruent condition, the interaction of group × message appeal is significant [F(1,72) = 5.784, p < 0.019, η2P = 0.074]. Specifically, the difference in N400 amplitude between the NFA and NFC groups was significant in the rational appeal condition, with the NFC group showing a larger N400 than the NFA group [F(1,72) = 6.024, p < 0.017, η2P = 0.077]. However, this difference was not significant in the emotional appeal condition [F(1,72) = 0.823, p < 0.367, η2P = 0.011]. The main effect of the group is nonsignificant [F(1,72) = 1.338, p < 0.251, η2P = 0.018]. The main effect of message appeal is nonsignificant [F(1,72) = 0.298, p < 0.587, η2P = 0.004]. Furthermore, in the incongruent condition, there was no significant main effect of the personality trait group [F(1,72) = 0.043, p < 0.837, η2P = 0.001], nor interaction effect of group × message appeal [F(1,72) = 0.193, p < 0.662, η2P = 0.003], and the main effect of message appeal was also nonsignificant [F(1,72) = 0.168, p < 0.683, η2P = 0.002].

Discussion

Phase 2 aimed primarily to evaluate how participants with either high NFC or high NFA personality traits receive rational and emotional message appeals differently when the message appeals either matched or mismatched their personality trait and processing style. We found that our experiment did not yield a personality-matching effect. Additionally, we observed that individuals with high NFC demonstrated greater resistance to matched persuasion, while those with high NFA were more receptive to rational appeals, partially supporting H3 and H4.

More specifically, the behavioral results showed higher D-score for pre-test trials and a lower D-score for post-test trials. As previously established, higher D-score indicates a stronger implicit bias (Greenwald et al. 1998). This suggests that exposure to persuasive stimuli led to a more favorable implicit attitude towards COVID-19 vaccination. However, the absence of a significant interaction effect precludes us from determining whether there was a matched or mismatched persuasion effect on behavioral outcomes. This may be because the IAT task, while effective in measuring implicit biases, may not capture nuanced changes in attitude resulting from specific persuasion strategies employed. Therefore, to verify our hypothesis, we conducted an analysis of the ERPs and found that there are differences in ERP components (N400 and P2) elicited by participants in the NFA and NFC groups, which supports H5.

First, the N400 results revealed that rational persuasion strategies resulted in lower cognitive conflict among NFA individuals, suggesting that they are more receptive to rational message appeals. This finding aligns with previous research indicating that individuals with weakly held attitudes are more susceptible to attitude change in response to persuasive messages, unlike those with strongly held attitudes (Clarkson et al. 2011). Therefore, rational persuasion strategies may be more effective in influencing the attitudes of NFA individuals who are less certain in their beliefs, as they are more open to being persuaded by messages that challenge their existing attitudes. In contrast, NFC had a greater cognitive conflict, indicating that they have higher attitude certainty and are less likely to be swayed in their attitudes. Therefore, when people are in higher elaboration states, the belief that their attitudes were biased by the personalized match can lead them to mentally correct for the bias and hold their original attitude (Wilson and Brekke, 1994).

The result showed that NFA had a higher P2 activation, suggesting that they allocate more attention to rational messages. It happens maybe because they typically adopt a peripheral route and only process the message shallowly in their prior experience. Encountering a more complex and rational message, therefore, requires more attention to process and comprehend. The finding that NFC participants had lower P2 activation and less attention allocation to rational appeal suggests that they usually adopt a central route and process messages carefully, so seeing new messages is not surprising. It also aligns with findings in prior research that individuals with high NFC tend to prefer complex tasks, carefully considering all available information before arriving at a decision, resulting in more in-depth processing (Cacioppo and Petty, 1982; Haugtvedt et al. 1992).

Generally discussion

We discovered that customizing message appeals (rational vs. emotional) to align with an individual’s personality traits (NFA vs. NFC) does not necessarily result in a persuasive advantage, challenging the effectiveness of the personality-content matching approach (Teeny et al. 2021; Haddock et al. 2008). Across two experiments, we observed that individuals with high NFC exhibited greater consistency between their implicit and explicit attitudes and displayed higher resistance to persuasion attempts. Conversely, those with high NFA showed greater susceptibility to external influences, resulting in larger discrepancies between their internal and expressed attitudes. Furthermore, individuals with high NFA were more receptive to rational persuasion that did not match their affective-based attitudes.

Since implicit attitudes are unconscious, uncontrolled cognitive states, it is possible for individuals to have an opposing implicit attitude even if they outwardly support an attitude object and act accordingly. Previous studies have shown that individuals’ implicit and explicit attitudes toward the COVID-19 vaccine exhibited inconsistencies (Simione et al. 2022) and primarily considered individual differences in gender and age (Colledani et al. 2021). Phase 1 extends prior research by showing that personality traits—the NFC and the NFA—could moderate individuals’ implicit attitudes, including inhibiting their negative bias. Our findings demonstrate that although participants self-reported explicit attitudes in support of the COVID-19 vaccine, their implicit attitudes showed the opposite, which is related to the personality trait where people with high NFC exhibit more consistent implicit and explicit attitudes than those with high NFA. This complements prior observations that implicit bias is personality-based (Aguinis et al. 2009).

This could be due to the fact that people with high NFC not only reflect on external message arguments but also reflect on their own thoughts guided by top-down goal-oriented control (Enge et al. 2008). Therefore, they engage in more pre-processing of messages and form their implicit attitudes accordingly, with their explicit attitude being more likely to be a genuine reflection of their implicit attitude. Our results provide initial evidence that individuals with high NFC tend to show internal consistency in their implicit and explicit attitudes. According to Conner et al. (2007), more habitual attitudes could be predicted more by implicit attitudes, while explicit attitudes become more influential in determining behavior for individuals who have a tendency to engage in thinking (high NFC). Our findings echo this argument, indicating that high NFC individuals are more likely to show consistency among their behaviors, implicit attitudes, and explicit attitudes.

Correspondingly, people with high NFA were more likely to show inconsistent implicit versus explicit attitudes. The high NFA also indicates that they give more weight to bottom-up attention before decision-making (i.e., taking the vaccine), so their explicit attitudes are less dependent on their prior processing of information and more influenced by the stimuli available to them in the current moment. This is more likely to change their explicit attitudes and show contradictions with their implicit attitudes.As implicit attitudes are formed through automatic subconscious pairings between an attitude object and related evaluations (Rydell and McConnell 2006), they are not influenced by one’s goals and ulterior motives. Therefore, implicit attitudes are more likely to reveal personal and socially undesirable biases than explicit attitudes (Fazio and Olson 2003). Consequently, even if a person’s explicit attitude changes, the implicit attitude can remain the same. Because individuals who are high in NFA are more likely to view emotions, rather than information, as useful when making judgments, it is very possible that these people could shift their explicit attitudes due to external emotional stimuli. In the context of our study, as getting vaccines is strongly recommended by the Chinese government and people were afraid of being infected by the virus, people with high NFA were very likely to change their explicit attitudes toward the vaccine from negative to positive, influenced by external factors. However, their implicit attitudes might remain negative. This could potentially explain why our results demonstrated that high NFA people showed more incongruent attitudes than high NFC people.

Additionally, Phase 2 builds upon the findings of Phase 1 by investigating how the influence of content-based matching may vary based on individuals’ personality traits, through the use of either a rational or emotional appeal. Specifically, the study aims to determine if different messages would amplify or reduce the effect of message appeal-based matching on individuals with high or low levels of certain personality traits. In essence, Phase 2 seeks to explore whether personalized matching is a more effective persuasion technique for some individuals than others. Research has demonstrated that message appeals are more effective when personalized to an individual’s personality traits, with rational message appeals eliciting more positive attitudes among NFC, while emotional message appeals produce more positive attitudes among NFA (Adler et al. 2016; Clarkson et al. 2011). However, Phase 2 suggests that NFC individuals are more resistant to matched persuasion, whereas NFA individuals are more receptive to rational appeals, contrary to their affective-based attitude orientation.

This finding can be explained by the fact that personalized matching can influence attitudes through different mechanisms depending on the recipient’s level of elaboration (DeBono 1987; Wilson and Brekke 1994). Attitude certainty significantly impacts the effectiveness of persuasion, which can either increase or decrease susceptibility to change, depending on an individual’s underlying cognitive structure. According to Clarkson et al. (2008), ambivalent attitudes are more prone to change than univalent attitudes, and attitude certainty amplifies this effect. Our results support this argument, as NFA individuals with low attitude certainty appear to benefit more from the persuasive advantage of mismatched (compared to matched) messages.

Despite the persuasive advantage of matched messages, high attitude certainty did not seem to enhance the effect for NFC individuals in Phase 2. This finding can be explained by the self-validation hypothesis, which suggests that individuals are more likely to be influenced by their own validated cognitions than by external persuasion (Petty et al. 2002). In Phase 2, NFC individuals may have had a strong attitude towards the COVID-19 vaccination topic, making it more resistant to external challenges, even personalized persuasion. Additionally, with COVID-19 being a global pandemic, people may have experienced message fatigue due to the constant bombardment of repetitive messages (Koh et al. 2020). This could have led to desensitization to COVID-19 vaccination information after an initial period of anxiety, increasing the likelihood of resistance.

In addition to these theoretical issues, our findings also have practical implications. Firstly, this study sheds new light on the underlying determinants of COVID-19 vaccine hesitancy and the effectiveness of persuasion strategies at the individual implicit cognitive level. This understanding can aid in the development of communication strategies for vaccination promotion initiatives. Prior research has shown that implicit cognitive processes are critical in shaping attitudes and subsequent decision-making processes (Bargh et al. 1996; Chen and Bargh 1999; Dovidio et al. 1997). Moreover, automatic associative processes have been shown to be the primary driver of individuals’ attitudes toward objects, and these processes can lead to discriminatory behavior (Fazio 2001; Nosek and Banaji 2001). It has also been found that the IAT effect is heightened under stereotype threat (Frantz et al. 2004), but can be reduced when participants employ a strategy to mitigate their biases (Fiedler and Bluemke 2005). Therefore, Phase 2 provides valuable insights into the conditions under which personalized persuasion strategies will be more effective based on personality traits such as NFA and NFC, which can moderate the acceptance of matched messages.

There are also some limitations in this study. Our study investigates the impact of persuasive messages on vaccine hesitancy and negative implicit attitudes toward vaccines. However, the constant barrage of information on COVID-19 vaccines may have led to information overload, potentially resulting in message fatigue and reduced effectiveness of persuasive strategies. In the future, studies should investigate the persuasive effects of social events with diverse themes to expand our understanding of how these events influence attitudes and behavior.

Secondly, given the observed group differences in response to rational but not emotional message appeals, future research should delve deeper into understanding why emotional appeals did not produce the same differential effects. It is possible that emotional appeals might universally affect both NFC and NFA individuals similarly, or there might be other underlying factors that were not captured in the current study.

Thirdly, the current study focuses only on personality differences based on the NFC and the NFA. Future explorations can be made on the impact of more types of individual differences, such as the big five personality traits or individual motivation.

Moreover, ERP is a tool with high time accuracy but poor spatial accuracy. Hence, methods like fMRI can also be employed in the future to explore the precise spatial information of brain activity during information processing.

Conclusion

To the best of our knowledge, this study is the first to utilize an ERP approach to investigate how personality differences impact implicit attitudes and the effectiveness of personalized matching toward the COVID-19 vaccine. In Phase 1, we found that there are differences in implicit cognitive patterns between people who are high in NFA and those high in NFC. Phase 2 demonstrates that personalized matching doesn’t guarantee a corresponding enhancement of persuasion, as it can influence people’s attitudes through different psychological processes depending on their level of elaboration. Both IAT and ERP results demonstrate that individuals with high NFC show more consistent implicit and explicit attitudes as they take more controlled top-down cognitive resources toward the target stimuli. This means that individuals with high NFC will automatically pre-process and reflect on external information in the central path and regulate their internal thoughts accordingly, resulting in a more consistent attitude system. In contrast, individuals with high NFA usually adopt a bottom-up processing pattern, with less careful processing of information evidence via the peripheral route, which elicits less lasting attitude changes and thus results in an inconsistent attitude system. The results support the claim that personality traits are a critical characteristic factor in attitude formation and attention allocation that can modify the extent of information processing.

These findings also provide helpful guidance in practice by offering a new understanding of the factors that influence COVID-19 vaccine hesitancy. Implicit attitudes play a vital role in guiding individual behaviors and preferences, and they are also associated with personality traits, thus becoming one of the main leverages to reduce vaccine hesitancy. Pandemic recovery programs should prioritize the development of interventions to change people’s implicit attitudes by applying different persuasive strategies to different personalities in order to promote overall vaccination rates.