Online word of mouth (WOM) is a key determinant of product sales (Rosario et al., 2016). It is therefore critical for marketers to understand the factors that shape consumers’ intentions to create and seek recommendations. In this research, we argue that individual differences in the need to evaluate (NE; Jarvis & Petty, 1996) influence both creating and seeking online WOM. Past research has shown that there are individual differences in the chronic need to engage in evaluative responding. Individuals high (vs. low) in the NE are more likely to have attitudes toward a variety of issues. Recent research has demonstrated that there are two additional aspects of the NE focused on interpersonal contexts, NE-expressing (i.e., the need to express one’s attitudes to others) and NE-learning (i.e., the need to acquire attitudes from others; Xu et al., 2021). Because online WOM is inherently interpersonal (Chen, 2017), we explore whether these two recently identified aspects of the NE can be used to make more nuanced predictions regarding consumers’ likelihood of engaging in two distinct types of online WOM, posting reviews and seeking recommendations. Our results have important implications for marketers who wish to understand the psychological drivers of online WOM, including helping marketers recognize which segments to target when encouraging different types of WOM in different contexts.

1 Theoretical background

1.1 Antecedents for engaging in online WOM

Past research has explored a variety of antecedents of WOM. For example, companies sometimes offer incentives (Godes & Mayzlin, 2009) or promotions that trigger WOM (Berger & Iyengar, 2013). Characteristics of the offering (Berger & Schwartz, 2011) and consumers’ satisfaction (Brown et al., 2005) also influence WOM. But what about characteristics of the consumer? Past research suggests that consumers choose to share WOM due to a basic human motive to self-enhance (Berger, 2014) and because doing so is inherently enjoyable (He et al., 2019). However, limited work has explored the role that individual differences play in the desire to share online WOM beyond mavenism (Feick & Price, 1987).Footnote 1 Consequently, we have limited knowledge of which segment(s) of consumers are likely to spread WOM. We propose that individual differences in NE-expressing and NE-learning can be especially effective at predicting the likelihood of engaging in different types of online WOM because these scales measure different motives that map onto various goals consumers can have when engaging in online WOM (e.g., self-expression and information acquisition; Xu et al., 2021).

1.2 NE and motivations behind providing online WOM for others

NE-expressing captures individual differences in the human tendency to express one’s evaluations (Xu et al., 2021). Prior work in marketing has identified market mavens as individuals who have information (including evaluations) specifically about the marketplace that they want to share with others (Feick & Price, 1987). In 2022, mavenism might often take the form of writing online product reviews. Both those high in NE-expressing and mavenism should be more likely to write online reviews for familiar products, but we propose that NE-expressing predicts the likelihood of writing reviews even when controlling for mavenism since NE-expressing captures a more generalized and relatively more fundamental driver of the human tendency to express all sorts of opinions, ranging from opinions about products to social and political issues.

However, we propose that those high in NE-expressing are not always more likely to share online WOM. In order to express opinions on novel products, high NE-expressing individuals must first develop their own evaluations. Thus, we propose novel products, which we define as either products from a completely new/unfamiliar product category or with novel/unfamiliar features (Moldovan et al., 2011), as an important boundary condition for the tendency of high NE-expressing individuals to share online word of mouth. In fact, we propose that when consumers encounter novel products, they are only likely to share online WOM when both NE-expressing and NE-learning are high. This is because NE-learning has been linked with an epistemic motive (the desire to develop a thorough understanding of a given target; Xu et al., 2021). Those high in NE-learning are thus more likely to develop knowledgeable evaluations that can then be expressed. Thus, we propose that NE-expressing does not always lead to more WOM; although it does so for familiar products, for novel products, a basic level of knowledge (more likely to be developed by those high in NE-learning) must precede the willingness of such consumers to engage in WOM. In contrast, NE-learning should not predict likelihood of writing reviews for familiar products since doing so does not provide a direct opportunity to seek others’ opinions. More formally:

  • H1a For familiar products, consumers higher (vs. lower) in NE-expressing are more likely to provide online product reviews, regardless of their NE-learning motivation.

  • H1b For novel products, consumers higher (vs. lower) in NE-expressing are more likely to provide online product reviews if their NE-learning motivation is also relatively high versus low.

1.3 NE and motivations behind asking for recommendations from others Online

In addition to sharing WOM, consumers sometimes seek online recommendations about products from others. Berger (2014) argued that consumers engage in such information acquisition when a decision is particularly risky or when trustworthy information is unavailable. However, to our knowledge, no prior work has explored individual differences in the tendency to seek recommendations from others. Given that people who score higher in NE-learning are motivated to learn about others’ opinions, we predict that they will also be more likely to seek online recommendations to gain insights into others’ evaluations. However, we argue that NE-learning does not always uniquely predict recommendation seeking behaviors. Instead, forum anonymity will determine whether seeking recommendations is driven by NE-learning alone or NE-learning and NE-expressing. We define an anonymous forum as one in which an individual’s identity is not linked to their posts/comments (e.g., Reddit). Consumers using anonymous forums to seek advice are likely driven solely by their desire to seek information, and hence consumers higher in NE-learning should be more likely to seek recommendations in such forums. In contrast, NE-expressing should not predict seeking recommendations in anonymous forums since doing so provides no opportunity for self-expression.

However, consumers can also seek recommendations using platforms where their posts are linked to their identities (e.g., Facebook’s Recommendation feature). We propose that, because NE-expressing taps into the value-expressive function of attitudes, which is linked to self-expression (Katz, 1960; Xu et al., 2021), NE-expressing will interact with NE-learning to predict seeking recommendations in identity-linked forums. Specifically, we predict that consumers higher in NE-expressing will be more motivated to seek recommendations online on identity-linked platforms even when NE-learning is relatively low because identity-linkage makes even a request for information an opportunity for self-expression (e.g., through self-enhancement; Berger, 2014). In contrast, in anonymous forums, NE-learning (i.e., a true desire for information) will drive recommendation seeking behaviors regardless of NE-expressing. Thus, we propose:

  • H2a In anonymous online forums, consumers higher (vs. lower) in NE-learning are more likely to seek recommendations, regardless of their NE-expressing motivation.

  • H2b On identity-linked platforms, NE-expressing will interact with NE-learning to influence consumers’ likelihood to seek recommendations, such that NE-learning will be a stronger predictor of recommendation seeking when NE-expressing is relatively low versus high.

2 Study one

In study 1, we tested hypothesis 1a while also measuring market mavenism.

2.1 Method

A total of 222 Amazon MTurk workers (42.3% females; Mage = 38.36) completed this study in exchange for $0.40. Participants first responded to the NE-expressing, NE-learning, and mavenism (Feick & Price, 1987) scales presented in a random order (see web appendix for all study stimuli). NE-expressing and NE-learning were evaluated using a five-point scale (1 = extremely uncharacteristic of me; 5 = extremely characteristic of me) while mavenism was assessed using a seven-point scale (1 = strongly disagree; 7 = strongly agree). All participants were then asked to recall a material product (priced at $20–30) that they had purchased from Amazon within the past year toward which they had a positive attitude.After the exclusion of 13 participants for recalling a negative purchase, the final sample was 209 (42.6% females; Mage = 38.34).Footnote 2

Participants indicated their attitude toward the product on four seven-point semantic differential scales (e.g., negative – positive) and then indicated their likelihood of writing an online review on Amazon on a seven-point scale (1 = not at all likely; 7 = extremely likely). All participants completed demographic measures in all studies.

2.2 Results and discussion

In line with prior research (Xu et al., 2021), overall NE-expressing and NE-learning scores were obtained, standardized, and treated as continuous variables in all studies, as was mavenism. Similar to previous work, the correlation between NE-expressing and NE-learning was 0.57 (p < 0.001). The correlation between mavenism and NE-expressing was 0.52 (p < 0.001), and the correlation between mavenism and NE-learning was 0.41 (p < 0.001).

We ran a multiple regression model with NE-expressing, NE-learning, mavenism, and all interactions as independent variables and likelihood of writing a review as the dependent measure. Participants’ attitudes toward the product were included as a covariate. Consistent with prior work (Feick & Price, 1987), there was a significant positive effect of mavenism on likelihood of writing a review, b = 0.74, t(200) = 4.81, p < 0.01, 95% CI: [0.4, 1.05]. Of more interest, there was also a significant positive effect of NE-expressing on likelihood of writing a review, b = 0.37, t(200) = 2.49, p = 0.01, 95% CI: [0.08, 0.71]. More importantly, this positive effect held when controlling for mavenism, suggesting that NE-expressing contributes unique variance when predicting the likelihood of writing online reviews. However, as expected, there was no significant effect of NE-learning (p = 0.49). No interactions were significant (ps > 0.42).

These results support our prediction that NE-expressing, a driver of people’s tendency to express in general, predicts likelihood of providing reviews for familiar products even when controlling for mavenism, a measure that assesses expertise in the marketplace specifically.Footnote 3

3 Study two

In study 2, we tested hypothesis 1a/b by asking participants to indicate their likelihood of providing a product review for a novel versus familiar product in a preregistered study (https://aspredicted.org/5FK_Q41).

3.1 Method

We employed a two-phase study design. In phase 1,702 MTurk workers (48.3% females; Mage = 40.67) completed the study in exchange for $0.35. Two days later, all participants were invited to return for a presumably unrelated study; 422 MTurk workers (49.2% females; Mage = 40.78) completed phase 2 in exchange for $0.20. In phase 1, participants completed either the NE-expressing or NE-learning scale. Then participants imagined they had purchased a novel (KOR ONE) or familiar (CamelBak) water bottle and were given 30 s to read a brief product description. Participants then indicated their likelihood of writing a review of the product using the same item from study 1. Next, participants completed manipulation checks assessing product novelty and familiarity (1 = not at all; 7 = extremely) and the attitude measures from study 1. In phase 2, participants completed the scale measure they did not complete in phase 1. Thus, they completed the NE-expressing and NE-learning scales in randomized order and two days apart (r = 0.23, p < 0.001).Footnote 4 Participants also indicated the approximate number of online product reviews (for any product on any website) they had written in the past six months (0; 1–2; 3–4; 5–6; 7–8; 9–10; 11 or more; coded as 1–7). These anchors were determined based on a pilot survey from the same population (n = 49) in which participants indicated the exact number of reviews they had written in the past six months (range: 0–16, M = 3.06, SD = 3.73).

3.2 Results and discussion

As expected, the KOR ONE water bottle was perceived as more novel (M = 4.75, SD = 1.42) and less familiar (M = 1.94, SD = 1.60) than CamelBak (Mnovel = 3.17, SD = 1.82; Mfamiliar = 3.55, SD = 2.33), t(420) = 10.02 and -8.32 respectively, ps < 0.001. There was no difference in product attitudes (p = 0.96).

We ran a multiple regression model with NE-expressing, NE-learning, product novelty condition (1 or -1), and all interactions as independent variables and likelihood of writing a review as the dependent measure. There was a significant positive effect of both NE-expressing, b = 0.77, t(414) = 9.15, p < 0.001, 95% CI: [0.61, 0.94] and NE-learning on likelihood of writing a review, b = 0.30, t(414) = 3.51, p = 0.001, 95% CI: [0.13, 0.46]. More importantly, there was a significant three-way interaction (i.e., NE-expressing X NE-learning X novelty), b = 0.20, t(414) = 2.82, p = 0.005. In the familiar product condition, consistent with hypothesis 1a, NE-expressing was the only significant predictor of providing a review, b = 0.89, t(202) = 7.23, p < 0.001 (Fig. 1 top panel). The interaction between NE-expressing and NE-learning was not significant (p = 0.80). However, in the novel product condition, consistent with hypothesis 1b, the NE-expressing X NE-learning interaction was significant, b = 0.36, t(212) = 3.75, p < 0.001 (Fig. 1 bottom panel). When the product was novel, higher NE-expressing led to a greater likelihood of providing a review but especially so when NE-learning was also high rather than low. A floodlight analysis (Spiller et al., 2013) indicated that for those who scored 2.25 and above on the NE-learning scale (80.6% of the sample), NE-expressing scores had a significant positive effect on their likelihood of writing a review. Thus, for novel products, NE-expressing predicts providing product reviews more when NE-learning is also high.Footnote 5

Fig. 1
figure 1

Predictive effect of NE-expressing and NE-learning on consumers’ likelihood of writing a review for a familiar (top panel) and novel water bottle (bottom panel). Note: NE-expressing is a continuous variable in all analyses. In the figure, we illustrated the predicted dependent measure at + 1 SD above and -1 SD below the mean of NE-expressing scores

Full size image

We also conducted a regression analysis replacing NE-expressing with participants’ past WOM behavior (r = 0.35, p < 0.001). The three-way interaction (i.e., past WOM behaviors X NE-learning X novelty) showed a similar pattern described above but was not significant, b = 0.06, t(414) = 1.22, p = 0.22. When including both this three-way interaction and the NE expressing X NE-learning X novelty interaction in the same model, the latter remained significant (p = 0.005).

Overall, these results support hypothesis 1a/b that consumers with higher NE-expressing do not engage in WOM for novel products unless they are also higher in NE-learning. We note that while all participants had the opportunity to learn about the product, high NE-learning participants might process the product information more deeply and therefore gained more product knowledge, giving those with high NE-expressing more to say.

4 Study three

In study 3 we sought to conceptually replicate the interaction between NE-expressing and NE-learning while (1) providing participants with an actual novel product with which to interact and (2) measuring actual WOM behavior.

4.1 Method

A total of 244 undergraduates (50% female, Mage = 19.71) completed this study for course credit. Participants completed the NE-expressing and NE-learning scales in randomized order (r = 0.49, p < 0.001). Then they were introduced to the novel product, a KOR ONE water bottle. All participants were given three minutes to physically examine the water bottle, while reading a description of its features. Participants were then presented with a choice; they could spend the next three minutes writing a review of the water bottle for subsequent lab participants or use the same amount of time writing about their day yesterday. Their choice served as our dichotomous dependent variable (177 participants, 72.54%, wrote the review).

4.2 Results and discussion

We ran a logistic regression in which NE-expressing and NE-learning scores were entered as predictors along with their interaction. The results revealed a significant interaction between NE-expressing and NE-learning, b = 0.35, wald = 4.74, p = 0.03, odds ratio of 1.42 (Fig. 2), that was in the same direction as for the novel product in Study 2.Footnote 6 For a novel product, higher NE-expressing again led to a greater likelihood of providing a review and especially when NE-learning was also high. When NE-learning was relatively low, however, high NE-expressing was associated with a reduced likelihood of writing a review. A floodlight analysis (Spiller et al., 2013) revealed that for those who scored 1.83 and below on the NE-learning scale (3.3%), higher NE-expressing scores had a significant negative effect on their choice of writing a product review.

Fig. 2
figure 2

Predictive effect of NE-expressing and NE-learning on consumers’ choice of writing a review for a novel water bottle. Note: NE-expressing is a continuous variable in all analyses. In the figure, we illustrated the predicted dependent measure at + 1 SD above and -1 SD below the mean of NE-expressing scores

Full size image

This study conceptually replicated the findings in the novel condition in study 2 with a behavioral measure.Footnote 7 Interestingly, in the current study, we also found that consumers who have the motivation to express but lack the motivation to learn about the novel product are least likely to write a review.

5 Study four

In study 4, we tested hypothesis 2 by manipulating the anonymity of the online forum and assessing participants’ likelihood to seek rather than share WOM.

5.1 Method

A total of 407 MTurk workers (44.0% females; Mage = 38.33) participated in this study for $0.35. Participants responded to the NE-expressing and NE-learning scales (r = 0.56, p < 0.001) in randomized order.Footnote 8 Participants imagined they were planning to make a purchase and were having trouble choosing which model to buy. They imagined finding an online forum for customers to discuss the brand that was either anonymous or required them to set up a profile linked to one of their social media accounts. Participants then rated their likelihood of using this forum to ask for recommendations using a seven-point scale (1 = extremely unlikely; 7 = extremely likely).

5.2 Results and discussion

We ran a multiple regression model with NE-expressing, NE-learning, platform anonymity (-1 or 1), and all interactions as independent variables and likelihood to seek recommendations as the dependent measure. There was a significant positive effect of NE-learning on likelihood of seeking recommendations, b = 0.56, t(399) = 5.60, p < 0.001, 95% CI: [0.36, 0.76]. The effect of NE-expressing on likelihood of recommendation seeking was also positive and significant, b = 0.32, t(399) = 3.20, p = 0.001, 95% CI: [ 0.13, 0.52]. Additionally, the significant positive effect of platform type on participants’ likelihood of seeking recommendations indicated that participants are less likely to seek recommendations on identity-linked forums, b =—0.26, t(399) = -2.88, p = 0.004, 95% CI: [-0.43, -0.08].

Importantly, there was a significant three-way interaction (i.e., NE-learning X NE- expressing X anonymity), b = -0.15, t(399) = -2.12, p = 0.035. In the anonymous condition, NE-learning was the only significant predictor of seeking recommendations, b = 0.75, t(203) = 5.15, p < 0.001 (Fig. 3 top panel). Neither the effect of NE-expressing (p = 0.42) nor the interaction between NE-expressing and NE-learning was significant (p = 0.83). However, in the identity-linked condition, the interaction between NE-expressing and NE-learning was significant, b = -0.28, t(196) = -2.88, p = 0.004 (Fig. 3 bottom panel): NE-learning was more strongly related to recommendation seeking when NE-expressing was low. A floodlight analysis (Spiller et al., 2013) indicated that for those who scored 2.98 and below on the NE-expressing scale (69.0% of the sample), their NE-learning scores had a significant positive effect on the likelihood of seeking a recommendation on an identity-linked forum.

Fig. 3
figure 3

Predictive effect of NE-learning and NE-expressing on consumers’ likelihood of seeking recommendations on an anonymous (top panel) and an identity-linked platform (bottom panel). Note: NE-learning is a continuous variable in all analyses. In the figure, we illustrated the predicted dependent measure at + 1 SD above and -1 SD below the mean of NE-learning scores

Full size image

Overall, these results support our hypothesis that, on an anonymous platform, consumers’ NE-learning scores predict likelihood of seeking recommendations. In contrast, seeking recommendations on an identity-linked forum resulted from an interaction between NE-learning and NE-expressing.

6 Study five

In study 5, our goal was to conceptually replicate the NE-learning X NE-expressing interaction on consumers’ recommendation seeking behaviors using a well-known non-anonymous social media platform.

6.1 Method

A total of 181 MTurk workers (49.7% females; Mage = 34.2) participated in this study for $0.50. Participants completed the NE-expressing and NE-learning scales presented in a randomized order (r = 0.59, p < 0.001) and then read a description of the Facebook Recommendation feature and imagined they were going to an unfamiliar city for a business trip. They subsequently indicated their likelihood of using this feature to get suggestions for local attractions/restaurants from their Facebook friends on a seven-point scale (1 = extremely unlikely; 7 = extremely likely). All participants then indicated their Facebook usage frequency (1 = never; 5 = multiple times a day).

6.2 Results and discussion

We ran a multiple regression model with NE-expressing, NE-learning, and their interaction as independent variables to predict likelihood of using the Facebook Recommendation feature. Participants’ Facebook usage frequency was included as a covariate. There was a significant positive effect of NE-learning on likelihood of using the Recommendation feature, b = 0.31, t(176) = 2.42, p = 0.02, 95% CI: [0.06, 0.56]. No significant effect of NE-expressing was found (p = 0.19). Replicating the finding in the identity-linked condition of study 4, there was a significant interaction between NE-expressing and NE-learning, b = -0.19, t(176) = -2.13, p = 0.03, 95% CI: [-0.36, -0.01] (Fig. 4). NE-learning was more strongly related to recommendation seeking when NE-expressing was low. Using the floodlight technique (Spiller et al., 2013), for consumers who scored 3.12 and below on the NE-expressing scale (61.33% of the sample), there was a positive effect of NE-learning on willingness to use the Recommendation feature.Footnote 9

Fig. 4
figure 4

Predictive effect of NE-learning and NE-expressing on consumers’ likelihood of seeking recommendations using the Facebook Recommendation feature. Note: NE-learning is a continuous variable in all analyses. In the figure, we illustrated the predicted dependent measure at + 1 SD above and -1 SD below the mean of NE-learning scores

Full size image

7 General discussion

In this research, we demonstrated that individual differences in NE-expressing and NE-learning predict consumers’ likelihood of engaging in different types of online WOM. Study 1 showed that for familiar products, NE-expressing uniquely predicted likelihood of providing online reviews over and beyond the classic concept of market mavenism. Furthermore, Studies 2 and 3 demonstrated that NE-expressing does not always lead to a higher likelihood of providing WOM. Although it does so for familiar products, for unfamiliar products, a basic level of knowledge (through the motivation to learn) must accompany the willingness of high NE-expressing consumers to write online reviews. Studies 4 and 5 demonstrated that consumers’ NE-learning scores do not always uniquely predict recommendation seeking. Although it does so on anonymous forums, on identity-linked forums, NE-learning is more predictive when NE-expressing is relatively low.

Theoretical implications

Our findings contribute to the growing body of literature on the antecedents of WOM (Berger, 2014) by focusing on individual differences in NE-expressing and NE-learning. We demonstrated that these fundamental motives that drive people to express and seek opinions can be useful in predicting consumers’ likelihood to create and seek online WOM.

We also contribute to the literature on online reviews by highlighting the different motivations behind people’s desire to provide reviews for familiar versus novel products. Our work suggests that WOM about familiar versus novel products helps to fulfill different personal goals in terms of expressing and learning evaluations.

Finally, our work also adds to the recommendation seeking literature, a relatively understudied form of WOM. Berger (2014) discussed but did not empirically test how identifiability of the communication channel influences people’s WOM motivations. Anonymous posting should lower people’s concern with how they present themselves. However, in an identity-linked platform, impression management motives should play a greater role (Eisingerich et al., 2015). Our research provides a direct test of this to disentangle the different motives involved when people seek recommendations.

Our findings contribute to the growing body of literature on the antecedents of WOM (Berger, 2014) by focusing on individual differences in NE-expressing and NE-learning. We demonstrated that these fundamental motives that drive people to express and seek opinions can be useful in predicting consumers’ likelihood to create and seek online WOM.

We also contribute to the literature on online reviews by highlighting the different motivations behind people’s desire to provide reviews for familiar versus novel products. Our work suggests that WOM about familiar versus novel products helps to fulfill different personal goals in terms of expressing and learning evaluations.

Finally, our work also adds to the recommendation seeking literature, a relatively understudied form of WOM. Berger (2014) discussed but did not empirically test how identifiability of the communication channel influences people’s WOM motivations. Anonymous posting should lower people’s concern with how they present themselves. However, in an identity-linked platform, impression management motives should play a greater role (Eisingerich et al., 2015). Our research provides a direct test of this to disentangle the different motives involved when people seek recommendations.

Practical implications

Marketers are increasingly interested in motivating consumers to engage in online conversations about their products (Godes & Mayzlin, 2009). Our research helps marketers recognize which segments (based on NE scores) to target when encouraging different types of WOM in different contexts. Importantly, in Study 2, NE-expressing was a better predictor of WOM about novel products than past WOM behaviors, presumably because NE-expressing taps into a more stable trait-level tendency of opinion expression (vs. past WOM behavior which, depending on the scope of measurement, might be influenced by other situational factors, reducing its predictive power). Thus, NE-expressing could be a more effective segmentation variable than past WOM behavior. For example, marketers could target people high in expressing motivation if their goal is to elicit more online reviews. However, to encourage WOM reviews for newly released products, marketers should not only target consumers with high tendencies to provide product reviews, but those who also have demonstrated an interest in exploring new products. Finally, our work suggests that to encourage recommendation seeking online, marketers should not only target consumers with high learning motives, but also those with high expressing motives in identity-linked settings like social media.