In 2023, we found the following excerpt from an investment company on their website. We have bolded the second person pronouns (and will do so in the remainder of this article):

There is an undeniable trend in the Netherlands for companies and organizations to address their customers with the informal pronoun jij instead of the formal pronoun u. The underlying idea is that the informal pronoun is more personal and expresses solidarity and trust, whereas the formal pronoun that is traditionally used can be considered a negative politeness marker, an expression of distance (Vismans, 2013). A student once told us that, for his work at a company, he was instructed to replace all formal pronouns of address on all pages of the company’s website with their informal counterparts. We assume that the fragment in (1) is the result of that same strategy.¹ The formal version of (1) is given in (2) (translation is the same as in (1)):

The authors who have Netherlandic Dutch as their first language agree that the use of the personal pronoun jij ‘you’ and the possessive pronoun jouw ‘your’ in (1) makes the text seem overly direct and, consequently, rude and blunt. For this reason, some first speakers of Dutch would prefer to be addressed with the formal personal pronoun u ‘you’ and the formal possessive pronoun uw ‘your’, as in (2). However, the reason for this preference is not (just) that the pronouns in (1) are informal, but rather (also) that they are the unreduced, ‘strong’ versions of the informal pronouns (Gruber, 2013; 2017). A third version of the excerpt with the reduced, ‘weak’ informal personal and possessive pronouns je ‘you/your’ is given in (3) (again, the translation is the same as in (1)):

Vermaas (2002) already noted the ongoing change from the formal pronoun u to the informal pronouns jij and je, and hypothesized that it was mainly due to an increasing use of the weak informal pronoun je as a general form of address: this pronoun sounds more neutral than jij, which emphasizes informality, and the formal pronoun u, which expresses distance. “Je could therefore be seen as an intermediate form – between u and jij” (Vermaas, 2002, p. 59, our translation).

We, i.e. the authors who have Netherlandic Dutch as their first language, agree that (3) sounds more natural and neutral, much friendlier, and more polite than (1), even though the weak pronoun je is not considered a polite form, unlike the formal pronouns u and uw in (2) (cf. Vismans, 2023 for an analysis of the complex relationship between theories of address and politeness). This raises the question whether differences in perception of these pronouns go beyond the distinction between formal and informal alone. The aim of the present study is to address this question. Section 2 reviews the literature to find out whether the three types of pronouns have different effects on addressees, in particular, in the context of service advertising. The literature review yields four hypotheses about the impact of these pronouns in service ads. Section 3 reports on an experiment that tested these hypotheses. Section 4 presents the results of the experiment. Section 5 reports some intriguing additional findings that we made when performing exploratory post-hoc analyses of the data. We believe that our findings may have important implications for future studies of the effects of linguistic variables on consumers’ attitudes. Section 6 discusses the results and concludes.

In Dutch, as in many other languages, second person pronouns can be used either deictically (i.e. referring to the addressee), which is considered their “normal” use, or generically, also called “impersonally” (e.g., Siewierska, 2004; Deringer et al., 2015; de Hoop & Tarenskeen, 2015; Kluge, 2016). An example of a generic use of the English second person pronoun you is given in (4) (Whitley, 1978, p. 18):

The pronoun you in (4) is used generically, as it does not exclusively refer to the addressee. Indeed, it is unlikely that you refers to the addressee at all, as they were presumably not alive when the speaker’s great-grandad was young. The mechanism of generalizing is omnipresent in conversations, and speakers “generalize intuitively, often on the basis of their own experiences” (de Hoop & Tarenskeen, 2015, p. 165). De Hoop and Tarenskeen (2015, p. 164) claim that even when a second person pronoun arguably refers to the speaker rather than the addressee (an example from Dutch can be found in (6) below), it is considered an instance of a generic pronoun, because “the situation is presented as a generalization over people” (de Hoop & Tarenskeen, 2015, p. 164). The generic use of second person pronouns is also found in marketing communication, as illustrated by Whitley’s (1978, p. 27) example in (5):

Bolinger (1979, p. 207), in response to Whitley (1978), points out that the second person pronoun in (5) has a personalizing effect and that the use of you enables speakers “to generalize and personalize at the same time”. We assume that the generic reading of a second person pronoun corresponds to its generalizing function, and the deictic reading to its personalizing function. The line between the two readings is not always clear. Kluge (2016) shows that in the case of potential ambiguity, addressees usually do not ask for further specification, but instead rely on contextual cues to resolve the issue. Third person generic pronouns, such as one in English or men ‘one’ in Dutch, do not have the deictic reading. Generic you, however, retains a connection to the addressee simply because it is a pronoun of address (Malamud, 2012). By using a second person pronoun, the reader is invited to place themselves in someone else’s shoes, which is characteristic for the generic reading of second person pronouns (cf. Malamud, 2012; de Hoop & Tarenskeen, 2015). Generically used second person pronouns have been shown to enhance resonance between people and ideas (Orvell et al., 2020).

The question is whether all three types of second person pronouns in Dutch, je, jij, and u, can refer equally well deictically and generically, and thus have both a personalizing and generalizing function. It has been proposed that only the weak informal pronoun je can get a generic reading in Dutch (Malamud, 2006). The reason would be that generic pronouns cannot bear stress, and je is a reduced, unstressed pronoun in Dutch. Whitley (1978) similarly claims that generic you in English cannot be stressed. However, this claim about English has already been refuted by Bolinger (1979, p. 195), who shares the following example with a stressed you receiving a generic interpretation: I’ve felt the same way sometimes. It’s all right for the professor to ignore the nobodies, but when the one he ignores is yóu, that means he’s unfair. Notably, Tarenskeen’s (2010, p. 75) Dutch translation of this sentence has the strong informal pronoun jij: Ik voel me soms ook zo. Het is niet erg als de professor de sukkels negeert, maar als jíj degene bent die wordt genegeerd, dan is het niet eerlijk. Tarenskeen concludes that the strong pronoun jij in Dutch can have a generic reading, pace Malamud (2006).

Gruber (2013, p. 132) discusses some of Tarenskeen’s (2010) corpus examples that illustrate the generic reading of the strong pronoun jij, such as (6), taken from an interview with a teacher of Dutch in the Netherlands.

Gruber (2013, p. 133) confirms that both je and jij in (6) “can easily be understood as general remarks about teachers”, which would be “irreconcilable with my [Gruber’s] generalization that Dutch will always only employ the weak pronoun in generic context.” Yet, she maintains that the strong pronoun jij does not get a generic reading here, because, she claims, this reading is acceptable only if the addressee is “part of the group the pronoun is generalizing over” (Gruber, 2013, p. 133). However, the addressee in (6), the interviewer, need not be a teacher, and in fact was almost certainly not a teacher. So, there is no evidence to support Gruber’s (2013; 2017) and Malamud’s (2006) claim that the strong informal pronoun jij cannot obtain a generic reading.

A difference between je and jij is that jij is overall much less frequent than je (Tarenskeen, 2010). The same holds for the formal pronoun u in Netherlandic Dutch. De Hoop and Hogeweg (2014) annotated the readings of all types of second person pronouns in a literary novel and found that 35 % of the informal weak pronoun je obtained a generic reading. U and jij, by contrast, hardly ever obtained a generic reading in the novel. It thus appears that je receives a generic reading more easily than jij and u. This is confirmed by a small post-hoc experiment reported in Sadowski et al. (2024). Given that the generic reading generalizes by definition, this difference between address pronouns may have consequences for the effectiveness of marketing communications. Orvell et al. (2019; 2020) show that generic you carries persuasive force. Thus, generic forms may improve consumers’ perception of the prevalence of an advertised product or service on the market and increase the pressure they experience to appreciate and purchase it (see also Rimal & Real, 2003, on the impact of perceived norms and group identity on human behavior). If the message is aimed at a larger, generic group of consumers, rather than at the individual, the message may be perceived as more persuasive, and the communication may consequently become more effective. Moreover, consumers may react more negatively to a deictically used pronoun, as the message is more likely to be interpreted as a directive in that case (Zemack-Rugar et al., 2017; Sadowski et al., 2024). We thus expect je to be more effective than jij and u in marketing communications (H1).

Research has found different or ambiguous effects of Netherlandic Dutch formal and informal pronouns of address in different communicative settings. On the one hand, Leung et al. (2023) and Schoenmakers et al. (2024) found a positive effect for the informal pronouns of address compared to the formal pronoun in a product advertising context. Moreover, Sadowski et al. (2024) found higher ratings for the informal pronouns of address in donation appeals. On the other hand, van Zalk and Jansen (2004), Jansen and Janssen (2005), and de Hoop et al. (2023) found an overall slightly positive effect of the formal pronoun in different types of text.

Importantly, van Zalk and Jansen (2004) and Jansen and Janssen (2005) compared the formal pronoun u to the weak informal pronoun je in their studies, while de Hoop et al. (2023) chose to compare the formal pronoun u to the strong informal pronoun jij, “because the reduced form is more neutral, whereas the full form puts more emphasis on the informal character” (de Hoop et al., 2023, p. 100). While Schoenmakers et al. (2024) also compared the strong informal pronoun of address to the formal one in a product adverting context, Sadowski et al. (2024) used both the strong and weak informal pronoun in one and the same slogan and compared them to the formal one.

These studies demonstrate that the specific context of pronoun use is crucial. In our study, we focus on service ads. Service advertising differs from product advertising, but they are also quite similar in terms of what makes them persuasive (Butler & Abernethy, 1994). The question is whether a formal or informal pronoun will lead to a higher rating for the service in question, which could also depend on the type of service. Traditionally, a difference is made between utilitarian and hedonic services, where the former are functional (e.g., financial services or healthcare) and the latter more related to customer experience, having fun (e.g., leisure or entertainment industries). Barcelos et al. (2018) argue that for hedonic services, it is better to use a human instead of the more traditional corporate voice in interactions with customers on social media. In contrast, for utilitarian services such as financial consulting or medical services, it is better to use a corporate than a human voice and keep more distance when interacting with customers. This could mean that in hedonic services marketing, informal pronouns of address, marking closeness and solidarity, would lead to higher evaluation (attitude towards the ad and service, and purchase intention) than formal pronouns, marking respect and distance (H2), and vice versa for utilitarian services marketing (H3).

Additionally, we assume that involvement plays a role in people’s appreciation of services advertising. Jansen and Janssen (2005) found that higher ratings of persuasive texts were given by readers who were more involved in the sense that they were interested in the topic of the text and agreed with the advice given in it. Marketing research has shown that the interaction between attitude towards an ad and product involvement can have a positive impact on brand image (e.g., Nagar, 2015). Schoenmakers et al. (2024) also measured participants’ product involvement in their study on product advertising. The results showed that participants who were more involved with a product were more appreciative of it as well as of its ad, and they had a higher purchase intention compared to less involved participants. Therefore, we expect higher ratings when participants feel more involved with the services advertised (H4).

We conducted a questionnaire in Qualtrics to test the four hypotheses put forward in Section 2. The study was approved by the Ethics Assessment Committee Humanities of Radboud University Nijmegen (ETC-GW number 2021-9598).

This section reports the results of statistical analyses based on the four experimental service ads: utilitarian (energy supplier and health insurance) and hedonic (wellness and cinema subscription). The dataset contained 10,248 response scores.

The model included all datapoints and had a response variable with five categories representing the participants’ Evaluation: “Minimum”, “Low”, “Middle”, “High”, and “Maximum” (see above). We used weakly informative, “liberal” Cauchy priors with center 0 and scale 2.5 for the fixed effects (Gelman et al., 2008), and the default priors for the other parameters provided in the brms package.

We began the process of model selection by including random intercepts for the individual participants and an interaction between Scale and Ad as random effects (cf. de Hoop et al., 2023), and all predictors (Pronoun, Age, Gender, Education, Category, and Involvement). Using the criterion WAIC, we estimated whether the initial model could be improved by incorporating a more complex random effects structure. We found that the model was substantially better when we added the random intercepts for Category nested in participants (Participant/Category). Moreover, we added two random slopes for the interaction between Scale and Ad: one for the Involvement scores and the other for the participants’ Age.

As far as interactions are concerned, we tested the theoretically important interaction between Pronoun and Category. We also used WAIC to identify other useful interactions, which led to the addition of an interaction between Involvement and Category. The regression terms are displayed in Table 1.

The coefficient estimates are on the logit scale. A positive estimate means that this regression term improves Evaluation, whereas a negative one shows that this term makes it worse. The boundaries of the 95 % credible intervals represent 2.5 % and 97.5 % quantiles of the posterior distributions. Finally, the posterior probabilities of b > 0 in the rightmost column show how consistent the posterior distributions are with a positive effect on Evaluation. If a distribution has near 100 % posterior probability of the effect being greater than 0, this means that we can be confident that this variable has a positive effect. If it is close to 0, we can believe that the effect is negative.

The intercepts are in fact “thresholds”, showing the log-odds of having a specific score category if all the predictors are at their base levels or equal to zero. The first intercept [1] shows the log-odds of being in category Minimum versus all other score categories. The second intercept [2] represents the log-odds of being in categories Minimum or Low versus the higher categories, and so on.

As far as the demographic variables are concerned, participants who only attended secondary school or senior secondary vocational college gave overall slightly higher scores compared to the reference level (university education). The posterior probabilities of these positive effects are 93.8 % and 89.5 %, respectively, which can be interpreted as relatively weak evidence. The effects of Age and Gender are close to zero and not convincing, judging from the posterior probabilities and credible intervals.

The other variables, Pronoun, Category, and Involvement, which are relevant for our theoretical hypotheses, are involved in interactions. Note that we also tested a three-way interaction between these variables, but this interaction was not supported by the data. Let us begin with the interaction between Pronoun and Category. Figure 3 displays the predicted probabilities of each of the five binned responses for all combinations of Pronoun and Category. The plot shows that the predicted probabilities of the Minimum, Maximum, and Middle Evaluation scores are similarly low for all pronouns and categories. The main differences are observed for the Low and High scores. The utilitarian service ads have higher probability to get low evaluations and lower probability to get high evaluations than the hedonic ones. Notably, all pronouns have very similar probabilities if a utilitarian service is advertised. However, the pronouns display substantial differences for the hedonic service ads, where the probability of a low score is the lowest if je is used and highest if u is used, with jij in the middle. As for the probabilities of a high score, the picture is reversed: the probability is the lowest if u is used and the highest if je is used, with jij in between.

Because of the interaction between Pronoun and Category, we need to compare the effects of the forms for each Category separately. For the hedonic services, je is more likely to have higher Evaluation scores than jij. The log-odds difference is positive: the median d = 0.317, with the 95 % Highest Posterior Density (HPD) interval between –0.113 and 0.729. This means that the evidence for this difference is not strong. By contrast, the difference between je and u in the hedonic service ads is positive and strongly supported by the data: d = 0.452 with the 95 % HPD from 0.024 to 0.846. The difference between jij and u is small and not credible, though: d = 0.14, the 95 % HPD from –0.303 to 0.58. As for the utilitarian service ads, the differences between the pronouns are very close to zero: for je vs. jij, d = –0.08 (95 % HPD –0.506 to 0.357), for jij vs. u, d = –0.053 (95 % HPD –0.485 to 0.361), and for je vs. u, d = 0.027 (95 % HPD –0.421 to 0.47).

The interaction between Involvement and Category is displayed in Figure 4. Higher involvement (transformed to fall in the range from 0 to 1) leads to lower chances of the ad receiving minimum and low scores and higher chances of it getting high and maximum Evaluation scores. The effect of Category differs depending on the level of Involvement. In comparison with the hedonic service ads, the utilitarian service ads are overall more likely to receive lower Evaluation scores and less likely to receive higher Evaluation scores. But if Involvement is very low, there is no difference between the two categories of ads. As Involvement increases, the difference becomes more pronounced.

Because the effects of the demographic variables (Gender, Age, Education) were not supported by the data, or supported only weakly, we also fitted a model without these variables. This enabled us to include data from all participants. We detected no substantial differences in the effects of the remaining variables. The results can be found in the document “Additional tests” in the online repository.

Below we summarize the results of our hypothesis testing. These will be further discussed in Section 6.

Our first hypothesis was that participants will respond more positively to service ads with je than with jij or u, due to the generic reading that je can obtain more easily than jij and u: je > jij, u. This hypothesis was supported by the categorical data (overall preferences), but only for the hedonic services, and only for je compared to u, not for je compared to jij. It might be that jij more than u has the potential to generalize, although this is not what Sadowski et al. (2024) found in their small post-hoc study concerning the differences between jij, je, and u. They showed that u acquires a generic reading more easily than jij, probably because grammatically singular u can also be used in the plural to address a group of addressees. However, the informal pronouns are by default used in Dutch ads and have been shown to contribute to a more positive evaluation of the ads (Leung et al., 2023; Schoenmakers et al., 2024). Therefore, it might be the combination of informality with the generalizing effect that distinguishes je (but not jij) from u in two respects, leading to a more positive response, albeit only for hedonic services.

That informality plays a role in hedonic service ads would be in line with our second hypothesis, which was that participants would respond more positively to hedonic service ads featuring an informal pronoun of address (jij or je) than to the formal pronoun u (jij, je > u), although only the contrast between je and u is sufficiently reliable. However, the absence of an interaction between the type of pronoun and the type of service (hedonic or utilitarian) suggests that both the second and third hypotheses, about hedonic and utilitarian service ads, respectively, are not confirmed by the data. The question is how to explain these findings. We assumed that the generalizing rather than the personalizing function of second person pronouns would be responsible for their success in a services marketing context. We find that it might rather be the combination of generalization and informality that makes the weak informal pronoun je preferred to the formal pronoun u in service advertising, because the difference between je and jij is not significant in any of our models. The difference between je and u is only found for hedonic service ads, for which we hypothesized that participants would prefer an informal pronoun of address to a formal one. The fact that no differences are found for utilitarian service ads may be because utilitarian service ads are subject to two conflicting constraints: the first hypothesis was that je would lead to a more positive response than jij and u in service ads, but the third hypothesis was that u would be more appreciated than the informal pronouns jij and je in utilitarian service ads.

The fourth hypothesis about the positive effect of involvement with a certain service is supported by the categorical and numeric data. The effect is stronger for hedonic services than for utilitarian ones. Overall, involvement scores are lower for utilitarian services, resulting in a lower rating regardless of pronoun type.

To summarize, the different types of pronouns seem to have little effect on the overall evaluation of the service ads, as our model of five coarse-grained response categories demonstrates. However, the picture changes if we zoom in on specific regions of the Evaluation scores, as reported in Section 5. When participants are more positive in their evaluation of the service ad (the numeric high scores data), they give higher scores to the ads with the weak informal pronoun je than with jij and especially u. The order of preference is: je > jij > u. In contrast, when participants are less positive in their evaluation (the numeric low scores data), the order of preference is reversed. In that case, there seems to be a preference for the formal pronoun u, which slightly boosts the scores, whereas the weak informal pronoun je has the lowest scores, and the strong informal pronoun jij is again in between: u > jij > je. Note, however, that the evidence supporting this reverse effect is weaker than in the case of the dataset with high scores. Still, it is notable that the regression coefficients reveal opposing patterns in the datasets with high and low scores.

Furthermore, our post-hoc analyses allowed us to detect some effects of the demographic variables. When we zoomed in on the higher Evaluation scores, we found that older participants and those with lower education tended to give (even) higher scores than younger and more educated participants. When we focused on the lower Evaluation scores, older and male participants tended to give (even) lower scores than younger and female participants. These findings suggest that older customers give more extreme (positive and negative) reactions to service ads. More research is needed to corroborate these results.

Schoenmakers et al. (2024) concluded that the use of a strong informal pronoun jij in a product ad is more effective than the formal pronoun u, which is better avoided. This cannot be generalized to all types of ads, however. For hedonic services, the use of the weak informal pronoun je seems the best choice, probably because this pronoun combines a generalizing function with informality. However, when advertisers want a specific ad to appeal to potential customers who do not like these services or ads in the first place, they might be better off choosing the formal pronoun u, which can mitigate the negative attitude and help to achieve a more positive evaluation of the service. Thus, our study shows for the first time that the effect of the pronoun choice depends on whether language users generally like or dislike the service ad in question. If they like it, the use of an informal pronoun enhances their appreciation, but if they do not like it, the use of the informal pronoun will only reduce their appreciation. This discovery, which emerged as a result of exploratory analyses, has important implications both for marketing practices and for future investigations of the effects of linguistic variables on consumers’ attitudes towards products or services. The effect of language use can be moderated by the overall appreciation of service or product ads. We hope that future studies will find appropriate instruments for capturing these complex relationships between linguistic and non-linguistic factors.