1
(I’m) Happy to Help (You):
The Impact of Personal Pronoun Use in Customer-Firm Interactions
2
In responding to customer questions or complaints, should firm agents indeed “put the customer
first”? We examine this question at the level of the language used in marketing sales and service
interactions, focusing on firm agents’ use of personal pronouns (e.g., “I”, “you”, “we”).
Customer orientation theory implies that firms will benefit if agents emphasize how “you” (the
customer) is served by “us” (or “we” the firm) in this setting. Field and lab studies reveal that
firm agents do show this linguistic manifestation of a customer orientation, and that this pronoun
use pattern is endorsed by managers, firm agents, and consumers. However, we demonstrate that
this pattern is sub-optimal. We report two distinct effects. First, we find null effects on
satisfaction, purchase intentions, and actual purchase volume with increased use of “you”
pronouns (emphasizing the customer). Second, we report positive effects on these same
outcomes with increased use of “I” pronouns by the firm agent (singular self-references) relative
to (a) no self-referencing, (b) less self-referencing, and (c) “we” use (plural self-references in
which the agent refers to themselves as part of the firm). These two effects are robust to multiple
covariates. We build on theory from linguistic psychology to predict and demonstrate that
perceptions of agent empathy and agency drive the positive “I” effect. These findings offer
valuable implications for marketers and enhance our conceptual understanding of how subtle
language variations impact social perceptions.
Keywords: language, customer orientation, social perception, personal selling, customer service
3
A central role of marketers is to manage the “speaking terms” of the relationship between
firms and their current or prospective customers (Duncan and Moriarty 1998; Vargo and Lush
2004). This dialogue spans everything from advertising and public relations to customer
relationship marketing, and occurs via traditional sales or service interactions with customers in
stores, as well as through an array of technology-mediated interactions (e.g., phone, Internet).
A vast literature has examined how marketers might optimize these customer-firm
interactions, with a focus on managing firm agent (i.e., front line service employee) actions and
relationships with customers. For example, customer satisfaction, purchase intentions, and sales
can be positively influenced by firm agent behaviors such as adaptive selling (Weitz, Sujan, and
Sujan 1986), mimicry (Tanner et al. 2008), flattery (Chan and Sengupta 2010), adopting the
firm’s customer orientation (Brady and Cronin 2001; Marinova, Ye and Singh 2008; Singh and
Sirdeshmukh 2000), and by developing customer loyalty to specific salespeople (vs. the firm)
(Palmatier, Scheer, and Steenkamp 2007). Similarly, firm agents can better resolve customers’
inquiries or complaints by responding quickly, apologizing, or offering compensation (e.g.,
Davidow 2003; Rust and Chung 2006; Zeithaml, Berry, and Parasuraman 1996). Customer-firm
interactions are also enhanced when firm agents communicate effectively through listening, turn
taking, and asking questions (Ingrain, Schwepker, and Hutson 1992; Schuster and Danes 1986).
Rather than focusing on actions, we ask what impact a firm agent’s specific words might
have in their interactions with customers. While this topic has been neglected in the literature,
marketers appear to recognize the importance of a firm agent’s words. Apple has “stop words”
that employees are prohibited from using, as well as specific scripts and phrases that employees
are encouraged to use with customers (Chen 2011). In contrast, Zappos eschews scripts for its
agents, instead allowing each to create a “personal emotional connection” when conversing with
customers (Hsieh 2010). Despite their different philosophies, both firms invest in managing firm
4
agent language use in customer interactions. In this vein, we offer an empirical examination of
how the language used by firm agents influences customer attitudes and behaviors.
To do so, we focus on personal pronouns; specifically, second-person singular (you, your,
yours), first-person plural (we, us, our), and first-person singular (I, me, mine) pronouns.1
Personal pronouns are important psychosocial indicators (Chung and Pennebaker 2007;
Pennebaker 2011) and have been conceptualized as linguistic signals of the speaker’s relative
emphasis on the self versus other in dyadic interactions (Goffman 1981; Fahnestock 2011). For
example, take a firm agent who wishes to express her delight in acting on the customer’s behalf.
Her speech may emphasize the customer who is the object of her attention (“Happy to help
you.”), the agent as part of the larger firm entity for which she works (“We are happy to help.”),
or the agent herself (“I am happy to help.”). Of course, the firm agent may also embed references
to both the agent and the customer in this utterance (“We are happy to help you.”) or may
exclude both parties (“Happy to help.”). While each of these variations convey the same situated
meaning, they subtly modify the emphasis on either the actor (the party who is happy to help) or
the recipient of action (the party who is being helped) by explicitly referencing specific
interaction parties.
Through a combination of surveys, experiments, and field data, this research examines
manager and customer beliefs regarding firm agent personal pronoun use, the actual use of such
pronouns by firm agents, and the impact of subtle variations in firm agent pronoun use on
customer attitudes and behaviors. We predict that the prevailing “customer comes first”
philosophy in the marketplace (i.e., customer orientation; Basch 2003; Saxe and Weitz 1982)
might lead firm agents to linguistically emphasize “you” (the customer) and “we” (the firm
For brevity, we refer to these personal pronoun categories as “I”, “you”, and “we” pronouns throughout the paper.
The category examples shown here are not exhaustive; for a full list, please see Appendix Table A1.
1
5
entity) in written customer-firm interactions, phenomena we indeed observe in over 1,000 such
interactions sampled from over 40 different firms in the field. Corroborating these findings, two
pilot studies reveal that managers and consumers believe that this pronoun emphasis is optimal.
However, subsequent experiments and field data studies show that—contrary to manager and
consumer intuitions—this pronoun use pattern is sub-optimal: firm agents’ linguistic emphasis
referencing the customer (“you”) and the firm entity (“we”) in customer-firm interactions offers
little benefit to firms. Instead, the present research reveals how and why firm agents who
linguistically emphasize themselves (“I”) generate improved customer attitudes and purchase
behaviors.
This research makes six main contributions. First, to our knowledge, this paper is the first
to explore the impact of subtle differences in the specific words used by marketing (sales or
service) agents in their interactions with customers. Second, while the linguistic psychology
literature has shown correlations between “I” pronoun use and the speaker’s own social status or
psychological state (Pennebaker 2011), we theorize and demonstrate that a speaker’s (e.g., firm
agent’s) use of “I” pronouns impacts the listener’s (e.g., the customer’s) perceptions of the
speaker. Third, we show that these social perceptions go on to affect the customer’s attitudes and
actual behaviors towards the firm entity that the agent represents. Fourth, in contrast to research
reporting positive effects of a particular kind of “we” pronoun use on customer-brand
relationships (Sela, Wheeler, and Sarial-Abi 2012), we demonstrate an important marketing
context in which increased use of self-referencing “I” pronouns may be beneficial. Fifth, we
offer a theory-driven explanation, complete with evidence for the psychological process, for the
source of these effects. Specifically, although some prior work suggests that a more self-focused
(self-referencing) firm agent should have negative effects on social perceptions, we show that
6
heavier self-referencing “I” pronoun use in our setting can lead to heightened perceptions that the
firm agent feels (empathy) and acts (agency) on the customer’s behalf.
Finally, beyond the conceptual and empirical contributions noted above, we offer
substantive insights that marketers can easily implement to enhance their selling and service
interactions with customers. In doing so, this research challenges conventional marketing
wisdom by showing how and why the linguistic manifestation of a customer orientation may not
be optimal for firms.
MORE THAN WORDS: THE IMPORTANT ROLE OF PERSONAL PRONOUNS
A growing body of marketing research demonstrates the importance of words. Subtle
variations in language use impact consumers’ self-control and motivation (Patrick and Hagtvedt
2012), senders and receivers of word of mouth (Moore 2012; Schellekens, Verlegh, and Smidts
2010), and consumers’ responses to advertising (Kronrod, Grinstein, and Wathieu 2011; Sela et
al. 2012).
Building on this work, we examine personal pronouns, which play an important role in
language as the “currency of social processes” (Pennebaker 2011). Beyond revealing a person’s
psychosocial state and intentions, the use of personal pronouns reflects the linguistic and
substantive contexts in which these pronouns appear (Fahnestock 2011). For example, “you”
pronouns are especially likely to occur when a speaker uses passive voice to draw attention to the
person or thing being acted upon (e.g., passive voice “Your order will be shipped soon” vs. active
voice “I will ship the order soon”; Johnson-Laird 1968; Penelope 1990). While a large literature
has examined the psychosocial meaning of personal pronoun use independent of linguistic and
substantive context (Chung and Pennebaker 2007; Pennebaker 2011), to consider their role in the
7
generation of social perceptions it is important to consider the specific context in which these
influential markers of social roles are used—in our case, customer-firm interactions.
We build on work examining the structure of social interactions to shed light on this issue.
Personal pronouns are said to contribute to the establishment of an interaction’s “participation
framework” (Goffman 1981), indicating the roles—and the relative importance—of the parties in
a given interaction (Fahnestock 2011; Goffman 1981; Schlenker and Weigold 1992). In dyadic
interactions, personal pronoun use helps establish this framework by identifying the principal
actor (typically, the grammatical subject; e.g., “I understand you.”) and the recipient of action
(the grammatical object or the object’s possessions; e.g., “I understand you.” or “I understand
your question.”). As alluded to earlier, in a given participation framework, the relative presence
or absence of referring pronouns may also serve to emphasize one or both parties (Fahnestock
2011; Goffman 1981). For example, relative to the utterance “I understand you,” which
references both parties, saying “You’re understood” emphasizes the grammatical object (“you”)
in the interaction, rather than the subject (“I”).
We predict that the linguistic emphasis (or de-emphasis) of one party in an interaction (the
firm agent) can have important effects on the attitudes and behaviors of the second party (the
customer). To set a foundation for this prediction, we must first establish the dominant
participation framework in our research setting. In a customer oriented interaction, both
linguistic (Fahnestock 2011, p. 147) and normative expectations dictate that the agent should
treat the customer as the recipient of the agent’s thoughts or actions. Normatively, the sales or
service agent’s role in a customer oriented firm is to be the frontline embodiment of the firm’s
focus on satisfying the customer’s needs (Basch 2003; Zablah et al. 2012). In this participation
framework, the firm or its agent(s) should most often be the grammatical subject (actor) and the
consumer or their purchase should most often be the grammatical object (recipient of action) of
8
the firm agent’s response. For example, a customer who requests assistance by saying, “I can’t
find a size six in this shirt.” is obviously more likely to hear a firm agent reply with, “We can
find that for you.” or, “Would you like me to find that for you?” than “You can find that for me.”
To verify whether this participation framework is indeed dominant in our customer-firm
interaction setting, we analyzed field data. Using real firm agent emails received from 40
different firms in reply to a bogus customer email (Study 2), two independent judges coded
whether each reference to the firm agent and the customer treated them as the grammatical
subject (actor) or as the grammatical object (recipient of action) in the dyad. A count of subject
references revealed that the firm agent was referred to as the subject (actor) more frequently than
the customer in 98% of the emails (39 of 40 emails). The firm agent referred to the self as the
subject (actor) in 89% of all first-person references, while 74% of all second-person references to
customers treated them as the object (recipient of action). Overall, the agent was referred to as
the grammatical subject 220 times versus only 39 times for the customer, a ratio of nearly 6 to 1.2
Given normative expectations in the managerial literature (Basch 2003; Zablah et al. 2012)
and the corroborating field data above, our studies focus on a typical customer sales and service
setting in which the firm agent is the actor and the customer is the recipient of that action. Within
this participation framework, we predict that the specific personal pronouns that firm agents use
and emphasize will be relevant to customer outcomes. But what pronoun emphasis is optimal?
We ask two separate questions. First, should firm agents emphasize “you” pronouns (e.g.,
“Happy to help you!” vs. “Happy to help!”) referencing the customer as the recipient of a
particular action? Second, should firm agents use “we” (e.g., “We’re happy to help!”) or “I”
2
A reversal of this grammatic relationship (i.e., the minority of cases when the firm agent spoke of the customer as
the subject and the firm or agent as the object) at the clause level most commonly entailed an invitation to the
customer for agent assistance. For example, “If you have any questions, please reply via email or call me at (…)”.
9
(e.g., “I’m happy to help!”) pronouns to emphasize themselves as the actor either plurally (as part
of a firm entity) or personally (as an individual)?
CUSTOMER ORIENTED INTERACTIONS
Customer orientation theory suggests that to maximize customer and firm outcomes, the
firm should prioritize the needs and wants of the customer. At the frontline of customer sales or
service interactions, this entails a heightened demonstration of the firm agent’s “concern for
others” (the customer) and downplayed “concern for the self” (the firm agent(s); Saxe and Weitz
1982, p. 344). A strong customer orientation is expected to enhance customer attitudes and
intentions (Brady and Cronin 2001) and has been empirically linked to increased sales
(Homburg, Hoyer, and Fassnacht 2002) and firm performance (Ramani and Kumar 2008). A
flood of managerial publications (e.g., Basch 2003; Evenson 2011) similarly advocate a strong
customer orientation. That this philosophy has attained the status of conventional marketing
wisdom is attested to, for example, by the popularity of Ray Kroc’s famous advice, “Always put
the customer first.”3
While customer orientation theory and the managerial literature suggest that firm agents
should put the customer first in everything they do, could managers and consumers believe that
what firm agents say should similarly follow this rule? A variety of managerial sources suggest
that agents should indeed speak in a manner indicating that the firm entity (“we”) is motivated to
satisfy “you” the customer. For example, Amazon’s bestselling book on language use in
customer service (Bacal 2011) consistently adds “you” (the customer) and “we” (the firm)
pronouns to its “exemplar phrases,” with relatively few “I” (agent) references. A handbook for
A Google Books search for books containing the exact phrase “Put the customer first” identified over 4,300 such
publications as of March 1, 2016.
3
10
customer service agent email communications presents examples recommending the use of
“you” and “we” pronouns to emphasize the customer and firm as ideal, rather than leaving these
two parties implicit (Rudick and O’Flahavan 2002, p. 75). A popular online writing blogger
suggests email service agents should “use lots of pronouns,”—in particular, “you” and “we” but
not “I” pronouns (Kurtz 2015).
Thus, the customer orientation concept and the above examples suggest the explicit use of
“you” (the customer) pronouns in customer-firm interactions, as well as the use of “we” (the
firm) pronouns over self-referencing “I” pronouns. Still, do managers actually believe that firm
agents should explicitly reference the customer (“you”) when speaking to them? Also, do
managers believe that firm agents should emphasize themselves as a party acting on behalf of a
firm entity (“we”) rather than his or her self individually (“I”)? Do frontline firm agents and
consumers share these beliefs? To address these questions, we conducted a pilot study of lay
beliefs about firm agent personal pronoun use when interacting with customers.
Pilot Study
American participants (N = 507) in a paid online panel were asked to imagine themselves
as a customer sales/service agent for a fictional online retailer called Shopsite.com. They were
presented with a customer email that asked about international shipping (see the Appendix for
full stimuli), and asked which of two responses would be most appropriate to say in their reply.
For half of the participants (N = 253), the only difference between the responses was whether the
firm agent emphasized the customer (“you” pronouns), or instead left references to the customer
implicit (no “you” pronouns). For the remainder (N = 254), the only difference was whether the
firm agent referred to himself or herself as part of the firm entity (“we” pronouns) or as an
individual (“I” pronouns). This design enabled independent tests of our expectations in regards to
11
lay theories of pronoun use: (1) that “you” pronouns referring to the customer would be preferred
over no explicit references to the customer (no “you”), and (2) that “we” pronouns would be
preferred over “I” pronouns when the agent referred to the firm “actor” in the interaction.4
Finally, we asked participants to indicate whether they currently or previously had worked in
paid employment in a sales or service role or as a manager of such individuals. This allowed us
to isolate what actual managers and firm agents believed about pronoun use.
Participants asked to decide whether the customer should be emphasized in their response
(“you” vs. no “you”) strongly preferred emphasizing “you” the customer in their reply (88.5%
vs. chance, 2(1, N = 253) = 1311.21, p < .001). The strong preference to explicitly reference
“you” the customer also held when analysis was limited to participants with customer sales or
service experience (87.4% vs. chance, 2(1, N = 175) = 780.05, p < .001) or managers of sales
and service employees (83.8% vs. chance, 2(1, N = 68) = 37.12, p < .001).
For participants in the condition manipulating firm agent self-referencing (“we” vs. “I”), a
strong majority preferred referring to themselves in the first person plural as “we” (85.8% vs.
chance, 2(1, N = 254) = 920.11, p < .001) rather than in the singular “I”. This preference for
“we” over “I” pronouns also held when analysis was limited to participants with customer sales
or service experience (88.3% vs. chance, 2(1, N = 179) = 118.79, p < .001) or managers of sales
and service employees (91.8% vs. chance, 2(1, N = 67) = 55.54, p < .001).
This pilot study suggests that, consistent with managerial theory, managers and consumers
believe that firm agents should (1) explicitly emphasize the customer (“you”), and (2) describe
themselves as part of the firm entity (“we”) rather than as an individual (“I”) when responding to
We also included a statement that successfully manipulated the perceived difficulty of addressing the customer’s
inquiry (high vs. low difficulty), but this factor failed to impact results in either the “you” versus no “you” (2(1) =
.01, p = .93) or “we” versus “I” conditions (2(1) = 2.30, p = .13).
4
12
a customer.5 However, it is not clear whether this explicit emphasis on “you” and “we” pronouns
is optimal. Nor is it clear how instead emphasizing “I” pronouns might affect customer and firm
outcomes. Below, we consider the potential impact of a firm agent’s emphasis on each of these
pronoun categories on customer attitudes and behavior. In each case, while the effects of a
speaker’s personal pronoun use on the listener are largely untested, the existing (largely
correlational) literature points to possible benefits and pitfalls for the use of each. We review this
work below and briefly predict how it is likely to apply in our interaction context.
POTENTIAL EFFECTS OF PERSONAL PRONOUN USE
IN CUSTOMER-FIRM INTERACTIONS
Firm Agent Use of “You” Pronouns
A firm agent emphasis on “you” pronouns (e.g., “How do those pants fit you?” vs. “How
do those pants fit?”) as prescribed by customer orientation theory, managerial sources, and our
pilot study might signal interest and involvement in the customer’s issues. Some prior research
supports this theorizing. Specifically, language higher (vs. lower) in “you” pronouns is correlated
with that the speaker’s focus on satisfying the interaction partner (Chung and Pennebaker 2007;
Pennebaker 2011). Individuals who are motivated to impression manage by accommodating
others’ desires use “you” pronouns at a higher rate (Ickes, Reidhead, and Patterson 1986).
Finally, increased use of “you” pronouns is thought to contribute to the perception that the
speaker views a situation from the audience’s point of view (Campbell, Riley, and Parker 1990).
This literature suggests that, consistent with managerial and lay beliefs, the more a firm agent
uses “you” pronouns to refer to the customer as the recipient of action in a customer-firm
interaction, the more positive an impact this might have on customer attitudes and behavior.
5
An additional pilot study using a different approach fully corroborates these results. See Pilot Study 2 in the Web
Appendix.
13
However, a firm agent’s use of “you” pronouns could also be detrimental. Heavy use of
“you” pronouns might be perceived as the firm agent abdicating responsibility or shifting it to
the customer (Bitner 1990; Buttny 1993; Simmons, Gordon, and Chambless 2005). This
inference might be made, for example, if an agent says, “I’m sorry your product isn’t working”
rather than, “I’m sorry the product isn’t working.” Thus, “you” pronoun use could be perceived
as accusatory and have a negative impact on customer attitudes and intentions towards the firm.
Firm Agent Use of “We” versus “I” Pronouns
The use of “we” pronouns could similarly have either positive or negative effects on
customers. “We” pronouns that exclude the audience (i.e., the majority of customer-firm
interactions we observe)6 may serve to relationally distance the firm agent from the customer
(Fitzsimons and Kay 2004; Sela et al. 2012) by making it seem like the agent serves a large,
impersonal entity (e.g., “We have a policy against that.”). Alternatively, the exclusive “we” may
signal an employee that strongly identifies as part of the larger firm entity, making it seem like
there is a veritable army rather than a single soldier fighting to serve the customer (e.g., “We can
make that happen.”; Pennebaker 2011, p. 227).
Conceptual support for predictions regarding the effect of a firm agent’s use of “I”
pronouns is also equivocal. On one hand, use of self-referencing “I” pronouns by the firm agent
might signal neglect of the customer (e.g., “I have a pair of those pants!”) or self-centered
motivations (e.g., “I think those pants are really stylish!”). Overall, “I” pronouns are said to
“We” pronouns can encompass the audience only (i.e., royal we; “We must eat our broccoli, children.”), can
include the speaker and audience (inclusive we, “We’ll figure this out together.”), or can exclude the audience
(exclusive we, “We’re happy to help you.”) (Inigo-Mora 2004; Wales 1996). Independent judges found that the
exclusive “we” was used by firm agents 100% of the time in Study 2 field data and 98% of the time in a sample (N =
100) of the Study 4 field data. Our conceptual and empirical focus is therefore on the exclusive “we,” which
distinguishes the present research from prior work examining the relationship-implying use of the inclusive “we”
(Fitzsimons and Kay 2004; Sela et al. 2012).
6
14
connote self-focus (Pennebaker 2011), and are used more when people are motivated to draw
attention to the self for purposes such as self-enhancement (Barasch and Berger 2014; Packard
and Wooten 2013; Rude, Gortner, and Pennebaker 2004). “I” pronouns are also used more
frequently by people high in traits linked to competitive self-interest (e.g., Machiavellianism;
Ickes, Reidhead and Pattterson 1986). This work suggests that “I” pronoun use by firm agents
during customer-firm interactions may not be beneficial for customer attitudes and behavior.
On the other hand, some research suggests that “I” pronoun use by firm agents might be
perceived positively. In addition to egotistic self-centeredness, “I” pronouns are used more when
a speaker is actively attempting to understand their interaction partner, which might signal
empathy to customers (e.g., “I hope you’re satisfied.” vs. “Hope you’re satisfied.”; Ickes et al.
1990; Wales 1996). If so, this signaling of a firm agent’s genuine benevolence and concern
should lead to enhanced customer satisfaction and repurchase intentions (Singh and Sirdeshmukh
2000).
“I” pronoun use has also been linked to the speaker’s personal involvement in a situation
(Scherwitz, Berton, and Leventhal 1978). If this behavior also signals involvement to the
listener, “I” pronoun use may signal agency to customers, such that the firm agent appears
willing and able to take autonomous responsibility or action on the customer’s behalf (e.g., “I
will take care of that right now.”; Ahearn 2001; Chung and Pennebaker 2007; Kashima and
Kashima 1998; Marinova et al. 2008), leading to more satisfied and loyal customers (Singh and
Sirdeshmukh 2000). Such evidence of a firm agent’s personal involvement may be particularly
impactful in our setting given the trend towards increasingly depersonalized customer-firm
interactions (van Osselaer 2016). Thus, we predict that increased use of “I” pronouns by the firm
agent may signal emotional (empathy) and behavioral (agency) engagement with the customer’s
inquiry or complaint, and could have a beneficial impact on customer attitudes and behaviors.
15
To sum, while managerial and lay beliefs recommend an emphasis on how “we” (the
firm”) can serve “you” (the customer), the academic literature on the speaker’s state of mind
when using personal pronouns is more equivocal, allowing us to make varying predictions on the
potential impact of a firm agent’s use of “you” pronouns when speaking to the customer as the
object of the firm agent’s attention.7 The literature also supports contrasting predictions on
whether firm agents should emphasize references to the firm or themselves personally as the
subject of the conversation by using more “we” or more “I” pronouns in their interactions with
customers. In addition, while our pilot study and literature review reveal some lay beliefs
regarding pronoun use in this setting, the relative prevalence of “real” firm agent pronoun use in
the marketplace is not clear. The present research sheds light on these issues.
SUMMARY OF STUDIES
To this point, an analysis of field data found that the dominant participation framework in
our customer-firm interaction setting entails a firm or firm agent referring to themselves as the
actor (“we” or “I” as grammatical subject) and to the customer as the recipient of action (“you”
as grammatical object). Further, a pilot study revealed that managers, firm agents and consumers
believe that the agent should optimally emphasize “you” the customer and “we” the firm when
responding to a customer’s inquiry.
In four subsequent studies, we examine (a) firm agent personal pronoun use in practice and
(b) how pronoun use by firm agents impacts consumer attitudes and behaviors.8 Consistent with
7
As discussed earlier, the context and relevant participation framework in which pronoun use occurs should play a
role in the direction of pronoun effects. The present research and the literature in which it is grounded address these
two factors by using a variety of controlled stimuli in lab studies, and through natural heterogeneity in language use
as well as statistical covariates controlling for context effects in field data.
8
We offer an additional study (Study 5) in the Web Appendix as further evidence of the robustness of these effects
across a variety of interaction stimuli.
16
our pilot study findings regarding customers’, managers’, and firm agents’ beliefs about what
should happen in the field, Study 1 uses field data from 40 firms to confirm that firm agents do
tend to use “you” (the customer) and “we” (the firm entity), while downplaying references to
“me” (the firm agent) in their interactions with customers.
Study 2 provides an initial causal test of the positive consequences for customer
satisfaction and purchase intentions when firm agents use self-referencing “I” pronouns rather
than referring to themselves as part of the firm entity (“we”).
Study 3 replicates this effect, and demonstrates the psychological processes through which
the positive effect of “I” pronoun use occurs (perceived empathy and agency). This study also
independently demonstrates a null effect for “you” pronoun references to the customer relative to
no “you” pronoun use. Finally, Study 3 accounts for the broader context of pronoun use by
testing a sentence-level factor that may moderate customer satisfaction and purchase intentions:
active versus passive voice, speaking to the generality of our findings.
Our final study contributes evidence of external validity for the impact of pronoun use on
real customer behavior, rather than attitudes. Study 4 uses field data—over 1,000 real customerfirm email interactions—to show a positive link between firm agents’ use of self-referencing “I”
pronouns in their dialogue with the customer and the customer’s subsequent purchase behavior,
controlling for numerous covariates. It also finds null effects for the firm agents use of “you”
(the customer) and “we” (the firm entity) pronouns endorsed by managers and consumers.
Taken together, our remaining studies support two distinct effects. First, using “you”
pronouns to emphasize the customer has no impact on customer attitudes or behaviors, relative to
no customer references (Study 3) or to fewer customer references (Study 4). Second, using “I”
pronouns to emphasize the firm agent as the subject in the interaction has a positive impact on
17
customer attitudes and behaviors relative to (1) “we” pronouns emphasizing the firm entity
(Study 2), no “I” pronouns (Study 3), or fewer “I” pronouns (Study 4).
Our empirical setting is technology-mediated customer service interactions (e.g., email), a
setting with strong managerial relevance. Nearly half of consumers report it as their preferred
way of interacting with firms (44%; Charlton 2011), and over 90% of U.S. firms offer emailbased customer-firm interactions, with a majority of these also providing other online
mechanisms such as live chat, social media, or web forums (Dutta 2012). Within this setting, we
assess both customer inquiries and complaints to test the two main types of customer-firm
interactions (Bolton 1998). We consider the generalizability of our findings to other customerfirm interaction settings in the general discussion.
STUDY 1: FIRM AGENT PERSONAL PRONOUN USE IN PRACTICE
Building on our pilot study, which demonstrated a belief among consumers and managers
that firm agents should use “you” and “we” pronouns, but not necessarily “I” pronouns, in
customer-firm interactions, Study 1 examines which personal pronouns firm agents actually
emphasize in practice in this setting.
Following customer orientation theory and the lay theories revealed in our pilot studies, we
expect to observe a high rate of “you” (the customer) and “we” (the firm) pronoun use in firm
responses to customers. The pilot studies also suggest that we should see higher use of “we” than
“I” pronoun use by firm agents in referring to the “actor” in the customer-firm interaction. We
test for this pronoun use pattern in the field, and further, we compare it to several natural
language samples.
Participants, Design and Procedure
18
We constructed two bogus customer emails—an inquiry and a complaint—and sent these
to a sample of twenty firms each (total target N = 40) randomly selected from the top 100 online
retailers of 2013 as identified by Internet Retailer, a leading trade magazine. The inquiry asked
about the company’s international shipping and returns policy, while the complaint expressed
frustration with the website’s usability for a touchscreen tablet (Appendix B).
One of the two bogus customer emails was sent to each of the 40 selected firms using a
webmail account created for this purpose. Two firms did not offer web/email-based customer
support and seven failed to respond within two weeks. Nine more firms were randomly selected
to achieve the predetermined sample size of twenty for each interaction type. All nine additional
firms responded, delivering a total sample of 40.
Results
We assessed personal pronoun use in two ways. First, for each firm response, we coded
whether or not each pronoun category was used (yes/no). Second, we used Linguistic Inquiry and
Word Count (LIWC; Pennebaker et al. 2007), an application common in research leveraging
linguistic psychology variables such as personal pronouns,9 to assess firm agents’ relative use of
the three pronoun categories of interest. For a given text (a firm agent response), LIWC counts
the incidence of words in each personal pronoun category, adjusting this count for the total
number of words in the target text. The LIWC statistic can be interpreted as the proportion of
words in a text that are represented in the dictionary for that pronoun category (Appendix Table
A1).
Firm Agent Personal Pronoun Use. “You” pronouns referring to the customer appeared in
nearly all (97.5%) of the firm responses. “We” pronouns referencing the agent as part of the firm
9
See Tausczik and Pennebaker (2010) for a review of the development of and psychometric properties of LIWC, as
revealed in its use in over 500 peer-reviewed publications (as of July 2015).
19
entity were similarly present in 100% of firm responses, while “I” pronouns referencing the firm
agent were present in fewer than half (45%) of the responses ((“I” vs. “you” 2(1) = 19.91, p <
.001; “we” vs. “I” 2(1) = 22.81, p < .001). The relative incidence of each of the three personal
pronoun categories did not vary across the inquiry and complaint samples (“you” pronouns: 95%
vs. 100%; 2(1) = 1.03, p > .30; “we” pronouns: 100% vs. 100%; 2(1) = 0, NS); “I” pronouns:
40% vs. 50%; 2(1) = .40, NS)
As for the LIWC statistics, “I” (M = 0.94) pronouns were used less frequently than “we”
(M = 4.83; t(39) = 6.68, p < .001) as well as “you” (M = 6.04; t(39) = 12.35, p < .001) pronouns
in the full sample. The relative absence of “I” pronoun use by the firm agent was sustained
whether firm agents were responding to the customer inquiry (MI = .80, Myou = 4.94, t(19) =
6.86, p < .001; MI = .80, Mwe = 3.28, t(19) = 3.77, p < .01) or complaint (MI = 1.07, Myou = 7.14,
t(19) = 12.40, p < .001; MI = 1.07, Mwe = 6.37, t(19) = 6.13, p < .001). Summary statistics are
provided in Table 2.
TABLE 2: PRONOUN USE IN REAL FIRM AGENT RESPONSES (STUDY 1)
Firm Agent Response to Customer
Mean*
SD
Min
Max
0.80
1.78
0.00
7.73
4.94
2.16
0.00
9.52
3.28
1.92
1.10
8.70
Bogus Customer
Communication
Inquiry
(N = 20)
LIWC Personal
Pronoun Categories
“I”
“You”
“We”
Complaint
(N = 20)
“I”
“You”
“We”
1.07
7.14
6.37
1.60
1.38
2.80
0.00
4.62
1.72
6.03
9.86
11.76
Total
(N = 40)
“I”
“You”
“We”
0.94
6.04
4.83
1.68
2.11
2.84
0.00
0.00
1.10
7.73
9.86
11.76
*LIWC statistics represent proportion of words in observed corpus.
20
Personal Pronoun Use in Comparison Samples. While the results above confirm that firm
agents frequently use “you” (the customer) and “we” (the firm entity) pronouns, but use “I” (the
firm agent) pronouns significantly less often, is this use pattern consistent with natural language?
To assess this question, we compared the sample obtained in this study to several linguistic
corpuses, including a large natural language dataset (N > 700,000) and an Internet-based
consumer forum. See Appendix Table A2 for personal pronoun use statistics for our focal and
comparison samples. The current sample of firm agent language use is inconsistent with these
natural language samples, with the firm agent language corpus showing a higher use of “you”
and “we” pronouns and a lower use of self-referencing “I” pronouns.
We next obtained other samples where, as with the current firm agent sample, a customer
orientation might be relevant. We used a sample of 95 telephone interactions from the online
retailer who provided the data in Study 4 and a sample of 80 advice columnist responses from
four major syndicated advice columns (Appendix Table A2). The current sample showed
language use consistent with these other customer orientation samples, with a similarly heavy
emphasis on “you” and “we” pronouns, and less use of “I” pronouns relative to natural language.
Discussion
In sum, consistent with our predictions and pilot studies, we see heavy use of “you” and
“we” pronouns in the language of real firm agents. Firm agents linguistically “put the customer
first” (more frequent “you” use). They also tend to refer to themselves as part of the firm entity
(more frequent “we” use) rather than as an individual (infrequent “I” use). This pattern of
personal pronoun use differs from the global natural language mean (Pennebaker et al. 2007) and
from consumers’ interactions with one another in an online forum (Appendix Table A2, sample
F), where “I” pronouns predominate. However, in samples where a participation framework
21
consistent with the customer orientation is relevant, we find a pattern similar to the current
study’s results, with heavier use of “you” and “we” pronouns and less use of “I” pronouns.
Thus far, through managerial theory, lay theories revealed in the pilot studies, and Study
1’s field results, we have shown that, firm agents are expected to and actually do: (1) emphasize
“you” (the customer) as the object; and (2) emphasize “we” (the firm entity) more than “I” (the
firm agent) as the subject. Our next study provides an initial test of how firm agents’ tendency to
refer to themselves as part of the firm entity (“we”) rather than as a singular self (“I”) impacts
consumer perceptions of the agent and firm.
STUDY 2: “I” VS. “WE” USING REAL FIRM AGENT RESPONSES
Study 2 examines whether the use of self-referencing “I” pronoun use by firm agents as an
alternative to “we” pronoun use referring to the agent as part of the firm affects customer
attitudes and intentions towards the firm in a relatively externally valid context. We used real
firm email responses from six of the companies sampled in Study 1 (three inquiries and three
complaints). The six firm agent responses were selected to cover a broad range of product
categories and to be representative of the base rates of personal pronoun use of the larger 40-firm
sample in Study 1. We then modified the actual firm responses to replace “we” pronouns with
“I” pronouns while minimizing potential changes in meaning.
Participants, Design and Procedure
Canadian undergraduates participated for partial course credit (N = 211). Participants were
asked to imagine themselves as the customer in each of two unrelated customer-firm
interactions—an inquiry and a complaint. Participants’ first names were collected and inserted
into the customer’s initial communication to the firm and into the firm’s response to enhance
involvement in the scenario. Each participant saw one of three inquiry interactions and one of
22
three complaint interactions (order was randomized). Participants evaluated each of these
independently, resulting in approximately 35 participants per condition.
Any information identifying the real firm was removed from the original agent responses,
and where appropriate, replaced with the fictional firm name Shopsite.com. Firm agent response
was manipulated by using either the original firm response from Study 1 or a modified version.
In the modified condition, “we” pronouns referencing the firm agent and firm as the grammatical
subject were replaced with “I” (the firm agent) pronouns when this did not change the meaning
of the sentence. For example, “We thank you for understanding” can be modified to “I thank you
for understanding,” but “We do not offer international shipping” cannot be modified to “I do not
offer international shipping,” as it is the firm that offers free shipping.10 Sample stimuli are
presented in the Appendix. This resulted in a mixed design, with interaction type (inquiry,
complaint) as a within-subjects factor, and firm replicate (1, 2, or 3) and firm agent response
(original, modified) as between-subjects factors across each of the two interaction types.
After reading the scenario, participants indicated their satisfaction with the firm agent and
their purchase intentions toward the firm. Satisfaction was assessed using three items (“I am
satisfied with my overall experience with this person,” “As a whole I am not satisfied with the
response provided by this person”, “How satisfied are you with the quality of service provided
by this person?”; 1 = not at all, 7 = very much;  = .77). Purchase intentions were measured with
three items (“In the future, I would purchase from Shopsite.com,” “If I was in the market for the
kind of product they sell, I would use Shopsite.com,” “In the future, I would not use
Shopsite.com again” (reverse item); 1 = strongly disagree, 7 = strongly agree;  = .80). Items
were adapted from Maxham and Netemeyer (2002).
A test of the prevalence of this and other specific boundary conditions under which “we” cannot be replaced with
“I” is reported in the General Discussion.
10
23
Results
Our primary independent variable of interest was firm agent response condition, where we
tested for changes in both satisfaction with the firm agent and purchase intentions from the
original “we” to the modified “I” firm response. We assessed the robustness of any effect across
customer-firm interaction type (inquiry and complaint) and across the six replicates.
An omnibus ANOVA indicated a significant main effect of firm agent response (original
vs. modified) on satisfaction with the agent (F(1, 405) = 39.95, p < .001) and on purchase
intentions (F(1, 404) = 25.95, p < .001). This model also revealed a marginal main effect across
replicates on satisfaction (F(5, 405) = 2.14, p < .10) and purchase intentions (F(5, 404) = 2.13, p
< .10). This variation across replicates is of little empirical interest given there was no interaction
of replicate with firm agent response condition for either dependent measure (Fs < 1). As such,
this marginal firm-level variation is not discussed further. There was also no effect of the order
in which the two interaction types were presented to participants on either satisfaction with the
firm agent (F < 1) or purchase intentions (F(1, 414) = 1.25, p = .26).
TABLE 3: IMPACT OF REPLACING “WE” WITH “I” PRONOUNS
IN REAL FIRM AGENT RESPONSES (STUDY 2)
Firm
A
B
C
D
E
F
Product Category
Apparel, lifestyle
Media, travel
Women’s apparel
Mass merchant
Apparel, outdoor
Automotive
Interaction
Complaint
Complaint
Complaint
Inquiry
Inquiry
Inquiry
Satisfaction with Firm Agent
Original Modified t-stat
Purchase Intention
Original Modified t-stat
4.33
4.26
4.79
4.42
4.79
4.28
4.38
4.05
4.51
4.56
4.52
4.36
5.23
5.13
5.59
5.51
5.50
4.96
2.50 *
2.68 **
2.37 *
3.50 **
2.31 *
2.15 *
5.05
4.68
5.21
5.34
5.40
4.76
2.25 *
1.71 +
2.33 *
2.22 *
3.05 **
1.19
** p < .01, * p < .05, + p < .10
The omnibus test revealed improved customer satisfaction and purchase intentions for the
modified “I” firm response (Msatisfaction = 5.32; Mintentions = 5.07) relative to the original “we” firm
24
response (Msatisfaction = 4.48; Mintentions = 4.40). We present means and simple comparisons for the
original and modified firm response for each of the six replicates in Table 3. As shown in the
table, the positive effect of “I” pronoun use in the modified response condition is revealed in
heightened customer satisfaction across all six firms, and in increased purchase intentions for
five of the six firms. We speculate that the non-significant result for purchase intentions in the
Firm F replicate is related to long purchase cycles for the automotive category.
Discussion
Study 2 demonstrates that simply changing the grammatical subject from “we” to “I” in real
firm agent responses improves satisfaction and purchase intentions after both inquiries and
complaints. We have suggested that these results may be due to perceptions of increased firm
agent agency and empathy, a mechanism we will test in Study 3. However, alternate mechanisms
could also explain the improvements for “I” relative to “we” pronoun use observed in this study.
First, if the modified (“I”) firm responses are viewed as more or less expected or typical than
the original (“we”) responses, our results might be explained by conversational norms or
expectancy violations (Kronrod et al. 2011; Kronrod and Danziger 2013). To assess this, we
conducted a post-test using the same participant pool, which confirmed that our modified
responses were not perceived to be less typical than the original responses (see Appendix
discussion and Tables A4 and A5). Second, it is possible that, rather than variation in pronoun
use, variation in other linguistic content across the six firm agent emails used is causing our
results. Study 3 uses experimenter-created stimuli to address this issue.
We also note that in this study, to provide a clean test of “I” versus “we” pronouns for the
grammatical subject (firm agent), we could not modify the emphasis on “you” pronouns
referencing the grammatical object (the customer) observed in real firm agent email responses.
25
Our next study uses more controlled stimuli to independently assess both of our two effects: (1)
how firm agent’s use of self-referencing “I” pronouns and (2) how firm agent’s use of “you”
pronouns referencing the customer influence customer attitudes and intentions. Given the
consistent results we have seen for firm agent pronoun use (Study 1) and participant responses
(Study 2) across customer inquiries and complaints, we focus on inquiries only in Study 3. Study
4 revisits both categories in the field.
STUDY 3: INCREASED “I” OR “YOU” PRONOUN USE BY FIRM AGENTS
In Study 3, we independently test how firm agent use of “I” (referencing the agent) and
“you” (referencing the customer) impacts customer attitudes and intentions relative to a nopronouns control condition. The no-pronouns condition can be thought of as representing cases
in which a party involved in a specific participation framework is not emphasized explicitly (e.g.,
“Happy to help!” vs. “Happy to help you!” or “I’m happy to help!”).
Study 3 also seeks evidence of our theorized process underlying the “I” effect. As
discussed earlier, we propose that perceived empathy and agency of the firm agent will mediate
the positive effect of increased “I” pronoun use on consumer attitudes and behavioral intentions.
This is because the presence of “I” pronouns may indicate that the firm agent feels involved
and/or concerned with the customer’s issue (empathy; Ickes et al. 1990; Singh and Sirdeshmukh
2000; Wales 1996), as well as suggesting that the firm agent will act on the customer’s behalf
(agency; Ahearn 2001; Chung and Pennebaker 2007; Kashima and Kashima 2003; Singh and
Sirdeshmukh 2000).
Further, Study 3 tests the generality of our effects. As noted earlier, we examine language
use within a specific participation framework that operates at the clause or sentence level. To
assess a possible clause or sentence level factor that may impact the generality of the observed
26
pronoun effects, we also manipulate active versus passive verb voice in Study 3. For example, a
firm agent could either say, “I will assist you in a moment” (active voice) or “You’ll have my
assistance in a moment” (passive voice). While the use of active or passive voice in
communications is a matter of style, active voice is commonly described as more lively and
direct. Notably, however, passive voice is recommended when the speaker wishes to call
attention to or focus on the recipient of action rather than the actor (Johnson-Laird 1968;
Penelope 1990). Such a focus may be desirable under a customer orientation, when the firm
agent wishes to linguistically emphasize the customer. Thus, it is not immediately clear whether
a firm agent’s pronoun references are affected by active or passive voice in customer-firm
interactions. The present study tests this potential moderator of pronoun use effects.
Participants, Design and Procedure
American participants (N = 326) from an online panel completed the study for a small cash
payment. Participants were asked to imagine themselves in one of six versions of a hypothetical
customer-firm interaction. This study used a scenario rather than the real firm responses from
Studies 1 and 2 to ensure that we could carefully manipulate pronoun use and verb voice while
changing as little other linguistic content as possible. In the scenario, the participant contacted
fictional retailer Shopsite.com to inquire about a product they had ordered but had not yet
received. Participants saw a firm agent response that either had no personal pronoun references,
“I” pronouns referencing the firm agent as the actor (grammatical subject), or “you” pronouns
referencing the customer as the recipient of action (grammatical object). This design allows us to
independently test the two effects of interest, (1) firm agent “I” (subject) references versus a
control, and (2) customer “you” (object) references versus a control. We also manipulated verb
voice across these conditions, resulting in a 3 (pronoun use: None, “I,” “You”) x 2 (verb voice:
active, passive) between-subjects design. Complete stimuli are presented in the Appendix.
27
After reading the firm agent’s response, participants reported their satisfaction with the
person who responded to their inquiry ( = .86) and their purchase intentions towards the firm (
= .92) using the measures from Study 2. Next, participants indicated the extent to which a series
of words described the firm agent’s empathy toward the customer (understanding, empathetic,
concerned;  = .94) and the agent’s agency on behalf of the customer (acts on my behalf, tries
hard to help, takes initiative for me;  = .96). All items used seven-point scales (1 = not at all, 7
= very much). Presentation order was randomized across the six items. Lastly, we collected and
tested the same language typicality items used in Study 2 (see Appendix for results).
Results and Discussion
An omnibus ANOVA for the satisfaction dependent measure revealed a main effect for
pronoun use condition (F(2, 320) = 18.19, p < .001), but no effect for verb voice (F(1, 320) =
.89, p = .35) and no interaction of pronoun use and verb voice (F(2, 320) = .78, p = .69). The
same pattern of results held for purchase intentions, with a main effect for pronoun use condition
(F(2, 320) = 17.25, p < .001) and null effects for verb voice (F(1, 320) = .47, p = .49) and the
interaction of pronoun use and verb voice (F(2, 320) = .50, p = .61). Verb voice had no impact
on our dependent variables. We thus collapse the verb voice conditions in the analysis below.
“You” pronouns. When compared to a no-pronouns control, the addition of “you”
pronouns emphasizing the customer as the recipient of action (object) in the firm agent’s email
had no effect on customer satisfaction (MNone = 4.28 vs. MYou = 3.97; F(1, 205) = 2.43, p = .12)
or purchase intentions (MNone = 4.08 vs. MYou = 3.94; F(1, 205) = .44, p = .51).
“I” pronouns. Compared to the no-pronouns control condition, the addition of “I”
pronouns emphasizing the firm agent as the actor (subject) increased customer satisfaction
(MNone = 4.28 vs. MI = 5.17; F(1, 209) = 19.33, p < .001) and purchase intentions (MNone = 4.08
vs. MI = 4.76; F(1, 209) = 9.97, p < .01). Condition means are summarized in Figure 1.
28
FIGURE 1: IMPACT OF “I” PRONOUNS (AGENT REFERENCES), “YOU” PRONOUNS
(CUSTOMER REFERENCES), AND NO PRONOUNS (IMPLICIT REFERENCES) ON
SATISFACTION AND PURCHASE INTENTIONS (STUDY 3)
6.00
5.50
5.17
5.00
4.76
None
4.50
4.00
4.28
3.97
4.08
"You"
3.94
"I"
3.50
3.00
Satisfaction
Intentions
Process. The firm agent empathy ( = .94) and agency ( = .96) measures were assessed
as mediators of the above relationship between the firm agent agent’s use of “I” pronouns (vs.
their absence) and each of the dependent measures (satisfaction with the firm agent, purchase
intentions). Confirmatory factor analysis supported empathy and agency as separate factors
(2(1) = 205.87, p < .001). We leveraged Preacher and Hayes’ (2008) PROCESS macro to
assess the two predicted mediators in parallel, contrasting the effect of “I” pronoun use by the
firm agent versus its absence (in the no pronouns control) as our independent variable and
satisfaction and purchase intentions as dependent measures.
The presence of “I” pronouns by the firm agent (vs. their absence) increased perceptions of
firm agent empathy (B = .89, t = 4.33, p < .001) and agency (B = 1.22 t = 5.29, p < .001).
Further, both mediators significantly predicted increased satisfaction with the firm agent (Bempathy
= .17, t = 2.12, p < .05; Bagency = .51, t = 7.03, p < .001) and increased purchase intentions
29
(Bempathy = .19, t = 2.18, p < .05; Bagency = .50, t = 6.36, p < .001). Bootstrapping with 5,000
resamples showed that the effect of “I” pronouns by the firm agent on customer satisfaction was
mediated by both empathy (CI: .01 – .37, p < .05) and agency (CI: .34 – .97, p < .05). Bootstrap
confidence intervals also supported mediation by empathy (CI: .03 – .39, p < .05) and agency
(CI: .35 – .97, p < .05) for the purchase intentions dependent measure.
Discussion
Study 3 provided a controlled test of the consequences of firm agent use of “I” pronouns.
Results revealed that the presence (versus absence) of self-references by the firm agent improved
attitudes toward the agent and behavioral intentions towards the firm. Process analysis confirmed
that enhanced perceptions of both empathy and agency drive the positive impact of “I” pronoun
use by firm agents.11
In contrast, separate tests examining references to the customer as the recipient of the firm
agent’s actions found that the presence (versus absence) of “you” pronouns did not enhance
satisfaction or purchase intentions. Variation in active versus passive voice also had no
observable effect on satisfaction or purchase intentions.
STUDY 4: PRONOUN USE IN REAL CUSTOMER-FIRM INTERACTIONS
To this point, we have demonstrated (1) a relative emphasis on “you” pronoun use, and (2)
a relative de-emphasis of “I” pronoun use (relative to “we” pronoun use) by real firm agents
relative to natural language in the field (Study 1). We have also provided causal evidence that
increases in “I” pronoun use by firm agents offer benefits in terms of customer satisfaction and
purchase intentions relative to no “I” pronoun use and to “we” pronoun use (Studies 2 and 3, plus
Web Appendix Study 5). In our last study, we leverage field data to provide evidence for the
11
These results were wholly replicated in another study using different stimuli. See Study 5 in the Web Appendix.
30
positive effect of firm agent “I” pronoun use on real purchase behavior. The use of field data
allows us to account for heterogeneity in individual (e.g., language complexity) and dyadspecific (e.g., interaction topic) firm agent pronoun use, a limitation not easily addressed with
researcher-produced stimuli in the lab.
Notably, this setting also offers an opportunity to control for heterogeneity in the
customer’s pronoun use. While our lab studies fixed the customer’s initial inquiry through
experimental design, in reality, since personal pronouns are used to establish the roles and/or
focus of a social interaction (i.e., the participation framework), the pronouns used by the
customer (e.g., “I’d like the bill please.” vs. “Can you bring the bill please?”) are likely to
interact with the pronouns used by firm agents in reply (Goffman 1981; Gordon, Grosz, and
Gilliom 1993). This study assesses and controls for this possibility by accounting for the
customer’s pronoun use (e.g., “I” pronouns) in their initial email inquiry to the firm as a main
effect and as an interaction term with the focal effect of the present research—the firm agent’s
use of the same personal pronoun category (e.g., “I” pronouns) in his or her response to the
customer. As in prior studies, we examine the effects of each pronoun category independently,
resulting in three separate models for analysis (one for each of “I,” “we,” and “you”).
Data
We obtained a data set that included both the linguistic content of actual customer-firm
interactions and customer purchases before and after these interactions. The data was obtained
from a large multi-category and multinational online retailer which primarily sells entertainment
and information products (books, music, movies). The firm wishes to remain anonymous. The
data set contains a random (nth select procedure) sample of customer-firm interactions (N =
31
2,098) in 200412 initiated by customers using a “contact us” link that appeared on every page of
the firm’s website. The firm operated its own telephone and email customer contact center. We
were provided with data at the customer-firm interaction level. The interaction began with a
customer-initiated communication that was not connected with a prior contact “ticket,” and was
considered closed after two weeks of no communication by either the customer or firm. Included
in the data were a timestamp and the full text of the customer’s initial communication, the firm
agent’s reply, and any additional emails for the interaction before the ticket was closed.
We were able to link 1,277 (60.9%) of these interactions to a purchase account either via
the customer’s email address, order information, or account information referenced in the text of
the customer’s message. For customers linked to a purchase account, the firm provided purchase
volume for one year (365 days) before the date of a given customer-firm interaction, and for one
year (365 days) after the date of the final communication of the same interaction. We observed
the date each purchase was made and the dollar value of each purchase net of taxes. The data
also included the customer’s first name and their billing residence, which were used to produce
demographic covariates.
The textual content of the observed customer-firm interactions underwent extensive
“cleaning” prior to submitting it to linguistic analysis. We removed generic headers and footers
generated by the customer or firm’s email application, as well as third-party marketing footers
(e.g., “Post your free ad now at Yahoo! Personals”). As is common in marketing contact centers,
the firm’s managers provided its agents with a selection of “boilerplate” content to adapt in
response to common inquiries (e.g., order status). As any boilerplate content used was heavily
integrated into the firm agent’s personalized response to the customer, it was not removed. We
The age of the data is driven by the firm’s decision to outsource its customer contact center in 2005. The firm was
not able to obtain complete customer-firm interaction transcripts from the third-party provider.
12
32
note that while any boilerplate language the firm agent elected to adapt to an individual customer
communication would not have been written by the responding agent, it was written by another
agent of the same firm (i.e., a more senior marketing or customer service agent).
The cleaned customer-firm interaction text was processed in LIWC to measure pronoun
use for each customer-initiated communication and the subsequent firm agent response. For
analysis, we considered only the initial customer communication and firm response to the
customer. We did this for several reasons. First, this replicates the structure of customer-firm
agent interactions examined in prior studies. Second, the modal interaction (77% of cases) was
two emails (customer inquiry, firm reply). Third, the modal number of interactions beyond the
first two emails was three (64% of remaining interactions), and the third email was most often a
simple “thank you” from the customer (67% of cases as identified by LIWC word counts).
Further, two student judges scored 89% of three-step email interactions as resolved, suggesting
there was little new information in the majority of third emails. Finally, it is beyond the scope of
this research to conceptualize and model dynamics in an extended time-series of interactions
between customers and firm agents. However, we include the total number of emails in the
interaction as a covariate to help control for the relative presence of boilerplate language (which
likely decreases with the number of total emails exchanged), and to help control for the
possibility that unresolved or more complex interactions (which likely increase with the number
of total emails exchanged) drove firm agent pronoun use.
Model
To carry out the analysis, we regress total purchases for the customer in a given
customer-firm interaction i for a defined time period after the customer-firm interaction (Pi,post)
on both the customer and firm agent’s use of one of the three personal pronoun categories (“I,”
“you,” or “we) and a set of interaction-level covariates. This can be specified as
33
Pi,post = Cust_Pronounic + Firm_Pronounic +  + zi + i
(1)
where Cust_Pronounic and Firm_Pronounic represent the LIWC pronoun category statistic for the
customer’s initial communication and firm agent reply, respectively, in customer-firm agent
interaction i, for LIWC pronoun category c (c = “I,” “you,” or “we”). The expected interactive
effect of the customer’s pronoun use on the use of the same pronoun by the firm agent (Gordon
et al. 1993) is captured by . We mean-center the personal pronoun statistics and model the
pronoun categories independently, resulting in three separate models (one for each of the three
pronoun categories). For example, the first model considers (a) the simple effect of the
customer’s use of “I” pronouns in their initial email to the firm, (b) the simple effect of the firm
agent’s use of “I” pronouns in his or her reply, and (c) the interactive effect of the customer’s
and the firm agent’s use of “I” pronouns (Table 4, Model 1).
As for the remaining model terms, zi is a vector of interaction-specific covariates, and i
captures idiosyncratic error. The interaction-specific covariates are as follows:
Purchase volume prior to the interaction. This covariate addresses omitted variable bias
(e.g., self-selection) by capturing customer heterogeneity in baseline purchase volume for
customer i for the same time period prior to the interaction (Pi,pre) as the time period observed for
the dependent measure (Pi,post). We report a 90-day purchase observation window before and
after the customer-firm interaction for our dependent measure and its (pre-period) control. This
34
offered the tightest observation window to the interaction event before purchase data becomes
exceedingly sparse.13
Several covariates help assess whether the difficulty, complexity, or severity of the
interaction topic explain pronoun use during the interaction and/or the customer’s purchase
behavior after the interaction.
Number of emails. We include the total number of emails in the customer-firm interaction
as a potential indicator of more difficult or complex interactions (> 2 emails).
Customer posemo and Customer negemo. LIWC’s two measures of positive and negative
emotion in language were captured as controls for the customer’s emotional tone, which could
indicate the severity of an interaction and drive the personal pronouns used in the agent’s
response (Chung and Pennebaker 2007).
Resolution. Two independent judges scored the extent to which they perceived the reason
for the customer’s initial email to the firm as resolved (1 = not at all resolved, 7 = very much
resolved). We used the mean score of the two judges (r = .43, p < .001).
Complaint. Two independent judges coded the customer’s initial communication in each
interaction as either a complaint or inquiry (inter-rater agreement = 91%, disagreement resolved
by a third judge) to replicate the two main categories of customer-initiated interactions assessed
in the lab studies. This is incorporated in the model as a dummy for complaints.
Compensation. This dummy variable captured whether the customer was offered a
financial incentive or compensation as a consequence of the interaction. Compensation always
entailed a $5 online coupon for a future purchase as a customer-service gesture.
13
Average annual purchase frequency is less than quarterly (1.99 purchases per year). Model results are fully
replicated using a 180-day window, and fall to non-significance at one year (365 days), suggesting that the effects of
personal pronoun use on customer purchases are moderated by other interaction events over time.
35
Reason. We include the firm’s four-level categorization of the customer’s reason for
initiating the interaction (order-related, website-related, multiple reason, or “other”). The agent
identifies each interaction reason based on the customer’s initial email. This provides another
control for the specific nature of the interaction (i.e., its difficulty, complexity, severity).
Region and Gender. We were able to produce two demographic covariates with the firm’s
data. We used ZIP and postal code data to produce a four-level geographic variable that captures
the firm’s major operating regions. This was done in order to manage the number of geographic
covariate terms in the model, since a large number of cities (N = 373) and states or provinces (N
= 32) were represented in the data. To produce a gender covariate, we used “genderizer”
software that codes first names as male (e.g., Samuel), female (e.g., Samantha) or unknown (e.g.,
Sam) using over 100,000 common first names. The customer was identified as either male or
female for 1,119 (87.6%) of the customer-firm interactions linked to transactional accounts. We
included a dummy term for female customers.
Summary statistics for all independent variables and covariates included in the equation are
provided in the Appendix (Table A6). Notably, personal pronoun use means for firm agents in
this data set were similar to those observed in Study 1’s 40-firm sample, with firm agents deemphasizing “I” pronouns relative to natural language use. In contrast, and consistent with
natural language use, customer language in the current sample emphasized “I” pronouns
(Appendix Table A2).
Covariance between the personal pronoun use independent variables and covariates was
modest, alleviating endogeneity concerns (all r values < |.25|). Correlation among the predictors
is presented (Table A7) and discussed in the Appendix. All predictors and covariates fell below
the Variance Inflation Factor threshold of 10 (Kutner et al. 2004), suggesting that the variance of
model coefficients was not substantially increased due to collinearity.
36
TABLE 4: RELATIONSHIP BETWEEN PRONOUN USE AND CUSTOMER PURCHASE
VOLUME IN REAL CUSTOMER-FIRM INTERACTIONS (STUDY 4)
(1) “I” Model
.
(2) “You” Model
(3) “We” Model
Estimate
SE
Estimate
SE
Estimate
Intercept
3720.9
(2339.0)
525.5
(2305.4)
1654.0
Cust_pronounc = “I”
Firm_pronounc = “I”
692.5
1236.3
226.1
c = “I”
SE
(2328.0) +
(117.3) ***
(267.1) ***
(49.4) ***
Cust_pronounc = “You”
Firm_pronounc = “You”
-483.4
68.3
-68.6
c = “You”
Cust_pronounc = “We”
Firm_pronounc = “We”
(139.2) ***
(127.6)
(38.5) +
-387.8
-129.8
79.8
c = “We”
(312.0)
(181.6)
(114.0)
Covariates
Pi,pre
# of emails
Customer posemo
Customer negemo
Complaint
Resolution
Compensation
Order reason
Website reason
Multi reason
Region 1
Region 2
Region 3
Female
R-squared
0.3
175.5
-192.8
-629.7
-307.0
305.2
-157.0
-672.9
-985.6
-955.5
-578.6
2296.8
-618.4
-171.9
0.19
(.0) ***
(198.9)
(133.6)
(238.6) **
(1147.2)
(263.2)
(3437.9)
(1035.2)
(1168.3)
(1993.2)
(1188.9)
(1640.8)
(1279.6)
(758.4)
0.3
222.4
-41.4
364.2
-82.5
-155.7
-550.5
-917.9
-807.2
-999.9
-981.5
1843.4
-913.6
-58.8
(.0) ***
0.3
(199.4)
147.7
(1157.7)
-187.5
(265.2) *
-571.7
(137.6)
-44.9
(3463.9)
344.2
(240.6)
-171.6
(1048.1)
-1036.0
(1192.6)
-1034.0
(2007.0)
-668.0
(1203.9)
-613.0
(1654.6)
2075.0
(1291.1)
-642.9
(758.1)
-45.8
0.18
0.17
*** p < .001, ** p < .01, * p < .05, + p < .10
Other reason and Region 4 are baselines for respective dummy sets.
(.0) ***
(202.2)
(135.8)
(242.2) *
(1164.0)
(268.3)
(3485.0)
(1042.0)
(1198.0)
(2024.0)
(1207.0)
(1663.0)
(1298.0)
(765.4)
37
Results
Table 4 presents the results of models assessing the effect of firm agent use of “I” (Model
1), “you” (Model 2) and “we” (Model 3) pronouns on post-interaction purchase behavior in real
customer-firm interactions. For the covariates, all three models revealed a significant positive
relationship between post-interaction purchases (Pi,post) and pre-interaction purchases (Pi,pre) and
a negative effect on post-interaction purchases for the customer’s use of negative emotion words
(Customer negemo). All other covariates were non-significant.
After accounting for the customer’s use of the same pronoun category in their initial email
(Cust_Pronounic) and the expected interaction between the customer and firm agent’s personal
pronoun use, we found a significant positive effect for firm agent use of “I” pronouns ( =
1236.35, t = 4.63, p < .001; Model 1). In the “you” model, we replicate the null effect for agent
references to the customer ( = 68.34, t = 0.54, p = .59; Model 2). We also find a null effect for
firm agent use of “we” pronouns ( = -129.80, t = -0.70, p = .47; Model 3). There was also
significant and positive interaction of the customer and firm agent’s use of “I” pronouns ( =
226.13, t = 4.58, p < .001; Model 1) and a marginal negative interaction of customer and firm
agent use of “you” pronouns ( = -68.59, t = -1.78, p = .08; Model 2). All other independent
variables were non-significant.
Discussion
Overall, the results suggest that increased firm agent use of “I” pronouns in replying to a
customer’s initial email is linked to increased purchase volume following this interaction. This
relationship is positively compounded when the firm agent uses more “I” pronouns at the same
time as the customer uses more “I” pronouns (i.e., the positive interaction for “I” pronoun use).
This finding is in contrast with what managers and consumers believe firm agents should do
38
(pilot studies) and what they actually do in practice (Study 1 and the present study). In short,
while firm agents linguistically emphasize the customer (“you” pronouns) and firm (“we”
pronouns), we replicate the findings of our lab studies and provide real-world evidence that
instead, only self-referencing “I” pronoun use by the firm agent has a positive relationship with
economic outcomes for the firm.
Of potential interest for future research is the negative main effect observed for customer
use of “you” pronouns in their initial email to the firm ( = -483.41, t = -3.47, p < .001). We
speculate that this result may suggest customer negativity or attribution of responsibility toward
the firm (e.g., “Your website doesn’t work” or “You need to resolve this”). The marginal negative
interaction between increased firm agent and customer use of “you” pronouns ( = -68.59, t = 1.78, p = .08) similarly hints at the possibility of a “blame game” assigning or denying
responsibility between the dyad’s participants (customer and firm agent).
Critically, the heterogeneous customer inquiries and complaints used in this study help
control for interactions between pronoun use, sentence construction, and the larger interaction
content in the customer’s initial email and the firm agent’s reply. This bolsters confidence that
the results of our earlier studies are not artifacts of the specific linguistic content used in labcontrolled experiments. However, the interactive effects also raise the potential for
multicollinearity. As noted earlier, variance inflation factors reveal that the potential for this
issue falls below the threshold of concern. We nonetheless urge caution in interpreting the
relative size of the significant coefficients for the pronoun use predictors.
A possible alternative explanation for this study’s results is that the use of “I” pronouns by
firm agents represents idiosyncratic deviations from boilerplate language written by senior firm
agents (e.g., managers who write the boilerplate might use fewer “I” pronouns than do the more
junior firm agents who use it). As boilerplate language is heavily integrated into the firm agents’
39
own language, it is not feasible to directly test this possibility at the interaction level. However,
boilerplate language should be less common in longer, more complex interactions that demand
more personalized responses. The Number of emails covariate provides a proxy for such
interaction complexity. If the idiosyncratic deviation from boilerplate explanation holds, we
would expect to find an interaction between firm agent “I” pronoun use and the number of emails
in a given customer-firm agent interaction. However, when this interaction term was added to the
full “I” pronoun model (Model 1), the positive effect of firm agent “I” pronoun use was
replicated (B = 1226.00, t = 3.63, p < .001), while the interaction term was non-significant (B =
20.77, t = .42, p = .67). This offers some evidence that idiosyncratic deviation from boilerplate is
unlikely to explain the results. That said, even if this factor partly explained the results, it merely
suggests that more senior firm agents should also modify their personal pronoun use in
boilerplate language.
Finally, we note the limitation of this study’s use of field data. As is common in field data
analysis, we cannot assert causality as we do not randomly assign customers to treatment and
control groups. However, several factors support the relationship we causally demonstrate in our
lab experiments. First, the temporal sequence of the independent variable (interaction language)
and dependent variable (post interaction purchases) rules out reverse causality. Second, our use
of panel data (customer purchases) before and after the customer-firm agent interaction event
controls for possible self-selection (Manchanda, Packard, and Pattabhiramaiah 2013). Third, the
inclusion of multiple covariates reduces the likelihood that interaction-level factors such as
interaction difficulty or severity explain the effect. That said, it is still possible that some
additional unobserved factor(s) help explain the relationship observed between firm agent
pronoun use and subsequent customer purchase behavior in the field.
40
GENERAL DISCUSSION
Prevailing managerial theory implies that frontline selling and service agents should
emphasize the firm entity’s (“we”) focus on the customer (“you”) in customer oriented
interactions. This intuition was verified in two surveys of managers and consumers, and its
practice was verified in the field (Studies 1 and 4). However, we found null effects for firm agent
use of these pronouns in customer interactions in both the laboratory and field (Studies 3 and 4).
In contrast, across three studies, we demonstrate an effect that is inconsistent with these
intuitions and practices: our data show enhanced customer attitudes, intentions, and purchase
behavior when firm agents increase their use of “I” pronouns (Studies 2-4).14 These findings are
consequential given manager and consumer beliefs that firm agents should be linguistically
focused on “you” the customer and “we” the firm, but not “I” the firm agent (Pilot Study, as well
as Pilot Study 2 in the Web Appendix), and given that this sub-optimal pronoun use pattern is
dominant in the field (Studies 1 and 4).
Study 3 supported our theory-driven account that the positive effect of self-referencing “I”
use by firm agents arises through enhanced perceptions of the extent to which the agent feels
(empathy) and acts (agency) on behalf of the customer.14 A final study leveraged field data to
offer evidence of a positive link between increased “I” pronoun use by firm-agents and actual
customer purchases after controlling for numerous covariates.
Implications
Taken together, these results suggest that the current pattern of pronoun use endorsed and
used by firm agents may be leading to sub-optimal outcomes for customers and firms. In addition
to contributing insight on the subtle impact of “I” pronoun use on social perceptions, behavioral
14
See also Web Appendix Study 5.
41
intentions, and actual behavior, we theorize and demonstrate the psychological mechanism
through which “I” pronoun use drives more positive social perceptions and behavioral outcomes
(perceived agency and empathy).
In addition to its contributions to the marketing, social psychology, and linguistic
psychology literatures, this research sheds light on how firms can improve their (literal)
“speaking terms” with customers. In contrast to prevailing managerial philosophies, firms should
train customer-facing employees to emphasize the self more in the standard participation
framework of customer-firm agent interactions. This might arise naturally if, for example, agents
are encouraged to think of themselves as being personally involved in the customer’s needs (“I”)
rather than as an impersonal agent of the firm entity (“we”). Straightforward language use
recommendations for firms, along with a few examples of these, are summarized in Table 5.
TABLE 5: EXAMPLES OF ACTUAL VERSUS RECOMMENDED PRONOUN
EMPHASES IN CUSTOMER-FIRM INTERACTIONS
42
Limitations and Future Research
We note that the current research focuses on customer-firm interactions via email or other
text-based forms of communication. However, we expect our results may hold across different
modes of communication. While prior work has established some differences between spoken
and written language (e.g., speed), communication goal (e.g., addressing customer inquiries or
complaints) is a stronger predictor of linguistic content than communication format (Halliday
1995). Indeed, in a sample of telephone-based customer-firm agent interactions (N = 95) from
the company that provided the data for Study 4, we found elevated levels of “you” use similar to
those observed in email interactions (Appendix Table A3, comparison sample D). Thus, while
there may be main effect differences, personal pronoun emphases should have similar effects
across channels. That said, it would be worthwhile for future research to test these effects
directly. It is also plausible that increased “I” pronoun use by other marketing agents of the firm,
such as celebrity endorsers or spokespeople, could enhance persuasion in other marketing
communications.
This research opens the door to a variety of further examinations of pronoun use (and more
broadly, language use) in customer-firm interactions. Researchers could identify moderator and
boundary conditions for the beneficial effects of “I” pronoun use we predict and observe.
Importantly, we do not argue that firm agents should use more “I” pronouns in all participation
frameworks and contexts. For example, it is likely that altering the participation framework
affects which personal pronoun emphases are optimal. Indeed, in an additional study (Web
Appendix Study 5), we found that reversing the participation framework of customer-firm agent
interactions, such that firm agents primarily treated the customer as the actor (grammatical
subject), resulted in poorer outcomes. As another example, consistent with related work on
language and relationships (Fitzsimons and Kay 2004; Sela et al. 2012) suggests that increased
43
firm agent “we” pronoun use may be beneficial if it refers to the agent and the customer
(inclusive “we”), rather than the firm entity (exclusive “we”) reference use observed in the field.
Also, as noted previously, subject references using the exclusive “we” cannot and should
not always be replaced by references to “I” the firm agent. This boundary is likely to occur in
cases where (a) the firm agent asserts a policy arising from a group/entity (i.e., the firm), or (b)
when action is taken by a group/entity rather than an individual. Independent judges examined a
sample (N = 200) from the Study 2 field data and found that this boundary occurred in less than
one in five (17.7%) of the “we” pronoun use occasions. Further, infrequent firm agent use of “I”
(relative to natural language) and the ease of its substitution or inclusion in the majority of the
interactions we observe, strongly suggests that firms should encourage agents to refer to the self
when interacting with customers more frequently than they do today.
Of course, the effects observed here may depend on English language conventions. Future
research could examine whether they hold for languages that have more complex pronoun
reference structures (e.g., French, German, Japanese; Kashima and Kashima 1998).
Having established the important consequences of firm agent pronoun use, new research
might also examine why firm agents avoid referring to the self (“I”) in the first place. While the
results of our pilot study and managerial publications (e.g., Bacal 2012, Rudick and O’Flahavan
2002) suggest that this may be strategic,15 we did not empirically demonstrate the source of such
a strategy. In addition to the apparently widely-held (and errant) belief that the pattern of
pronoun use we observe in the field is optimal, agents may be explicitly trained to emphasize
“you” and we” references and avoid self-referencing (“I"). Alternatively, the firm agent’s
tendency to refer to “we” the firm rather than “me” may be driven by the strength of the agent’s
own organizational identification (Smidts, Pruyn, and Riel 2001), a desire to evince the firm’s
15
See also Pilot Study 2 in the Web Appendix.
44
orientation towards the customer (Saxe and Weitz 1982; Zablah et al. 2012), and/or a sense of
closeness or commitment to their employer (Agnew et al. 1998). Future research examining the
drivers of firm agent pronoun emphasis may fruitfully examine such issues.
What’s more, the present research opens the door to examination of the customer’s use of
personal pronouns in their communications to firm agents as a potential indicator of the
customer’s mindset. “You” pronoun use by the customer might indicate blame and anger toward
the firm, while more “I” pronouns might indicate self-reflection and sadness or disappointment
(Rude, Gortner, and Pennebaker 2004). More broadly, examining customer pronoun use as an
attitudinal or behavioral signal—or as a consequence of marketing mix elements—may offer a
prime avenue for future research. For example, while firms are increasingly interested in
assessing consumer sentiment by analyzing the valence of firm-related consumer chatter online
(Henschen 2012), measures of customer’s pronoun use in Study 4 were stronger predictors of
purchase outcomes than LIWC’s positive and negative emotion measures, which are similar to
measures used in sentiment analysis.
Finally, future work could examine other language categories that may either directly
impact marketing outcomes or moderate the effect of “I” pronoun use in marketing
communications. While our effects were robust to natural variation in the language that
accompanied the personal pronoun categories we consider (Studies 2 - 4), explaining language
(Moore 2012), temporal focus (Chen and Lurie 2013; Pennebaker, Mehl, and Niederhoffer
2003), or specific emotion word use (e.g., anger, anxiety; Yin, Bond, and Zhang 2014), could
play an important role in customer outcomes following firm agent or other marketing
communications.
45
REFERENCES
Agnew, Christopher R., Paul A. M. Van Lange, Caryl E. Rusbult, and Christopher A. Langston
(1998), “Cognitive interdependence: Commitment and the mental representation of close
relationships,” Journal of Personality and Social Psychology, 74(4), 939-954.
Ahearn, Laura M. (2001), “Language and agency,” Annual Review of Anthropology, 109-137.
Bacal, Robert (2011), Perfect Phrases for Customer Service, second edition, New York, NY:
McGraw-Hill.
Barasch, Alixandra, and Jonah Berger (2014), “Broadcasting and Narrowcasting: How
Audience Size Impacts What People Share,” Journal of Marketing Research, 51(3),
286-299.
Basch, Michael D. (2003). Customer Culture: How FedEx and Other Great Companies Put the
Customer First Every Day. London, UK: Financial Times Press.
Bitner , Mary Jo (1990), “Evaluating Service Encounters: The Effects of Physical Surroundigns
and Employee Responses,” Journal of Marketing, 54 (April), 69-82.
Bolton, Ruth N. (1998), “A dynamic model of the duration of the customer's relationship with a
continuous service provider: the role of satisfaction,” Marketing Science, 17(1), 45-65.
Brady, Michael K., and J. Joseph Cronin (2001), “Customer orientation effects on customer
service perceptions and outcome behaviors,” Journal of Service Research, 3(3), 241-251.
Buttny, Richard (1993), Social Accountability in Communication, Newbury Park, CA: Sage.
Campbell, Kim S., Kathryn Riley and Frank Parker (1990), “You-Perspective: Insights from
Speech Act Theory”, Journal of Technical Writing and Communication, 20(2), 189-199.
Chan, Elaine, and Jaideep Sengupta (2010), “Insincere flattery actually works: A dual attitudes
perspective,” Journal of Marketing Research, 47(1), 122-133.
Charlton, Graham (2011), “44% prefer email for customer service: survey,”
http://econsultancy.com/ca/blog/7999-44-prefer-email-for-customer-service
Chen, Adrien (2011), “Working at the Apple Store: Tales from the Inside,”
46
http://gawker.com/5812622/working-at-the-apple-store-tales-from-the-inside.
Chen, Zoey, and Nicholas H. Lurie (2013), “Temporal Contiguity and Negativity Bias in the
Impact of Online Word-of-Mouth,” Journal of Marketing Research, 50(4), 1-51.
Chung, Cindy and James W. Pennebaker (2007), “The psychological functions of function
words,” in Social Communication, K. Fiedler, ed., Chicago, IL: Psychology Press.
Duncan, Tom, and Sandra E. Moriarty (1998), “A communication-based marketing model for
managing relationships,” Journal of Marketing, 62(2), 1-13.
Dutta, Sumair 2012, “Trends in Customer Service,” http://voxeo.com/wpcontent/uploads/2013/01/TrendsCustSvc-Aberdeen-2012.pdf
Evenson, Renee (2011), “Customer Service Management Training 101: Quick and Easy
Techniques that Get Results.” New York: AMACOM (American Marketing Association).
Fahnestock, Jeanne (2011). Rhetorical Style. The Uses of Language in Persuasion. New York,
NY: Oxford University Press.
Fitzsimons, Gráinne M., and Aaron C. Kay (2004), “Language and interpersonal cognition:
Causal effects of variations in pronoun usage on perceptions of closeness,” Personality
and Social Psychology Bulletin, 30(5), 547-557.
Goffman, Erving (1981). Forms of Talk. Philadelphia, PA: University of Pennsylvania Press.
Gordon, Peter C., Barbara J. Grosz, and Laura A. Gilliom (1993), “Pronouns, names, and the
centering of attention in discourse,” Cognitive Science, 17(3), 311-347.
Halliday, M. A. K. (1985), Spoken and Written Language. Australia: Deakin University Press.
Henschen, Doug (2012), “Sentiment Analysis: How Companies Now Listen to the Web,”
Information Week, http://www.informationweek.com/it-leadership/sentiment-analysishow-companies-now-listen-to-the-web/d/d-id/1104836.
Homburg, Christian, Wayne D. Hoyer, and Martin Fassnacht (2002), “Service orientation of a
retailer’s business strategy: dimensions, antecedents, and performance
outcomes,” Journal of Marketing, 66(4), 86-101.
Hsieh, Tony (2010), “Zappos’ CEO on Going to Extremes for Customers,” Harvard Business
47
Review, 88(7), 41-45.
Ickes, William, Susan Reidhead, and Miles Patterson (1986), “Machiavellianism and selfmonitoring: As different as “me” and “you”,” Social Cognition 4(1), 58-74.
Ickes, William, Linda Stinson, Victor Bissonnette, and Stella Garcia (1990), “Naturalistic social
cognition: Empathic accuracy in mixed-sex dyads,” Journal of Personality and Social
Psychology, 59(4), 730-42.
Inigo-Mora, Isabel (2004), “On the use of the personal pronoun we in communities,” Journal of
Language and Politics, 3(1), 27-52.
Ingrain, Thomas, Charles H. Schwepker, and Don Hutson (1992), “Why Salespeople Fail,”
Industrial Marketing Management, 21, 225-230.
Johnson-Laird, P. N. (1968), “The interpretation of the passive voice,” Quarterly Journal of
Experimental Psychology, 20, 69–73.
Kashima, Emiko S., and Yoshihisa Kashima (1998), “Culture and Language: The Case of
Cultural Dimensions and Personal Pronoun Use,” Journal of Cross-Cultural Psychology,
29(3), 461-486.
Kronrod, Ann and Shai Danziger (2013), “’Wii Will Rock You!” The Use and Effect of
Figurative Language in Consumer Reviews of Hedonic and Utilitarian Consumption,”
Journal of Consumer Research, 40(4), 726-739.
Kronrod, Ann, Amir Grinstein, and Luc Wathieu (2011), “Enjoy! Hedonic Consumption and
Compliance with Assertive Messages,” Journal of Consumer Research, 39 (June), 51-61.
Kurtz, Jill (2015), “How to Write Email Templates That Don’t Make Customer Service Agents
Sound Like Robots,” http://www.ewriteonline.com/how-to-write-email-templates-thatdont-make-customer-service-agents-sound-like-robots.
Kutner, Michael, Christopher Nachtsheim, John Neter, and William Li (2004). Applied Linear
Regression Models, 4th edition. New York, NY: McGraw-Hill.
Manchanda, Puneet, Grant Packard, and Adithya Pattabhiramaiah (2013), “Social Dollars: The
Economic Impact of Customer Participation in a Firm-Sponsored Online Customer
48
Community,” Marketing Science, 34 (3), 367-387.
Marinova, Detelina, Jun Ye and Jagdip Singh (2008), “Do Frontline Mechanisms Matter? Impact
of Quality and Productivity Orientations on Unit Revenue, Efficiency, and Customer
Satisfaction,” Journal of Marketing, 72 (March), 28-45.
Maxham, James G., III and Richard G. Netemeyer (2002), “A Longitudinal Study of
Complaining Customers’ Evaluations of Multiple Service Failures and Recovery
Efforts,” Journal of Marketing, 66(4), 57-71.
Moore, Sarah G. (2012), “Some Things Are Better Left Unsaid: How Word of Mouth Influences
the Storyteller,” Journal of Consumer Research, 38(6), 1140-1154.
Packard, Grant, and David B. Wooten (2013), “Compensatory knowledge signaling in
consumer word-of-mouth,” Journal of Consumer Psychology, 23(4), 434-450.
Palmatier, Robert W., Lisa K. Scheer, and Jan-Benedict Steenkamp (2007), “Customer loyalty
to whom? Managing the benefits and risks of salesperson-owned loyalty,” Journal of
Marketing Research, 44(2), 185-199.
Patrick, Vanessa M., and Henrik Hagtvedt (2012), “‘I Don’t’ versus ‘I Can’t’: When
Empowered Refusal Motivates Goal-Directed Behavior,” Journal of Consumer
Research, 39, 371-81.
Penelope, Julia (1990). Speaking freely. New York: Pergamon Press.
Pennebaker, James W. (2011). The secret life of pronouns: What our words say about us.
New York, NY: Bloomsbury Press.
Pennebaker, J. W., Chung, C. K., Ireland, M., Gonzales, A., & Booth, R. J. (2007). The
development and psychometric properties of LIWC2007. Austin, TX, LIWC.net.
Pennebaker, James W., Matthias R. Mehl, and Kate G. Niederhoffer (2003), “Psychological
aspects of natural language use: Our words, our selves,” Annual Review of Psychology,
(54)1, 547-577.
Preacher, Kristopher J., and Andrew F. Hayes (2008), “Asymptotic and resampling strategies
for assessing and comparing indirect effects in multiple mediator models,” Behavior
49
Research Methods, 40(3), 879-891.
Ramani, Girish, and Vipin Kumar (2008), “Interaction orientation and firm performance,”
Journal of Marketing, 72(1), 27-45.
Rude, Stephanie, Eva-Maria Gortner, and James Pennebaker (2004), “Language use of
depressed and depression-vulnerable college students,” Cognition & Emotion, 18(8),
1121-1133.
Rudick, Marilynne and Leslie O’Flahavan (2002), Clear, Correct, Concise E-Mail: A Writing
Workbook for Customer Service Agents, Silver Spring, MD: E-Write.
Rust, Roland T., and Tuck Siong Chung (2005), “Marketing models of service and
relationships,” Marketing Science, 25(6), 560-580.
Saxe, Robert, and Barton A. Weitz (1982), “The SOCO scale: a measure of the customer
orientation of salespeople,” Journal of Marketing Research, 19(3), 343-351.
Schellekens, Gaby A. C., Peeter Verlegh, and Ale Smidts (2010), “Language Abstraction in
Word of Mouth,” Journal of Consumer Research, 37(2), 207-23.
Scherwitz, Larry, Kent Berton, and Howard Leventhal (1978), “Type A behavior, selfinvolvement, and cardiovascular response,” Psychosomatic Medicine, 40(8), 593-609.
Schlenker, Barry R., and Michael F. Weigold (1992), “Interpersonal Processes Involving
Impression Regulation and Management,” Annual Review of Psychology, 43, 133-168.
Schuster, Camille P., and Jeffrey E. Danes (1986), “Asking Questions: Some Characteristics of
Successful Sales Encounters,” Journal of Personal Selling & Sales Management, 6(1),
17-27.
Sela, Aner S., Christian Wheeler, and Gülen Sarial-Abi (2012), “We Are Not the Same as You
and I: Causal Effects of Minor Language Variations on Consumers’ Attitudes toward
Brands,” Journal of Consumer Research, 39(3), 644-661.
Simmons, Rachel A., Peter C. Gordon, and Diane L. Chambless (2005), “Pronouns in Marital
Interactions,” Psychological Science, 16 (12), 932-936.
Singh, Jagdip and Deepak Sirdeshmukh (2000), “Agency and Trust Mechanisms in Consumer
50
Satisfaction and Loyalty Judgments,” Journal of the Academy of Marketing Science, 28
(1), 150-167.
Smidts, Ale, Ad Th. H. Pruyn and Cees B. M. Van Riel (2001), “The Impact of Employee
Communication and Perceived External Prestige on Organizational Identification,”
Academy of Management Journal, 44 (5), 1051-1062.
Tanner, Robin J., Rosellina Ferraro, Tanya L. Chartrand, James R. Bettman, and Rick Van
Baaren (2008), “Of chameleons and consumption: The impact of mimicry on choice and
preferences,” Journal of Consumer Research, 34(6), 754-766.
Tausczik, Yla R., and James W. Pennebaker (2010), “The psychological meaning of words:
LIWC and computerized text analysis methods,” Journal of Language and Social
Psychology, 29(1), 24-54.
van Osselaer, Stijn M. J. (2016), “The Power of Personal,” presidential address at the 2016
Society for Consumer Psychology Winter Conference, St. Pete Beach, Florida.
Vargo, Stephen L., and Robert F. Lusch (2004), “Evolving to a new dominant logic for
marketing,” Journal of Marketing, 68(1), 1-17.
Wales, Katie (1996). Personal pronouns in present-day English. Cambridge University Press.
Weitz, Barton A., Harish Sujan, and Mita Sujan (1986), “Knowledge, motivation, and adaptive
behavior: a framework for improving selling effectiveness,” Journal of Marketing, 50,
174-191.
Yin, Dezhi, Samuel Bond, and Han Zhang (2014), “Anxious or angry? Effects of discrete
emotions on the perceived helpfulness of online reviews,” MIS Quarterly, 38(2), 53960.
Zableh, Alex R., George R. Franke, Tom J. Brown, and Darrell E. Bartholomew (2012), “How
and When Does Customer Orientation Influence Frontline Employee Job Outcomes? A
Meta-Analytic Evaluation,” Journal of Marketing, 76 (May), 21-40.
Zeithaml, Valarie A., Leonard L. Berry, and Ananthanarayanan Parasuraman (1996), “The
behavioral consequences of service quality,” Journal of Marketing, 60(2), 31-46.
51
APPENDIX
PILOT STUDY STIMULI
Imagine you are a customer service employee for an online retailer called Shopsite.com. Below
is an email from a customer.
Hello, I am wondering about Shopsite.com’s international shipping policies. Can I buy
something on my credit card and have it shipped to a person in a different country as a
gift? How would returns work in that case?
It is very easy [very difficult] for your company to handle this kind of issue.
Which of the following responses would be most appropriate to write to the customer in reply?
[Condition 1 response options]
[ ] Happy to help answer this question. Unfortunately, international shipping and returns
aren’t offered at this time. Thanks for understanding.
[ ] Happy to help answer your question. Unfortunately, international shipping and returns
aren’t offered at this time. Thank you for understanding.
[Condition 2 response options]
[ ] I am happy to help answer this question. Unfortunately, international shipping and
returns aren't offered at this time. My thanks for understanding.
[ ] We are happy to help answer this question. Unfortunately, international shipping and
returns aren't offered at this time. Our thanks for understanding.
52
STUDY 1 STIMULI
Inquiry Email
Hello, I am wondering about your return policy. Can I buy something on my credit card and have
you ship it to a person in a different country as a gift? If so, how would you handle the return if
she doesn’t like it? She shouldn’t have to pay to return it. If it’s not free, can return shipping go
on my credit card so she doesn’t have to pay? Thanks in advance for the help. --Bob
Complaint Email
Hello, I’m having trouble finding a product at your website. I never write letters like this but
your site interface is clunky and hard to navigate on a touchscreen tablet. I think you should also
consider changing the fonts to something easier to read. The links were hard to find and took me
several tries to click on. The search engine doesn’t seem to work, so I have to use the category
links to find anything. --Bob
TABLE A1: LIWC PERSONAL PRONOUN CATEGORIES
“I”
“you”
“we”
I
Id
I'd
I'll
Im
I'm
Ive
I've
Me
Mine
My
Myself
You
Youd
You'd
Youll
You'll
Your
Youre
You're
Yours
Youve
You've
Lets
Let's
Our
Ours
Ourselves
Us
We
We'd
We'll
We're
Weve
We’ve
53
LANGUAGE SAMPLE COMPARISONS (STUDY 1)
For the real firm agent email responses collected in Study 1, we tested whether the use of
“I,” “you,” and “we” pronoun emphasis is consistent with base rates in other settings.
For a global natural language comparison, we leverage means and standard deviations from
a large dataset (N > 700,000) that reports global rates of personal pronoun use from over 70
linguistic psychology studies (Pennebaker et al. 2007). As summarized in Table A2 (sample A),
firm agents in the present study were significantly more likely to use “you” and “we” pronouns,
and significantly less likely to use “I” pronouns than this global sample. Pennebaker and
colleagues (2007) report means suggesting that modality (e.g., live oral (Table A2, sample B) vs.
written online (Table A2, sample C)) does not produce variations in the global distribution of
pronoun use across “I,” “you,” and “we” pronouns.
We also consider three comparison samples in which the speaker is, like the firm agent in
our participation framework, responding to another individual. One might expect that this
interaction setting would differ from the global mean, with pronoun use shifting towards the
party who initiated the conversation (“you”). Consistent with this thinking, we find an emphasis
on “you” pronouns (relative to “I” and “we”) in firm-agent oral responses to customer inquiries
in a sample of 95 telephone interactions from the online retailer who provided the data for Study
4 (Table A2, sample D) and a sample of 80 advice columnist responses from four major
syndicated advice columns16 (Table A2, sample E). However, consumer responses to posts made
by other consumer participants at an Internet-based parenting forum about baby products (Table
A2, sample F) retain an “I” emphasis, following a pronoun use distribution similar to the natural
language mean. Taken together, these three samples suggest that a “you” emphasis may be
normative in customer-service settings involving professional or expert responses to others (i.e.,
firm agents or advice columnists), while casual conversation retains an emphasis on “I”
pronouns, even when the speaker is responding to the needs of another (Table A2, sample F).
16
20 random samples from each of Ask E. Jean (Elle magazine), Miss Manners (Washington Post), Dear Prudence
(Slate), and Ask a Dude (Hairpin.com). Pronoun use means did not vary significantly across these advice columns.
54
TABLE A2: PRONOUN USE OF REAL FIRM AGENTS
VERSUS COMPARISON SAMPLES (STUDY 1)
Corpus
Source
LIWC
Pers.
Pronoun
Category
LIWC
Statistic
Mean
Welch t-stat
(Focal vs.
Comparison Sample)
Focal Sample
Firm agent responses to
bogus customer emails
(N = 40)
Study 1
"I"
"You"
"We"
0.94
6.04
4.83
A. English language
global mean
(N = 721,726)
Pennebaker
et al. (2007)
"I"
"You"
"We"
5.72
1.18
0.76
B. Oral conversation in unstructured real world settings
(N = 2,014)
Pennebaker
et al. (2007)
"I"
"You"
"We"
6.30
3.94
1.09
-- ^
-- ^
-- ^
C. Writing by Internet-based
bloggers and posters
(N = 9,537)
Pennebaker
et al. (2007)
"I"
"You"
"We"
6.42
1.23
0.88
-- ^
-- ^
-- ^
D. Oral firm agent responses to real
customer telephone inquiries
(N = 95)
Firm used
in Study 4
"I"
"You"
"We"
3.78
7.03
1.09
9.54 ***
17.57 ***
0.54
E.
Written responses to questions
asked of syndicated advice
columnists (N = 80)
Four syndicated
advice columns
"I"
"You"
"We"
1.59
5.67
0.20
1.99 *
11.91 ***
-2.67 **
F.
Written responses to questions
posted on Internet-based forums
(N = 108)
Online parenting
forum
"I"
"You"
"We"
5.21
2.51
0.82
5.85 ***
3.97 ***
1.74 +
Comparison Samples
All LIWC statistic values correspond to proportions (out of 100).
^Standard deviations were not reported for corpus sub-groups in Pennebaker et al (2007).
*** p < .001, ** p < .01, * p < .05, + p < .10
17.99 ***
-14.57 ***
-9.06 ***
55
REAL AND MODIFIED FIRM AGENT RESPONSES (STUDY 2)
A comparison of personal pronoun use by firm agents from the six firm responses selected
for Study 2 from the full sample of 40 is presented in panel A of Table A3, which shows no
significant differences in the rate of use of these pronouns between the full (Study 1) and reduced
(Study 2) samples. Panel B of the same table shows the impact on LIWC statistics of increasing
“I” (relative to “we”) pronoun use in the modified firm response condition, where there is a
significant increase in “I” pronouns (5.32 vs. 1.07; t(12) = 2.39, p < .05) and a significant
reduction in “we” pronouns relative to the original response (1.49 vs. 5.73; t(12) = -3.00, p <
.05). Note also that the LIWC statistics for the modified versions remain within the distribution
of pronoun use in natural language reported earlier in Table A2, and indeed appear closer to
these distributions than the original firm response versions.
TABLE A3: SUMMARY STATISTICS FOR REAL AND MODIFIED FIRM AGENT
RESPONSES (STUDY 2)
A
LIWC Personal
Pronouns Category
Full Sample
(N = 40)
Sub-sample
(N = 6)
Welch's
t-stat
"I"
"You"
"We"
0.94 (1.68)
6.04 (2.11)
4.83 (2.84)
1.07 (1.73)
7.23 (1.75)
5.73 (2.39)
0.17
1.51
0.84
LIWC Personal
Pronouns Category
Firm agent response condition
Original
Modified
(N = 6)
(N = 6)
t-stat
"I"
"You"
"We"
1.07 (1.73)
7.23 (1.75)
5.73 (2.39)
2.39 *
0.00
-3.00 *
B
Standard deviations reported in parentheses.
** p < .01, * p < .05, + p < .10
5.32 (3.03)
7.23 (1.75)
1.49 (0.80)
56
LANGUAGE TYPICALITY CHECKS (STUDIES 2 AND 3)
It is possible that any positive effects of increasing “I” pronoun use could stem from
making firm language more consistent with natural language in the modified firm response, since
Study 1 shows that unmodified firm agent language differs significantly from natural language
samples. In this case, our modified firm responses would conform to natural language use norms,
which could lead firm responses to be viewed as more typical and generate positive responses.
However, an alternative prediction can also be made. If, as indicated in the pilot study,
consumers have expectations that firms should talk about “you” and “we” but not “I” in
customer-firm interactions, our modified firm responses could violate language norms (Kronrod
and Danziger 2013; Kronrod et al. 2011). This may lead them to be perceived as atypical and
generate negative responses, opposite to our predictions.
We examined these possibilities in two pretests comparing typicality across the stimuli
used in Studies 2 and 3. Pretest participants were asked to indicate whether the language of a
single email response stimuli was typical, expected, and standard (1 = not at all, 7 = very much;
Kronrod et al. 2011). Pretest sample size and Cronbach’s Alpha for the three-item scale are
shown in Table A4. For Study 2, contrasts of original versus modified stimuli found no
differences in typicality with one exception, for which the modified stimuli was marginally more
typical than the original firm agent response (Table A5). For Study 3, an omnibus ANOVA of
the three level pronoun condition (none, “I,” “You”) and 2 level verb voice (active vs. passive)
revealed a significant main effect for verb voice (F(1, 320) = 7.01, p < .01) but no effect for
pronoun condition (F(2, 320) = 1.03, p = .36) or the interaction of voice and pronoun condition
(F(2, 320) = .13, p = .88). Active voice was more typical (M = 5.24) than passive voice (M =
4.81). However, as discussed in Study 3, verb voice had no effect on our dependent measures.
57
TABLE A4: LANGUAGE TYPICALITY PRETEST SAMPLE SIZES
AND SCALE RELIABILITIES (STUDIES 2 AND 3)
Study
2
3
N
376
326

0.87
0.92
TABLE A5: LANGUAGE TYPICALITY MEANS
AND STATISTICAL TESTS (STUDIES 2 AND 3)
Study 2
Original
Email
Typicality
4.95
5.74
5.74
6.33
5.84
5.85
Firm
email
A
B
C
D
E
F
Modified
Email
Typicality
5.84
5.94
5.48
6.03
6.20
5.56
t-stat
(row contrasts)
0.39
-0.08
0.95
1.58
-1.68 +
1.30
Study 3
Pronouns
None
I
You
None
I
You
Verb voice
Active
Active
Active
Passive
Passive
Passive
** p < .01, * p < .05, + p < .10
Typicality
5.41
5.03
5.32
4.79
4.71
4.85
F-test
(omnibus)
7.06 **
58
STUDY 3 STIMULI
Pronoun
Verb
Condition Voice
Firm Agent Email Stimuli
None
Active
The order is leaving the warehouse. It will arrive in 3-5 days. Apologies for the unacceptable delay
experienced. For further assistance, just reply to this email.
I
Active
I found that the order is leaving the warehouse. It will arrive in 3-5 days. My apologies for the unacceptable
delay experienced. If I can provide further assistance, just reply to this email.
You
Active
Your order is leaving the warehouse. It will arrive in 3-5 days. Apologies for the unacceptable delay you've
experienced. If you need further assistance, just reply to this email.
None
Passive
The order was found to be leaving the warehouse. It will arrive in 3-5 days. The delay experienced is
unacceptable. Apologies. If further assistance can be provided, just reply to this email.
I
Passive
The order was found by me to be leaving the warehouse. It will arrive in 3-5 days. The delay experienced is
unacceptable. My apologies. If further assistance can be provided by me, just reply to this email.
You
Passive
Your order was found to be leaving the warehouse. It will arrive in 3-5 days. The delay you experienced is
unacceptable. Apologies. If further assistance can be provided to you, just reply to this email.
59
TABLE A6: SUMMARY STATISTICS FOR REGRESSION MODEL TERMS (STUDY 4)
Mean
SD
Min
Max
Full sample Pronoun Use (N = 2,098)
Customer
"I"
"You"
"We"
5.38
3.63
0.94
4.86
3.01
1.65
0
0
0
30.00
25.00
14.29
Firm agent
"I"
"You"
"We"
1.93
6.19
3.12
3.85
3.30
3.25
0
0
0
26.32
19.30
10.00
Transactional account sample Pronoun Use (N = 1,277)
Customer
"I"
"You"
"We"
5.68
3.55
0.90
4.95
2.94
1.64
0
0
0
30.00
25.00
14.29
Firm agent
"I"
"You"
"We"
1.83
6.28
3.11
3.78
3.27
2.44
0
0
0
26.32
19.15
11.76
559.10
3.00
3.83
1.07
5.75
0.12
0.01
0.60
0.24
0.04
0.53
0.29
0.08
0.65
1536.53
1.92
2.64
1.46
1.52
0
2
0
0
1
0
0
0
0
0
0
0
0
0
35850.00
22.00
25.00
16.67
7
1
1
1
1
1
1
1
1
1
Model covariates
Purchase volumei,pre
# of emails
Customer posemo
Customer negemo
Complaint
Resolution
Compensation
Order reason
Website reason
Multi reason
Region 1
Region 2
Region 3
Female
60
STUDY 4 ANCILLARY ANALYSIS
ASSESSING MULTICOLLINEARITY IN PERSONAL PRONOUN USE BY
CUSTOMERS AND FIRM AGENTS BY INTERACTION
As expected, the correlation matrix of personal pronoun use by the customer and firm
agent in a given interaction indicate significant interactive effects (Table A7). Specifically, if the
customer initially emphasized the self (the firm and/or its agent), the firm agent’s reply reflected
this conversational focus through a personal pronoun emphasis on the customer (the firm or
themselves).
Variance inflation factors for the full regression models in Study 4 indicate the
consequences of potential multicollinearity are not severe. All predictors and covariates fell
below the VIF threshold of 10 (Kutner et al. 2004), suggesting that the variance of model
coefficients was not substantially increased due to collinearity.
TABLE A7: CORRELATIONS IN PRONOUN USE
IN REAL CUSTOMER-FIRM INTERACTIONS (STUDY 4)
1
2
3
4
5
6
Customer “I:
Customer “You”
Customer “We”
Firm agent “I”
Firm agent “You”
Firm agent “We”
1
2
3
4
5
1.00
-0.50
-0.56
-0.52
0.45
0.43
1.00
0.53
0.51
-0.37
-0.40
1.00
0.57
-0.44
-0.49
1.00
-0.51
-0.59
1.00
0.43
All correlations are significant at p < .01.
61
WEB APPENDIX
PILOT STUDY 2:
LAY THEORIES ABOUT DESIRED AND ACTUAL FIRM AGENT PRONOUN EMPHASIS
American participants (N = 498) in a paid online panel were asked what a firm employee
(agent) should talk about in response to customer inquiries or complaints for each of the three
personal pronoun categories (“I”, “you”, and “we”). Specifically, participants rated the extent to
which firm employees should talk about: (1) “your” (the customer’s) question or complaint, (2)
how “we” (the firm) can address the question or complaint, and (3) how “I” (the employee) can
address the question or complaint (1 = Not at all, 7 = Very much; question order randomized).
Participants then indicated whether they had currently or previously been employed as (a)
managers of other people, (b) managers of customer sales or service representatives, and/or (c)
customer sales or service representatives themselves, either in-person or remotely (e.g., internet,
phone). Finally, participants reported their tenure in these roles in years and the maximum
number of people they had supervised in that role.
Results indicated that for the full sample and for each employment history sub-category,
participants believed firm agents should focus more on “you” (the customer) than “I” (the firm
agent), and more on “we” (the firm) than “I” (the firm agent). There were no differences in the
extent to which participants believed firm agents should focus on “you” (the customer) versus
“we” (the firm). Neither the participant’s years of experience in a given customer service or
managerial role, nor the number of people they had managed, were related to these beliefs (see
Table W1).
TABLE W1: DESIRED PRONOUN EMPHASIS IN CUSTOMER-FIRM INTERACTIONS
(PILOT STUDY 2)
Firm agents SHOULD talk about…
Sample
N
Managers (excl. customer service)
Customer service managers
In-person service representatives
Remote service representatives
None of the above (“consumers”)
Full sample
122
94
226
113
154
498
…how “I” (the
firm agent)
can address
the subject.*
5.20
5.35
5.22
5.19
5.10
5.17
A
A
A
A
A
A
…”your” (the
customer's)
subject.*
6.09
5.83
5.96
6.01
5.86
5.94
B
B
B
B
B
B
...how “we”
(the firm) can
address the
subject.*
5.98
5.98
6.08
5.97
5.84
5.96
Row differences p < .05 are indicated by different alphabetical superscripts.
*See text for exact question wording. All means are on a seven-point scale; 1 = not at all, 7 = very much.
B
B
B
B
B
B
62
In addition to responding to the question on who firm agents should talk about in
response to customer inquiries or complaints presented in the survey reported in the paper, we
asked Study 2 participants to indicate the extent to which firm agents do talk about each of the
three pronoun categories (1 = not at all, 7 = very much). The pattern of results is the same.
Participants believe firm agents talk about how “we” (the firm) can address “your” (the
customer’s) needs more than firm agents talk about how “I” personally can address the subject
(Table W2).
TABLE W2: PERCEIVED PRONOUN EMPHASIS IN CUSTOMER-FIRM INTERACTIONS
(PILOT STUDY 2)
Sample
n
Managers (excl. customer service)
Customer service managers
In-person service representatives
Remote service representatives
None of the above ("consumers")
Full
122
94
226
113
154
498
Firm agents DO talk about…
…how "I" (the
...how "we"
…"your" (the
firm agent)
(the firm) can
customer's)
can address
address the
subject.
the subject.
subject.
5.20
5.17
4.81
4.52
4.72
4.73
A
A
A
A
A
A
6.09
5.86
5.68
5.44
5.61
5.59
B
B
B
B
B
B
5.98
6.14
5.86
5.86
5.58
5.73
B
B
B
B
B
B
Row differences p < .05 are indicated by different alphabetical superscripts.
Taken together, and consistent with the pilot study reported in the main paper, Pilot Study
2 suggests that managers, firm agents, and consumers believe that firm agents both should and
do in fact emphasize “you” (the customer) and “we” (the firm) in their interactions with
customers, and speak relative less about how “I” (the firm agent) can address the customer’s
inquiry.
63
STUDY 5:
THE IMPACT OF PRONOUN USE ON CUSTOMER PERCEPTIONS WHEN
REVERSING THE CUSTOMER-FIRM PARTICIPATION FRAMEWORK
As discussed in the paper, firm agent use of self-referencing “I” pronouns as the
grammatical subject and firm agent use of “you” pronouns to refer to the customer as the
grammatical object were the dominant participation framework pattern observed in our data. The
basic premise of customer sales and service interactions and customer orientation theory further
suggest that this framework is normative. That is, the firm agent is an actor (subject) who
attempts to assist, satisfy, or sell products to the customer (object).
In this ancillary study, we test the impact of firm agent use of “you” pronouns referring to
the customer as subject rather than object in their communications. We orthogonally manipulate
whether the firm agent and customer are treated as the object or subject of a firm agent
statement. We expect that treating the customer as subject will produce lower satisfaction and
purchase intentions than treating the firm agent as subject. We also expect that treating the firm
agent as object will produce lower satisfaction and intention outcomes than treating the customer
as object. In short, reversing the dominant participant framework of our interaction setting
should have negative effects.
Participants, Design and Procedure
Undergraduate students (N = 258) participated for partial course credit. Participants were
asked to imagine themselves as a customer who had emailed a firm to inquire about how to
return a product they had purchased at the firm’s website. Participants were presented with one
of four different versions of the firm agent’s response to their inquiry (see Table W3), which
either presented the firm agent as subject (condition 1), the customer as subject (condition 2), the
firm agent as object (condition 3), or the customer as object (condition 4).
Best efforts were made to modify the stimuli to avoid changes in meaning beyond the
subject/object relationship. We note, however, that this is difficult to achieve when the
participation framework is reversed due to grammatical subject or object re-assignment. As
effects for active versus passive verb voice were null in Study 3, and similar meaning across
conditions was difficult to achieve in the present study without altering verb voice, we do not
fully control for it here. This should also be of minimal concern because the contrasts that test
our predictions do not compare across stimuli with different verb voices. Specifically, our central
contrasts are between (a) the two conditions that do not use passive voice at all (conditions 1 and
2), and (b) the two conditions that include a single clause using passive voice (conditions 3 and
4).
After participants read the firm agent’s response to their inquiry, they indicated their
satisfaction with the agent ( = .83) and their purchase intentions towards the firm ( = .84). We
then collected empathy ( = .91), agency ( = .91), and language typicality ( = .87). All items
are the same as those used in Studies 2 and 3.
Results and Discussion
No variation in language typicality was found across the four conditions (F < 1).
Planned contrasts revealed significantly higher satisfaction (4.81 vs. 3.98; F(1, 85) = 6.69,
p = .01) and purchase intentions (4.74 vs. 3.96; F(1, 84) = 6.58, p = .01) when the firm agent
64
rather than customer was presented as the actor (grammatical subject) in the firm agent’s
response. There was no difference in customer satisfaction or intentions based on whether the
customer or firm agent was treated as the recipient of action (grammatical object; Fs < 1, ps >
.85).
We also replicated the underlying drivers (empathy, agency) of the positive effect of firm
agent “I” pronouns on customer satisfaction and intentions. Similar to our results in Study 2, the
positive impact of making the firm agent the subject (vs. the customer as subject) on our
dependent measures was mediated by both the perceived empathy (satisfaction CI: .36 - .99, p <
.05; intentions CI: .22 - .84, p < .05) and agency of the agent (satisfaction CI: .01 - .62, p < .05;
intentions CI: .05 - .84, p < .05).
Overall, this pattern of results suggests that the firm agent should generally be presented as
the grammatical subject when responding o a customer inquiry such as the one operationalized
here. The finding that firm agent references to “you” (the customer) as the grammatical object
were no better than references to “me” (the agent) as grammatical object to some extent
replicates the findings of prior studies: a null effect for a customer-oriented “you” emphasis. This
result also leaves open the possibility that firm agents may sometimes refer to themselves as the
recipient of action by the customer without adverse effects on customer attitudes and intentions.
TABLE W3: STUDY 5 STIMULI
Condition
Firm
Agent
Customer
Firm Agent Email Stimuli
1
Subject
--
If I can have your username, I can look into the
account to find the order and return options
through the website. Once I’ve done this, I
should be able to return the product right away.
2
--
Subject
If you have your username, you can look into the
account to find the order and return options
through the website. Once you’ve done this, you
should be able to return the product right away.
3
Object
--
If your username is available, the account can be
looked into to find the order and return options
through the website. Once this is done for me,
the product should be able to be returned to me
right away.
4
--
Object
If your username is available, your account can
be looked into to find your order and return
options through the website. Once this is done
for you, the product should be able to be returned
for you right away.