“Make it Happen!”

Verbs and Agency

Verbs are words that often describe an action. This function is so deeply embedded in our conception of verbs that it remains attached to the grammatical class beyond and above the conveyed meaning. Accumulated evidence documents an embodied link between verbs and denoting activity, as prototypical verbs are associated with actions (Vigliocco et al., 2011). This link extends also to responsiveness towards motivating messages. Girls were more likely to engage in scientific research when encouraged with the use of a verb “Let’s do science” rather than an identity (noun-related) slogan “Let’s be scientists!” (Rhodes et al., 2019).

Importantly, the role of verbs as a grammatical category extends beyond marking or evoking activity, as it reaches the social potential of action, namely agency: a basic dimension of human perception signaling dynamism and goal achievement (for an overview, see Abele & Wojciszke, 2014). This meta-semantic effect of verbs as a grammatical category – extending beyond the content of what is said – has been termed verb-agency link (Formanowicz et al., 2017) and demonstrated in studies using a pseudo-word paradigm. Pseudo-verbs (e.g., to frol), adjectives (frollative), and nouns (a frol) were created to investigate whether the mere grammatical category can signal the social dimension of the agency. Consistently, pseudo-verbs were evaluated as more agentic than pseudo-adjectives and pseudo-nouns. The meta-semantic effect of verbs was specific to agency and did not extend to ratings of communion – which along agency constitutes the second primary dimension of social perception related to maintaining social relationships (Abele & Wojciszke, 2014). The effect of verbs on the agency but not communion likely suggests that activity is related to both verbs and agency; however, does not reflect a core content of the communion construct. Perceptions of communal qualities (such as friendly, caring) are seen as relatively stable and not easily susceptible to change (Uchronski et al., 2013). That is, wordings “to protect” versus “protection” are likely to be seen as similarly communal; however, the first may be seen as more agentic than the latter due to the usage of the verb.

Moreover, research on the verb-agency link has also provided evidence that verbs, as the grammatical carriers of agency, are more frequently used to describe (stereotypically) agentic than non-agentic targets (Formanowicz et al., 2017). Similarly, eye-tracking studies show that nouns addressing stereotypically agentic roles (e.g., blacksmith) are expected to serve as agents in a sentence, thus be paired with a verb more so, than stereotypically non-agentic role nouns (e.g., beautician; Esaulova et al., 2015). This suggests that the verb-agency link can extend beyond marking agency of individual words and transfer to sentence subjects paired with verbs. For instance, the subject of the sentence “she acts” versus “she is active” is likely to be seen differently with the former seen as more agentic than the latter due to the use of the verb. Altogether, verbs not only evoke agentic associations, but agentic social targets are also paired with increased use of verbs – suggesting a strong verb-agency link. This link may serve as a possible pathway explaining the role of verbs not only in evoking actions but also in marking message effectiveness.

Agency and Message Effectiveness

Priming studies attest to a clear link between making salient the concept of agency and agency-related behaviors in terms of striving, efficiency, and performance. A meta-analytical review found a small, yet robust effect (d = 0.35) of using goal-related words as primes and subsequent goal-oriented behavior (Weingarten et al., 2016). For example, participants solving an achievement-related word puzzle (including words such as “strive” or “master”) put more effort into a subsequent task than participants solving a neutral word task (including words such as “hat” or “window” (Bargh et al., 2001). In a different set of studies (Albarracín et al., 2008), participants primed with action words (e.g., “engage” or “action”) had faster reaction times and better recall of a text than participants primed with non-action words (e.g., “still” or “pause”). Accordingly, the primed agency was linked to higher goal-directed behaviors and performance across studies. However, all of these results focus on semantics rather than comparing grammatical categories of similar meaning (such as verbs vs. adjectives or verbs vs. nouns).

Along similar lines, studies examining a more direct link between agency and message effectiveness found that evoking a personal or group level agency in message recipients (e.g., Eckel et al., 2017; Kessler et al., 2017; Whillans et al., 2017; Whillans & Dunn, 2018) was positively related to participants’ behavioral intention or actual behavior. Referencing recipients’ personal agency by highlighting choice (Eckel et al., 2017), individual gains (White & Peloza, 2009), or a need for an individuals’ effort (Whillans & Dunn, 2018) increased message effectiveness. In some studies; however, the sensitivity to messages evoking personal agency was particularly evident among people for whom the personal agency was assumed to be relevant (those in power and being affluent; Kessler et al 2017; Whillans et al., 2017). These participants reacted with more engagement to messages formulated as “Come forward and take individual action” (Whillans & Dunn, 2018). Those less affluent were more responsive to messages formulated in a way that highlighted communion (Whillans et al., 2017).

Whether participants responded more to messages highlighting agency or communion may depend on the fit between individual values or motivations and the message framing (Cesario et al., 2004; Lee & Aaker, 2004). The individual motivations could be shaped by social class and economic standing (Whillans & Dunn, 2018). Having control over one’s economic situation may correspond to being sensitive towards messages highlighting individuals’ agency, whereas having less control may necessitate turning towards others, which is related to the working class being more collectivist (Markus, 2017), and also likely being affected more by communal appeals. Including a social class variable is thus an important moderator of examining message effectiveness as indicated in previous studies (Whillans et al., 2017, see also Whillans & Dunn, 2018).

One thing that is, however, common for everybody engaging in social actions is that people want to be effective, they want to make a difference. Depending on their social standing they may choose whether this difference can be achieved by their own actions or whether it requires a collective effort as discussed above. For prosocial campaigns, communal or pro bono intentions can be taken for granted and what may matter is whether the organization is able to deliver its promises. The agency of the organization putting the action forward is, therefore, of utmost importance. None of the studies presented above assessed the agency of the organization promoting the campaign. It is not known, therefore, whether agency ascribed to the message sender will also affect message effectiveness and likely recipients’ actions. We propose here that agentic messages should increase message effectiveness by implying that the message sender is agentic and can act upon his or her goals. If message effectiveness is driven by the perception of the message sender as agentic, then this should occur largely independent from the participant’s own social standing.

Yet, all of the studies in the agency-message effectiveness domain were conducted in reference to prosocial entities, thus leaving open the question of whether the same principle applies to the effectiveness of messages put forward by business organizations. One could speculate that efficiency and striving are taken for granted for business companies. Accordingly, it may matter more in the evaluation of message effectiveness whether the company is also oriented toward others. For example, entrepreneur campaigns were more successful on the crowdfunding platform when they included more prosocial words (Pietraszkiewicz et al., 2017). In order to make their messages effective, business companies may therefore want to appear not even more agentic but rather communal.

Overview of the Present Research

Together, the literature suggests that verbs prompt agency and that agency improves message effectiveness. Three issues were, however, neglected. The first focal issue regards the effectiveness of verbs in communication outcomes. Applying syllogistic reasoning one may expect that given the two before-mentioned premises (verbs prompt agency and agency improves message effectiveness), verbs should logically enhance the effectiveness of the messages. In the first archival study, we test whether verbs (vs. adjectives) are in general seen as more agentic. In this way, we replicate prior findings on the verb-agency link in a broader context and extend previous findings on pseudo-words (see Formanowicz et al., 2017) to inferences from real words. Furthermore, in three experiments, we test whether the use of verbs affects message effectiveness. Based on prior research on the verb – agency link (Formanowicz et al., 2017), and on agency and message effectiveness (e.g., Eckel et al., 2017; Whillans & Dunn, 2018), we predict that slogans using verbs will be perceived as more agentic and more effective than slogans relying on adjectives as the focal words.

The relative lack of knowledge regarding the role of grammar in affecting message effectiveness contributes to the second neglected issue that is the interplay of grammar and message content. We focus on the two types of agency that can be conveyed through language. The first is the semantic agency, referring to what we say, for instance by using words representing the core concept of the agency such as accomplishment or goal (Pietraszkiewicz et al., 2019). The second is the grammatic agency, referring to how we talk about things, for example, by using verbs (Formanowicz et al., 2017). These two types of agency in language were intertwined in the prior studies, and so it is not known, whether their references in language affect message effectiveness to a similar extent. Indeed, past research focusing on recipients has found different effectiveness with foci on personal agency versus communion (e.g., acting as part of a community vs. highlighting personal agency; see Eckel et al., 2017; Kessler et al., 2017; Whillans & Dunn, 2018; Whillans et al., 2017). In the present experiments, we rather focused on the semantics in the message of the slogan (referring to acting/goal achievement – agency vs. helping/collaboration – communion). This allowed us to disentangle the importance of grammatical cues and semantic cues of agency linked to the message sender/agent for message effectiveness.

Importantly, however, the relationship between semantics and grammar may be additionally contingent on the contextual cues in the NGO (Studies 2 and 3) and in the corporate context (Study 4). There may be a default assumption for NGOs to be communal and for corporate entities to be agentic. It is possible that communication of those entities is most efficient when framed along dimension other than the default one. For prosocial campaigns, where communion can be taken for granted semantic agency might be particularly relevant. For business actions, where goal-orientation can be assumed, it may be that communion is a key predictor of message effectiveness. It is not known, how the grammatical agency can add to that picture. Assuming its subtle and implicit role, it may work independently of semantics, increasing any message effectiveness due to its relationship with activity and taking actions. However, it may also increase message effectiveness where the agency is of utmost importance or disrupt the effectiveness of the messages (for a lack of fit) where communion plays the most important role. Beyond the obvious implications for communication practices, the answer to this question has theoretical relevance, as it would address the role of grammar in relation to content, providing further insights into the interplay between semantic and meta-semantic properties of verbs.

Finally, the third neglected issue refers to how the perception of the message sender contributes to message effectiveness. So far studies examining agentic content in persuasive messages focused on messages evoking a sense of personal agency among message recipients. However, messages can also convey information about the message sender. To the extent to which they are seen as agentic or capable of implementing their proposed actions, the messages they put forward may also be seen as more persuasive. We address this question both in NGO and corporate contexts. Given the focus on the message sender and message effectiveness, we do not hypothesize moderation by participants’ economic standing – which was apparent when message effectiveness was examined with references to message recipients.

Study 1 was an archival analysis of existing word ratings for a sample of verbs and adjectives, allowing to address first the basic question of whether real verbs are in general evaluated as more agentic than adjectives. Studies 2 and 3 were experiments with a similar design in which we varied grammatical cues (verb vs. adjective) and semantical cues with agentic versus communal messages (e.g., “We act” vs. “We help”) to investigate comprehensively whether using verb-slogans and semantic agency translate into higher perceived effectiveness in NGO context. Study 4 applied a similar conceptual design in a corporate context. We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the experiments. The data used in this research and supporting information are available on our Online Science Framework (OSF) project site at https://osf.io/bgrsy/.

Method

Two datasets were merged to provide information about the meaning associated with verbs and adjectives. An evaluation of 13,915 English words (Warriner et al., 2013) provided information on valence (measured on a 9-point scale unhappy to happy), arousal (calm to excited), and dominance (controlled to in control). The latter scale provides a good approximation of agency, as the agency is related to having control over one’s own actions (Abele & Wojciszke, 2014; Bandura, 1989). We also added a measure of how abstract versus concrete the word is (5-point scale), taken from Brysbaert et al.’s (2014) archive of 40,000 English words. The resulting matched dataset was further limited only to those words that can be unambiguously classified as either verb (e.g., involve or consolidate) or adjective (e.g., addictive or additional). This strategy allowed us to obtain 950 unique verbs and 2,115 unique adjectives.

Results

To test whether the grammatical category is related to agency ascriptions of words, we applied regression analysis. In the first step, we entered measures of valence (B = 0.57; SE = 0.01), arousal (B = −0.04; SE = 0.01), and concreteness (B = −0.05; SE = 0.02) – full results of this analysis are presented in the top section of Table 1. In the second step, the grammatical category was added to the set of predictors (with verbs coded as 1 and adjectives coded as 0). As presented in the lower section of Table 1, above and beyond the effects of covariates, verbs are evaluated as more agentic than adjectives (B = 0.31; SE = 0.03).

Table 1 Regression model for word ratings and the effect of grammatical category for Study 1

	R2 or ΔR2	B	SE B
Note. *p < .05, **p < .01, ***p < .001.
Step 1	.60***
Intercept		2.68***	0.09
Concreteness		−0.05**	0.02
Valence		0.57***	0.01
Arousal		−0.04**	0.01
	F(3, 3,061) = 1526.61; p < .001
Step 2	.02***
Intercept		2.64***	0.09
Concreteness		−0.08***	0.02
Valence		0.56***	0.01
Arousal		−0.03*	0.01
Adjectives vs. Verbs		0.31***	0.03
	ΔF(1, 3,060) = 145.69; p < . 001

Table 1 Regression model for word ratings and the effect of grammatical category for Study 1

View as image HTML

Method

Participants

All participants were recruited through Prolific (https://www.prolific.co) at a rate of £7.80 per hour. They were adult US citizens and their native language was English. From the pool of completed questionnaires, we have excluded 14 participants based on the attention-check question: “To continue click No” (Wiener et al., 2014) and a seriousness check (Aust et al., 2013, addressing how serious participants were in filling the study. Finally, 59 participants were excluded based on another attention-check question examining the verbatim recollection of the presented short slogan. Please note that for Studies 2 and 3, we employed a very conservative measure of attention with the verbatim recall of the presented slogan due to the acknowledged need to check data quality in such paid online platforms. Therefore, for the experimental condition “We help”, for instance, participants with responses such as “We care” or “We are helping” were excluded based on the criterion to let only exact wording pass. The final sample consisted of 255 participants (112 women, 142 men, and 1 person not indicating their gender, M_age = 33.67, SD_age = 11.44; basic information about participants in Studies 2–4 are presented in Table 2). We relied on G*Power 3.1 software (Mayr et al., 2007) to calculate the sample size, using a power of 80% benchmark for a medium effect size of f = 0.25 (Cumming, 2014).

Table 2 Participant demographic characteristics across studies

Study	2	3	4
Note. For income, the scale ranges from $10,000 and below to $150,000 and above, with $10,000 intervals in between. For subjective status, the scale ranges from 1 (= bottom of the social ladder) to 10 (= top of the social ladder). For political views, the scale ranges from 1 (= very liberal) to 7 (= very conservative).
N	255	194	272
% Female	43.9	43.3	58.8
% Caucasian	77.3	74.7	80.1
% College graduates	61.9	64.9	62.9
Mdn age (range)	31 (18–77)	33 (21–70)	34 (18–69)
Mdn income (range)	$40 K to $50 K (< $10 K to > 150 K)	$30 K to $39,999 K (<$10 K to > 150 K)	$40 K to $50 K (< $10 K to > 150 K)
Mdn subjective status (range)	5 (1–10)	5 (1–10)	5 (1–10)
Mdn political views (range)	3 (1–7)	3 (1–7)	3 (1–7)

Table 2 Participant demographic characteristics across studies

View as image HTML

Procedure and Materials

Participants were informed that the study concerned the effectiveness of NGO campaigns and more specifically perceptions of succinct slogans that would form the basis for more specific campaigns in the future. Specifically, the instruction read as follows:

“As previously mentioned, we are also investigating perceptions of brief slogans in this preliminary study. We draw from a collection of several short slogans that will form the basis for more specific campaigns in the future. However, for this study, we are only interested in the perception of the basic slogans.”

Participants were randomly assigned to view a slogan employing either agentic or communal content (referring to acting or helping) based on frequently used adjectives in agency and communion scales (specifically referring to acting or helping, Abele & Wojciszke, 2007), with the message expressed with the use of a verb or an adjective. Thus, the study consisted of four experimental conditions: two slogans employing verbs (“We act,” N = 63 or “We help,” N = 67) and two slogans employing adjectives (“We are active,” N = 66 or “We are helpful,” N = 59). For a graphical demonstration of the manipulation and materials used in Studies 2 and 3, see Figure 1. The G.C.Y.W. sign appeared in the upper left corner of each succinct slogan. It was fictitious and did not refer to any existing NGO.

Measures

Participants were first asked how indicative the slogan was for the NGO’s agency and communion. Agency was measured with three items (e.g., “This NGO strives to achieve its goals even if obstacles have to be conquered”). Communion was also measured with three items (e.g., “This NGO cares about being integrated in the community”). To assess the perceived effectiveness of the message we used two items (“Based on this slogan, how successful would you estimate this campaign to be?” and “Based on this slogan, do you believe this campaign is an outright failure?” – reverse coded). All the questions were assessed with a scale ranging from 1 (= not at all) to 9 (= very much), the obtained results were averaged, and combined to form indicators of agency, communion, and message effectiveness respectively. The reliability coefficients, means, standard deviations, and correlation coefficients of the measures used in Study 2 are presented in Table 3.

Table 3 Reliability coefficients, means, standard deviations and correlation coefficients for the variables used in Study 2

	α	M	SD	2	3
Note. ***p < .001.
1. Agency	.90	5.60	2.02	.50***	.66***
2. Communion	.87	6.00	1.86		.52***
3. Effectiveness	.85	5.67	2.10

Table 3 Reliability coefficients, means, standard deviations and correlation coefficients for the variables used in Study 2

View as image HTML

At the end of the study, participants were asked attention-check questions and also to provide demographic information concerning their gender, age, ethnicity, as well as social status, and political views. Social status was measured using three variables: self-reported income, subjective status, and education.

Results

The Electronic Supplemental Material (ESM 1) contains all information regarding tests of interaction effects between grammatical agency, semantic agency, social status variables, and political views. The findings were nonsignificant and as more peripheral to the current research, they will be briefly summarized and discussed in the General Discussion.

Agency

To examine whether the experimental manipulations affected the perceptions of agency, we conducted a 2 (grammatical form: verbs vs. adjectives) × 2 (semantic content: agentic vs. communal) analysis of variance (ANOVA) on the measure of perceived agency. The analysis revealed the expected main effect of the grammatical form, F(1, 251) = 15.07, p < .001, η_p² = .06, 95% CI [.01, .12], indicating that NGOs using verbs in slogans (M = 6.01; SD = 1.94) were seen as more agentic than NGOs using adjectives in slogans (M = 5.17; SD = 2.03). Moreover, there was also the expected significant effect of the semantic content, F(1, 251) = 39.05, p < .001, η_p² = .14, 95% CI [.06, .21], indicating that NGOs using agentic words in slogans (M = 6.30; SD = 1.94) were indeed seen as more agentic than NGOs using communal words in slogans (M = 4.89; SD = 1.86). The interaction was not significant, F(1, 251) = 0.41, p = .53. The results for the agency indicate that both grammatical and semantic content independently contributed to the ascriptions of the agency. Considering the effect sizes, semantic content not surprisingly affects the attributions of agency more strongly. Yet, the role of grammar is still present, above and beyond the meaning of the message, in line with the meta-semantic role of verbs predicted by the verb-agency link. The means and standard deviations for all the dependent variables across all conditions are presented in Table 4.

Table 4 Descriptive statistics for effectiveness, agency, and communion indicators in Study 2

	Semantic agency				Semantic communion
	Verb		Adjective		Verb		Adjective
Variable	M	SD	M	SD	M	SD	M	SD
Agency	6.84	1.56	5.79	2.13	5.24	1.95	4.49	1.67
Communion	5.78	1.88	5.25	2.02	6.66	1.47	6.32	1.75
Effectiveness	6.76	1.86	5.68	2.07	5.40	2.02	4.80	1.98

Table 4 Descriptive statistics for effectiveness, agency, and communion indicators in Study 2

View as image HTML

Indirect Effect

To test for the hypothesized indirect effect of grammatical form on perceived message effectiveness via agency ascriptions, we conducted a mediation analysis using Hayes’ (2013a, 2013b) macro and 1,000 bootstrap procedure. We included communion as a second parallel mediator to examine, whether the effect of grammar is limited to the agency only. As semantic content did not interact with grammatical form in neither of the variables, it was included as a covariate in the model (see Figure 2). The model explained 52% of the variation in effectiveness scores, R² = .52; p < .001. As expected, the indirect effect via agency was significant (point estimate = 0.32; Boot CIs [0.13, 0.57]), the indirect effect via communion, however, was not (point estimate = 0.20; Boot CIs [−0.01, 0.42]).

Method

Measures

All measures used in Study 3 were identical to those used in Study 2 with one exception: To assess effectiveness, we used one additional item in comparison to Study 2: “Based on this slogan, how persuasive do you perceive this campaign could be?” – with the scale ranging from 1 (= not at all) to 9 (= very much). Similarly to Study 2, at the end of the questionnaire, participants provided demographic information concerning their gender, age, ethnicity, as well as social status, and political views. The reliability coefficients, means, standard deviations, and correlation coefficients of the effectiveness indicator and agency and communion scales are presented in Table 5.

Table 5 Reliability coefficients, means, standard deviations, and correlation coefficients for the variables in Study 3

	α	M	SD	2	3
Note. ***p < .001.
1. Agency	.84	6.19	1.75	.41***	.63***
2. Communion	.88	6.43	1.75		.54***
3. Effectiveness	.84	5.88	1.92

Table 5 Reliability coefficients, means, standard deviations, and correlation coefficients for the variables in Study 3

View as image HTML

Results

Agency

To examine whether the experimental manipulations affected the perceptions of agency, we conducted a 2 (grammatical form: verbs vs. adjectives) × 2 (semantic content: agentic vs. communal) ANOVA on the measure of perceived agency. The analysis revealed the expected main effect of the grammatical form, F(1, 190) = 4.98, p = .03, η_p² = .03, 95% CI [.00, .08], indicating that NGOs using verbs in slogans (M = 6.48; SD = 1.52) were seen as more agentic than NGOs using adjectives in slogans (M = 5.87; SD = 1.93). Moreover, there was also the expected significant effect of the semantic content, F(1, 190) = 13.12, p < .001, η_p² = .07, 95% CI [.01, .14], indicating that NGOs using agentic words in slogans (M = 6.65; SD = 1.46) were indeed seen as more agentic than NGOs using communal words in slogans (M = 5.75; SD = 1.89). The interaction was not significant, F(1, 190) = 0.60, p = .44. The results for the agency indicate that both grammatical and semantic content independently contributed to the ascriptions of the agency. The means and standard deviations for all the dependent variables across all conditions are presented in Table 6.

Table 6 Descriptive statistics for agency, communion, and effectiveness indicators in Study 3

	Semantic agency				Semantic communion
	Verb		Adjective		Verb		Adjective
Variable	M	SD	M	SD	M	SD	M	SD
Agency	6.80	1.37	6.45	1.55	6.12	1.60	5.39	2.10
Communion	5.87	1.61	5.72	1.57	7.24	1.44	6.85	1.90
Effectiveness	6.27	1.68	5.62	2.03	6.03	1.86	5.53	2.06

Table 6 Descriptive statistics for agency, communion, and effectiveness indicators in Study 3

View as image HTML

Indirect Effect

To test for the hypothesized indirect effect of grammatical form on perceived message effectiveness via agency ascriptions, we conducted a mediation analysis using Hayes’ (2013a, 2013b) macro and 1,000 bootstrap procedure. We included communion as a second parallel mediator to examine, whether the effect of grammar is limited to the agency only. As semantic content did not interact with grammatical form in neither of the variables, it was included as a covariate in the model – see Figure 3. The model explained 49% of the variation in effectiveness scores, R² = .49; p < .001. As expected, the indirect effect via agency was significant (point estimate = 0.26; Boot CIs [0.03, 0.52]), the indirect effect via communion, however, was not (point estimate = 0.11; Boot CIs [−0.08, 0.33]).

Method

The design of Study 4 was similar to the one applied in Studies 2 and 3 with the following differences. First, we created a business rather than an NGO context. Participants were instructed as follows:

Second, in Studies 2 and 3 we used only one short sentence per condition to manipulate grammatical agency and semantic content. In Study 4, we used four phrases per condition. For a graphical demonstration of the manipulation and materials used in Study 4, see Figure 4. Third, given that some of the newly developed materials contained phrases that overlapped with the content of items previously used to measure agency and communion, we included new items for the purpose of measurement of these dimensions. Finally, given the high exclusion rate based on the previously used verbatim recollection of the presented slogans, we now presented participants with a recognition task. They were presented with four sentences (of which two had appeared in the actual manipulation). They were asked for each independently to indicate whether they have seen this sentence in the campaign material. This allowed us to compute an incremental measure across the four recognition items and qualify to the study those participants who responded above chance level (75% of correct answers or more).

Measures

To assess effectiveness, we used the same measures as in Study 3. Measures of agency and communion were adapted to correct for the fact that the experimental material contained some of the phrases used in the agency and communion items employed in our previous studies (goal achievement or cooperation). Therefore, the agency was measured with two items (“This company is performance-oriented” and “This company strives for efficiency”), so as communion (“This company is community-oriented” and “This company cares for employees”) – with the scales again ranging from 1 (= not at all) to 9 (= very much). Similarly to Studies 2 and 3, at the end of the questionnaire, participants provided demographic information concerning their gender, age, ethnicity, as well as social status, and political views. The reliability coefficients, means, standard deviations, and correlation coefficients of the effectiveness indicator and agency and communion scales are presented in Table 7.

Table 7 Reliability coefficients, means, standard deviations and correlation coefficients for the variables in Study 4

	α	M	SD	2	3
Note. *p = .05; ***p < .001.
1. Agency	.76	6.49	1.84	.12*	.47***
2. Communion	.90	5.77	2.19		.47***
3. Effectiveness	.90	6.10	1.80

Table 7 Reliability coefficients, means, standard deviations and correlation coefficients for the variables in Study 4

View as image HTML

Results

Agency

To examine whether the experimental manipulations affected the perceptions of agency, we conducted a 2 (grammatical form: verbs vs. adjectives) × 2 (semantic content: agentic vs. communal) ANOVA on the measure of perceived agency. The main effect of the grammatical form was not significant, F(1, 268) = 0.96, p = .33, η_p² = .004, 95% CI [.00, .03]. However, the effect of the semantic content was significant, F(1, 268) = 6.70, p = .01, η_p² = .02, 95% CI [.001, .07], indicating that companies using agentic words in materials (M = 6.77; SD = 1.72) were indeed seen as more agentic than companies using communal words in materials (M = 6.20; SD = 1.91). The interaction was not significant, F(1, 268) = 0.02, p = .90. The results for the agency in Study 4 indicate that only semantic content contributed to the ascriptions of the agency. The means and standard deviations for all the dependent variables across all conditions are presented in Table 8.

Table 8 Descriptive statistics for agency, communion, and effectiveness indicators in Study 4

	Semantic agency				Semantic communion
	Verb		Adjective		Verb		Adjective
Variable	M	SD	M	SD	M	SD	M	SD
Agency	6.89	1.61	6.65	1.83	6.29	1.95	6.10	1.89
Communion	4.53	1.72	4.29	2.06	7.30	1.49	7.01	1.51
Effectiveness	6.00	1.72	5.60	1.82	6.65	1.71	6.14	1.83

Table 8 Descriptive statistics for agency, communion, and effectiveness indicators in Study 4

View as image HTML

Future Directions

As indicated anecdotally in the introduction, slogans employing verbs apparently did affect many people in their voting or purchasing decisions. The verb slogans used in real life could either evoke a sense of agency in recipients as in previous studies (Whillans et al., 2017; Whillans & Dunn, 2018) or inform about a message sender as being efficient and likely worth the promise they were making, as in the Studies 2 and 3 here. Importantly, in our studies, both agency and communion were associated with effectiveness, but the indirect effect of verbs on message effectiveness occurred only for the agency and only for the NGO context. This suggests that in some contexts, verbs can indirectly affect persuasiveness by stressing the agency component of the persuasion target. That this happened in NGO rather than incorporate context, may imply that verbs signal the agency of the message sender predominantly for non-typically agentic targets (where agency associations are less chronically accessible). This hypothesis requires further testing; however, in a similar vein, it was predominantly girls that responded positively to the verb–noun manipulation (Rhodes et al., 2019; for the opposite pattern of results of verbs versus nouns see Bryan et al., 2011, 2014; for a null result see Gerber et al., 2016; and for the discussion on verbs and agent nouns see Formanowicz et al., 2017). Given that in Study 4, we did not observe the indirect effect, it is likely that (under certain conditions) verbs independently affect message content without affecting attributions to sender, and that for agentic targets (such as companies) evaluations of the agency are to a lesser extent affected by subtle grammatical cues.

Contextual effects of agency on message effectiveness obtained for NGO targets as obtained in our Studies 2 and 3 may further inform studies on prosocial communication. Despite traditional associations linking communion with prosocial behaviors, the agency also affects prosocial tendencies (Gagné, 2003; Weinstein & Ryan, 2010), thus benefitting not only an individual but also a community. The nuanced relationship between the Big Two and prosociality in personality is best explained considering the social motives perspective (Gebauer et al., 2014). In this perspective, the agency is linked to standing out and establishing one’s uniqueness. Communion on the other hand contributes to fitting in, establishing one’s connectedness with others. In line with this model, communal traits were most predictive of prosocial behaviors in sociocultural contexts that favor prosocial tendencies as expressed by volunteering interests. On the contrary, the agency predicted prosocial behaviors in the sociocultural context, where such behaviors were uncommon (Gebauer et al., 2014). The only study on message effectiveness that resonates with the nuanced perspective of agency and communion also showed that whether it is agency or communion that affects message effectiveness depends on the context (White & Peloza, 2009). When participants were held publicly accountable for their behavior, communal appeals were more successful than agentic appeals for donations. However, when participants’ decision whether to make a donation was to remain private, they responded more to agentic than to communal appeals. Accordingly, all studies investigating the role of agency in message effectiveness were conducted in a private context, when an individual makes a decision of whether to make a donation or not (White & Peloza, 2009). This raises the question of whether the agency affects message effectiveness only under certain conditions. The main avenue for future research is thus to investigate the boundary conditions for the agency – message effectiveness effect. Our studies add that the contextual effects of agency and communion need to include a grammatical aspect.

Limitations of the Current Research

Importantly, in our studies, we did not find susceptibility to persuasion by either semantically agentic or communal slogans to be dependent on participant’s social standing as had been found previously (Whillans et al., 2017). This can be interpreted in different ways. First of all, the relationship between social class and receptiveness to message content is not entirely replicable (Whillans & Dunn, 2018). Second, previous manipulations of communion (Whillans et al., 2017) can be in fact seen as manipulations of collective agency. For instance, Whillans and Dunn (2018) created a communal message through the following phrase: “Join your community and support a common goal.” Although this message is clearly referencing communal content by evoking community and a sense of commonness, it is also addressing the core of the agency that is goal orientation (Abele & Wojciszke, 2014). Given that efficacy is in the core of collective action and likely also of prosocial engagement (van Zomeren, 2013), participants in previous studies (Whillans et al., 2017) were possibly affected by agency appeals (either personal or collective) depending on their social standing. In our studies, the manipulation of communion was orthogonal to the manipulation of agency – and we found no difference in the responsiveness to the content of the message between participants differing in social class. This may indicate that in previous studies it wasa collective agency rather than communion that increased working-class engagement in social action.

However, the lack of interactive effects of political views and social class with semantic content and grammatical agency need to be interpreted with caution as our interactions are likely underpowered (Kraus & Callaghan, 2016). Furthermore, prior studies on message effectiveness in relation to social class (Whillans et al., 2017) did not employ Mturk nor Prolific samples – and their critical thresholds of income whereby agency affected message effectiveness exceeded $90,000 USD. Both relatively low power to detect interactions and the lower household income typically obtained in Mturk samples may preclude obtaining the income by agency interactions observed in past research.

An important limitation of the current research is that variables measuring agency and communion were always assessed in the current studies prior to message effectiveness. In this way manipulations of agency and communion as well as of grammatical form, could affect directly following ratings of the big two dimensions, which, once activated, could affect the results on message effectiveness. Future studies should carefully disentangle the order effects.

Conclusion

Can grammar be social and effective? Yes, it can. The present results underscore the verb-agency link and extend it to increased message effectiveness. Moreover, the present studies provide new insights into the role of agentic cues (grammatical and semantic) for message effectiveness more generally. They open several new research avenues to be taken regarding the presumed subtlety of grammatical cues compared to semantic cues, message-person fit, message-content fit, and reactance. Importantly, the present studies attest to a verb-message effectiveness link that may work irrespective of semantics for communication. Message recipients extract information that goes beyond the content of what is said, by subtly exploiting grammatical information – let’s make it happen.