The aim of this study is to contribute to the restoration of a lacuna in recent psychosociological literature relating to the very status of the source of influence as an independent variable in the experimental study of social influence processes. The present experiment is in line with a new research track pointing to the semantic ‘slippage’ that has gradually been introduced in this field of study and, with a few rare exceptions (cf. Stroebe, 2010), has remained unnoticed. Namely, a habit has been adopted for a long while by most researchers interested in the phenomenon of social influence, consisting of discussing theoretically the minority or majority source of influence, but studying experimentally the effects of the social support allegedly attributed to its message by its recipients, that is to say, the participants.
We argue that this ‘slippage’ is not confined to a simple methodological clumsiness but is fraught with theoretical and epistemological consequences. These consequences would be: the obsolescence of the notion of the group in favor of that of consensus; putting aside the normative criterion in defining the source of influence and putting forward its numerical equivalent; the progressive abandonment of the study of the role of this same source in favor of that of the recipients of the message of influence; and the implicit substitution of the paradigm of influence (Asch, 1952; Moscovici, 1976) by that of social support (Allen, 1975).
Social Influence of Groups
The original Asch experiment (1951) questioned for the first time the autonomy of the individual and his/her resistance to group pressures. It was followed by that of Kelman (1958), a pioneer in the study of various forms of conformism, and then the early experiments conducted by Moscovici (Moscovici & Lage, 1976; Moscovici & Lage, 1978; Moscovici, Lage & Naffrechoux, 1969; Moscovici & Faucheux, 1972; Moscovici & Nève, 1971; Moscovici & Nève, 1973) on the influence exerted by active minorities. All used an experimental design simulating in laboratory the real or symbolic confrontation of participants with a group of confederates (or their representative, in the case of Kelman) representing the source of influence. The same went for Nemeth’s experiments (Nemeth, Swedlund & Kanki, 1974; Nemeth, Wachtler & Endicott, 1977) and Gabriel Mugny’s first-ever study (Mugny, Pierrehumbert & Zubel, 1972–1973) on negotiation styles. Things were quite clear at the time: on the theoretical level, authors were discussing the influence exerted by a minority or majority group and, on the methodological level, they were experimentally studying, in laboratory, exactly what they had proposed to do.
It should be noted, however, that there was a particularly salient hiatus between method and theory in contrasting minority and majority influence within the same experiment, notably when applying the classic design introduced by Serge Moscovici: the answers (green) of the source being always incorrect (since the stimuli were clearly blue). Thus, they certainly satisfied the normative criterion of the definition of the status of a minority source, but in no way that of a majority source which enjoyed, of course, a numerical superiority but had, in addition, the disadvantage of formulating always incorrect responses and thus, normatively speaking, holding minority positions. This asymmetry was restored thanks to the experimental paradigm shift observed in Gabriel Mugny’s work on negotiation styles (Mugny, 1975; Mugny, 1982): the use of opinion tasks rather than perceptual tasks actually allowed for the normative definition of the status of the source, both minority and majority. In these experiments, the operationalization of the source’s status was ensured through the explicit reference to its ‘group’ aspect (majority vs minority) and through the corresponding anti- or pro-normative content of the discourse held by this source (majority vs minority) of influence.
The reported discrepancy between theoretical discourse and experimental illustration in research using an ‘objective’ task has, however, prompted some researchers to propose a more complete definition of social entities undertaking a process of influence: combining normative and numerical criteria, Maass and Clark proposed in the early 1980s the notion of ‘double minorities’; that is, social entities both numerically restricted and supporting positions which countered dominant norms (Maass, Clark & Haberkorn, 1982; Maass & Clark, 1983; Maass & Clark, 1984). Undoubtedly, this notion has the merit of facilitating the rapprochement between epistemic questioning and socio-political reality. Unfortunately, few experimental studies have followed this promising track.
On the contrary, the methodological trick used by Personnaz (1981), who replaced the minority or majority (numerically) group of accomplices physically present in the experimental condition with a single accomplice supposed to represent 18.2% or 81.8%, respectively, of the general population, has met with unprecedented success. For example, the experiments conducted with Moscovici using the spectrometer method to study indirect minority influence (Moscovici & Personnaz 1980; Moscovici & Personnaz, 1986; Moscovici & Personnaz, 1991) have adopted this particular operationalization of the status of the source of influence. While having obvious practical advantages, the use of this method does not go without epistemic consequences.
Group Influence and Social Consensus: Size and Intensity
Despite the absence of epistemic vigilance, this methodological ‘trick’ proved to be very popular and has been largely used, including in experimental paradigms involving the use of opinion tasks which, however, could have rendered easier the normative manipulation of the status of the source of influence (thanks to the manipulation of the pro- or anti-normative orientation of the content of its message). This was, for example, the case of the experiments carried out by the Genevan team: the influence inducing text was presented as having been written by a minority or majority group (cf. Mugny et al., 1991; Mugny & Papastamou, 1980; Mugny & Papastamou, 1982–1983; Mugny & Pérez, 1985; Papastamou & Mugny, 1985; Papastamou & Mugny, 1987; Pérez & Mugny, 1987); and the percentage of the consensus allegedly obtained among general population was also communicated to the participants.
This way of operationalizing the source was mainly adopted by researchers interested in studying social influence from a persuasion perspective (Chaiken, 1987; Chaiken, Liberman & Eagly, 1989; Petty & Cacioppo, 1986). This attempt to merge these two theoretical universes allowed for the mutual enrichment through the use of respective methodological and theoretical tools (cf. Bohner, Frank & Erb, 1998; De Dreu & De Vries, 1993; Martin & Hewstone, 2001a; Martin & Hewstone, 2001b; Martin & Hewstone, 2003; see also Briñol & Petty, 2009; Erb et al., 1998). The common denominator was the questioning of differentiating effects in terms of minority and majority influence. It is worth stressing that the genetic model of influence (Moscovici, 1976; Moscovici, 1980) attributed them to the psychosocial specificities of the majority or minority status of the source, each of which would implicate different socio-cognitive mechanisms (preferential activation of social comparison in the first case, of cognitive validation in the second one). However, for proponents of persuasion models, these effects would rather be due to the socio-cognitive functioning of the participants (and influence targets), whose activation would not depend on factors directly related to the influence process as much as on the participants themselves: their previous attitudes (cf. Erb et al., 2002), their personal interest (cf. Trost, Maass & Kenrick, 1992), or their need for cognition (see Haugtvedt & Petty, 1992).
Therefore, in this line of research, the transition from the status of the source to that of its consensus allegedly granted by the target took place in a relatively clear manner. The operationalization of the status of the source is effectively ensured by the sole manipulation of the percentage explicitly attributed to it by the experimenter. However, as noted by Stroebe (2010), this methodological choice has denuded the source-recipients relationship approach of significant dimensions present in natural conditions in which recipients encounter ‘real’ groups that support and diffuse normatively and ideologically charged points of view (Pérez & Molpeceres, 2018; Chryssochoou, 2018). Among these dimensions, the absence of the intensity of the consensus attributed to a persuasive message is particularly striking. Yet Moscovici has argued that this dimension would provide particularly interesting information in a process of social influence such as, for example, the cohesion between the members of the source; their commitment and involvement in their influence activity; their conviction in the ideas they defend; and their consistency.
In keeping with the tradition of attitude change studies, this research line is, moreover, solely interested in experimental populations whose positions are far distant from the ones held in the persuasive messages diffused by the source. This is the case, for example, of the works on voluntary euthanasia (the theme also used in the experiment reported in this paper), strongly rejected by the source of influence but widely accepted by the experimental population (Gardikiotis, Martin & Hewstone, 2005; Martin & Hewstone, 2003). This practice obscures the important role of intergroup relations in influence processes and impoverishes the dynamics of the psychosocial field they delimit. This is exemplified in the modulating effects produced by the social and ideological distance separating the source from the target of influence (Mugny, Kaiser & Papastamou, 1983; Souchet et al., 2006): notably, they activate sometimes unexpected resistances to minority influence, such as the receptivity of participants ideologically or/and socially close to the source, negatively affecting its psychologizing perception (Papastamou, Mugny & Kaiser, 1980; Papastamou & Mugny, 1987).
The Ambiguous but Decisive Role of the Participants
This is not to underestimate the interest of social psychologists in the differentiated psychosocial functioning of the participants as recipients of influence in the experimental condition. Indeed, we are aware 1) that the active role of the targets of the message of the source stipulated by Moscovici (1976; 1980) represented one of the main distinguishing features of his influence model; 2) that their differentiated perception modes of the source allowed for the distinction between behavior style and negotiation style as strategies of influence (Alvaro & Crano, 2017; Mugny, 1982); 3) that these same diverse modes also led to the theoretical emergence and experimental illustration of psychologization as a strategy of resistance to the influence exerted by active minorities (Papastamou, 1986; Papastamou, Mugny & Pérez, 1991–1992); 4) that taking into account the participants’ initial attitudes allowed for the approach of social influence in terms of psychosocial identification (Gabarrot & Falomir-Pichastor, 2017; Mugny et al., 1984); and 5) that it also allowed for experimental testing of the hypothesis of the correspondence between participants’ initial positions and social contexts of influence interaction (Buchs et al., 2002; Mugny et al., 2001; Mugny et al., 2002; Quiamzade et al., 2002), as well as 6) of the hypothesis of the preponderant role of the interest of the influence target in the thematic challenge of the social interaction (Falomir-Pichastor et al., 2011; Invernizzi et al., 2003).
These studies take into account the critical role of the participants as socio-cognitively active targets in the psychosocial interaction engaged. However, they do so in a somewhat restrictive manner. In these studies the ‘thoughts’ of the participants as targets of influence during the social interaction, concerning various aspects of the source, the message, even themselves, have a rigorously instrumental character. Whether invoked a posteriori by the researcher or injected into the experimental interaction as independent variables, these thoughts are treated as socio-cognitive variables of intra-individual nature that differentiate individuals one from another.
This is also the case in works on metacognition of targets that focus merely on the cognitive elements of the participants’ attitudes, such as certainty about their initial attitudes, the benefits of uncertainty in their regard (Tormala & Rucker, 2018), their self-validation (Brinol & Petty, 2009), or the legitimacy of the means used to resist change in their attitudes (Tormala et al., 2007). These studies, of course, clarify some aspects of the thoughts the participants would generate in order to explain pretty specific aspects of their behavior during the experimental interaction. However, since these metacognitions are expressed and collected after the actual experimental phase, they could easily be suspected of carrying rationalizing elements; this would certainly undermine their reliability as behavioral indexes, but would enhance their informational value concerning influence in terms of representational object.
This is an unknown aspect of these works which we think should be highlighted. Indeed, many anchor points have long and repeatedly been established between influence and social representations. In particular, there has been discussion about: a) social representations of behavioral styles (Mugny & Papastamou, 1984); b) social influence as an epistemologically favorable ground for the study of social representations (Mugny et al., 2009); and c) social representations as the appropriate theoretical instrument for the study of social influence (Papastamou & Prodromitis, 2012) or the unification of these two fields of research into one (Papastamou, 2019). No one has, however, discussed yet the metacognitive thoughts of those participating in influence experiments in representational terms.
Informative Value of Consensus and Social Influence
Ultimately, what seems rather problematic to us is not the displacement as such of social psychologists’ interest from the source of influence to its recipients, but the way in which this change of interest has been integrated in literature; on the one hand, the intra-individual dimension of the participants’ psychosocial functioning seems to have been privileged, on the other hand, the interactive involvement of participants in the construction of influence in terms of social representation has been neglected, and finally, the potential significance of the latter in the unfolding of the influence process has been underestimated.
This last point is exemplified in the work of Tormala et al. (2009): indeed, the influence exerted by the source increases when the arguments used correspond to the expectations of the participants, i.e., when the minority makes use of weak arguments and the majority of strong arguments. However, the authors, speaking here of the hypothesis of matching effect between the status of the source of influence and the quality of the arguments, propose by all accounts an epistemic analysis of intra-individual level and do not seem to take into account the intervention of the ideological or representational level (see Doise, 1982). Other works seem to provide findings in the sense of our point of view, without explicitly referring to social representations. Studies on the potential impact of information regarding the size of the alleged consensus of the source on the influence the latter exerts on the participants are another example. They show, in particular, that the majority or minority aspect of the message might constitute the basis on which the recipient would construct the representation of the numerical force of the fictional group transmitting this message. This representation would in turn serve as a point of reference for judgments formulated later by the target of influence (Erb et al., 2006). Apart from this function of explicit consensus, these authors suggest that there is an inferred consensus which would be related to the subjective representation of the way in which the majority of people would behave, judge or think in a given condition or regarding a specific theme (Erb & Bohner, 2010).
Furthermore, we know that different types of information about consensus in general generate different effects on the influence exerted by the source. It has been experimentally demonstrated that the majority exerts more influence when it is presented ‘only in terms of percentage’ compared to that obtained in the case it is described by means of a ‘combination of label and percentage’. On the contrary, the minority is found to be more influential when it is defined ‘only in terms of label’ than when its definition is based on a combination of both, even though under ‘label only’ conditions majority and minority influences cannot be distinguished (cf. Gardikiotis, Martin & Hewstone, 2005; Gardikiotis, Martin & Hewstone, 2010; Martin, Gardikiotis & Hewstone, 2002). The fact that ‘large numbers’ and ‘restricted entities’ seem to be the sources that generate the most social influence, raises once more the question of the meanings activated by the terms ‘majority’ and ‘minority’.
Three remarks should be made here. First, the very essence of the operationalization of the status of the source through the consensus attributed to it is clearly representational. Second, there is a certain confusion emerging from research on social influence, concerning the information that experimenters communicate to participants about the source of influence: there is explicit reference, sometimes, to the size of the source as a social group; sometimes, to the consensus accorded to the message of the source as a social group; sometimes, to the consensus accorded to the message of the source (without mentioning its group aspect); sometimes, finally, to the consensus accorded to a message devoid of source. These various types of information differ from one another and may activate in participants equally different social representations of the symbolic or real interaction. Indeed, reflecting on the source of a message is not the same as reflecting on the impact that the message of this source would have on the target, or reflecting on the impact of a message devoid of source. The third remark has to do with the one-dimensional way in which the notion of consensus is introduced in experimental conditions, that is to say, by referring only to its size. The informative value of the intensity of consensus in terms of influence remains, so far, untapped. The objective of this research is to remedy this issue.
Overview of the Research
To that end, we carried out a between-subjects factorial design: 2 (consensus status: majority vs minority) × 3 (orientation of the consensus evaluation: non-evaluation, size evaluation, intensity evaluation).
On the basis of the rationale outlined above, we could argue that the semantic slippage from the status of the source to its presumed consensus in the eyes of the target would transform the representational field of the influence process entailing specific consequences at the level of its outcome. First of all, it would lead the experimental population to reflect on social interaction in terms of communicative effectiveness; it would thus make salient the ultimate, but usually concealed, objective of the influential interaction: to exert influence on other people. This semantic slippage would thus induce different reading dimensions of the experimental condition in which participants are involved. In the condition of the evaluation of the size of the consensus, they would be made aware of the effectiveness of the source but more specifically of the acceptation of the extent of its impact; thus, the majority source would be the one whose influence would be most likely. In the consensus intensity evaluation condition, participants would still be informed of the notion of the effectiveness of the source, but this time in the sense of the power of its impact (i.e., the ‘strength’ with which people believe and support the persuasive message). They would therefore be sensitive to the commitment of influenced people, which would lead them to perceive active minorities as the most likely sources of influence. Simplifying, we could consider that the first case corresponds to the functionalist model, while the second corresponds to the genetic or interactionist model of influence. If this is indeed the case, then when the participants are led to conceptualize the influence condition (to which they participate) in terms of size of the consensus, they function in the way targets of a majority influence usually do (i.e., they conform to the dictates of the norm by submitting to the power of large numbers). When, on the other hand, they are led to think of this same condition of influence in terms of intensity of the consensus, they function rather as targets exposed to a minority influence (i.e., they follow the guidelines of the norm of originality, seduced by the attractions of novelty).
Let us note that in the conditions of non-evaluation of the consensus, the participants are placed in a condition similar to that in which they are placed in most of the experiments on influence conducted during the last decades; they are only informed of the majority or minority aspect of the consensus that the message of the source has supposedly received from their peers. It should also be noted that manipulation of the orientation introduced in the consensus evaluation conditions is meant to activate different meanings and elaborations that will ultimately determine the outcome of the influence process. This is the reason why we predict a mediation of the effect of the independent variables on the influence exerted, operated by the evaluative dimensions of the behavior of the source during the social interaction, such as its originality, consensual aspect, persuasion, extremism or realism (Eagly & Chaiken, 1993; Mugny, 1982; Papastamou, 1986; Petty & Cacioppo, 1986).
With regard to direct influence, in the conditions of non-evaluation, the usual majority superiority should be confirmed. This effect should also appear in the conditions of evaluation of the size of the consensus, but diminish or disappear in the conditions of evaluation of the intensity of the consensus. The indirect influence should also differ as a function of the evaluation or not of the consensus supposedly obtained by the source from its target. According to the genetic model of influence (Moscovici, Mugny & Papastamou, 1981; Mugny, 1982; Papastamou, 1986) in the conditions of non-evaluation, the indirect influence should be more important on the part of the minority than the majority. In the condition of evaluation of the consensus (especially of its intensity) the indirect minority influence should be accentuated since this evaluation is supposed to further activate social representations of influence compatible with the specificities of the influence exerted by active minorities. For the same reasons, the majority influence should be attenuated in these conditions compared to that obtained in the absence of any consensus evaluation.
The experiment was conducted in the context of social psychology courses at a University in Athens. The students participated voluntarily in the research and were asked to distribute a number of ‘booklets’ to be filled in by people of their circle of acquaintances.
Participants and Experimental Design
Two hundred and forty-six people participated in the experiment (120 men and 126 women). 74 participants were between 18–24 years old, 59 between 25–34 years old, 48 between 35–44 years old, 44 between 45–54 years old and 21 over 55 years old). They were randomly assigned to one of six experimental conditions 2 (consensus status: majority vs minority) × 3 (orientation of the consensus evaluation: non-evaluation, evaluation of the intensity of the consensus, evaluation of the size of the consensus) between subjects factorial design. On average, participants were in favor of voluntary euthanasia (M = 3.42, SD = 1.34, on a 5-point scale: 5 = for euthanasia) and the right to abortion (M = 3.56, SD = 1.36, on a 5-point scale: 5 = for abortion) and rather against the right to suicide (M = 2.61, SD = 1.45, on a 5-point scale: 5 = for suicide). The experiment was conducted in one single session but consisted of three different phases. In the first phase, participants answered a few pre-test questions related to the experimental task. In the second phase, independent variables were introduced. In the third phase, the participants responded to the post-test (influence measurements) as well as to certain questions that were supposed to capture their reactions to the experimental condition.
Material and Procedure
Phase 1: Pre-Test
The participants filled in the first ‘booklet’ (consisting of 3 items) by expressing their degree of agreement on a 5-point scale (5 = most favorable attitude) with regard to three issues (one item per issue): legalization of voluntary euthanasia (M = 3.42, SD = 1.33), the right to abortion (M = 3.56, SD = 1.36) and the right to suicide (M = 2.61, SD = 1.45).
Phase 2: Experimental Phase
The issue of influence was ‘the legalization of voluntary euthanasia’ already used in previous studies (cf. Martin, Gardikiotis & Hewstone, 2002). After completing the pre-test questionnaire, participants received a second ‘booklet’ containing the influence inducing text composed of either strong or weak arguments against legalization of voluntary euthanasia, developed according to the standard procedures (Eagly & Chaiken, 1993: 311; Gardikiotis, Martin & Hewstone, 2005; Martin et al., 2002; Martin & Hewstone, 2003; Petty & Cacioppo, 1986: 133). This text, presented in the form of a newspaper article (written by a social group), laid out the results of a fictitious opinion poll on the subject, according to which, in function with the experimental condition, the majority or the minority (with no further arithmetic details) of the representative sample interviewed agreed with this text and were therefore unfavorable to voluntary euthanasia. Participants were invited to read the text and answer a number of questions on it. It was at this point that the manipulation of the second independent variable was introduced, pertaining to the orientation of the evaluation of the consensus (majority or minority) granted to the positions held by the source concerning voluntary euthanasia. In the consensus size evaluation conditions, participants were asked to estimate the size of this majority vs. minority consensus on a 7-point scale (1 = the majority vs. minority consensus of the sample surveyed who agreed with these ideas is very small; 7 = very large); they therefore had to estimate the size of the consensus that the surveyed sample would have given in majority or in minority to the influence inducing text. In the consensus intensity evaluation conditions, participants were asked to estimate the degree to which this majority or minority of the questioned sample agreed with the influence inducing text on a 7-point scale; (1 = the majority vs. minority barely agreed, 7 = strongly agreed); they therefore had to estimate the intensity of the consensus that the questioned sample would have given in majority vs. minority to the influence inducing text. In the non-evaluation of the consensus conditions, participants received no additional information and went directly on phase 3 of the experiment. It is important to note here that measurements of the size or the intensity of the consensus obtained by the source did not serve as a manipulation check of the corresponding independent variable, but simply consisted of making the target think of the influence process in terms of intensity or size of the consensus obtained.
Phase 3: Post-Test
A. Influence Measurements
Participants were asked to express their attitudes on three semantic differential 7-point scales (bad-good, senseless-intelligent, harmful-beneficial), concerning the legalization of voluntary euthanasia, the right to abortion and the right to suicide. The text being unfavorable to the legalization of voluntary euthanasia, the bipolar scales were recoded so that point 7 reflected maximum influence.
A1. Measurement of Direct Influence
We constructed a composite index of the three items concerning voluntary euthanasia in order to assess direct influence (α = 0.924, M = 3.94, SD = 1.76) since these items were reproduced in the influence inducing text.
A2. Measurement of Indirect Influence
We also constructed two other indexes: one composed of the three items concerning the right to suicide (α = 0.934, M = 3.68, SD = 1.91) and another one composed of the three items concerning the right to abortion (α = 0.905, M = 5.05, SD = 1.82). Attitudes towards the right to suicide and towards the right to abortion were significantly correlated with each other (r = 0.442, p < 0.0001), but also with attitudes towards the right to euthanasia (r = 0.457, p < 0.0001 and r = 0.422, p < 0.0001, respectively). These findings permitted to consider that these three attitudes shared the same organizing principle, the right to self-determination of one’s body. However, since attitudes towards the right to suicide and the right to abortion were not included in the influence inducing text, they were treated as measurements of indirect influence. We therefore constructed a fourth index composed of the two previous indexes (suicide and abortion) which finally constituted our measurement of indirect influence (α = 0.879, M = 4.36, SD = 1.59).
B. Complementary Measurements
The experiment ended with the participants responding to a series of questions about the text they had just read, using the 7-point bipolar scales. These scales were constructed based on various dimensions which, as previously stated, are drawn from the persuasion and influence literature aiming at the evaluation of the effectiveness of a persuasive message. These scales were as follows: ‘The text you just read, in general: (1) defends conflictual/consensual ideas, (2) is original/trivial, (3) is supported by the majority/minority of the public opinion of the Greek population, (4) is absolutely not/quite convincing, (5) is extreme/moderate, (6) is interesting/uninteresting, (7) is absolutely incomprehensible/comprehensible, (8) is rare/current, (9) totally unrealistic/realistic.’
Participants’ responses to the three pre-test items were subjected to an analysis of variance that revealed a single statistically significant effect on the item concerning voluntary euthanasia (M = 3.42, SD = 1.34; consensus × evaluation interaction: F (2, 240) = 3.55, p < 0.03, η2 = 0.029). As a result, participants’ initial attitude towards voluntary euthanasia was introduced as an additional independent variable in the analyses presented below.
Age of Participants
The 2 × 3 analysis of variance applied to the data pertaining to the main demographic variable in this study, the age of participants, revealed a statistically significant simple effect (M = 34.37, SD = 0.82, F (2.240) = 6.88, p < 0.001, η2 = 0.054) of the orientation of consensus evaluation variable (non-evaluation: M = 33.62, SD = 1.50; intensity evaluation: M = 31.19, SD = 1.46; size evaluation: M = 38.30, SD = 1.31). In this respect, it should be noted that in the following analyses the age of the participants was therefore introduced as a covariant.
Direct and indirect influence
In order to assess direct and indirect influence we proceeded to analyses having as independent variables the status of the consensus (majority/minority), the oriented evaluation of the consensus (non-evaluation, intensity evaluation, size evaluation), as well as the initial position of participants regarding voluntary euthanasia; the age of participants was introduced in the analysis as a covariant. We used the PROCESS macro v.3 by Hayes (model 3; Hayes, 2013; Hayes & Montoya, 2017) choosing 10,000 bootstraps.
As far as direct influence is concerned, the results revealed a tendentially significant triple interaction F (2, 233) = 2.54, p < 0.08. The decomposition of this interaction shows that the interaction between the two independent variables (status × evaluation) is significant only for participants with a favorable attitude towards voluntary euthanasia (84th percentile of the distribution, F (2, 233) = 3.36, p < 0.04 (cf. Figure 1). In fact, among the participants opposed to the message of the source, the direct influence exerted by the minority consensus in conditions of intensity evaluation (M = 2.73) is, as expected, significantly higher than that obtained in conditions of non-evaluation of the minority consensus (M = 1.82) (b = 0.48, SE = 0.21, t = 2.35, p < 0.02); however, it does not differ significantly from the condition of evaluation of the size of the consensus (M = 2.18) (b = –0.06, SE = 0.22, t = –0.29, ns). Still among participants far distant from the positions held by the source, this same direct influence exerted by the minority consensus in the condition of the evaluation of its intensity (M = 2.73) is higher than that obtained by its majority consensus (M = 1.99), but only when the participants are asked to evaluate the intensity of the consensus allegedly obtained by the source message among the population (b = 0.74, SE = 0.37, t = 2.02, p < 0.04).
Finally, in the non-evaluation conditions, the direct influence of the majority (M = 2.57) is tendentially higher than that exerted by the minority (M = 1.82, b = –0.75, SE = 0.45, t = –1.67, p < 0.10).
As far as indirect influence is concerned, our hypotheses seem to be partially confirmed. In fact, while the triple interaction is statistically significant, F (2, 233) = 3.58, p < 0.03, the expected interaction between status and orientation of the evaluation is only observed in participants with an initial favorable attitude towards voluntary euthanasia (84th percentile of the distribution, F (2, 233) = 6.33, p < 0.001). The decomposition of this interaction shows, however, that this time it is the majority consensus that exerts the highest indirect influence (M = 4.42) compared to the minority consensus (M = 2.58) in the condition of non-evaluation of the consensus, b = –1.84, SE = 0.56, t = –3.29, p < 0.001 (cf. Figure 2).
In addition, the indirect influence exerted by the majority in the non-evaluation condition is significantly higher than that observed in the condition of intensity evaluation (M = 2.97, b = –1.45, SE = 0.52, t = –2.78, p < 0.006) as well as in the condition of size evaluation (M = 3.41, b = –1.00, SE = 0.47, t = –2.14, p < 0.03) In other words, it would seem that when participants are engaged in the evaluation of the consensus (regardless of the dimension of the consensus being evaluated), they are more resistant to indirect majority influence than when they are not asked to make an evaluation of this presumed consensus.
Things are quite different with regard to the indirect influence of the minority. In fact, when it comes to the conditions of minority consensus, among distant participants, it is always the non-evaluation condition that produces a weaker indirect influence (M = 2.57, b = 1.16, SE = 0.52, t = 2.20, p < 0.028) compared to the condition of size evaluation and, in a tendentially significative manner, compared to that recorded in the condition of intensity evaluation (M = 3.51, b = 0.94, SE = 0.50, t = 1.85, p < 0.064). Between the two conditions of evaluation of the intensity and the size of the minority consensus, no significant difference is observed (b = 0.21, SE = 0.47, t = 0.51, p < 0.61). In other words, it would seem that when participants are asked to evaluate the consensus obtained (regardless of the dimension being evaluated), they are more indirectly influenced by the minority than when they do not proceed to any evaluation of the presumed consensus.
Perception of the argumentative quality of the influence inducing text
Factor analysis (with varimax rotation) applied to the data relating to the characteristics of the arguments used by the source in its influence inducing text (total variance explained: 61.09%) identified 3 factors. The first factor (explained variance: 27.55%, eigenvalue: 2.48) can be summarized as describing the perception of the consensual aspect of the arguments since the factor in question is more saturated by the following items: moderate (0.743), realistic (0.742), convincing (0.685), comprehensible (0.611) and consensual (0.541). The second factor (explained variance: 21.04%, eigenvalue: 1.89) would describe the perceived originality of the argument to the extent that it is especially saturated by the items: original (–0.781), rare (–0.743) and interesting (–0.635), while the third factor (variance explained: 12.50%, eigenvalue: 1.12), highly saturated by a single item: defended by the minority of public opinion (0.927), would rather refer to the perception of the minority aspect of the argumentation of the source (M = 3.56, SD = 1.54). On the basis of this factor analysis, two indexes were constructed: an index of the perceived consensual aspect (α = 0.698, M = 4.58, SD = 1.04) and one of the perceived originality (α = 0.586, M = 3.70, SD = 1.18) of the source’s arguments.
In order to examine the mediating role of the three perceptual dimensions of the message content in the production of direct and indirect influence by the interaction between the status of the consensus, its oriented evaluation, and the initial attitude towards voluntary euthanasia, we performed a moderated mediation analysis using PROCESS macro (model 11; Hayes, 2013; Hayes, 2017). We used bootstrapping (10,000 bias correction iterations) with 95% CI. The analysis conducted for the direct influence demonstrated that only one dimension, that relating to the perception of the originality of the message plays a mediating role in the interaction between the independent variables. The analysis revealed that it is only for participants strongly in favor of voluntary euthanasia (84th percentile of the distribution) placed in the conditions of the consensus intensity evaluation that the impact of the consensus status is mediated by the dimension of the originality of the message, Effect = 0.33, BootSE = 0.18, 95% CI [0.03, 0.72], index of conditional moderated mediation = 0.27, BootSE = 0.10, 95% CI [0.01, 1.07]. In other words, among ‘distant’ participants, the superiority of the minority status of the consensus in the production of direct influence, compared to that produced by the majority status, requires a comparatively higher evaluation of the message as original, interesting and rare.
The goal of the present experiment was to examine the effects of the orientation of the evaluation of the majority or minority consensus allegedly attributed to a message on the influence this latter can exert. Its innovative aspect consists of the explicit focusing of the participants on constituent elements other than the target of influence. We predicted that directing participants’ attention to the source and its characteristics (identificatory, ideological, normative or other) would not be equivalent to focusing on the size of the consensus it is supposed to obtain (and the connotations of resulting numerical strength), nor on the intensity of this same consensus and the meanings it may entail. While at the conceptual level the dimension of consensus intensity has played an important role in the construction of the theory of active minorities, it has been relatively neglected at the empirical level.
The results of this experiment partially confirm our hypotheses. First of all, data relating to the direct and indirect influence of the source show that only participants with a certain ‘distance’ from the positions of the source are sensitive to the manipulation of the independent variables. They also demonstrate the expected rupture between non-evaluation and consensus evaluation (regardless of the orientation of the latter), both directly and indirectly. On the contrary, the predicted differentiation of the effects of the consensus evaluation according to its orientation has not been satisfactorily confirmed.
In particular, in conditions of non-evaluation (similar to the ‘classic’ conditions of influence), the majority consensus implies tendentially more direct influence than does the minority consensus, a result compatible with those encountered in the psychosocial literature, however, the majority consensus also engenders a higher indirect influence than that of the minority consensus. This is an unusual but understandable result. The text of influence in this experiment is indeed composed of strong and weak arguments; however, according to Chaiken and Maheswaran (1994), this confers on the text an ambiguity that would prompt a more elaborate processing of the message in the majority than in the minority condition, which would explain its increased indirect influence.
However, the most interesting result is that this response pattern changes radically in the consensus evaluation conditions.
A) In terms of direct influence, when the participants are asked to evaluate any dimension of the support supposedly obtained by the source, they no longer reproduce the usual effect of the superiority of the direct majority influence: either this effect disappears (size evaluation conditions) or it is reversed (intensity evaluation conditions). It seems that evaluating the consensus obtained by the source would make the communicative aspect of any influence process more visible. By emphasizing the concession of social support to the positions of the source, we would effectively highlight its effectiveness as a communication agent. In the case of the minority consensus, this could also partly put aside its usual obstacles (such as high psychosocial cost or relative social invisibility) that traditionally ruin its chances of exerting a direct influence. If, in addition, the participants evaluate the intensity of the support granted to the positions of the source, they will be made more aware than usual of the personal commitment of the recipients of the message, as well as the consistency with which they would have expressed their support. This is what is suggested in the first place by the increased direct influence in the intensity evaluation conditions. It is also particularly suggested by the finding that the superiority of direct minority influence, compared to that produced by the majority status, is mediated by the comparatively higher evaluation of the message as original, interesting and rare.
B) In terms of indirect influence, we equally observe that in the evaluation conditions—regardless of the dimension concerned—the capacity of the majority consensus to induce an indirect influence superior to that obtained by the minority consensus disappears. We also observe that this leveling of the indirect impact seems to be the product of two complementary effects: the diminution of indirect influence by the evaluation of the majority consensus on the one side, the increase of indirect influence by the evaluation of the minority consensus on the other side. Perhaps in the first case, the evaluative task of the majority consensus distracted the participants in non-evaluation conditions from the supposedly accentuated elaboration of the persuasive message (because of its ambiguity due to its strong and weak arguments), which turned out to diminish the message and consequently weaken its indirect influence. Perhaps, also, that in the second case, the evaluative procedure of the minority consensus would encourage the participants to reflect more on the reasons for the communicative effectiveness of the source and, probably, lead them to an increased processing of the message, and thus to a higher indirect influence. These interpretations merit to be empirically tested in future research.
The present experiment does not solve all the theoretical and methodological issues it raises. Two of its major effects, the superiority of direct minority influence and the superiority of indirect majority influence, although understandable, are quite unusual in literature. They suggest that these ‘obstacles’ traditionally set before the two antagonistic social entities in any influence process, are not necessarily unavoidable. They obviously require a careful examination of the conditions of their appearance, as well as other experimental confirmations, before we can be sure of their robustness. They also require additional adjustments, theoretical as well as methodological, which will refine these results and clarify some not yet fully elucidated points, such as: a) the experimental control of functional consequences of the social support accorded to the source which probably are different from those of the source’s group essence, or b) the testing of the superiority of the indirect influence of the majority, by comparing the communicative effectiveness of strong, weak, or compound argumentation.
This experiment, however, also suggests that the evaluation of the intensity and size of the consensus are two very distinct axes of the connotations activated in a process of social influence. We consider that the multiplication of experimental research on various forms of processing which the recipients of influence would perform, could complement, and even renovate, the plethoric and virtually saturated production of works centered on the only component of the source of influence. The development of a research line evolving around the recipients of the message of the source could thus contribute to highlighting the societal aspect of influence, somewhat neglected in recent years (Papastamou, Gardikiotis & Prodromitis, 2017). Indeed, the activation of variables relating to the socio-cognitive functioning of the recipients could introduce them as protagonist societal entities in the social interaction of the processes of influence. In fact, let us not forget that until now, researchers traditionally considered the recipients of influence as a simple aggregate of individuals and, as a result, the interest they showed in the target was mainly confined to examining its multiple and varied individual differences (relating to knowledge, internal coherence-consistency, social approval, motivational preferences, etc.; for an almost exhaustive review, see Briñol & Petty, 2005).
On the contrary, the psychosocial perspective we propose here implies that the target of influence could and should henceforth be considered as the object of representational elaborations by the participants, just like the source of influence, as we have had the opportunity to suggest on several occasions (Mugny & Papastamou, 1984; Papastamou & Prodromitis, 2012).
Data Accessibility Statement
Data used in the present research has been made available and can be accessed through the following link: https://osf.io/7v2qk/?view_only=401698de2bf548a1900b401ba50fe4de.