Journal of Experimental Social Psychology 85 (2019) 103853

Contents lists available at ScienceDirect

Journal of Experimental Social Psychology

journal homepage: www.elsevier.com/locate/jesp

This paper has been recommended for acceptance by Ernestine Gordijn

T
Why people abandon groups: Degrading relational vs collective ties uniquely
impacts identity fusion and identification
⁎
Ángel Gómeza, , Alexandra Vázqueza, Lucía López-Rodríguezb, Sanaz Talaifarc,
Mercedes Martíneza, Michael D. Buhrmesterd, William B. Swann Jr.c
a
Universidad Nacional de Educación a Distancia, UNED, Madrid, Spain
b
Universidad de Almeria, Spain
c
University of Texas, Austin, USA
d
University of Oxford, UK

A R T I C LE I N FO A B S T R A C T

Keywords: Six studies explored the mechanisms that diminish allegiance to social groups. Results showed that degrading
Identity fusion either collective ties (i.e., sentiments toward the group as a whole) or relational ties (i.e., sentiments toward
de-fusion individual group members) lowered identity fusion with the group (Studies 1–3 & 6). Lowered fusion, in turn,
Extreme behavior explained the tendency for weakened collective and relational ties to reduce endorsement of pro-group action,
Collective ties
and this effect replicated cross-sectionally and longitudinally (Studies 2, 4 & 6). Additional evidence revealed
Relational ties
that attenuated group-related agency seemed to mediate the relationship between weakened identity fusion and
diminished commitment to help the group (Studies 3, 4 & 6). Although degrading collective ties reduced overall
group identification (Studies 1–3 & 6), degrading relational ties did not reliably do so (Studies 1–6). Instead,
degrading relational ties only reduced scores on a single component (ingroup solidarity) of a multidimensional
measure of group identification (Studies 5 & 6). Hence, measures of identity fusion are equally sensitive to
relational and collective ties while measures of identification emphasize collective ties. The results replicated
whether we considered country (Studies 1–5) or gender (i.e. females, Study 6) as the focal social group. These
findings therefore highlight the unique properties of fusion and identification and help explain why identity
fusion predicts extreme pro-group behavior with greater fidelity than group identification.

1. Why people abandon groups: the toll of degraded relational as fellow group members (“relational ties”). Together, these strong alle-
well as collective ties on identity fusion giances trigger feelings of group-related agency that motivate pro-
group actions. Efforts to reduce fusion and its consequences should
People sometimes make extraordinary sacrifices for their group, therefore consider these dual forms of alignment with one's group and
including even sacrificing their own lives. This phenomenon is pro- the feelings of group-related agency they inspire. To contextualize these
blematic when, for example, malevolent actors sacrifice themselves in assertions, we begin by noting how they build upon previous for-
the service of killing others. Such instances call for developing ways to mulations.
reduce the psychological forces that inspire such extreme behaviors.
Here, we focus on reducing identity fusion, a form of alignment with 1.1. Identity fusion as an index of alignment with groups
groups that predicts extreme pro-group behaviors with exceptionally
high fidelity (for original statements of identity fusion theory, see To a greater degree than classic theories of group relations (Tajfel &
Gómez et al., 2011; Swann Jr., Gómez, Seyle, Morales, & Huici, 2009; Turner, 1979; Turner, Hogg, Oakes, Reicher, & Wetherell, 1987),
Swann Jr., Jetten, Gómez, Whitehouse, & Bastian, 2012). Identity fu- identity fusion theory (Swann Jr. et al., 2012) emphasizes the intra-
sion refers to a visceral sense of oneness with a group. Fusion is not only group dynamics that cause people to make extreme sacrifices for the
marked by strong allegiance to the group category and the goals and group. Most important, while acknowledging the importance of col-
values it represents (“collective ties”), but also by strong allegiance to lective ties in predicting pro-group behavior, fusion theory also

⁎
Corresponding author.
E-mail address: [email protected] (Á. Gómez).

https://doi.org/10.1016/j.jesp.2019.103853
Received 16 February 2019; Received in revised form 22 July 2019; Accepted 22 July 2019
Available online 31 July 2019
0022-1031/ © 2019 Elsevier Inc. All rights reserved.
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

emphasizes the motivational role of relational ties and the personal self. relational ties into two new measures of identification. Note, however,
In so doing, fusion theory departs from social categorization theory's that face validity does not guarantee construct validity (Cronbach &
assumption that group functioning is a uniquely group-level (as op- Meehl, 1955) and evidence that these scales are actually related to
posed to interpersonal-level) process that is motivated by perceptions of relational ties is currently lacking.
ingroup similarity and differentiation from the outgroup. Specifically, Such forays into the role of the personal self and relational ties in
the relational ties principle of fusion theory embraces the interpersonal group relations (e.g., Baray et al., 2009; Hornsey & Jetten, 2004, 2005;
level of group functioning, arguing that group members value and be- Jans, Postmes, & van der Zee, 2011; Postmes & Jetten, 2006; Postmes,
come attached to fellow group members and this occurs even if they are Spears, Lee, & Novak, 2005; Van Zomeren, Saguy, & Schellhaas, 2013)
not personally acquainted with them. Hence, identity fusion theory is may cause social identity theorists to impugn our juxtaposition of fu-
concerned with intragroup relations as well as intergroup relations. sion theory against classic theories of group identification instead of
The more inclusive approach taken by identity fusion theory has led recent informal revisions of those theories. We do so because the classic
to the development of measures that capture not only collective ties to theories have never been formally revised. As a result, the early state-
the group but also relational ties to other group members and group- ments continue to be regarded as the theories of group identification.
related agency (Gómez et al., 2011). In contrast, measures of group Moreover, incorporation of relational ties into social identity formula-
identification – the standard measure of alignment with groups – have tions continues to be left to the discretion of individual investigators. In
historically emphasized allegiance to the group as a whole rather than addition, the emphasis on the sovereignty of collective ties in the classic
to its individual members. This likely explains why fusion measures theories continues to shape the construction of measures of identifica-
have proven to be uniquely associated with both relational ties and the tion even to this day. The result is that even the newer measures of
personal self. For example, fusion scores predict the strength of rela- identification (e.g., Leach et al., 2008) emphasize collective ties over
tional ties (e.g., Vázquez, Gómez, & Swann, 2017) and relational ties, in relational ties. This means that manipulations designed to degrade
turn, mediate the link between fusion and sacrifice for the group (e.g., collective ties should diminish identification but manipulations de-
Buhrmester, Fraser, Lanman, & Whitehouse, 2014; Swann Jr. et al., signed to degrade relational ties should not. We examine this possibility
2014). Similarly, the influence of fusion on pro-group behavior is am- in our research.
plified by increasing the salience of the personal self either through The central question we address here is how one might lower fusion
physical exercise (Swann Jr., Gómez, Huici, Morales, & Hixon, 2010) or and its correlates. Although there is no research addressing this issue
writing about the self (Gómez et al., 2011; Swann Jr. et al., 2009). directly, one series of studies suggest that fusion can indeed change. In
Given that fusion theory emphasizes three key constructs (collective particular, major socio-political events in Spain (a scandal involving the
ties, relational ties and the personal self) whereas classic theories of royal family and threats of succession by a prominent province) did
group processes emphasize only one of these constructs (collective ties), diminish fusion by eroding collective ties to the country (Vázquez et al.,
it is not surprising that fusion theory spawned measures that are 2017). Nevertheless, although the negative events eroded collective ties
stronger predictors of extreme pro-group behavior than measures of and fusion, they did not decrease relational ties or willingness to fight
identification. For example, fusion out-predicts identification when the and die for the group. Furthermore, because Vázquez et al. (2017) ex-
outcome is endorsement of fighting and dying for ingroup members amined the correlates of naturally occurring events, the causal impact
(Gómez et al., 2011; Swann Jr. et al., 2009), self-sacrifice to save group of those events is not firmly established.
members in variations of the trolley dilemma (Gómez et al., 2011; To fill the foregoing gaps in the literature, we adopted an experi-
Swann Jr. et al., 2014; Swann Jr., Gómez, Dovidio, Hart, & Jetten, mental approach to lowering alignment with the participant's country.
2010), or donating personal funds to group members under duress The general goal of these experiments was to observe the effects of
(Buhrmester et al., 2014; Swann Jr., Gómez, Huici, et al., 2010; for systematically compromising collective and relational ties to the group
reviews, see Fredman et al., 2015; Gómez & Vázquez, 2015; Swann Jr & on identity fusion, group identification, and subsequent group-related
Buhrmester, 2015). agency and pro-group behaviors.
To be sure, in recent years, researchers have begun to explore the We expected that degrading collective or relational ties would lower
ways in which the relational ties of group members promote identifi- identity fusion (Study 1) and that lowered fusion would mediate the
cation with groups (e.g., Brewer & Gardner, 1996; Prentice, Miller, & effect of weakened relational and collective ties on pro-group behavior
Lightdale, 1994), especially within non-Western cultures (e.g., Brewer (Study 2). Also, we anticipated that lowering identity fusion by com-
& Yuki, 2007; Yuki, 2003; Yuki & Takemura, 2013). Researchers promising relational or collective ties would reduce group-related
working within the social identity tradition have shown that identifi- agency, and such reductions would, in turn, mediate the effects of fu-
cation with the group category decreases when relational ties to fellow sion on pro-group behavior (Studies 3–4). Furthermore, we predicted
group members are jeopardized by intra-group friction (Bhappu & that these previous effects would be extended to actual commitment
Crews, 2005), lack of cooperation (Griffith, 2002), or violations of in- behaviors such as commitment to help the group (Studies 4–6). In
tragroup trust and differences between the individual and other group contrast, we anticipated that only degrading collective ties would lower
members (e.g., Becker & Tausch, 2014; Becker, Tausch, Spears, & overall group identification scores (Studies 1–6). Furthermore, we ex-
Christ, 2011; Glasford, Pratto, & Dovidio, 2008; Zagenczyk et al., pected that, as in past studies, the direct effects of fusion on pro-group
2013). Others have noted the importance of the personal self in align- behavior would be stronger than the direct effects of identification.
ments to groups (e.g., Baray, Postmes, & Jetten, 2009; Brewer & Several features of our methodologies deserve comment. First, in
Gardner, 1996), in that identification weakens when people recognize light of past evidence that enduring alignments to important groups
that the ingroup prototype deviates from personal values (Becker & such as one's country tend to persist over time (Swann Jr. et al., 2012),
Tausch, 2014; Ellemers, van den Heuvel, de Gilder, Maass, & Bonvini, we assumed that our manipulations would influence state versions of
2004; Elsbach & Bhattacharya, 2001) or when group members fail to group alignment (how participants feel toward the group at the moment)
verify the individual's (negative as well as positive) personal identities but not necessarily trait versions (how people feel toward the group in
(Simon & Stürmer, 2003; Swann, 2012; Swann Jr, Milton, & Polzer, general). To test this assumption, we measured both state and trait
2000). Furthermore, some have noted that the identification of group fusion and identification in Study 1. Second, we employed both be-
members can translate into their relational identification with leaders tween-subjects and within-subjects designs to assess change in fusion.
(Steffens, Haslam, & Reicher, 2014) and have noted the implications of Using between-subjects designs (Studies 1–3 and 5–6), we inferred
these processes for collective action (Drury & Reicher, 2009). Re- change by comparing fusion scores across experimental conditions
searchers (e.g., Ellemers, Kortekaas, & Ouwerkerk, 1999; Leach et al., during a single slice of time. Using a within-subjects design (Study 4),
2008) have even begun to incorporate face-valid items designed to tap we observed change directly by measuring the predictor variables

2
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

(fusion and identification), introducing a manipulation, and observing friends against the foreign inhabitants (blacks normally) who live in our
the impact of the manipulation on the outcome measures (fusion, country, I have felt a lot of shame and indignation when I have seen this
identification, agency and endorsement of pro-group behavior). Third, type of behavior displayed by close friends”; “When I was in a restau-
while Studies 1–3 focus on the impact of our manipulations on en- rant with some friends some other Spaniards sitting there started to
dorsement of pro-group behavior, to determine if our effects would insult others because they were immigrants. It was very embarrassing.”
generalize to overt behavior, in Studies 4, 5 and 6 we measured parti- Participants in the control condition described their two last trips to
cipants' commitment of time to help the group. Fourth, to strengthen their workplace or university.
the results of our mediational hypotheses, in Study 5 we manipulated To check on the effectiveness of the manipulations, we asked two
the potential mediator, personal agency, identified in Studies 3 and 4. independent judges who were blind to the goals of the research to
Fifth, to provide stronger tests of our prediction that our single measure evaluate the examples of degraded ties generated by participants. After
of fusion would be more predictive of our outcome measures than reading a description of the meaning of relational and collective ties,
identification, we employed four widely used measures of identifica- judges scored the narrative as “1” when there was agreement between
tion, including measures that have not been used in previous studies of what participants were encouraged to contemplate (e.g., degraded re-
identity fusion. And sixth, to reinforce the generalizability of our con- lational ties) and what they were actually contemplated (e.g. degraded
clusions, we tested whether our effects would generalize to two dif- relational ties vs. degraded collective ties), “0” when there was dis-
ferent group identities, country (Studies 1–5) or gender (Study 6). We agreement (e.g. the manipulation encouraged them to report degrade
report all measures, manipulations, and exclusions in these studies. relational ties and they instead reported degraded collective ties), or no
Although we did not determine a particular sample size a priori, all number when they wrote something unrelated to the manipulation. The
studies were open for a week and then closed and no additional data judges' ratings were closely associated (kappas = 0.843 and 0.842 in
were collected subsequently. Participants in all the studies provided the relational and collective ties conditions, respectively). In the rela-
their consent before participating in each study. tional ties condition, Judge1 detected agreement fully 91.4% of the
time (with only 5.17% of participants judged as referring to collective
2. Study 1: does degrading collective and relational ties decrease ties); Judge 2 detected agreement 90.5% of the time (only 6.03% re-
alignment with the group? ported information referred to collective ties). In the collective ties
condition, judge 1 detected agreement 91.1% of the time (only 3.57%
Whereas identity fusion should vary as a function of both collective reported information referred to relational ties); judge 2 considered
ties to the group as well as relational ties to members of the group, that 90.2% of participants reported examples relative to collective ties
group identification should vary as a function of collective ties only. To (only 5.36 reported information referred to relational ties). These
test this hypothesis, we asked participants to recall situations that made findings affirm the effectiveness of the manipulation.
them question either their collective ties to the group or their relational As a secondary manipulation check, we had participants responded
ties to individual members of the group. Then, we measured feelings of to two, 3-item scales adapted from a previous study (Vázquez et al.,
alignment (i.e., fusion and identification) with the group, with parti- 2017). One scale measured their collective ties with Spain: “Right now,
cipants randomly assigned to complete either trait or state versions of I feel strong ties to my country,” α = 0.89; the other measured their
alignment. We expected that compromising either collective or rela- relational ties to other Spaniards: “Right now, I feel strong ties to the
tional ties would lower identity fusion but that only compromising members of my country,” α = 0.86. The manipulation checks were
collective ties would lower identification with the group. Furthermore, successful in all Experiments (see Table 1 of supplemental material).
we anticipated that these changes in alignment to the group would be After completing the manipulation checks, depending on condition,
reflected in measures of state, but not trait, fusion and identification. participants completed either the trait or state versions of the measures
of alignment with the group (i.e., fusion and identification). The trait
2.1. Method versions were identical to the scales used in previous research. For the
state versions of the measures of alignment, we modified the trait
2.1.1. Participants version by adding “Right now” to the state version. For example, we
We recruited participants using the snowball technique wherein modified the trait fusion item “My group is me” to be “Right now I feel
undergraduates enrolled in an online distance learning psychology that my group is me”.
course from UNED asked their acquaintances to participate. A total of Within both the trait and state conditions, participants completed,
370 Spanish volunteers participated online (55.9% women, in random order, the verbal fusion scale and the three measures of
Mage = 36.39 years, SD = 11.85). identification (this randomization procedure was followed in all ex-
periments included in this report). All measures were internally con-
2.1.2. Procedure sistent, including the verbal fusion scale (αs = 0.80 and 0.88, trait and
Participants were randomly assigned to each condition in a 3 (type state, respectively), Mael and Ashforth (1992), αs = 0.84 and 0.87, and
of tie: collective, relational, control) X 2 (measure of alignment: Trait Ellemers et al. (1999) scale, αs = 0.83 and 0.79. Alpha for the single-
vs. State) between-subjects design. In the collective ties condition, parti- item scale (“Right now I identify with my group”) devised by Postmes,
cipants described two of their country's actions that made them ques- Haslam, and Jans' (2013) obviously could not be computed. Responses
tion their relationship with their country. Examples generated by the to all scales ranged from 0 (totally disagree) to 6 (totally agree).
participants included: “The way the Government and the politicians are
framing the conflict of the Independence of Catalonia” or “The repeated 2.2. Results and discussion
cases of political corruption in the national or regional governments.”
Participants in the degraded relational ties condition described two We performed a single analysis on the outcome variables using the
interactions with fellow countrymen that made them question their Lavaan package in R software. We entered condition (type of tie: de-
relationship with the members of their country. Participants reported graded collective ties vs. degraded relational ties vs. control), measure
examples of degrading relational ties with individuals ranging from of alignment (trait vs. state) and the 2-way interaction as predictors of
those whom they did not know personally to family members or close the regression analysis. As the condition had three levels, we created
friends (e.g. “A negative interaction that I remember having had was to two dummy codes with the control condition as the comparison group.
go with a friend for the fiestas of San Isidro in Madrid and he insulted The first dummy code (D1) compared the control condition with the
an Ecuadorian boy just for not being Spanish. At that time I felt very far degraded collective ties condition, whereas the second (D2) compared
from my Spanish friend”; “On some occasions I have felt racism of the control condition with the degraded relational ties condition.

3
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

Table 1
Experiments 1–4. Correlations.
Fusion M&A Postmes Ellemers Fight die Agency

Exp.1 Fusion – 0.50 0.54 0.36

M&A 0.55 – 0.62 0.40
Postmes 0.49 0.59 – 0.62
Ellemers 0.43 0.41 0.68 –
Exp.2 Fusion – 0.56 0.53 0.62 0.43
M&A – – 0.50 0.47 0.24
Postmes – – – 0.56 0.20
Ellemers – – – – 0.23
Fight die –
Exp.3 Fusion – 0.57 0.46 0.53 0.40 0.33
M&A – – 0.47 0.63 0.28 0.18
Postmes – – – 0.58 0.20 0.18
Ellemers – – – – 0.17 0.18 Fig. 1. State fusion and identification scores as a function of the Type of
Fight Die – 0.29
Manipulation (Study 1).
Agency –
Exp.4 Fusion – 0.39 0.40 0.38
M&A 0.53 – 0.67 0.57 2.2.1. Changes in fusion
Postmes 0.60 0.55 – 0.71
The regression on fusion scores yielded two significant interaction
Ellemers 0.47 0.46 0.70 –
Fight die 0.46 0.48 0.27 0.23 effects, B = −1.08, p < .001, 95% CI [−1.68, −0.48] for Interaction
Agency 0.48 0.43 0.24 0.21 0.66 1 (D1xVersion), and B = −1.12, p < .001, 95% CI [−1.71, −0.52]
Hours 0.33 0.23 0.12⁎ 0.10+ 0.38 0.38 for Interaction 2 (D2xVersion). As shown in Fig. 1, state fusion was
lower in the degraded collective ties condition, B = −0.91, p < .001,
Notes: * p = .029; + p = .062. The remainder correlations have a p < .010.
95% CI [−1.35, −0.49], M = 1.83, SD = 1.04, and in the degraded
M&A = Mael and Ashforth (1992).
In Experiment 1 correlations below the diagonal correspond to the trait version
relational ties condition, B = −0.92, p < .001, 95% CI [−1.34,
of fusion and identification measures, and correlations above the diagonal −0.49], M = 1.82, SD = 1.09, relative to the control condition,
correspond to the state version of the measures. M = 2.74, SD = 1.48. Trait fusion did not differ across conditions,
In Experiment 4 correlations below the diagonal correspond to the second however, Bs = 0.16 and 0.20, ps = 0.456 and 0.351 for the Interaction
wave, and correlations above the diagonal correspond to the first wave. 1 and Interaction 2, respectively.
The main effect of measure of alignment was marginally significant,
Table 2 B = 0.36, p = .076, 95% CI [−0.04, 0.76], indicating that state fusion
Experiment 5. Correlations. tended to be lower than trait fusion, M = 2.16, SD = 1.31 vs. M = 2.49,
1. 2. 3. 4. 5. 6. 7.
SD = 1.20.

1. Fusion –
2. Solidarity 0.55⁎⁎ – 2.2.2. Changes in identification
3. Satisfaction 0.45⁎⁎ 0.67⁎⁎ – The regression on Mael and Asforth's scale only yielded a significant
4. Centrality 0.45⁎⁎ 0.51⁎⁎ 0.57⁎⁎ – effect of the interaction between D1 and measure of alignment,
5. Self-stereotyping 0.44⁎⁎ 0.46⁎⁎ 0.54⁎⁎ 0.55⁎⁎ –
6. Homogeneity 0.30⁎⁎ 0.23⁎⁎ 0.33⁎⁎ 0.29⁎⁎ 0.47⁎⁎ –
B = −0.87, p = .013, 95% CI [−1.56, −0.19]. State identification was
7. Fight die 0.43⁎⁎ 0.20⁎⁎ 0.10⁎ 0.19⁎⁎ 0.22⁎⁎ 0.10⁎ – lower in the degraded collective ties condition than in the control
8. Hours 0.41⁎⁎ 0.14⁎⁎ 0.05 0.13⁎⁎ 0.10⁎ 0.11⁎ 0.37⁎⁎ condition, B = −0.78, p = .001, 95% CI [−1.26, −0.30], M = 2.74,
SD = 1.29 and M = 3.52, SD = 1.46, respectively. However, trait
⁎⁎ ⁎
Notes: p < .010, p < .050. identification was similar in both the degraded collective ties and
control conditions, B = 0.09, p = .713. Unlike fusion, identification
Interaction 1 refers to the interaction between Dummy 1 and measure was not affected by the manipulation of relational ties.
of alignment, whereas Interaction 2 refers to the interaction between The regression on the Postmes et al.'s (2013) measure only yielded a
Dummy 2 and measure of alignment. We will refer only to significant or significant effect of the interaction between D1 and measure of align-
marginal effects below but report all other effects as well as the means ment, B = −1.32, p = .001, 95% CI [−2.10, −0.54]. State identifi-
and standard deviations for each condition in the supplemental mate- cation was lower in the degraded collective ties condition than in the
rial (Tables 2 and 3). Table 1 shows the correlations between the out- control condition, B = −1.15, p < .001, 95% CI [−1.70, −0.60],
come variables for each study. M = 2.25, SD = 1.62 and M = 3.41, SD = 1.90, respectively. Trait
identification was similar in both the degraded collective ties and

Table 3
Experiment 6. Correlations.
1. 2. 3. 4. 5. 6. 7. 8.

1. Fusion –
2. Solidarity 0.47⁎⁎ –
3. Satisfaction 0.32⁎⁎ 0.25⁎⁎ –
4. Centrality 0.23⁎⁎ 0.32⁎⁎ 0.53⁎⁎ –
5. Self-stereotyping 0.20⁎⁎ 0.18⁎⁎ 0.19⁎⁎ 0.38⁎⁎ –
6. Homogeneity 0.12 0.23⁎⁎ 0.15⁎ 0.23⁎⁎ 0.44⁎⁎ –
7. Agency 0.38⁎⁎ 0.14⁎ 0.06 −0.01 0.23⁎⁎ 0.25⁎⁎ –
8. Fight die 0.51⁎⁎ 0.27⁎⁎ 0.16⁎⁎ 0.09 0.04 0.09 0.38⁎⁎ –
9. Hours 0.31⁎⁎ 0.23⁎⁎ 0.12⁎ 0.12⁎ 0.12 0.15⁎⁎ 0.38⁎⁎ 0.50⁎⁎

⁎⁎ ⁎
Notes: p < .010, p < .050.

4
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

control conditions, however, B = 0.17, p = .557. Once more, the ma- questionnaire online.
nipulation of degraded relational ties had no impact on identification.
The regression on the Ellemers et al.'s scale yielded a significant 3.1.2. Procedure
effect of the interaction between D1 and Version, B = −0.72, p = .004, As in Study 1, the experimenter assigned participants to one of three
95% CI [−1.20, −0.23]. State identification was lower in the degraded conditions: degraded collective ties, degraded relational ties, or control.
collective ties condition than in the control condition, B = −0.59 After the manipulations, participants completed the manipulation
p = .001, 95% CI [−0.93, −0.25], M = 3.08, SD = 0.89 and M = 3.67, checks, αs = 0.90 vs. 0.85 for the collective and relational ties scales
SD = 1.15, respectively. However, trait identification was similar in respectively.
both the degraded collective ties and control conditions, B = 0.12, After completing the manipulation checks, participants responded
p = .478. The effect of D2 was significant, B = 0.36, p = .037, 95% CI to the state versions of the fusion and identification scales used in Study
[0.02, 0.70], indicating that identification was slightly higher in the 1, αs > 0.82. Finally, they completed Swann Jr. et al.'s (2009) 7-item
degraded relational ties condition than in the control condition, measure of willingness to fight/die for the group (e.g., “I would fight
M = 3.98, SD = 0.94 and M = 3.81, SD = 1.11, respectively. Finally, someone physically threatening another Spaniard”), α = 0.81.
the effect of measure of alignment was marginally significant,
B = −0.29, p = .084, 95% CI [−0.61, 0.04], indicating that state 4. Results and discussion
identification tended to be lower than state identification, M = 3.48,
SD = 1.01 vs. M = 4.10, SD = 1.00. As in Study 1, we performed a single analysis on the outcome
variables using the Lavaan package in R software. Since the predictor,
2.2.3. Sensitivity power analysis type of tie, had three levels, two dummy codes were created as in the
Sensitivity power analyses allow determining the minimum effect previous study. The non-significant effects are specified in the supple-
size that a certain study could detect based on the sample size recruited mental material (Table 4).
and the alpha level and power specified. We performed a sensitivity
power analysis using GPower (Faul, Erdfelder, Lang, & Buchner, 2007)
assuming an alpha significance criterion of 0.05. With a sample size of
370 participants and five predictors, we could detect a minimum effect 4.1.1. Fusion
size of f2 = 0.035 with 80% power. The regression on fusion yielded significant effects of both D1
(collective ties vs. control), B = −0.76, p < .001, 95% CI [−0.94,
2.2.4. Summary −0.58], and D2 (relational ties vs. control), B = −0.59, p < .001,
Consistent with prediction, the results indicated that having parti- 95% CI [−0.77, −0.41]. Participants in both the degraded collective
cipants describe actions that compromised their relationship with the ties, M = 1.68, SD = 1.14, and degraded relational ties conditions,
country as a whole (e.g. collective ties) or other members of their M = 1.85, SD = 1.31, expressed weaker fusion than participants in the
country (e.g. relational ties) diminished their feelings of state fusion control condition, M = 2.45, SD = 1.27.
with the group relative to a control condition. In contrast, compro-
mising collective ties—but not relational ties—diminished state iden- 4.1.2. Identification
tification with the group. The regressions on the identification scales only yielded significant
As expected, attenuations in fusion and identification emerged on effects of D1, B = −0.37, p < .001, 95% CI [−0.54, −0.19] for Mael
state but not trait measures. This finding is unsurprising given that and Asforth's scale, B = −0.56, p < .001, 95% CI [−0.75, −0.36] for
people can draw on a lifetime of experiences when they respond to trait Postmes et al.'s scale, and B = −0.31, p < .001, 95% CI [−0.46,
measures as compared to that moment when they respond to state −0.17] for Ellemers et al.'s scale. Identification was weaker in the
measures. In light of these findings, in the remaining experiments we degraded collective ties condition than in the control condition. None of
only included state measures of fusion and identification. the effects of D2 (control vs. degraded relational ties) was significant.

3. Study 2: does the reduction of collective and relational ties 4.1.3. Fight/die for the group
diminish willingness to sacrifice for the group? The regression on willingness to fight and die yielded significant
effects of both D1, B = −0.49, p < .001, 95% CI [−0.60, −0.37], and
Study 1 provided evidence that there exist distinct strategies for D2, B = −0.40, p < .001, 95% CI [−0.52, −0.29]. Participants in the
lowering identity fusion and identification. Nevertheless, although our degraded collective ties, M = 0.73, SD = 0.66, and degraded relational
manipulations of relational and collective ties were systematically re- ties conditions, M = 0.81, SD = 0.76, expressed less willingness to fight
lated to fusion and identification, it is unclear whether these manip- and die for the group than those in the control condition, M = 1.22,
ulations would likewise influence the theoretical consequences of SD = 0.99.
alignment with groups. Study 2 was designed to replicate evidence that
compromising relational and collective ties would lower fusion and also 4.1.4. Indirect effects
one of its consequences, endorsement of fighting and dying for the To test whether the experimental condition could affect willingness
group. Since identity fusion is a powerful predictor of pro-group be- to fight/die for the group via lowered identity fusion, we conducted a
havioral intentions (Gómez et al., 2011; Swann Jr. et al., 2009), we mediational test using the Lavaan package in R software. D1 and D2
expected an indirect effect via fusion between our manipulation and were entered as predictors, identity fusion was the potential mediator
willingness to fight and die for the group. and willingness to fight/die for the group was the outcome variable.
This model was based on previous research revealing that fusion is a
3.1. Method powerful predictor of willingness to fight and die for the group (Gómez
et al., 2011; Swann Jr. et al., 2009). As shown in Fig. 2, the indirect
3.1.1. Participants effect via identity fusion was significant for D1, B = −0.19, p < .001,
As in Study 1, we utilized the snowball technique to recruit parti- 95% CI [−0.25, −0.14], and for D2, B = −0.15, p < .001, 95% CI
cipants. Because data collection fell over a long holiday weekend, [−0.20, −0.10], suggesting that compromising collective or relational
participation rate was unusually high, resulting in much larger sample ties could weaken willingness to fight/die for the group by decreasing
than participated in Study 1 (N = 1151 Spaniards; 62.6% women, feelings of identity fusion with the country. Parallel analyses in which
Mage = 37.11 years, SD = 12.34). Participants completed the we sequentially substituted each of the three measures of identification

5
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

Fig. 2. Indirect effect of compromising collective and relational ties on willingness to fight/die via identity fusion (Study 2).

for the measure of fusion revealed that indirect effects via D1 (control 5.1.2. Procedure
vs. degraded collective ties) were significant although smaller than the As in the foregoing studies, the experimenter assigned participants
indirect effects of fusion (see Table 5 of supplemental material). to one of three conditions: degraded collective ties, degraded relational ties,
or control. After the manipulations, participants completed the manip-
ulation checks, αs = 0.89 and 0.90, respectively.
4.1.5. Sensitivity power analysis
After completing the manipulation checks, participants responded
We performed a sensitivity power analysis using GPower (Faul
to the state version of the fusion and identification scales used in the
et al., 2007) assuming an alpha significance criterion of 0.05. With a
previous experiment, αs > . 83. They then completed the measure of
sample size of 1151 participants and two predictors, we could detect a
agency. We used five items based on Haggard and Tsakiris's (2009)
minimum effect size of f2 = 0.008 with 80% power.
discussion of the agency construct (see also Gómez et al., 2011; Swann
Jr., Gómez, Huici, et al., 2010). With reference to Spain, participants
4.1.6. Summary rated their agreement with a series of statements relevant to control of,
As in Study 1, compromising either collective or relational ties and responsibility for, the group (e.g., “I am able to control what my
weakened identity fusion relative to a control group. Furthermore, the group does in the same way that I control what I do,” “I usually feel
impact of the manipulations on diminutions in willingness to fight/die responsible for what my group does,” α = 0.78. Then they completed
for the group might occur via reductions in fusion. In contrast, com- Swann Jr. et al.'s (2009) 7-item measure of willingness to fight/die for
promising collective ties, but not relational ties, weakened identifica- the group, α =. 78.
tion. Moreover, the impact of collective ties (but not relational ties) on
willingness to fight/die for the group seems to be related with reduc- 5.2. Results and discussion
tions in identification. However, the indirect effects of fusion on will-
ingness to fight/die for the group were stronger than those of identifi- As in Experiments 1–2, we performed a single analysis on the out-
cation. come variables. The non-significant effects are specified in the supple-
mental material (Table 6).
5. Study 3: does agency mediate the impact of decreased fusion on
diminutions in willingness to fight/die for the group? 5.2.1. Fusion
The regression on fusion yielded significant effects of both D1
The foregoing results indicate that compromising both collective (collective ties vs. control), B = −0.92, p < .001, 95% CI [−1.24,
and relational ties weakened fusion and such diminutions, in turn, −0.59], and D2 (relational ties vs. control), B = −0.61, p < .001,
might weaken willingness to fight/die for the group. In contrast, only 95% CI [−0.94, −0.27]. Participants in the degraded collective ties,
diminutions in collective ties diminished identification and willingness M = 1.90, SD = 1.42, and degraded relational ties conditions,
to fight/die for the group. Although these findings illustrate the unique M = 2.22, SD = 1.59, expressed less fusion than participants in the
role of relational ties in fusion, they fail to provide evidence for another control condition, M = 2.82, SD = 1.47.
key variable that has been uniquely associated with fusion: group-re-
lated agency. Study 3 was designed to provide such evidence. Fused 5.2.2. Identification
individuals usually convert their elevated personal agency into en- The regressions on the identification scales only yielded significant
dorsement of pro-group behavior (e.g., Gómez et al., 2011; Swann Jr., effects of D1, B = −0.51, p = .002, 95% CI [−0.82, −0.19] for Mael
Gómez, Huici, et al., 2010). Thus, we explored the possibility of a serial and Asforth's scale, B = −0.39, p = .027, 95% CI [−0.74, −0.04] for
mediation effect in which degradations in relational and collective ties Postmes et al.'s scale, and B = −0.55, p < .001, 95% CI [−0.81,
diminished fusion and, in turn, group-related agency, and group-related −0.30] for Ellemers et al.'s scale. Identification was significantly
agency would mediate the effect of diminutions in fusion on willingness weaker in the degraded collective ties condition than in the control
to fight/die for the group. condition. None of the effects of D2 was significant.

5.1. Method 5.2.3. Agency

The regression on agency yielded significant effects of both D1,
5.1.1. Participants B = −0.40, p < .001, 95% CI [−0.59, −0.20], and D2, B = −0.36,
As in Study 1, we utilized the snowball technique to recruit 458 p < .001, 95% CI [−0.56, −0.16]. Participants in the degraded col-
Spaniards (59% women, Mage = 37.14 years, SD = 13.63). Participants lective ties, M = 0.80, SD = 0.71, and degraded relational ties condi-
completed the questionnaire online. tions, M = 0.84, SD = 0.79, expressed weaker feelings of agency for the

6
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

Fig. 3. Indirect effect of compromising collective and relational ties on willingness to fight/die via identity fusion and agency (Study 3).

group than participants in the control condition, M = 1.20, SD = 1.11. (see Table 7 of supplemental material).

5.2.4. Fight/die for the group 5.2.6. Sensitivity power analysis

The regression on willingness to fight and die yielded significant We performed a sensitivity power analysis using GPower (Faul
effects of both D1, B = −0.43, p < .001, 95% CI [−0.62, −0.23], and et al., 2007) assuming an alpha significance criterion of 0.05. With a
D2, B = −0.28, p = .007, 95% CI [−0.49, −0.08]. Participants in the sample size of 458 participants and two predictors, we could detect a
degraded collective ties, M = 0.91, SD = 0.79, and degraded relational minimum effect size of f2 = 0.021 with 80% power.
ties conditions, M = 1.05, SD = 1.01, expressed less willingness to fight
and die for the group than participants in the control condition, 5.2.7. Summary
M = 1.34, SD = 0.93. As expected, Study 3 replicated Studies 1–2 in that decreasing either
the perception of collective or relational ties weakened identity fusion.
5.2.5. Indirect effects In addition, as in Study 2, the experimental manipulations presumably
To determine if the effect of our manipulations on willingness to diminished willingness to fight and die for the group by diminishing
fight/die for the group could be serially mediated by identity fusion and fusion. Moreover, decreasing collective or relational ties weakened
agency, we tested the mediation model depicted in Fig. 3 using the feelings of agency, which, in turn, seemed to mediate the link between
Lavaan package in R software. This model was based on previous evi- fusion and willingness to fight and die. In contrast, identification was
dence that the effect of fusion on willingness to fight and die for the diminished by compromising collective but not relational ties. Finally,
group is mediated by personal agency (Gómez et al., 2011; Swann Jr., there was a significant serial indirect effect in which degradations in
Gómez, Huici, et al., 2010). The two dummy codes were entered as relational and collective ties weakened fusion and, in turn, group-re-
predictors, identity fusion was the first potential mediator, agency was lated agency, which in turn mediated the effect of diminutions in fusion
the second potential mediator and willingness to fight/die for the group on willingness to fight/die for the group.
was the outcome variable. This model had three distinct indirect effects
for each dummy code: 1) the effect via fusion and agency serially 6. Study 4: does compromising relational ties reduce commitment
(condition > fusion > agency > fight/die), 2) the effect via fusion to pro-group action?
alone (condition > fusion > fight/die), and 3) the effect via agency
alone (condition > agency > fight/die). Results indicated that all these Studies 1–3 offer converging evidence that degradations of both
indirect effects were significant. Relative to the control group, under- relational and collective ties weakened fusion but only compromising
mining collective ties (Dummy 1) seemed to weaken willingness to fight collective ties diminished identification. Nevertheless, the foregoing
and die for the group via 1) identity fusion and agency serially, findings were limited because we used between-subjects designs in
B = −0.03, p = .006, 95% CI [−0.05, −0.01], 2) identity fusion which participants were observed in a single slice of time. Hence, we
alone, B = −0.18, p < .001, 95% CI [−0.26, −0.10], and 3) agency could only infer change indirectly by noting the pattern of responses of
alone, B = −0.04, p = .044, 95% CI [−0.08, −0.001]. Relative to the participants in different experimental conditions. To observe change in
control group, undermining relational ties (Dummy 2) seemed to fusion directly, in Study 4 we sampled university students. This group
weaken willingness to fight and die for the group via 1) identity fusion could be readily incentivized to participate in a multi-wave study.
and agency serially, B = −0.02, p = .018, 95% CI [−0.03, −0.003], 2) Note that the Study was conducted during a period of extreme po-
identity fusion alone, B = −0.12, p = .002, 95% CI [−0.19, −0.04], litical volatility: between October 1st, 2017, when the president of the
and agency alone, B = −0.04, p = .038, 95% CI [−0.08, −0.002]. Catalan Government organized illegal voting for the independency, and
Parallel analyses in which we sequentially substituted each of the December 21th, after the arrest of the primary politicians who rebelled
three measures of identification for the measure of fusion revealed against the central government. Capitalizing on this explosive political
significant indirect effects of D1 (control vs. collective) on willingness issue (threats of secession by Spain's most prosperous province,
to fight/die for Mael and Asforth's and Ellemers et al.'s scales, but not Catalonia), we measured how many hours participants volunteered to
for the Postmes et al.'s item (only the indirect effect of agency was work with a team that was attempting to reduce conflict among
significant). The indirect effects of D2 (control vs. relational ties) via Spaniards. In addition to this pro-group behavior, we included the same
identification alone and via identification and agency were not sig- measure of pro-group behavior used in our earlier studies (willingness
nificant for any scale. Only the indirect effect via agency was significant to fight and die for the group).

7
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

A final feature of Study 4 was that we simplified the design by and waves was not significant either, Wilks' Lambda = 1.00, F
dropping the degraded collective ties condition that we had included in (2,354) = 0.81, p = .922, ηp2 = 0.00. That is, our manipulation had no
Studies 1–3. Given that we observed the effects of degrading collective impact on reductions in identification from wave 1 to wave 2.
ties in three previous experiments, it would have been redundant to
demonstrate this effect yet again. 6.2.3. Feelings of agency
The regression on agency yielded a significant effect of the two-way
6.1. Method interaction, B = −0.32, p < .001, 95% CI [−0.45, −0.18]. This in-
teraction was due to the tendency for identity fusion at wave 1 to
6.1.1. Participants predict agency at wave 2 in the control condition, B = 0.34, p < .001,
Spanish Psychology upper-class undergraduates from UNED com- 95% CI [0.27, 0.40], but not in the degraded relational ties condition,
pleted a two-wave Study online (N = 398 in wave 1 and N = 357 in B = 0.02, p = .754, 95% CI [−0.10, 0.14]. In addition, the main effect
wave 2; 72.8% female, Mage = 33.06, SD = 10.48 in wave 2) for course of condition was significant, B = −0.42, p < .001, 95% CI [−0.60,
credit. In the first wave participants completed the trait versions of the −0.24], such that agency was lower in the degraded relational ties
measures of fusion and identification (Ellemers et al., 1999; Mael & condition than in the control condition. Finally, a significant main ef-
Ashforth, 1992; Postmes et al., 2013), αs > 0.83. fect of fusion at wave 1 emerged, B = 0.34, p < .001, 95% CI [0.27,
Five weeks later we randomly assigned participants to either the 0.40], indicating that fusion at wave 1 predicted agency at wave 2.
degraded relational ties or control conditions used in Studies 1–3. After
the manipulations, participants completed the manipulation check, 6.2.4. Fight/die for the group
α = 0.91. They then responded to the state version of the fusion and The regression on willingness to fight and die yielded a significant
identification scales used in the previous experiments, αs > 0.84 (see two-way interaction, B = −0.28, p < .001, 95% CI [−0.39, −0.19].
Table 1 for the correlations between these measures at wave 1 and wave Identity fusion at wave 1 predicted willingness to fight and die at wave
2), the same measure of agency that in Study 3, α = 0.80 and Swann Jr. 2 in the control condition, B = 0.31, p < .001, 95% CI [0.25, 0.37],
et al.'s (2009) 7-item measure of willingness to fight/die for the group, but not in the degraded relational ties condition, B = 0.03, p = .512,
α = 0.80. 95% CI [−0.06, 0.13]. In addition, the main effect of condition was
Finally, participants learned that our research team was developing significant, B = −0.35, p < .001, 95% CI [−0.50, −0.19], such that
a program designed to reduce conflict and improve relations between willingness to fight and die was lower in the degraded relational ties
fellow Spaniards. Noting that as upper classmen their experience would condition than in the control condition. Finally, a significant main ef-
be potentially beneficial to the group, we asked to them if they would fect of fusion at wave 1 emerged, B = 0.31, p < .001, 95% CI [0.25,
like to collaborate with our team. They were informed that their task 0.37], in that fusion at wave 1 predicted willingness to fight and die at
would consist on evaluating proposals to resolve the conflicts between wave 2.
the Spaniards and make suggestions for improvement. Participants then
indicated how many hours they committed to work on this task. 6.2.5. Commitment to help the group
The regression on the number of hours participants committed to
6.2. Results and discussion help the group resolve conflicts yielded a significant two-way interac-
tion, B = −0.30, p = .002, 95% CI [−0.49, −0.11]. Identity fusion at
To examine the change in fusion and identification we performed wave 1 predicted commitment to help the group at wave 2 in the
repeated measures analysis. To examine the change on the outcome control condition, B = 0.24, p < .001, 95% CI [0.11, 0.36], but not in
variables (agency, fight and die, and commitment to help the group) as the degraded relational ties condition, B = −0.06, p = .378, 95% CI
a function of condition and prior levels of alignment with the group, we [−0.21, 0.08]. In addition, the main effect of condition was significant,
conducted a single analysis. In this latter analysis, condition (0 control, B = −0.42, p < .001, 95% CI [−0.67, −0.17], such that commitment
1 degraded relational ties), fusion (or identification) in wave 1 (cen- to help the group was lower in the degraded relational ties condition
tered) and the 2-way interaction were entered as predictors. For a priori than in the control condition. Finally, a significant main effect of fusion
reasons, commitment to help the group was treated as an ordinal at wave 1 emerged, B = 0.24, p < .001, 95% CI [0.11, 0.36], in-
variable. The non-significant effects are specified in the supplemental dicating that fusion at wave 1 predicted commitment to help the group
material (Table 8). at wave 2.

6.2.1. Change in fusion 6.2.6. Analyses of identification

We conducted a repeated measure ANOVA to test whether the ex- Parallel analyses in which we sequentially substituted each of the
perimental manipulation modified identity fusion between waves 1 and three measures of identification for the measure of fusion at wave 1
2. The interaction between the manipulation and waves was significant, showed no significant effects of the interaction between condition and
F(1,355) = 23.08, p < .001, ηp2 = 0.06. Fusion significantly decreased identification. Only the main effects of condition and identification
between waves 1 and 2 in the degraded relational ties condition, were significant (see Tables 9–11 of supplemental material).
M = 2.35, SD = 1.28 vs. M = 1.50, SD = 1.05, F(1,355) = 78.58,
p < .001, ηp2 = 0.18. In the control condition the difference between 6.2.7. Indirect effects
fusion at wave 1 and wave 2, M = 2.46, SD = 1.16 vs. M = 2.28, To determine if the interactive effect of our manipulation and fusion
SD = 1.25, was only marginal, F(1,355) = 3.29, p = .070, ηp2 = 0.01. at wave 1 on the outcome measures (willingness to fight/die for the
group and commitment to help the group) would be serially mediated
6.2.2. Change in identification by identity fusion at wave 2 and agency, we conducted a mediation test
We conducted a repeated measures MANOVA on the identification using Lavaan. As in Study 3, we based this model on previous evidence
scales to check whether the degraded relational ties manipulation that fusion influences pro-group activity via personal agency (Gómez
compromised identification between waves 1 and wave 2. The effect of et al., 2011; Swann Jr., Gómez, Huici, et al., 2010). We treated the
waves was significant, Wilks' Lambda = 0.72, F(1,355) = 140.34, interaction between fusion at wave 1 (centered) and the experimental
p < .001, ηp2 = 0.28, such that identification decreased from wave 1 manipulation as the predictor, identity fusion (centered) at wave 2 as
to wave 2. However, the interaction between condition and waves was the first potential mediator, and agency (centered) as the second po-
not significant, Wilks' Lambda = 1.00, F(1,355) = 1.10, p = .295, tential mediator. Fusion at wave 1 and the experimental manipulation
ηp2 = 0.003. The 3-way interaction between type of scale, manipulation (0 control, 1 degraded relational ties) served as covariates.

8
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

Fig. 4. Indirect effect of compromising collective and relational ties on willingness to fight/die and commitment to help the group via identity fusion and agency
(Study 4).

Commitment to help the group was treated as a continuous variable however: 1) via identity fusion and agency serially, B = 0.07, p = .043,
because the ordered factors were not supported for mediational models 95% CI [0.002, 0.14], 2) identity fusion alone, B = 0.11, p = .095, 95%
in Lavaan. Thus, results regarding commitment to help the group must CI [−0.02, 0.23], and 3) agency alone, B = −0.03, p = .698, 95% CI
be interpreted cautiously. [−0.20, 0.13].
As shown in Fig. 4, this model had three distinct indirect effects for
each outcome variable: 1) the effect via fusion and agency serially 6.2.8. Sensitivity power analysis
(interaction > fusion > agency > outcome), 2) the effect via fusion We performed two sensitivity power analyses using GPower (Faul
alone (interaction > fusion > outcome), and 3) the effect via agency et al., 2007) considering an alpha significance criterion of 0.05 and a
alone (interaction > agency > outcome). Results of the analysis of sample size of 357 participants (in wave 2). The first analysis referred to
willingness to fight and die for the group indicated that all the indirect changes in fusion and identification, that were tested with repeated
effects via identity fusion at wave 2 and agency were significant while measures ANOVAs. Given that we had two groups and two measure-
controlling for fusion at wave 1 and the experimental manipulation. ments, the sensitivity analysis yielded a minimum effect size of f
Relative to the control condition, undermining relational ties seemed to (V) = 0.191 with 80% power. The second analysis was related to the
reduce willingness to fight and die for the group via 1) identity fusion regressions on agency, fight and die and commitment to help the group.
and agency serially, B = −0.07, p < .001, 95% CI [−0.10, −0.04], 2) Given that we had three predictors, the sensitivity analysis yielded a
identity fusion alone, B = −0.04, p = .017, 95% CI [−0.08, −0.01], minimum effect size of f2 = 0.031 with 80% power.
and 3) agency alone, B = −0.09, p = .007, 95% CI [−0.15, −0.02].
All the indirect effects were significant in the control condition: 1) via 6.2.9. Summary
identity fusion and agency serially, B = 0.08, p < .001, 95% CI [0.05, The results of Study 4 replicate Study 3 using a half-longitudinal
0.11], 2) identity fusion alone, B = 0.05, p = .012, 95% CI [0.01, 0.09], design in which the measures of alignment with the group were col-
and 3) agency alone, B = 0.08, p = .003, 95% CI [0.03, 0.13]. How- lected more than a month prior to the experiment. Specifically, com-
ever, indirect effects were smaller or not significant in the degraded promising relational ties weakened fusion (but not identification) and
relational ties condition: 1) via identity fusion and agency serially, weakened fusion apparently mediated diminutions in willingness to
B = 0.02, p = .027, 95% CI [0.002, 0.03], 2) identity fusion alone, fight/die for the group as well as donations of one's time to the group.
B = 0.01, p = .080, 95% CI [−0.001, 0.02], and 3) agency alone, We also replicated the serial indirect effect from Study 3. That is, de-
B = −0.01, p = .697, 95% CI [−0.05, 0.03]. gradations in relational ties seemed to weaken fusion and, in turn,
Fig. 4 also shows that the analysis of commitment to help the group group-related agency, which in turn might mediate the effect of di-
yielded similar results. All indirect effects through identity fusion at minutions in fusion on willingness to fight/die for the group and do-
wave 2 and agency were significant while controlling for fusion at wave nations of one's time to the group.
1 and the experimental manipulation. Relative to the control condition,
undermining relational ties seemed to reduce volunteering to resolve 7. Study 5: does compromising relational ties and feelings of
group's conflicts via 1) identity fusion and agency serially, B = −0.28, agency reduce commitment to pro-group action?
p = .001, 95% CI [−0.45, −0.11], 2) identity fusion alone,
B = −0.42, p = .028, 95% CI [−0.80, −0.04], and 3) agency alone, The results of the foregoing studies offer converging evidence for
B = −0.37, p = .018, 95% CI [−0.68, −0.06]. All the indirect effects our hypotheses. Nevertheless, we included a fifth study to strengthen
were significant in the control condition: 1) via identity fusion and two aspects of our findings. First, to buttress our contention that group
agency serially, B = 0.35, p < .001, 95% CI [0.16, 0.55], 2) identity related agency mediated the effects of our manipulation in Studies 3
fusion alone, B = 0.53, p = .022, 95% CI [0.08, 0.98], and 3) agency and 4, in Study 5 we manipulated agency. Second, studies 1–4 used
alone, B = 0.34, p = .010, 95% CI [0.08, 0.60]. Indirect effects were three popular measures of identification that were developed in the
smaller or not significant in the degraded relational ties condition, tradition of social identity theory's emphasis on collective ties. In recent

9
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

years one set of authors, Leach et al. (2008), have adopted a broader 7.2. Results and discussion
conceptualization of identification stemming from a Lewian con-
ceptualization of identification as a bond between group members Table 2 shows the correlation coefficients for all the outcome vari-
(Lewin, 1948). Although most of the items in Leach's scale focus on ables.
collective ties, it also includes a relational ties component dubbed We performed two regressions using the Lavaan package in R soft-
“ingroup solidarity” that is closely related to identity fusion (Bortolini, ware. In the first regression, the predictor was relational ties (1
Newson, Natividade, Vázquez, & Gómez, 2018). To determine if de- Relational ties, 0 Control) and the outcome measures were fusion and
grading relational ties would diminish scores on the ingroup solidarity identification (the agency manipulation occurred after the measures of
component of the Leach scale, we included this scale in Study 5. fusion and identification and was thus not included as a predictor). In
As in previous studies, we expect the relational ties manipulation to the second regression, the predictors were Relational Ties (1 Relational
influence fusion, willingness to fight and die and commitment to help ties, 0 Control), Agency (1 Degrading agency, 0 Control) and the two-
the group but not the overall measures of identification. We also ex- way interaction; the outcome measures were willingness to fight and
pected, however, that degrading relational ties would diminish scores die and commitment to help the group. Commitment to help the group
on the ingroup solidarity component of the Leach scale. Finally, as in was coded as an ordered factor before conducting the analyses, as in
Study 4, we expected that degrading agency would reduce scores on the Experiment 4. The non-significant effects are specified in the supple-
outcome measures, willingness to fight and die and commitment to help mental material (Table 13).
the group.
7.2.1. Changes in fusion
The regression on fusion yielded a significant effect of the relational
7.1. Method
ties manipulation, B = −0.44, p = .001, 95% CI [−0.69, −0.19], such
that fusion was lower in the degraded relational ties condition,
7.1.1. Participants
M = 1.80, SD = 1.17, than in the control condition, M = 2.24,
As in the previous studies, we utilized the snowball technique to
SD = 1.36.
recruit 410 Spaniards (63.7% women, Mage = 34.73, SD = 12.63). Five
participants did not follow the instructions for the manipulation or the
7.2.2. Changes in identification
control conditions and were dropped from the analyses (including these
A regression revealed that degrading relational ties had no impact
participants in the analyses did not change our conclusions). The final
on overall identification, B = −0.17, p = .135. Follow-up analyses of
sample included 405 participants (64% women, Mage = 34.82,
the individual subscales of the Leach scale yielded only one significant
SD = 12.62). Participants completed the questionnaire online.
effect for ingroup solidarity, B = −0.36, p = .011, 95% CI [−0.64,
−0.08], such that solidarity was lower in the degraded relational ties
7.1.2. Procedure condition, M = 2.46, SD = 1.44, than in the control condition,
The experimenter assigned participants to either the degraded re- M = 2.82, SD = 1.41. No significant effects of the relational ties ma-
lational ties or control conditions used in Studies 1–4. After the ma- nipulation emerged for the remainder subscales of the Leach scale,
nipulation, participants completed the manipulation check of relational Bs = from −0.12 to −0.10, ps = from 0.042 to 0.537.
ties, α = 0.84. After completing the manipulation check, participants
responded to the state versions of the fusion scale, αs > 0.88, and to the 7.2.3. Fight/die for the group
state version of the Leach et al. (2008) identification scale, αs > 0.91. The regression on fight and die yielded significant effects of the
After that, participants were randomly assigned to either a degraded relational ties manipulation, B = −0.40, p = .029, 95% CI [−0.76,
agency condition or a no information control condition. In both con- −0.04], and the agency manipulation, B = −0.31, p = .030, 95% CI
ditions, participants learned that they would be taking a test of visual [−0.60, −0.03]. Participants in the degraded relational ties condition,
perception. They were instructed to watch the screen and capture as M = 1.02, SD = 0.86, were less willing to fight and die for the group
many details as possible during 4 s. We presented 10 screens where than those in the control condition, M = 1.27, SD = 1.30. Additionally,
names of several countries or persons appeared in different color, size participants in the degraded agency condition, M = 1.06, SD = 0.96,
and/or order. After each screen participants had to answer a question were less willing to fight and die for the group than those in the control
regarding the information just presented (e.g., “How many times did condition, M = 1.24, SD = 1.26. The effect of the interaction was not
the word Spain appear in black in the previous list?”). Once finished, all significant.
participants read the following information: “You have just completed
the Group Self-Control and Personal Test (GSPT), designed to estimate 7.2.4. Commitment to help the group
the degree to which people control their responses”. They further The regression on commitment to help the group yielded a sig-
learned that several studies have discovered that personal control cor- nificant effect of the relational ties manipulation, B = −0.42, p = .015,
relates with the control that people exercise over the groups to which 95% CI [−0.77, −0.08], and a marginal effect of the Agency manip-
they belong. Participants in the degraded agency condition received ulation, B = −0.30, p = .053, 95% CI [−0.61, 0.00]. Participants in
feedback telling them that they scored low on the test: only 20 points the degraded relational ties condition, M = 1.00, SD = 2.14, were less
out of 100. They learned that such results indicate that they normally committed to the group than those in the control condition, M = 1.86,
have little control over the events of their life, that low control has SD = 2.98. Additionally, participants in the degrading agency condi-
advantages and disadvantages, and that they would receive more in- tion, M = 1.12, SD = 2.27, were less committed to the group than those
formation at the end of the study. Participants in the no information in the control condition, M = 1.78, SD = 2.95. The effect of the inter-
condition learned that they would receive their scores at the end of the action was not significant.
study.
After the manipulation, participants responded to the Swann Jr. 7.2.5. Indirect effects
et al.'s (2009) 7-item measure of willingness to fight/die for the group, To determine if degraded relational ties influenced our outcome
α = 0.86, and to the same outcome measure used in Study 4 to de- measures (willingness to fight/die for the group and commitment to
termine how many hours they would commit to evaluating proposals to help the group) through identity fusion depending on group-related
resolve the conflicts between the Spaniards. Finally, as a manipulation agency, we tested the mediation model depicted in Fig. 5 using Lavaan.
check, participants completed the same measure of group-related The relational ties manipulation was entered as the predictor, identity
agency included in Studies 3–4, α = 0.80. fusion (centered) as the potential mediator, and the agency

10
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

Fig. 5. Indirect effect of compromising relational ties on willingness to fight/die and number of hours via identity fusion depending on the agency manipulation
(Study 5).

manipulation as the moderator of the relationship between fusion and 7.2.7. Summary
the outcome variables. As in Experiment 4, commitment to help the Experiment 5 replicates and extends the results of previous studies
group was considered as a continuous variable because the ordered in that compromising relational ties weakened fusion but not overall
factors were not supported for mediational models in Lavaan. Thus, identification scores. Furthermore, providing participants with in-
results regarding commitment to help the group must be interpreted formation indicating that their group-related agency was low reduced
cautiously. fusion which, in turn, might diminish their willingness to fight and die
As shown in Fig. 5, all the indirect effects via identity fusion were as well as their efforts to help the group. Interestingly, these indirect
significant. However, the effects were higher in the control as compared effects were smaller when participants were led to perceive themselves
to the degraded agency condition. Degrading relational ties seemed to as low in agency as compared to a no information condition. However,
reduce willingness to fight and die for the group through identity fusion the indirect effects via solidarity were not significant, except for the
in both the control condition, B = −0.22, p = .001, 95% CI [−0.35, indirect effect on willingness to fight and die when no information
−0.08], and in the degraded agency condition, B = −0.10, p = .008, about agency was provided.
95% CI [−0.18, −0.03]. Undermining relational ties seemed to reduce One potential limitation of the previous studies could be the nature
commitment to help the group as compared to the control condition via of our manipulations. Regarding the manipulation designed to degrade
identity fusion both in control condition, B = −0.47, p = .001, 95% CI relational ties, one could argue that our manipulation did not capture
[−0.77, −0.18], and in the degraded agency condition, B = −0.22, the idea of seeing ingroup members “as if they were their own families”.
p = .010, 95% CI [−0.39, −0.05]. Regarding the manipulation designed to degrade collective ties, one
Given that solidarity presented a similar pattern of results as fusion, could argue that we may have manipulated negative perceptions or
we tested the same mediational analysis considering solidarity as the feelings toward some sub-group such as the government or other in-
potential mediator of the effect of ties on willingness to fight and die stitutions, rather than toward the group as a whole. To address these
and commitment to help the group. Undermining relational ties seemed concerns, we conducted an additional study.
to reduce willingness to fight and die for the group as compared to the
control condition via solidarity only in the no information condition,
8. Study 6: does the impact of degrading relational and collective
B = −0.07, p = .039, 95% CI [−0.13, −0.004], but not in the de-
ties generalize to different manipulations and to a new group
grading agency condition, B = −0.04, p = .124, 95% CI [−0.09, 0.01].
category?
The indirect effects on commitment to help the group via solidarity was
only marginal in the no information condition, B = −0.12, p = .063,
The results of the Studies 1–5 offer converging evidence for our
95% CI [−0.25, 0.01], and not significant in the degraded agency
predictions. Nevertheless, we added a sixth study to extend four aspects
condition, B = −0.05, p = .353, 95% CI [−0.14, 0.05].
of our results. First, we sought to generalize our previous findings to
another group category: females. Second, we developed new strategies
for degrading relational and collective ties to determine if our findings
7.2.6. Sensitivity power analysis
were not an artifact of the specific manipulations we employed. Third,
We performed two sensitivity power analyses using GPower (Faul
we attempted to degrade relational ties by asking participants to report
et al., 2007) considering an alpha significance criterion of 0.05 and a
specific examples involving close friends or family members. Fourth, we
sample size of 405 participants. The first analysis referred to the re-
attempted to degrade collective ties by asking participants to report
gressions on fusion and Leach et al.'s subscales, where we only had one
examples of negative actions by other ingroup members with whom
predictor (relational ties manipulation). The minimum effect to be de-
they were unacquainted that made them feel very far from the group in
tected with 80% power was f2 = 0.019. The second sensitivity analysis
general and that made them question their relationship with the group.
was related to fight and die and commitment to help the group con-
sidering three predictors (relational ties manipulation, agency manip-
ulation and its interaction). The analysis revealed that we could detect a 8.1. Method
minimum effect size of f2 = 0.027 with 80% power.
8.1.1. Participants
As in studies 1–5, we utilized the snowball technique to recruit 276

11
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

females (91.7% Spaniards, Mage = 38.97 years, SD = 11.55). M = 2.22, SD = 1.58, and degraded relational ties conditions,
Participants completed the questionnaire online. M = 2.47, SD = 1.39, expressed less fusion than participants in the
control condition, M = 3.33, SD = 1.27.
8.1.2. Procedure
As in the Studies 1–3, the experimenter assigned participants to one 8.2.2. Identification
of three conditions: degraded relational ties, degraded collective ties or The regression on the Leach et al.'s scale only yielded a significant
control. In the relational ties condition, participants were asked to report effect of D1 (collective ties vs. control), B = −0.30, p = .029, 95% CI
specific examples in which they felt that they had suffered from de- [−0.56, −0.03]. The effect of D2 (relational ties vs. control) was not
gradations in their relational ties to other females who were “close significant, B = −0.20, p = .132, 95% CI [−0.46, 0.06]. Only the so-
friends or family members” (e.g., “One of my sisters, when I was lidarity subscale of Leach's identification measure yielded a significant
pregnant, cruelly questioned my reasons why I had decided to become a effect of D1, B = −0.56, p = .008, 95% CI [−0.97, −0.15]. The effect
mother. She did it knowing that it would be an issue that would be of D2 was marginal, B = −0.38, p = .055, 95% CI [−0.77, 0.01].
painful for me, since I was very insecure economically at that time and I Participants in the degraded collective ties, M = 3.76, SD = 1.35, and
was afraid of making ends meet”). In the collective ties condition, par- degraded relational ties conditions, M = 3.87, SD = 1.43, expressed
ticipants were asked to report examples of negative actions by other less solidarity than participants in the control condition, M = 4.29,
females with whom they were unacquainted that made them feel very SD = 1.15. None of the effects of D1 or D2 were significant for any of
distant from other females in general and that made them question their the other subscales, ps > 0.10.
relationship with the group (e.g., “A university professor took ad-
vantage of her power over a student to obtain sexual favors”). As in 8.2.3. Agency
previous studies in this manuscript, in the control condition participants The regression on agency yielded significant effects of both D1,
described their two last trips to their workplace or university. B = −0.66, p < .001, 95% CI [−0.95, −0.37], and D2, B = −0.64,
As checks on the manipulations, participants responded to two, 3- p < .001, 95% CI [−0.93, −0.35]. Participants in the degraded col-
item scales. One scale measured their relational ties to other females lective ties, M = 0.85, SD = 0.95, and degraded relational ties condi-
(e.g.,“Right now, I feel strong ties to all other individual females”) tions, M = 087, SD = 0.89, expressed weaker feelings of agency for the
α = 0.901; the other scale measured collective ties to other females: group than participants in the control condition, M = 1.51, SD = 1.17.
(e.g. “Right now, I feel strong ties to females as a group”) α = 0.942.
After completing the manipulation check, participants responded to 8.2.4. Fight/die for the group
two measures of alignment with females, including the state version of The regression on willingness to fight and die yielded significant
the fusion scale, α = 0.91, and the state version of Leach et al.'s (2008) effects of both D1, B = −0.74, p < .001, 95% CI [−1.02, −0.46], and
identification scale, αs > 0.85. They then completed the 5-item mea- D2, B = −0.58, p < .001, 95% CI [−0.85, −0.31]. Participants in the
sure of agency with reference to females (e.g., “I am able to control degraded collective ties, M = 1.98, SD = 0.95, and degraded relational
what females do,” α = 0.86, and Swann Jr. et al.'s (2009) 7-item ties conditions, M = 2.19, SD = 0.90, expressed less willingness to fight
measure of willingness to fight/die for females (e.g. “I would sacrifice and die for the group than participants in the control condition,
my life if it saved another female's life”), α = 0.74. Finally, we included M = 2.75, SD = 0.87.
the same measure of commitment to help the group used in Studies 4–5,
but in this case the goal was to reduce gender violence. Participants 8.2.5. Commitment to help the group
committed themselves to work on this task for a time duration of their The regression on commitment to help the group yielded significant
choosing. effects of both D1, B = −0.65, p = .001, 95% CI [−1.04, −0.26], and
D2, B = −0.58, p = .002, 95% CI [−0.94, −0.21]. Participants in the
8.2. Results and discussion degraded collective ties, M = 1.06, SD = 1.62, and degraded relational
ties conditions, M = 1.26, SD = 1.97, expressed less commitment to
Table 3 shows the correlations between the outcome measures. help the group than participants in the control condition, M = 5.81,
Identity fusion with women correlated significantly with agency, will- SD = 12.49.
ingness to fight and die for women and commitment to help the group.
The components of identification with the group were significantly 8.2.6. Indirect effects
related to some of these outcome measures but these relationships were To determine if the effect of our manipulations on willingness to
inconsistent across subscales. fight/die for the group and commitment to help the group were serially
As in previous experiments, below we report significant effects for mediated by identity fusion and agency, we tested the mediation model
the relation between our predictors and the outcome variables. We depicted in Fig. 6 using the Lavaan package in R software. D1 and D2
report non-significant effects in the supplemental material (Table 15). were entered as predictors, identity fusion was the first potential
As the condition had three levels, we created the same two dummy mediator, agency was the second potential mediator and willingness to
codes with the control condition as the comparison group, as we did in fight/die and commitment to help the group were the outcome vari-
Studies 1–3. The first dummy code (D1) compared the control condition ables. Nationality was entered as a covariate. This model had three
with the degraded collective ties condition, whereas the second (D2) distinct indirect effects for each dummy code: 1) the effect via fusion
compared the control condition with the degraded relational ties con- and agency serially (condition > fusion > agency > fight/die/com-
dition. As 8.3% of participants were not Spaniards, most belonging to mitment), 2) the effect via fusion alone (condition > fusion > fight/
different South-American countries, and there might be cultural dif- die/commitment), and 3) the effect via agency alone (condition >
ferences that could affect the results, nationality was included as a agency > fight/die/commitment). Results indicated that all these in-
covariate. We conducted the analyses with and without this covariate direct effects were significant for fight/die for the group and commit-
and the results were the same. ment to help the group. As in Experiment 5, commitment to help the
group was considered as a continuous variable because the ordered
8.2.1. Fusion factors were not supported for mediational models in Lavaan. Thus,
The fusion regression yielded significant effects of both D1 (col- results regarding commitment to help the group must be interpreted
lective ties vs. control), B = −1.08, p < .001, 95% CI [−1.50, cautiously.
−0.65], and D2 (relational ties vs. control), B = −0.89, p < .001, Relative to the control group, undermining collective ties (Dummy
95% CI [−1.31, −0.47]. Participants in the degraded collective ties, 1) seemed to weaken willingness to fight and die for the group via 1)

12
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

Fig. 6. Indirect effect of compromising collective and relational ties on willingness to fight/die and commitment to help the group via identity fusion and agency
(Study 6).

identity fusion and agency serially, B = −0.04, p = .010, 95% CI to help the group. In contrast, these serial indirect effects via identifi-
[−0.08, −0.01], 2) identity fusion alone, B = −0.27, p < .001, 95% cation subscales were not significant in any case. Thus, our findings
CI [−0.40, −0.14], and 3) agency alone, B = −0.07, p = .031, 95% CI replicated the earlier studies in this report despite our having modified
[−0.14, −0.01]. Relative to the control group, undermining relational the manipulations of degrading collective or relational ties and focusing
ties (Dummy 2) seemed to weaken willingness to fight and die for the on a different group.
group via 1) identity fusion and agency serially, B = −0.04, p = .013,
95% CI [−0.06, −0.01], 2) identity fusion alone, B = −0.22,
9. General discussion
p < .001, 95% CI [−0.34, −0.10], and 3) agency alone, B = −0.08,
p = .021, 95% CI [−0.14, −0.01].
Our findings provide a fresh perspective on how to most effectively
Relative to the control group, undermining collective ties (Dummy
diminish alignment with one's group. A series of experiments indicated
1) seemed to weaken commitment to help the group via 1) identity
that manipulations designed to degrade relational ties weakened
fusion and agency serially, B = −0.49, p = .004, 95% CI [−0.83,
identity fusion but not overall group identification. In contrast, ma-
−0.16], 2) identity fusion alone, B = −0.90, p = .017, 95% CI
nipulations designed to compromise collective ties weakened both fu-
[−1.65, −0.16], and 3) agency alone, B = −0.80, p = .020, 95% CI
sion and identification. Moreover, a study using a half-longitudinal
[−1.47, −0.13]. Relative to the control group, undermining relational
design showed that compromising relational ties decreased identity
ties (Dummy 2) seemed to weaken commitment to help the group via 1)
fusion when we compared participants' scores before and after the
identity fusion and agency serially, B = −0.41, p = .006, 95% CI
manipulation. Furthermore, compromising both relational and collec-
[−0.70, −0.12], 2) identity fusion alone, B = −0.75, p = .022, 95%
tive ties weakened group-related agency, willingness to fight and die for
CI [−1.39, −0.11], and agency alone, B = −0.85, p = .012, 95% CI
the group, and commitment to work on behalf of the group. Our results
[−1.51, −0.19].
also revealed that degrading group-related agency diminished will-
We conducted parallel analyses in which we sequentially sub-
ingness to fight and die for the group and efforts to help the group. A
stituted each of the five subscales of identification for the measure of
final study supported the generality of the foregoing effects by showing
fusion. The serial indirect effects (via identification and agency) were
that they generalized to different manipulations and a different group
not significant for any subscale, nor for fight/die nor commitment to
(gender as compared to country).
help the group. Although some significant indirect effects of D1 and D2
A key finding here is that collective ties, relational ties and group-
via identification or agency alone emerged, the pattern was inconsistent
related personal agency were intimately interrelated among strongly
across outcome measures (see Tables 16 and 17 of supplemental ma-
fused individuals. For example, identity fusion apparently mediated the
terial).
impact of compromising collective or relational ties on reductions in
group-related agency. Likewise, group-related agency seemed to med-
8.2.7. Sensitivity power analysis iate the effect of identity fusion on willingness to fight and die for the
We performed a sensitivity power analyses using GPower (Faul group and donations of one's time to the group. Moreover, a serial-
et al., 2007) considering an alpha significance criterion of 0.05, a mediational analysis suggested that identity fusion and group-related
sample size of 276 participants and three predictors (two dummy agency might account for the effects of collective ties (Study 3) and
variables and nationality). The minimum effect to be detected with 80% relational ties (Studies 3–4) on willingness to fight and die as well as
power was f2 = 0.04. commitment of one's time to support the group. Furthermore, experi-
mentally reducing group related agency diminished subsequent will-
8.2.8. Summary ingness to fight and die for the group and amount of time participants
The results of Experiment 6 replicate and extend the results of pledged to support the group. Finally, these effects emerged whether
previous studies in that compromising collective and relational ties we focused on gender or country as the group under scrutiny.
toward one's gender group weakened fusion. Both manipulations also Our evidence suggesting that our measures of identification may
reduced the solidarity subscale of Leach's measure but only the col- mediate the effect of collective ties, but not relational ties, on pro-group
lective ties manipulation significantly reduced the other subscales and behavior is consistent with social identity theory's (Tajfel & Turner,
total scores. Furthermore, reduced fusion seemed to undermine agency 1979) assumption that collective ties underlie and motivate pro-group
and, in turn, participants' willingness to fight and die as well as efforts behavior. It is noteworthy that this result emerged for three distinct

13
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

measures of identification. Our findings therefore address the concern 9.2. Theoretical and practical implications
that previous research on identity fusion has relied almost exclusively
on Mael and Ashforth's (1992) measure of identification as a rival to Although many contemporary social identity theorists have ven-
measures of identity fusion and has overlooked more recently devel- tured beyond the original theory by acknowledging the importance of
oped measures featured in the research we present here, such as the one relational ties in motivating group behavior, our evidence indicates that
developed by Ellemers et al. (1999), Postmes et al. (2013), and Leach four of the most commonly used measures of identification were rela-
et al. (2008). tively insensitive to manipulations of relational ties. In the future, re-
Our findings also indicate that fusion is qualitatively different from searchers who are interested in measures of group alignment that
identification. Witness, for example, that in our studies, identity fusion capture relational as well as collective ties should either use fusion
was related to collective ties, relational ties and group-related agency measures or develop measures of identification that are more balanced
while identification was primarily related to collective ties. We should in assessing collective and relational ties.
add two qualifiers to this generalization. First, the relational ties ma- To be sure, Leach et al. (2008) have recently developed a measure of
nipulation did produce several marginal effects on the Ellemers' scale, identification that is designed to capture relational ties. Although the
perhaps because 2 of its 10 items allude to relational ties (see Supp. Leach scale has the virtue of being broader than earlier measures of
Materials). Second, in the last two studies: degrading relational ties did identification, it should not be regarded as a substitute for our measure
reduce scores on one of 5 components of Leach et al.,’s measure (but not of identity fusion. At least four considerations support this conclusion.
the total score nor the other 4 components of identification in Leach's First, although one component of the Leach scale seems closely related
scale). This is significant because the Leach scale was specifically de- to fusion, the fusion scale has elements that Leach's scale does not. For
signed to be representative of previous research on identification. example, in addition to featuring items that assess connection to the
Apparently, when researchers deliberately attempt to create re- group (e.g., “I have a deep emotional bond to my group”), the fusion
presentative measures of identification, they accord relational ties a scale also features items that assess reciprocal strength (e.g., “I make
relatively minor role. my country strong” and “I am strong because of my country”). Second,
Our evidence that measures of fusion tap collective and relational the inclusion of reciprocal strength items in the fusion scale is im-
ties equally well whereas measures of identification favor collective ties portant, as it is likely that these items help explain why fusion is a
may help explain why past researchers have consistently found that better predictor of endorsement of extreme pro-group behavior than
fusion measures predict endorsement of extreme pro-group behavior identification, even when the Leach scale is used to measure identifi-
with greater fidelity than measures of identification (for reviews, see cation. For example, in our studies, fusion was consistently more
Fredman et al., 2015; Gómez & Vázquez, 2015; Swann Jr & Buhrmester, strongly related to endorsement of fighting and dying for the group
2015). That is, the predictive advantage of fusion measures may stem (r's = 0.44 & 0.51 for Studies 5 and 6 respectively) than the leach total
from the fact that fusion is uniquely linked to collective ties, relational score (r's = 0.22 & 0.21 for Studies 5 and 6 respectively) or each of the
ties and group-related agency. five Leach subscales (r's < 0.23 and 0.28 for Studies 5 and 6 respec-
tively), with ps < 0.001 for all comparisons between correlations in-
volving the fusion versus the leach scale. Third, even though elements
9.1. Limitations of the Leach scale were associated with fusion, overall the scale is still
primarily a measure of collective ties. Note, for example, that Leach
One limitation of the experiments reported here is that we were able total scores were lowered when collective ties were compromised but
to diminish state, but not trait, fusion. This finding likely reflects the not when relational ties were compromised (in contrast, fusion scores
fact that people draw support for their cognitive structures from both were lowered when either relational or collective ties were compro-
the immediate social environment and chronically activated re- mised). Fourth, the relative expansiveness of the Leach scale comes at a
presentations of the self and group. To produce more enduring changes cost. Of the commonly used measures of identification, at 18 items the
in fusion and behavior, it will likely be important to develop inter- leach scale is roughly twice as long as its competitors. Our data as well
ventions that are designed to reinforce temporary fusion changes as a decade of previous research suggests that our 7-item fusion scale is
through changes in the social environments that sustain people's iden- a relatively efficient measure that predicts a wide range of important
tities. For example, whereas one-off reminders that one's group suffers outcomes. It is especially attractive for researchers interested in a re-
from discord may lower fusion temporarily, repeated evidence of dis- latively brief measure that assesses an even mixture of relational as well
cord may be required to produce relatively permanent renunciations of as collective ties and predicts extreme pro-group behavior with rela-
the group. The latter possibility is supported by recent evidence that tively high fidelity.
highly salient and sustained challenges to the integrity of the group can Regarding more practical concerns, governments have recently be-
lower scores on trait measures of fusion (Vázquez et al., 2017). come increasingly alarmed by the steady increase in attacks launched
Although temporary changes in state fusion are limited in some by fringe groups. In response, researchers and practitioners have begun
respects, they are nevertheless significant for at least two reasons. First, to work to develop ways of encouraging members to sever their ties to
they are useful in identifying mechanisms that future researchers may such groups. It is important to ask which theoretical orientation pro-
amplify so as to produce changes in trait fusion. Second, despite their vides the optimal framework for such deradicalization programs. For
brevity, short-lived responses can be highly influential. Witness, for example, if one takes social identity theory as the point of departure,
example, that the premise underlying the popularity of measures of one's deradicalization programs will emphasize collective ties, focusing
implicit attitudes today is that people's instantaneous, unthinking re- on issues such as losing faith in the ideologies that the group represents
actions to stimuli uniquely reflect important underlying sentiments (Bjørgo, 2005; Demant, Slootman, Buijs, & Tillie, 2008), or methods
toward the attitude object (Payne, Vuletich, & Lundberg, 2017). It is (Harris, 2015; Horgan, 2009). Nevertheless, there is widespread re-
thus important to understand the causes and consequences of such cognition that deradicalization requires more than challenging the
fleeting reactions. ideology or values of the group. For example, some have noted that
Finally, our evidence of indirect effects must be interpreted cau- even when people have become disillusioned with the group's ideology,
tiously. Although we found consistent results across Studies 2–6 that ties of friendship and loyalty to group members may still inspire loyalty
were also consistent with previous research (Gómez et al., 2011; Swann to the group (Bjørgo, 2005). Our findings support this emphasis on the
Jr., Gómez, Huici, et al., 2010), future studies should explore the pro- significance of relational as well as collective ties to the group and rival
posed paths longitudinally to establish causality unequivocally. groups. In particular, our data indicate that one strategy for defusing
people from a group is to challenge their conviction that the group is a

14
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

harmonious “family” or to increase the salience of rival groups. Such study of organizational disidentification and the NRA. Organization Science, 12,
challenges will erode their assumption that the group will uniquely 393–413. https://doi.org/10.1287/orsc.12.4.393.10638.
Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). G* power 3: A flexible statistical
stand by them, insulate them from betrayal and bathe them in support power analysis program for the social, behavioral, and biomedical sciences. Behavior
for their core values and self-views. Sowing such doubts may cause Research Methods, 39, 175–191. https://doi.org/10.3758/BF03193146.
them to rethink their willingness to tether their fate to the group and Fredman, L. A., Buhrmester, M. D., Gomez, A., Fraser, W. T., Talaifar, S., Brannon, S. M.,
& Swann, W. B., Jr. (2015). Identity fusion, extreme pro-group behavior, and the path
make sacrifices for it. to de-fusion. Social and Personality Psychology Compass, 9, 468–480. doi:https://
doi.org/10.1111/spc3.12193.
Open practices Glasford, D. E., Pratto, F., & Dovidio, J. F. (2008). Intragroup dissonance: Responses to
ingroup violation of personal values. Journal of Experimental Social Psychology, 44,
1057–1064. https://doi.org/10.1016/j.jesp.2007.10.004.
Open data and open materials are available at: https://osf.io/ Gómez, Á., Brooks, M. L., Buhrmester, M. D., Vázquez, A., Jetten, J., & Swann, W. B., Jr.
rwuya/?view_only=6e25d904b89c463992ceb267477cb5a0 (2011). On the nature of identity fusion: Insights into the construct and a new
measure. Journal of Personality and Social Psychology, 100, 918–933. doi:https://
There is sufficient information for an independent researcher to
doi.org/10.1037/a0022642.
reproduce the reported results and methodology. Gómez, Á., & Vázquez, A. (2015). The power of ‘feeling one’ with a group: Identity fusion
and extreme pro-group behaviours. International Journal of Social Psychology, 30,
Acknowledgements 481–511. https://doi.org/10.1080/02134748.2015.1065089.
Griffith, J. (2002). Multilevel analysis of cohesion's relation to stress, well-being, iden-
tification, disintegration, and perceived combat readiness. Military Psychology, 14,
We acknowledge funding support from the Spanish Ministry of 217–239. https://doi.org/10.1207/S15327876MP1403_3.
Science and Innovation RTI2018-093550-B-I00 to Ángel Gómez; the Haggard, P., & Tsakiris, M. (2009). The experience of agency: Feelings, judgments, and
responsibility. Current Directions in Psychological Science, 18, 242–246. https://doi.
European Research Council [grant number 694986] to Michael D. org/10.1111/j.1467-8721.2009.01644.x.
Buhrmester; and Army Research Office (W911nf-17-1-0385) and Harris, K. J. (2015). Leaving ideological social groups behind: A grounded theory of psycho-
National Science Foundation (1528851) to William B. Swann Jr. logical disengagement. Dissertation Edith Cowan University. Retrieved from http://ro.
ecu.edu.au/theses/1587.
Horgan, J. (2009). Individual disengagememt: A psychological analysis. In T. Bjørgo, & J.
Appendix A. Supplementary data Horgan (Eds.). Leaving terrorism behind: Individual and collective disengagement (pp. 17–
29). New York, NY: Routledge.
Hornsey, M., & Jetten, J. (2005). Loyalty without conformity: Tailoring self-perception as
Supplementary data to this article can be found online at https:// a means of balancing belonging and differentiation. Self and Identity, 4, 81–95.
doi.org/10.1016/j.jesp.2019.103853. https://doi.org/10.1080/13576500444000236.
Hornsey, M. J., & Jetten, J. (2004). The individual within the group: Balancing the need
to belong with the need to be different. Personality and Social Psychology Review, 8,
References
248–264. https://doi.org/10.1207/s15327957pspr0803_2.
Jans, L., Postmes, T., & Van der Zee, K. I. (2011). The induction of shared identity: The
Baray, G., Postmes, T., & Jetten, J. (2009). When I equals we: Exploring the relation positive role of individual distinctiveness for groups. Personality and Social Psychology
between social and personal identity of extreme right-wing political party members. Bulletin, 37, 1130–1141. https://doi.org/10.1177/0146167211407342.
British Journal of Social Psychology, 48, 625–647. https://doi.org/10.1348/ Leach, C. W., Van Zomeren, M., Zebel, S., Vliek, M. L., Pennekamp, S. F., Doosje, B., ...
014466608X389582. Spears, R. (2008). Group-level self-definition and self-investment: A hierarchical
Becker, J., & Tausch, N. (2014). When group memberships are negative: The concept, (multicomponent) model of in-group identification. Journal of Personality and Social
measurement, and behavioural implications of psychological disidentification. Self Psychology, 95, 144–165. https://doi.org/10.1037/0022-3514.95.1.144.
and Identity, 13, 294–321. https://doi.org/10.1080/15298868.2013.819991. Lewin, G. W. (Ed.). (1948). Resolving social conflict. London: Harper & Row.
Becker, J., Tausch, N., Spears, R., & Christ, O. (2011). Committed dis(s)idents: Mael, F., & Ashforth, B. E. (1992). Alumni and their alma mater: A partial test of the
Participation in radical collective action fosters disidentification with the super- reformulated model of organizational identification. Journal of Organizational
ordinate group but enhances political identification. Personality and Social Psychology Behavior, 13, 103–123. https://doi.org/10.1002/job.4030130202.
Bulletin, 37, 1104–1116. https://doi.org/10.1177/0146167211407076. Payne, B. K., Vuletich, H. A., & Lundberg, K. B. (2017). The bias of crowds: How implicit
Bhappu, A. D., & Crews, J. M. (2005). The effects of communication media & conflict on bias bridges personal and systemic prejudice. Psychological Inquiry, 28, 233–248.
team identification in diverse teams. In R. H. Spraque, Jr. (Ed.), Proceedings of the https://doi.org/10.1080/1047840X.2017.1335568.
Annual Hawaii International Conference on System Sciences (pp. 50). Postmes, T., Haslam, S. A., & Jans, L. (2013). A single-item measure of social identifi-
Bjørgo, T. (2005). Reducing recruitment and promoting disengagement from extremist cation: Reliability, validity, and utility. British Journal of Social Psychology, 52,
groups: The case of racist sub-cultures. In C. Benard (Ed.). A future for the young: 597–617. https://doi.org/10.1111/bjso.12006.
Options for helping middle eastern youth escape the trap of radicalization (pp. 1–30). Postmes, T., Spears, R., Lee, A. T., & Novak, R. J. (2005). Individuality and social influ-
Santa Monica, CA: RAND National Security Research Division. ence in groups: Inductive and deductive routes to group identity. Journal of
Bortolini, T., Newson, M., Natividade, J. C., Vázquez, A., & Gómez, Á. (2018). Identity Personality and Social Psychology, 89, 747–763. https://doi.org/10.1037/0022-3514.
fusion predicts endorsement of pro-group behaviours targeting nationality, religion, 89.5.747.
or football in Brazilian samples. British Journal of Social Psychology, 57, 346–366. Prentice, D., Miller, D., & Lightdale, J. (1994). Asymmetries in attachments to groups and
https://doi.org/10.1111/bjso.12235 Early view. to their members: Distinguishing between common-identity and common-bond
Brewer, M. B., & Gardner, W. (1996). Who is this “we”? Levels of collective identity and groups. Personality and Social Psychology Bulletin, 20, 484–493. https://doi.org/10.
self representations. Journal of Personality and Social Psychology, 71, 83–93. https:// 1177/0146167294205005.
doi.org/10.1037/0022-3514.71.1.83. Postmes, T., & Jetten, J. (2006). Individuality and the group: Advances in social identity.
Brewer, M. B., & Yuki, M. (2007). Culture and social identity. In S. Kitayama, & D. Cohen Thousand Oaks, CA, US: Sage Publications, Inc.
(Eds.). Handbook of cultural psychology (pp. 307–322). New York: Guilford. Simon, B., & Stürmer, S. (2003). Respect for group members: Intragroup determinants of
Buhrmester, M. D., Fraser, W. T., Lanman, J. A., Whitehouse, H., & Swann, Jr., W. B. collective identification and group-serving behavior. Personality and Social Psychology
(2014). When terror hits home: Identity fused Americans who saw Boston bombing Bulletin, 29, 183–193. https://doi.org/10.1177/0146167202239043.
victims as “family” provided aid. Self and Identity, 14, 253–270. doi:https://doi.org/ Steffens, N. K., Haslam, S. A., & Reicher, S. D. (2014). Up close and personal: Evidence
10.1080/15298868.2015.992465. that shared social identity is a basis for the ‘special’ relationship that binds followers
Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. to leaders. The Leadership Quarterly, 25, 296–313. https://doi.org/10.1016/j.leaqua.
Psychological Bulletin, 52(4), 281. 2013.08.008.
Demant, F., Slootman, M., Buijs, F., & Tillie, J. (2008). Decline and disengagement: An Swann, W. B., Jr., & Buhrmester, M. D. (2015). Identity fusion. Current Directions in
analysis of processes of deradicalisation. Amsterdam, Netherlands: IMES, University of Psychological Science, 24, 52–57. https://doi.org/10.1177/0963721414551363.
Netherlands. Swann, W. B., Jr., Milton, L. P., & Polzer, J. T. (2000). Should we create a niche or fall in
Drury, J., & Reicher, S. D. (2009). Collective psychological empowerment as a model of line? Identity negotiation and small group effectiveness. Journal of Personality and
social change: Researching crowds and power. Journal of Social Issues, 65(4), Social Psychology, 79, 238–250. https://doi.org/10.1037//0022-3514.79.2.238.
707–726. https://doi.org/10.1111/j.1540-4560.2009.01622.x. Swann, W. B., Jr., Buhrmester, M. D., Gómez, A., Jetten, J., Bastian, B., Vázquez, A., ...
Ellemers, N., Kortekaas, P., & Ouwerkerk, J. W. (1999). Self-categorization, commitment Zhang, A. (2014). What makes a group worth dying for? Identity fusion fosters
to the group and group self-esteem as related but distinct aspects of social identity. perception of familial ties, promoting self-sacrifice. Journal of Personality and Social
European Journal of Social Psychology, 29, 371–389. https://doi.org/10.1002/(SICI) Psychology, 106, 912–926. https://doi.org/10.1037/a0036089.
1099-0992(199903/05)29:2/3. Swann, W. B., Jr., Gómez, Á., Buhrmester, M. D., López-Rodríguez, L., Jiménez, J., &
Ellemers, N., Van Den Heuvel, H., De Gilder, D., Maass, A., & Bonvini, A. (2004). The Vázquez, A. (2014). Contemplating the ultimate sacrifice: Identity fusion channels
underrepresentation of women in science: Differential commitment or the queen bee pro-group affect, cognition, and moral decision making. Journal of Personality and
syndrome? British Journal of Social Psychology, 43, 315–338. https://doi.org/10. Social Psychology, 106, 713–727. https://doi.org/10.1037/a0035809.
1348/0144666042037999. Swann, W. B., Jr., Gómez, Á., Dovidio, J. F., Hart, S., & Jetten, J. (2010). Dying and killing
Elsbach, K. D., & Bhattacharya, C. B. (2001). Defining who you are by what you're not: A for one's group: Identity fusion moderates responses to intergroup versions of the

15
Á. Gómez, et al. Journal of Experimental Social Psychology 85 (2019) 103853

trolley problem. Psychological Science, 21, 1176–1183. https://doi.org/10.1177/ van Zomeren, M., Saguy, T., & Schellhaas, F. M. (2013). Believing in “making a differ-
0956797610376656. ence” to collective efforts: Participative efficacy beliefs as a unique predictor of
Swann, W. B., Jr., Gómez, Á., Huici, C., Morales, F., & Hixon, J. G. (2010). Identity fusion collective action. Group Processes & Intergroup Relations, 16(5), 618–634. https://doi.
and self-sacrifice: Arousal as a catalyst of pro-group fighting, dying, and helping org/10.1177/1368430212467476.
behavior. Journal of Personality and Social Psychology, 99, 824–841. https://doi.org/ Vázquez, A., Gómez, Á., & Swann, W. B. (2017). Do historic threats to the group diminish
10.1037/a0020014. identity fusion and its correlates? Self and Identity, 16, 480–503. https://doi.org/10.
Swann, W. B., Jr., Gómez, Á., Seyle, D. C., Morales, J., & Huici, C. (2009). Identity fusion: 1080/15298868.2016.1272485.
The interplay of personal and social identities in extreme group behavior. Journal of Yuki, M. (2003). Intergroup comparison versus intragroup relationships: A cross-cultural
Personality and Social Psychology, 96, 995–1011. https://doi.org/10.1037/a0013668. examination of social identity theory in North American and East Asian cultural
Swann, W. B., Jr., Jetten, J., Gómez, Á., Whitehouse, H., & Bastian, B. (2012). When contexts. Social Psychology Quarterly, 66, 166–183. https://doi.org/10.2307/
group membership gets personal: A theory of identity fusion. Psychological Review, 1519846.
119, 441–456. https://doi.org/10.1037/a0028589. Yuki, M., & Takemura, K. (2013). Intergroup comparison and intragroup relationships:
Swann, W. B., Jr. (2012). Self-verification theory. In P. Van Lang, A. Kruglanski, & E.T. Group processes in the cultures on individualism and collectivism. In M. Yuki, & M.
Higgins (Eds.), Handbook of theories of social psychology (pp. 23–42). Sage: London. Brewer (Eds.). Culture and group processes (pp. 38–65). New York, NY: Oxford
Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. University Press.
Austin, & S. Worchel (Eds.). The social psychology of intergroup relations (pp. 33–47). Zagenczyk, T. J., Cruz, K. S., Woodard, A. M., Walker, J. C., Few, W. T., Kiazad, K., & Raja,
Monterey, CA: Brooks-Cole. M. (2013). The moderating effect of Machiavellianism on the psychological contract
Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M. S. (1987). breach–organizational identification/disidentification relationships. Journal of
Rediscovering the social group: A self-categorization theory. Oxford: Blackwell. Business and Psychology, 28, 287–299. https://doi.org/10.1007/s10869-012-9278-1.

16