This paper examines linguistic integroup bias, infrahu-manization, and agency in the context of Central and Eastern European inter-ethnic historical conflicts. We tested the hypothesis that people show more implicit lan-guage bias (e.g.: linguistic intergroup bias, infrahuman-ization, and agency) in relation to inter-ethnic, historical conflicts which are perceived as on-going (e.g.: Serbia vs. Croatia, Hungary vs. Slovakia) than in relation to in-ter-ethnic historical conflicts which are perceived termi-nated or non-existing (e.g.: Hungary vs. Lithuania). We also expected that national identification influences these linguistic biases. Our study provided support for the hy-pothesis that people show more implicit language bias in relation to conflicts, as well as to national identification.
Key words: national identification, inter-ethnic conflicts, linguistic intergroup biases
1. Theoretical background
One of the main and most current research areas in social psychol-ogy is the investigation of inter-group conflicts, which is closely relat-ed to the study of stereotypes and prejudice. In the last fifty years there has been a tendency of investigating stereotypes and prejudice based upon implicit measures which assess attitudes that respondents may not be willing to express directly (Wittenbrink and Schwartz, 2007).
In-group favoritism is another related theory which has been a cen-tral topic ever since Sherif and Sherif (1937) conducted their famous experiments about group processes. The phenomena of favoritism have been measured in these experiments (Sherif et al., 1961, 1988), just as in Tajfel’s subsequent inter-group paradigm (Tajfel, 1981, 1982;
Tajfel and Turner, 1979, 1986) that relied on behavioral measures, e.g.
the allocation of rewards. Early social cognition research dealt with favoritism in an attributive framework as a type of attribution error (Pettigrew, 1979). Contemporary social cognition paradigms study cognitive and emotional processes that underlie favoritism. Haslam et al. (2005), Bain, Kashima, and Haslam (2006) have provided evi-dence that in-groups are conceived more human than out-groups. In-group members endow their own In-group with human essences (such as intelligence, sentiments, language, and positive and negative so-ciability) and values (Leyens et al. 2000), whereas they deprive out-groups from these qualities. Several studies presented examples how people discriminate on the basis of intelligence (Crocker et al., 1998), language (Giles and Coupland, 1991), sociability (Fiske et al. 2002), values (Bain et al., 2006) and sentiments (see the infrahumanization paradigm below). This is in accord with anthropologic observations.
Erikson (1959) described the phenomenon of pseudospeciation, when a tribe uses its own name for denoting the concept of man. Social rep-resentation studies also showed that groups tended to think of them-selves in cultural terms whereas they applied natural terms for out-groups (Perez et al., 2007).
2. Infrahumanization
Another major paradigm is infrahumanization. The infrahumaniza-tion hypothesis suggests a preferential attribuinfrahumaniza-tion of human essences to the in-group. One of the human characteristics – according to Leyens and colleagues (2000) – is the ability to feel secondary emotion while primary emotions are considered to not be uniquely human emotions.
In several studies Leyens and colleagues investigated the phenomenon of infrahumanization and they were led to the conclusion that people attribute more secondary emotion to the in-group than to the out-group (Leyens et al., 2000; Paladino et al., 2002).
For example, using the Implicit Association Test, Paladino et al.
(2002) found Belgian students associated Belgian names with second-ary emotions and North African names with primsecond-ary emotions (com-patible set) more quickly than Belgian names with primary emotions and North African names with secondary emotions (incompatible set).
The type of emotion also has an effect on empathic behavior. In the study of Vaes, Paladino, Castelli, Leyens, and Giovanazzi (2003), in-dividuals were more helpful with those in-group members who pressed positive or negative secondary emotion than those who ex-pressed primary emotion.
3. Linguistic Intergroup Bias
Although it is not directly related to the issues of humanness, an-other leading paradigm of social cognition research has turned to lan-guage use and implicit semantics when searching for cognitive media-tors of in-group favoritism: Linguistic Intergroup Bias (LIB, Maas, Salvi, Arcuri and Semin, 1989). LIB is based on the Linguistic Cat-egory Model (Semin and Fiedler, 1988; 1991). LCM offers a four-fold distinction among descriptive action verbs (DAVs), interpretative action verbs (IAVs), state verbs (SVs), and adjectives (ADJs). Four different levels of abstraction are represented by the four different
lin-guistic categories. At the most concrete level are DAVs, such as kick, kiss, and telephone, that refer to a single, objective, neutral description of observable action with a clear beginning and end. No interpretation of the action is involved. At the second level of abstraction are IAVs, such as to help or to cheat, which describes a general class of behav-iors without identifying the specific behavior to which they refer in a given context (including various possible behavioral acts), but refers to a specific action with clear beginning and end. An interpretation beyond the mere description is provided. A psychological state (emo-tional, affective, mental, etc.) can be described by SVs (such as like, hate,and admire). A state verb goes beyond specific behaviors or situa-tions and does not have a clear beginning and end. At the highest level of abstraction are ADJs, such as aggressive or creative. ADJs are high-ly abstract dispositions or characteristics of a person. They have a low contextual dependence as they provide generalizations across events, situations and objects. The same behavioral episode can be encoded at different levels of abstraction. To demonstrate the same event, one can say „Person A hits Person B”, „Person A hurts Person B”, „Person A hates Person B” or „Person A is aggressive”. What is important is that the codification at a higher or lower level of abstraction has a number of implications for interpersonal and intergroup relations. Semin and Fiedler (1988) have demonstrated that the abstract statements imply greater temporal stability and provide more information about the per-son and less about the situation than are concrete statements.
With the application of the linguistic category model, Maass, Sal-vi, Arcuri, and Semin (1989) introduced the concept of the linguistic inter-group bias. They suggest that the same behavioral episodes are encoded at different levels of abstraction depending on whether such behaviors have positive or negative connotations and whether they are performed by in-group or out-group members. In particular, they argue that positive in-group and negative out-group behaviors are described with abstract predicates. In contrast, the same positive behavior dis-played by an out-group member and negative behavior disdis-played by an in-group member is described with concrete predicates. Thus, an
in-group member’s positive behavior and out-group member’s nega-tive behavior may be interpreted as temporarily stable and informa-tive about the actor’s characteristics, whereas the out-group member’s positive behavior and in-group member’s negative behavior may be interpreted as a specific instance that has high contextual dependence.
These patterns of lexical choices are termed the linguistic inter-group bias. The linguistic intergroup bias has been demonstrated in various studies of a number of inter-group contexts (eg. Maass et al, 1989;
Rubini et al., 2007; Anolli et al., 2006; Tanabe and Oka, 2001; Rubini and Menegatti, 2008).
4. Agency
Thinking about in-group favoritism, we think that agency is a po-tential candidate for both attributive and stereotype-forming reasons so as to positively discriminate one’s own group from other groups and to implement this discrimination into communication (Semin, 2000). For attribution it is evident that we may expect more agencies for in-group good deeds and fewer agencies when the in-group has committed something wrong. Similarly, an out-groups’ wrongdoings would be depicted and interpreted with higher level of agency, where-as agency in their positive actions would be decrewhere-ased.
Agency – the ability of a person or group to act effectively (Hamil-ton, 2007) – is a major domain of social perception and identity stud-ies. The extent of agency includes the responsibility for an act (Val-lacher and Wegner, 1989), self-efficacy (Schwarzer and Jerusalem, 1995), the influence on others and the environment, and the ability of attaining goals (Spencer-Rodgers et al., 2007; Hamilton, 2007).
Whereas several studies presented evidence that linguistic patterns provide information about a speaker’s representations of agency and of the depicted events (Semin and Fiedler, 1991; Maass et al., 1989;
Pólya, László, and Forgas, 2005), linguistic mediators of agency have never been experimentally examined. In this study we set out to study
the role of agency in in-group favoritism through linguistic patterns; a person or a nation can describe its life story or history so that the actor is actively shaping the events, or, on the contrary, as being simply a passive recipient of the happenings . Similarly, the individual’s or the group’s capacity to cope actively with challenges reflects important aspects of individual and group (national) identity. Agency is repre-sented in the texts by syntactic structures (active voice versus passive voice, which vary the role of the agent and the recipient in a sentence) and by the rate of active and passive expressions. Szalai and László (2006, 2007) developed a dictionary of verbs whose meta-semantic or connotative content is heavily loaded with either activity or passivity.
Seven judges categorized 10,000 verbs according to activity or passiv-ity. Eventually, a list of 1,200 verbs was categorized either into the ac-tive or passive category. The acac-tive verb category consisted of action verbs (e.g., fights, attacks, hits, moves), verbs changing the activation contour (speeds up, stands up), and verbs of starting or enhancing ac-tivity (e.g., begin, accelerate, strengthen, initiate). The passive verb category contains state verbs (e.g., exists, unfolds), verbs of happen-ings (e.g., happens, unfolds), verbs of changing the activation con-tour into passive direction (e.g., lies down, slows down) and verbs of stopping or decreasing activity (e.g., stops, ends, concludes). A large set of verbs, e.g., mental verbs or verbs describing changes in physi-ological states were not categorized. Using this dictionary, Szalai and László (2007) developed a computer program with the help of which they investigated the in-group and out-group asymmetry phenomena in Hungarian history textbooks and narratives of the Hungarian peo-ple. Results show there exists an agency bias; namely, the agency is higher for Hungarians in positive stories than in negative ones, and in contrast it is higher for out-groups in negative stories than in positive ones.
5. National identification
In the literature on national identity, there is no consensus on the definition and measurement of the concept. However, the majority of the studies mention two kinds of identification which have different labels but have a similar content. Several studies dealt with the ho-mogeneity of identification with one’s nation and found that it has at least two aspects. Kosterman and Feshbach (1989) used the terms patriotism and nationalism for the two modes of national identifica-tion, where patriotism refers to the positive attachment to one’s naidentifica-tion, while nationalism includes the feeling of superiority of one’s nation to other nations. Roccas and Klar (2006) used the terms attachment and glorification, which are similar to the previous categories; attach-ment corresponds to patriotism whereas glorification corresponds to nationalism. A new five-dimensional model of in-group identification has been introduced by Leach et al. (2008), which can also be useful in measuring national identification. The findings of Branscombe and her colleagues (2004) suggest that national identification has a mediating role in inter-group emotions such as collective guilt.
Viki and Calitri (2008) investigated the relation of national iden-tification and infrahumanization in their study. The main issue of the study was whether the differential attribution of secondary emotions to the in-group refers to the infrahumanization of the out-group or to the suprahumaniztaion of the in-group. According to their argument, if the differential attribution of secondary emotions to the in-group is related to the nationalism factor of national identification, then it sup-ports the infrahumanization hypothesis; however, if it is related to the patriotism factor, then it supports the suprahumanization hypothesis.
6. Historical context
Studies of in-group favoritism increasingly and more explicitly ap-proach real life inter-ethnic relations. They attempt to explore
cogni-tive and emotional factors in terms by which the appraisal of inter-group relations occurs. They are also interested in factors which may moderate group perception, such as group entativity (other variables that appear to be related to entativity are interaction among members, importance of membership, shared tasks, and common social catego-ries) (Hamilton and Sherman, 1996). For instance, ongoing or termi-nated conflicts may substantially influence inter-group perception. In their study, Roccas, Klar, and Liviatan (2006) found that the percep-tion of one’s own and of another group was a funcpercep-tion of the intensity of the conflict.
7. Our study
Our aim in this study was to test if the above mentioned phenom-ena are observable in a special experimental setting which makes the historical context salient. Historical contexts are probably particular-ly sensitive to the effects of national identification. In our study we wanted to test whether their impact could be observed in all implicit measures of in-group favoritism, and if they did, whether the type of conflict (terminated versus ongoing) would moderate these effects.
8. Hypothesis
Based on the above considerations and previous research the fol-lowing hypotheses have been formulated:
1. Infrahumanization will be stronger among those who have a higher score on the national identification scale and also by those who perceive a conflict still ongoing with a particular nation.
2. Linguistic intergroup bias will be stronger among those who have a higher score on the national identification scale and also by those who perceive a conflict still ongoing with a particular nation. For conflicts or cooperation between out-groups, LIB would occur for
groups with which the subjects’ group is still involved in an ongo-ing conflict.
3. Agency bias will be stronger among those who have higher score on the national identification scale and also by those who perceive a conflict still ongoing with a particular nation. For conflicts or cooperation between out-groups, the agency bias would occur for groups with which the subjects’ group is still involved in an ongo-ing conflict.
9. Method
Participants
In this study, 248 university students (of which 71 were male) par-ticipated voluntarily. The participants’ average age was 21,1 years old, with a standard deviation of 2,56. All of them were Hungarian citizens and they were from both the faculties of sciences and humanities.
Procedure
Each student received a small booklet with instructions, followed by eight photographs with two, three, or four titles - depending upon the investigated paradigm - and they were asked to choose the title they would give to the picture. They were told that there was no right or wrong answer. Afterwards they were asked to fill out a seven-point Likert-scale questionnaire about the national identification. Originally the questionnaire had six items, three about patriotism and three about nationalism. Finally, the participants answered two questions about the perceived termination of the historical conflict in question and about the possibility of a future conflict with all of the participating nations. The whole procedure of the experiment took 15 minutes for each participant; they filled out the booklets in groups but were seated separately.
Stimuli
Eight photographs were included in the study. Four pictures showed situations where characters experienced certain positive and negative emotions. Four pictures showed situations which could be interpreted in historical terms. These pictures depicted situations where members of a national group caused harm to another group (street massacres and deportations) or performed some socially positive deed (offering assistance with food or collecting aid during a revolution). The group identity of the participants in the events was not identifiable. We var-ied the nationality of the characters appearing on the photos. The fol-lowing pairings were applied:
For massacres: Hungarian soldiers killed Romanians; Romanian soldiers killed Hungarians; Hungarian soldiers killed Serbians; Ser-bian soldiers killed Hungarians; and Russian soldiers killed Poles.
For deportations: Hungarians deported Germans; Slovaks deported Hungarians; Slovaks deported Germans; and Croatians deported Ser-bians.
For help with food: Hungarian helped Romanians; Romanians helped Hungarians; Russians helped Poles; and Poles helped Russians.
For collecting aid: Hungarians helped Romanians; Romanians helped Hungarians; Russians helped Poles; and Poles helped Russians.
In each case we varied either the linguistic abstraction or the agen-cy appearing in the picture captions.
Examples of the different abstraction levels include the following:
1. After the Second World War, Hungarians deported ethnic Germans who lived in Hungary (interpretative verb, relatively low abstrac-tion).
2. Hungarians were vengeful towards ethnic Germans after the Sec-ond World War (adjective, relatively high abstraction).
Examples of the different levels of agency include the following:
1. The Germans who lived in Hungary were deported after the Sec-ond World War (low Hungarian agency).
2. The Hungarians deported the Germans who lived in the country after the Second World War (high Hungarian agency).
In the case of the emotional pictures subjects had to choose be-tween three captions. One was without emotion, one with a primary emotion and one with a secondary emotion. The four emotions we applied in the study were: happy, cheerful, sad, and disheartened. We applied the following nationalities for the characters in the pictures:
Hungarian, Romanian, Slovak, Serbian, Polish, and Lithuanian. For example:
A Hungarian soldier stands at the grave of his comrades (no emotion).
A sad Hungarian soldier stands at the grave of his comrades (primary emotion).
A mourning Hungarian soldier stands at the grave of his comrades (secondary emotion).
10. Results
Preliminary Analysis
We conducted a factor analysis on the items measuring national identification. The analysis produced only one reliable factor which contained the three patriotic items with the following factor load-ings: 0.81, 0.76, and 0.70. The reliability of this factor was acceptable (Cronbach alpha > 0.7). The items of this factor were: “I am glad to be born a Hungarian”; “I don’t like to live in Hungary”; and “My heart beats faster when I hear Hungarian words in abroad.”
We had two questions about the perceived conflicts with the par-ticipating nations. We conducted a correlation analysis which resulted in a significant correlation between the two questions (correlation co-efficients ranged from 0.5 to 0.7). For the main analysis we added the scores of the two questions (minimum 2 – ongoing conflict, maximum 14 – terminated conflict). Table 1 shows the results:
Perceived Conflict with the Participating Nations
Romania 5.46
Slovakia 6.23
Serbia 7.87
Russia 9.75
Croatia 10.82
Poland 12.24
Lithuania 12.38
Table 1: Perceived Conflict with the Participating Nations
Main Analysis
We conducted a logistic regression analysis in order to reveal which variables have an effect on the different linguistic categories.
Related to the infrahumanization hypothesis, the results show that the predictor variables – the intensity of national identification and the perceived termination of the conflict – did not affect the selection of the titles. We also tested the phenomenon of infrahumanization itself, and we did not find a difference in the amount of titles with secondary emotions selected for the in-group compared with the out-group.
Related to the linguistic intergroup bias hypothesis, we found the following results applying the logistic regression analysis. In the case of the “massacre” picture, when the picture caption read that
Hungar-ians killed RomanHungar-ians, we observed a significant relationship between national identification and linguistic abstraction level of the selected title, β= -0.232, S.E.= 0.115; χ2 (1)= 4.036; p<0.05. For the same pic-ture, in the Hungarian-Serbian version, the regression analysis also yielded a significant relationship between national identification and the linguistic abstraction level, β= -0.635, S.E.= 0,293; χ2 (1)= 4.692;
Hungar-ians killed RomanHungar-ians, we observed a significant relationship between national identification and linguistic abstraction level of the selected title, β= -0.232, S.E.= 0.115; χ2 (1)= 4.036; p<0.05. For the same pic-ture, in the Hungarian-Serbian version, the regression analysis also yielded a significant relationship between national identification and the linguistic abstraction level, β= -0.635, S.E.= 0,293; χ2 (1)= 4.692;