Self-Censorship Orientation : Scale Development , Correlates and Outcomes

Self-censorship is defined as intentionally and voluntarily withholding information from others in absence of formal obstacles. We conducted cross-sectional and longitudinal research to develop a quantitative measure of individuals’ Self-Censorship Orientation (SCO) and investigated its correlates and outcomes in the context of the intractable Israeli-Palestinian conflict. Stage 1 investigated the factor structure of the scale and its convergent and discriminant validity in a representative sample (N = 499). Findings revealed two negatively related factors representing preferences for self-censorship and for disclosure of information. The factors were distinct frommeasures of similar constructs and correlated as expected with variables representing conservatism, ingroup commitment and universalistic values. In Stage 2, participants were re-surveyed five months later to establish test-retest reliability and predictive validity. SCO factors assessed at Stage 1 predicted readiness to conceal or reveal information portraying the ingroup’s conduct in the conflict negatively beyond all Stage 1 measures. The SCO scale provides a reliable and valid instrument for future investigations of self-censorship and its individual and societal implications.


Non-Technical Summary
Background Self-censorship is an intentional voluntary choice to withhold information from others, even when no formal obstacles prevent the dissemination of this information.
Self-Censorship Orientation 332 For example, Maksudyan (2009) analyzed translations of history books about World War I into Turkish in Turkey.
She found that writers, translators and editors censored themselves when addressing information regarding what is considered by most of the world as the Armenian genocide during this period.They did so in order to avoid public censorship and/or in order to achieve approval from the formal institutions.As another example, Antilla (2010) noted that the news media in the United States practice self-censorship in reporting scientific findings about the effects of human actions on global warming.
In this article we focus on self-censorship in societal contexts.We address cases in which it is clear that the selfcensored information may contribute to improved societal functioning and/or prevent wrongdoings within the society.
Nonetheless, individuals withhold the information from other society members and outsiders, which can affect the well-being of societies and their members.One example of such a context is intergroup conflict, which is the focus of our research and is discussed elaborately in what follows.
Individuals may differ in their inclination to practice self-censorship.Therefore, it is of considerable interest to assess individuals' propensity toward self-censorship.The purpose of the present research was to develop an instrument to assess individuals' Self-Censorship Orientation (SCO) and to situate this construct among other established psychological constructs.Toward this aim, we examine the correlates of SCO and its consequences.

The Concept of Self-Censorship
We define self-censorship of information as intentionally and voluntarily withholding information from others in the absence of formal obstacles (Bar-Tal, 2017).Self-censorship typically occurs in situations in which the actor possesses new information that is unknown to others, believes that this information is valid, and believes that it has implications for others.By self-censoring, the individual prevents others' free access to the information and informally regulates the flow of information in society.Self-censorship then becomes a barrier to the dissemination of information that may be useful to the group.When individuals self-censor extensively, the functioning of democratic societies may be disrupted (for an extensive discussion see Bar-Tal, Nets-Zehngut, & Sharvit, 2017).
The present conception of self-censorship focuses on withholding information rather than opinions.Information is viewed as factual knowledge that can be based on personal experience, reports from others, research or documents.It refers to something that actually happened and that is considered verified and valid.i Of importance is individuals' belief that the information is based on evidence and does not depend on personal views.Opinions are considered as going beyond the evidence and reflecting subjective inferences, attributions, evaluations and impressions.Their expression is excluded from our definition of self-censorship.Legal scholars make a similar distinction between information and opinions (Barendt, 2005;Schauer, 1982).The emphasis on information distinguishes our conception of self-censorship from previous conceptions.Specifically, Hayes, Glynn, andShanahan (2005a, p. 298, 2005b, p. 443) define self-censorship as "withholding of one's true opinion from an audience perceived to disagree with that opinion" and maintain that conformity is a form of self-censorship.In our view, focusing on opinions rather than information does not allow sufficient distinction between the concept of self-censorship and the concept of conformity, which we discuss extensively in the following section.
Additionally, our definition suggests that self-censorship is practiced only when there are no formal obstacles to revealing information and the decision to withhold information is made voluntarily.The definition therefore excludes cases in which there is an institutionalized external obstacle that prevents dissemination of information, such as formal censorship (e.g., De Baets, 2002;Jansen, 1988;Tribe, 1973).
A number of different but not mutually exclusive motivations may drive individuals to self-censor (see Nets-Zehngut, Pliskin, & Bar-Tal, 2015).A salient motivation is protection of the ingroup based on the assumption that the information may hurt the group or its cause if revealed.More specifically, Bar-Tal (2017) suggests that self-censorship often reflects resolution of a dilemma between conflicting goals or needs.On the one hand, individuals are motivated to share information with others.In addition to the pro-social function of providing important information to others, sharing of information contributes to social coordination, to forming a meaningful understanding of events, and to relief of unpleasant emotions.On the other hand, individuals often identify strongly with social groups, which become sources from which they derive part of their self-concept (Tajfel, 1978).Moreover, in some cases individuals feel that they share a common fate with the group and experience concern for the welfare of the collective (David & Bar-Tal, 2009).Consequently, they are motivated to maintain a positive image of the group and to prevent harm to the group.Hence, when individuals are in possession of information that may have negative implications for a group with which they identify, they experience a dilemma between the need to share the information and their motivation to protect the group.In some cases, they may decide to resolve the dilemma by selfcensoring the information.
Societies often employ social sanctions against members who disseminate information that may be damaging to the group and reward members who display loyalty.This can encourage group members to self-censor in order to protect their own self-interest as well as that of the group.In other cases, however, concern for the group may lead individuals to reveal information even if it is damaging to the group's image, as a way of improving the group's functioning (Packer, 2008).

Self-Censorship Versus Conformity
Self-censorship must be distinguished from the related concept of conformity.Conformity is the tendency to change one's beliefs, attitudes, or behaviors to match those of the majority of others (Cialdini & Goldstein, 2004).This change is in response to pressure by group members, which can be concrete (e.g., expression by others of what should be said) or imagined (e.g., based on past experiences).The literature differentiates between private conformity (also acceptance or conversion) and public conformity (also compliance).The former takes place when individuals truly accept the beliefs, attitudes or behaviors of the majority viewing them as truthful.The latter takes place when individuals express the views of the majority publically, but continue to hold different views privately.
Individuals may conform because they do not want to appear deviant (normative conformity), or because they want to have accurate and valid knowledge (informational conformity; Deutsch & Gerard, 1955).
Several features distinguish conformity from our conception of self-censorship.First, conformity can involve either expressing or withholding opinions or information, whereas the present conception of self-censorship focuses only on withholding information.Second, individuals who conform believe that they are in the minority and withhold their beliefs either because they question their validity or because they are concerned about negative reactions by the majority.Such concern is also central to the conception of self-censorship proposed by Hayes et al. (2005aHayes et al. ( , 2005b)), who found that individuals high (versus low) on their measure of self-censorship are more sensitive to the opinion climate of a group when deciding whether to express their opinions (Hayes, Uldall, & Glynn, 2010).
However, in our conception of self-censorship individuals may withhold information when they are in the majority, if they believe that revealing the information could hurt their group.More generally, in cases of conformity the primary motivation is to protect the self, while in the case of self-censorship, according to the present conception, a central motivation is often to protect the group.
Self-Censorship in a Context of Intergroup Conflict Bar-Tal (2017) proposed that the likelihood of self-censorship can be a function of individual differences, as well as the social context and specific circumstances.In the present research, we examine individual differences in the readiness to self-censor and their implications in the specific context of an intractable intergroup conflict.
Studies suggest that the context of intractable conflict is especially conducive to self-censorship (Bar-Tal et al., 2017;Hameiri, Sharvit, Bar-Tal, Shahar, & Halperin, 2017).Societies involved in intractable conflict develop conflict-supporting narratives, which serve important functions when facing violence and war.These functions include justification of the group's behavior, mobilization of society members to participate in the conflict and maintaining a positive group identity (Bar-Tal, 2013).Dissemination of information that contradicts these conflictsupporting narratives is discouraged because it is seen as detrimental to the group's cause in the conflict (Bar-Tal, Oren, & Nets-Zehngut, 2014).In this context, self-censorship is often viewed by authorities and segments of society as necessary in order to protect the ingroup and its interests, leading to the development of norms that encourage self-censorship.Hameiri, Sharvit, et al. (2017) conducted a longitudinal study to investigate antecedents and consequences of support for self-censorship during a violent confrontation in the context of the intractable Israeli-Palestinian conflict.
The study included a measure referring to free flow of information in the media (reverse-scored) as well as a specific measure referring to support for self-censorship regarding harmful actions by Israelis toward Palestinians in the course of a violent confrontation (Operation Pillar of Defense; OPD).Notably, the measures in this study referred to support for self-censorship by others rather than respondents' willingness to self-censor themselves.
The findings showed that support for self-censorship during OPD was predicted by Right Wing Authoritarianism (RWA), ethnocentrism, siege mentality, and political orientation assessed prior to the operation, and was negatively related to support for conciliatory measures assessed after the operation.This suggested that support for self-censorship during violent confrontation can become a barrier to conflict resolution (see also Hameiri, Bar-Tal, & Halperin, 2017).
Additional studies examining self-censorship in the context of a violent intergroup conflict were conducted by Shahar, Hameiri, Bar-Tal, and Raviv (2018).In these studies, Israeli-Jewish participants were presented with hypothetical scenarios in which they might receive information about harmful conduct by members of their own group toward Palestinians, and were asked to indicate to whom they would be willing to reveal the information.There was no assessment of participants' general orientation toward self-censorship independent of content and role.
The findings showed that participants were more willing to disclose the information to individuals who were socially close to them than to more distant persons, particularly outgroup members.In addition, the second study found that participants were more willing to self-censor the information when they were hypothetically assigned to the role of a soldier than the role of civilian, and more willing to self-censor when the perpetrators of the harmful acts were soldiers rather than civilians.The present research aims to extend past findings by examining the implications of individual differences in a general orientation toward self-censorship, regardless of a specific outlet, topic or role, for the willingness to withhold or reveal information related to a violent intergroup conflict.

The Present Research
Measuring a general orientation toward self-censorship behavior is a challenge, because not everyone has necessarily experienced situations in which they needed to decide whether to disseminate or withhold information, and it is difficult to create such situations artificially.Therefore, we decided to focus on assessing the perceived appropriateness of self-censorship and the readiness to do so in certain situations.Our underlying assumption is that individuals who support the use of self-censorship are also more likely to practice it.The work of Hayes et al. (2005aHayes et al. ( , 2005b) ) on self-censorship similarly focused on the willingness to self-censor (WTSC), though as mentioned, their conception of self-censorship is different from ours.
The construction of the scale involved two stages.In the first stage we created a measure of SCO and collected data from a representative population sample in order to investigate the factor structure of the scale, determine its final item composition and test its correlates and implications in a context of violent intractable conflict.The second stage took place five months after the first.In that stage we re-surveyed as many of the participants from the first stage as possible in order to investigate the test-retest reliability and predictive validity of the scale, again in the context of an intractable conflict.
Stage 1: Developing the Self-Censorship Orientation

(SCO) Scale
The scale construction process at Stage 1 was theoretically driven and based on the definition of self-censorship provided earlier.With respect to correlates, we first looked at the WTSC scale (Hayes et al., 2005a).As mentioned earlier, the WTSC assesses willingness to withhold one's opinions, whereas our measure of SCO concerns the readiness to withhold valid information.Accordingly, we expected our measure of SCO to be distinct from WTSC.
Nevertheless, because some of the factors underlying the willingness to withhold opinions and the willingness to withhold information may be similar, we expected the two measures to be positively correlated.
In establishing the discriminant validity of the WTSC, Hayes et al. (2005a) demonstrated that is was related to, but distinct from, several other indicators including argumentativeness (Infante & Rancer, 1982) and fear of negative evaluation (Leary, 1983).The former refers to individuals' tendency to seek out and enjoy arguments, and the latter to experiencing distress in situations in which one is evaluated negatively.We expected our measure of SCO to also be distinct yet related to these measures.The expectation of a positive correlation between selfcensorship and fear of negative evaluation is consistent with previous findings showing that whistle-blowing and breaking silence behaviors are associated with low self-monitoring (Jos, Tompkins, & Hays, 1989;Premeaux & Bedeian, 2003), which is a tendency to be guided by situational cues of social appropriateness (Snyder, 1974).
Other measures used by Hayes et al. (2005a), such as shyness and self-esteem, were considered less relevant to our conception of SCO and therefore not included.
Our conception suggests that a major driving force behind SCO is a desire to protect one's ingroup, including its image, its norms, and its dominant narratives.Accordingly, we expected SCO to be positively related to indicators of commitment to the group, such as group identification and blind patriotism (Schatz, Staub, & Lavine, 1999), as well as indicators of adherence to group norms and narratives, such as conformity and firmly entrenched narrative closure (FENCE; Klar & Baram, 2016).Blind patriotism is defined as: "an attachment to country characterized by unquestioning positive evaluation, staunch allegiance, and intolerance of criticism" (Schatz et al., 1999, p. 151).
FENCE was proposed as an individual-difference construct characterizing motivations to protect the historical group narrative (Klar & Baram, 2016).The desire to protect the ingroup, in turn, should increase with increased concern for the group's safety.Hence, we expected SCO to be positively related to indicators of individual fear and collective angst (Wohl, Branscombe, & Reysen, 2010).The association of self-censorship with fear and angst is also consistent with previous research in the organizational context indicating that silence at work is related to neuroticism (Nikolaou, Vakola, & Bourantas, 2008).
Right-wing authoritarianism (RWA; Altemeyer, 1996) refers to a personality type which advocates a conservative view of the world, including adherence to traditional values and closure to new ideas.RWA is associated with adherence to convention and conformity (Feldman, 2003) and with perception of the world as dangerous (Duckitt & Sibley, 2009).In the organizational context, it has been suggested that authoritarianism may be associated with employee silence (Timming & Johnstone, 2015).Thus, we expected SCO to be positively related to RWA as well.
Finally, we included measures of universalistic values, support for abstract democratic values, and constructive patriotism.Universalistic values focus on the understanding, appreciation, tolerance, and protection of all people, as well as equality, social justice, broadmindedness, and world peace (Schwartz, 2007).Abstract democratic values include freedom of expression, equality, and justice (Shamir & Sullivan, 1985).Constructive patriotism refers to individuals' emotional affinity to a group accompanied by a motivation to resist the group's actions when they betray fundamental values (Schatz et al., 1999).We expected SCO to decrease when commitment to values other than group protection increases.Accordingly, we expected SCO to be negatively related to support for universal values, abstract democratic values and constructive patriotism.A negative relationship between self-censorship and general values is consistent with research in the organizational domain indicating that voice and whistle-blowing are associated with strong commitment to moral rules, as well as duty and conscientiousness (Jos et al., 1989;Nikolaou et al., 2008;Tangirala, Kamdar, Venkataramani, & Parke, 2013) An additional aim of Stage 1 was to conduct a preliminary test of the predictive power of SCO in the context of a violent intractable conflict (which was explored further in Stage 2).We expected that individuals high on SCO would be less willing to reveal to outsiders specific information that portrays their ingroup's conduct in the conflict negatively and the rival group's behavior positively, compared to information that portrays the ingroup positively and the outgroup negatively.

Method Sampling and Procedure
We administered an online survey to a nationwide sample of 499 Israeli Jewish participants between the ages of 18 and 64.The surveys were distributed by the research firm Midgam Project (MP) and participants were randomly selected from Midgam's opt-in panel.Participants who completed the survey had the following socio-demographic characteristics: Mean age 39.00 years (SD = 13.58);51% female; 51% considered themselves secular, 29% were observant, 15% were religious and 6% were Ultra-Orthodox; 26% had at least a college degree; 39% had some college or professional education, and 35% had 12 or fewer years of schooling; 57% regarded themselves as having a rightist political orientation; 24% regarded themselves as centrists, and 18% regarded themselves as leftists.The sample's socio-demographic characteristics largely represented the distribution in the relevant population (Central Bureau of Statistics, 2013).

Measures
Willingness to reveal specific information about ingroup and outgroup behaviors in an intergroup conflict -This measure was administered at the beginning of the survey in order to ensure that participants' responses are not affected by the assessment of SCO or any other variables.We informed participants that one purpose of the study was to assist the EU in creating a website with information about conflicts around the world, including the Israeli-Palestinian conflict.We selected the EU as the website creator for two reasons.On the one hand, it is a positive reference group for Israelis.Israel has numerous agreements with the EU in different domains and aspires to become an EU member.On the other hand, the EU is often critical of Israel's policies and actions regarding the occupied Palestinian territories.Therefore, Israelis are likely motivated to influence the views of EU members on the Israeli-Palestinian conflict and may apply self-censorship in service of this goal.
We presented the participants with eight images and asked them to indicate the extent to which each image portrayed a typical situation in the Israeli-Palestinian conflict, and the extent to which it was appropriate for the website (1 = "not at all" and 6 = "absolutely").Half of the images portrayed Palestinians as aggressors and/or Israelis as victims (consistent with the dominant Israeli-Jewish narrative), and the rest portrayed Israelis as aggressors and/or Palestinians as victims (contradicting the dominant narrative) (see also Shahar et al., 2018).We computed an index of appropriateness of narrative-consistent images (α = .85)and an index of appropriateness of narrative contradicting images (α = .89).
The self-censorship orientation scale (SCO) -In order to construct the scale, we began by formulating a mapping sentence describing the presumed characteristics of individuals who are high or low on SCO as derived from the theoretical definition of self-censorship proposed by Bar-Tal (2017) and reviewed earlier (see Table 1).
Table 1 Mapping Sentence of Self-Censorship Orientation (SCO)

A person low in SCO will… A person high in SCO will…
Oppose censorship of information even if it might challenge his/her worldview and/or portray his/her ingroup negatively; Support censorship of information that might challenge his/her worldview and/or portray his/her ingroup negatively; Do so on the basis of universal values and motivations, including freedom of expression and information, believing that disclosing the information would contribute to the group; Do so on the basis of values and motivations of loyalty to the group, group protection and maintenance of group unity, as well as loyalty to personal beliefs and a desire to protect them; Respond with positive emotions to the disclosure of such information; Respond with negative emotions to the disclosure of such information; Perceive the information as representative of the group, even if it challenges the individual's worldview, and believe that concealing it would be harmful to the group; Perceive the information as not representative of the group, potentially harmful to it, or challenging the individual's worldview; Form positive judgements of others who disclose such information to various bodies; Form negative judgements of others who disclose such information; If coming across such information, will be motivated to disclose it; If coming across such information, will not disclose it and will act to conceal it; Oppose sanctions against persons or bodies that disclose such information Support sanctions against persons or bodies that disclose such information On the basis of the mapping sentence, the six authors generated through discussion 51 statements referring to beliefs, attitudes, emotions, motivations and behaviors reflecting high and low SCO.Next, we asked eight judges, who had previously conducted research on self-censorship, to review the statements and indicate the extent to which they reflected components of the mapping sentence.The judges noted that some statements were inconsistent with the theoretical reasoning, and others were unclear or could be interpreted in different ways.Following the judges' comments, we discarded some statements, changed the wording of others, and added new items.
We then asked the judges to review the items again and repeated the process, ending with 31 statements that were selected for administration to participants.
Many of the statements referred to a group to which respondents belong.Because individuals may identify with numerous groups, we asked participants, before responding to the statements, to specify the group with which they most identify out of several options (religious group, state, nation, ethnic group, workplace, local community, or political party/organization). Participants were then instructed to respond to all subsequent statements in reference to the group specified using a 6-point scale ranging between 1 = "not at all" and 6 = "to a great extent."The same rating scale was used for all the following measures as well.
Willingness to self-censor (WTSC) -This scale, developed by Hayes et al. (2005a), refers to the tendency to withhold one's opinions.The scale includes eight items (e.g., "It is difficult for me to express my opinion if I think others won't agree with what I say") with α = .81.

Argumentativeness -
We used an abbreviated version of the scale developed by Infante and Rancer (1982) comprised of seven items (α = .60;sample item: "Arguing over controversial issues improves my intelligence").
Fear of negative evaluation -We used nine items (α = .86)from the short form of the Fear of Negative Evaluation Scale (Leary, 1983).A sample item is "I worry about what other people will think of me even when I know it doesn't make any difference".
Group identification -We assessed identification in two ways, using measures created for the present research.
First, participants rated, in reference to the group to which they are most attached (as indicated earlier), the extent to which belonging to this group was important to them, and the extent to which they were emotionally attached to the group (α = .87).This measure could refer to different groups for different participants.In addition, because the outcome measures in the present study addressed the context of the Israeli-Palestinian conflict, participants' identification as Jewish Israelis was of particular interest.In order to assess it, participants responded to four items (α = .92)referring to their Jewish Israeli identification specifically (e.g., "How important to you is belonging to the Jewish nation?").
Patriotism -We used three blind patriotism items (α = .68;e.g., "I support Israel's policies for the very reason that they are the policies of my country") and three constructive patriotism items (α = .65;e.g., "I express my love for Israel by supporting efforts at positive change") from the measure developed by Schatz et al. (1999).
Conformity -This scale was adapted from the Social Conformity versus Autonomy scale (Feldman, 2003).The original scale items are comprised of pairs of statements, one representing a preference for individual autonomy and one for social conformity.For the present study, we selected eight autonomy statements (e.g., "People can only develop their true potential in a fully permissive society") and six conformity statements (e.g., "Obeying the rules and fitting in are signs of a strong and healthy society"), and asked the participants to rate their agreement with each statement.After reverse scoring the autonomy items, the scale yielded acceptable internal consistency (α = .77).

Firmly entrenched narrative closure (FENCE) -This scale assesses the value that individuals place on pro-
tecting their group's historical narrative in the context of an intergroup conflict (Klar & Baram, 2016).FENCE was found to be negatively related to openness to counter-narratives and quality of recall of information related to in-tergroup conflict.In the present study we used five items from the FENCE scale (α = .65;sample item: "Having many opinions about the conflict weakens us in the face of our enemies").
Collective angst -We used three items (α = .69)from the scale developed by Wohl and Branscombe (2009).
A sample item is: "I feel anxious about the future of our state and society".
A sample item is "No matter what a person's political beliefs are, they are entitled to the same legal rights as anyone else".

Factorial Validity and Reliability of the SCO Scale
To test the factorial validity of the scale, participants were randomly divided into two groups.The first group was used to derive a preliminary version of the scale by means of exploratory factor analysis (EFA), and the second group was used to confirm this version of the scale using confirmatory factor analysis (CFA).A first EFA was thus conducted with the 31 items in a random sample of 250 participants using maximum likelihood and oblimin rotation.
In a simulation study by MacCallum, Widaman, Zhang, and Hong (1999), a sample size of 200 yielded convergent and admissible solutions in EFA in at least 95% of the cases where the item to factor ratio was 10:3, and in 100% of the cases where the ratio was 20:3.Hence, our sample size of 250 had a very high likelihood of yielding an admissible solution for an item to factor ratio greater than 10:3 (or more than 3 items per factor).
The EFA yielded four factors accounting for 53% of the variance.However, the fourth factor included only one item, which cross-loaded on the first factor as well.Hence, we decided to discard this item.Next, we proceeded to eliminate items with cross-loadings, as well as those with weak loadings below .30(Tabachnick & Fidell, 1989).This left only two items in the third factor, which had originally included three items.The common element of the two remaining items seemed to be a similar grammatical structure rather than a meaningful theoretical construct.
Therefore, we decided to eliminate all items loading on the third factor.At the end of the elimination process, we were left with 14 items loading on two factors.In the first factor, labeled "self-censorship", we retained six items referring to a readiness to conceal valid information that portrays the group negatively and contradicts one's beliefs.
In the second factor, labeled "disclosure", we retained eight items referring to a willingness to disclose valid information about problematic group conduct.
A second EFA (with maximum likelihood and oblimin rotation) was then conducted with the 14 items that remained after the exclusion process based on the results of the original EFA.Results revealed a two-factor solution with eigenvalues of 5.62 and 1.39 representing 40.11% and 9.95% of the variance, respectively.The oblimin factor rotation revealed that all the items referring to withholding information loaded strongly on the "self-censorship" factor, and all the items referring to disclosure of information loaded strongly on the "disclosure" factor (all loadings > .50).No cross-loadings greater than .30were observed.Table 2 shows descriptive statistics of the retained items and the results of the second EFA.The results of the EFA appear to suggest the presence of two factors in the SCO scale.However, previous research has shown that factor loadings of this sort may be an artifact of item-wording (Biderman, Nguyen, Cunningham, & Ghorbani, 2011;Marsh, Scalas, & Nagengast, 2010).In order to test whether the two factors obtained in the EFA reflect different measurement methods or substantive content differences, we conducted confirmatory factor analyses (CFA) with AMOS 23 software to model a multi-trait multi-method matrix (Bagozzi & Yi, 1990) using the second random sample of 249 participants.We tested the fit of four alternative models: (1) a model with all items loading on one "self-censorship" factor; (2) a model with one substantive "self-censorship" factor and reversescored items also loading on an uncorrelated method factor; (3) a model with one substantive "self-censorship" factor and two method factors (reverse-scored items loading on one method factor and straightforwardly worded items on another)  3. Note.CFI = Comparative Fit Index; IFI = Incremental Fit Index; NFI = Normative Fit Index; RMSEA = Root Mean Square Error of Approximation.
The one factor model had poor fit to the data, while the three remaining models had good fit.However, in Models 2 and 3 many of the items had weak loadings on the supposed substantive factor, and none of the loadings on this factor were significant.In Model 2, the factor loadings of the reversed scored items on the supposed substantive factor ranged between -.21 and .01(all ps > .90).In Model 3, the factor loadings of all items but one on the supposed substantive factor ranged between -.15 and .20 (all ps > .90).Conversely, in Models 2 and 3 all the loadings on the supposed method factor(s) were strong and significant.In Model 2, the factor loadings of the reversed scored items on the supposed method factor ranged between .50 and .69(all ps > .001).In Model 3, the factor loadings of the reversed scored items on their respective method factor ranged between .53 and .73(all ps > .001),and the factor loadings of the straightforwardly worded items on their respective factor ranged between .51 and .75(all ps > .001).Adopting Model 2 or 3 would suggest that some or all of the items in the scale are assessing primarily method and little substance.If this were the case, there would be little point in assessing these items and we would not expect them to correlate systematically with any other measures or with expected outcomes of SCO.
A more likely interpretation of these results is that the supposed "method" factors in Models 2 and 3, on which items loaded strongly, are in fact content factors.Indeed, Model 4 with two substantive content factors had good fit to the data, and in this model, all the item loadings on their respective factors were significant and greater than .50.Hence, it appears that a model with of two distinct substantive factors ("self-censorship" and "disclosure") represents the most reasonable interpretation of the findings.Model 4 is presented in Figure 1.Note.Coefficients correspond to the arrow on their right and item numbers correspond to Table 1.All coefficients were significant (p < .05).
Inter-item correlations can be found in Appendix 1 and full CFA results in Appendix 4. Finally, we tested the internal consistency of the two factors, which was found to be high for both (self-censorship: α = .83,disclosure: α = .86).

Correlates of SCO
Table 4 displays the zero-order correlations between the SCO factors and other measures.The two SCO factors were moderately and negatively correlated with each other.Moreover, consistent with our expectations, the selfcensorship factor was related positively but weakly to WTSC, supporting our contention that the two measures are distinct.In addition, this factor was related strongly and positively to blind patriotism, conformity and RWA.
The self-censorship factor was also related positively, though less strongly, to FENCE, identification as Jewish and identification with the group to which one is most attached.Finally, the self-censorship factor was negatively related to support for abstract democratic values, constructive patriotism, support for universalistic values and argumentativeness.The findings regarding the disclosure factor are almost a mirror image of those regarding the self-censorship factor.Disclosure was related weakly and negatively to WTSC, supporting the distinction between the scales.In addition, disclosure was related positively to support for abstract democratic values, constructive patriotism and support for universalistic values, and marginally related to argumentativeness.Disclosure was negatively and moderately related to conformity, RWA, blind patriotism and FENCE, and negatively and weakly related to identification with the group to which one is most attached.Contrary to expectations, neither factor was related to general fear, fear of negative evaluation, or collective angst.
Given that the different measures were also highly inter-correlated, we wanted to test which of them would remain a significant predictor of the SCO factors after controlling for all others.Hence, we regressed each factor on the other measures (entered simultaneously).The results of the multiple regressions are shown in Table 5. Blind patriotism and conformity significantly predicted both factors, and were positively related to self-censorship and negatively related to disclosure.In addition, RWA and WTSC were significantly and positively related to selfcensorship, but not to disclosure.Conversely, constructive patriotism, support for universalistic values and support for abstract democratic values were significantly and positively related to disclosure, but not to self-censorship.
None of the other measures significantly predicted the SCO factors.

SCO and Attitudes Toward Disseminating Specific Information About Ingroup and Outgroup Behaviors in the Context of an Intergroup Conflict
Next we sought to test whether the SCO factors would predict the willingness to reveal specific narrative-consistent and narrative-contradicting information about an intergroup conflict even after controlling for other measures.In order to do so, we regressed the measures of appropriateness of narrative-consistent and narrative-contradicting images on the SCO factors as well as all the other measures (entered simultaneously).The regression model for narrative-contradicting images was significant (R 2 = .21,p < .001)and the disclosure factor of the SCO scale significantly predicted the rated appropriateness of these images for the EU website above and beyond the other measures (B = .15,SE = .06,β = .12,p = .014).The self-censorship factor was not a significant predictor in this model.The regression model for narrative-consistent images was significant too, but the predictors accounted only for a small percentage of the variance (R 2 = .05,p < .001).This is possibly due to a ceiling effect, as most participants agreed that the narrative-consistent images were appropriate (M = 4.86, SD = 0.90).Neither factor of the SCO scale was a significant predictor of the appropriateness of narrative-consistent images.The complete results of the regression are presented in Appendix 2.

Discussion
The findings of Stage 1 demonstrated that the SCO scale is best conceptualized as a two-factor solution that consists of support for self-censorship and support for disclosure of information.Stage 1 also explored the nomological network of the two factors by examining their correlates.As predicted, the SCO factors were only weakly related to WTSC, which affirms the distinction between our conception of self-censorship and the conception previously suggested by Hayes et al. (2005aHayes et al. ( , 2005b)).
The existence of two factors in the SCO scale suggests that self-censorship is a complex phenomenon involving more than a single axis on which one can be low or high.This raises the possibility that active concealment is not necessarily the direct opposite of active disclosure of information.The opposite to both of these options may be to remain passive in face of the information.In other words, when individuals receive information that they believe to be valid but potentially threatening to their group, high tendency to self-censor could lead them to actively seek to conceal the information, while high tendency to disclose could lead them to actively disseminate the information.
When neither tendency is particularly high, individuals may simply do nothing with the information.It is worth noting that the two-factor structure was obtained in a sample that responded to all 31 prototype items.To strengthen our confidence in this structure, future studies could attempt to replicate it in a sample that is exposed only to the selected subset of 14 items.
In keeping with our expectations, scores on the self-censorship factor were positively related to blind patriotism, conformity, and RWA.This supports the proposition that a major motivation that drives self-censorship is the desire to protect the ingroup's positive image and to uphold and protect group norms.Conversely, scores on the disclosure factor were positively related to constructive patriotism and to universalistic and democratic values, and negatively related to blind patriotism and conformity.This suggests that the willingness to disseminate information that might reflect negatively on one's group could be driven by a motivation to change the group for the better and to protect the rights and welfare of all persons.
Interestingly, neither factor of the SCO scale correlated with fear or with angst.It is possible however, that fear or angst might lead to self-censorship in specific situations in which the negative consequences of revealing information are clearly evident.This direction could be explored in future research.
Overall, the results demonstrate that the SCO scale with its two factors is reliable and has good convergent and discriminant validity.There are also some preliminary indications for the scale's predictive validity in the context of an intergroup conflict.Individuals who reported support for disclosure of information also supported presentation of images on a website that contradict the dominant narrative of the conflict and present the ingroup negatively.
In sum, the results of Stage 1 support our conceptualization of self-censorship as reflected in the two factors of the SCO scale and shed light on the nomological networks of the factors.
the Stage 2 measures of readiness to withhold or disseminate information about the Kafer Kassem killing on the SCO factors as well as all other measures assessed at Stage 1 (entered simultaneously).The overall regression models for both criteria were significant (withhold information: R 2 = .29,p < .001,disseminate information: R 2 = .24, p < .001).More importantly, the self-censorship factor of the SCO scale significantly predicted the readiness to withhold information about the Kafer Kassem killing above and beyond all the other Stage 1 measures (B = .20,SE = .09,β = .12,p = .036).The disclosure factor of the SCO scale did not significantly predict this outcome (B = -.14, SE = .11,β = -.07,p = .208).However, the disclosure factor significantly predicted the readiness to disseminate information about the Kafer Kassem killing above and beyond all other measures (B = .33,SE = .11,β = .17,p = .003),whereas the self-censorship factor did not significantly predict this outcome (B = -.02,SE = .10, β = -.01,p = .811).The complete results of the regression are presented in Appendix 3.

Discussion
In line with our expectations, participants' SCO factor scores at Stage 1 were robustly correlated with their SCO factor scores at Stage 2. Furthermore, Stage 2 demonstrated the predictive validity of the SCO factors.In keeping with the theoretical definition of self-censorship, we found that the higher one's self-censorship score, the more one is ready to withhold information that impinges negatively on the group's image in the context of an intergroup conflict five months later.In contrast, the higher one's disclosure score, the more one is willing to reveal information that portrays the ingroup negatively in the context of a conflict five months later.The findings showing that each of the two factors in the SCO scale predicted different conflict-related outcomes at Stage 2 are consistent with the interpretation of the two factors as representing distinct dimensions rather than polar opposites.In general, the results validate the contention that support for self-censorship or disclosure of information is a stable individual characteristic with predictive validity for future outcomes related to the readiness to withhold or disseminate information in the context of an intractable violent conflict.

General Discussion
The present research sought to develop a new instrument to assess individuals' orientation toward self-censorship of information, examine this construct in relation to relevant constructs and demonstrate its predictive utility, particularly in the context of a violent intractable conflict.Our measure focused on self-censorship or disclosure of information that is considered valid and that may have implications for one's ingroup.Factor analyses revealed that individuals' orientations toward restricting or expanding the dissemination of such information vary along two distinct but related dimensions.The first dimension, labeled "self-censorship", reflects the tendency to conceal information that is seen as threatening.The second dimension, labeled "disclosure", reflects the tendency to disseminate critical information.While the two dimensions appear to be polar opposites, the two-factor solution that emerged both in exploratory and in confirmatory factor analysis suggests that this is not the case.The emergence of two factors suggests the possibility that both dimensions reflect taking some kind of action regarding the information that one possesses, whether the action aims to conceal the information or reveal it.The low end of each dimension may reflect a preference to remain passive and not take any action to reveal or conceal the information.
Interestingly, Timming and Johnstone (2015) make a similar observation regarding silence and voice in organizational contexts: …it remains unclear whether silence should be viewed as the opposite of voice…On the one hand, it is possible that employees make a simple choice between expressing their views (voice) or keeping quiet (silence)… On the other hand, while voice can be viewed as a deliberate choice, silence might not be a conscious decision, but the result of more general feelings of disengagement, psychological withdrawal or simply having nothing to say.The absence of intentional silence, defined as the deliberate withholding of information, does not mean the presence of voice behavior.(p.157) Our findings also confirmed that our measure of SCO is distinct from other measures of similar constructs.Importantly, the findings indicate that the two SCO factors had only low correlations with the WTSC measure developed by Hayes et al. (2005aHayes et al. ( , 2005b)).This affirms the distinction between our conception, which focuses on self-censorship of information, and the definition proposed by Hayes et al. (2005a), which focuses on self-censorship of opinions.It is worth noting that measures of constructs similar to self-censorship, such as silence, were also developed for organizational settings (Brinsfield, 2013;Knoll & van Dick, 2013).However, these measures ask respondents to refer specifically to work or organization settings, whereas our measure is applicable to many different settings outside of the organizational context.Because our focus in the present research was on implications of SCO in the context of large-scale intergroup conflicts, we did not include measures referring to organizational or work contexts in our research.
Our findings also indicated that the SCO factors had moderate correlations with conformity, supporting our contention that although related, self-censorship and conformity are distinct phenomena.Furthermore, we proposed that a central motivation for self-censorship of information, other than self-interest, is protection of the group to which one belongs.In keeping with this expectation, we found that the self-censorship dimension was positively related, and the disclosure dimension negatively related, to variables reflecting commitment to one's group, such as blind patriotism, and to variables reflecting conservatism and adherence to accepted ideas and norms, such as RWA.
Conversely, the self-censorship factor was negatively related, and the disclosure dimension positively related, to variables reflecting commitment to freedom of information and expression, such as universalistic and democratic values, and to variables reflecting willingness to be critical of one's group, such as constructive patriotism.
Finally, our findings affirm the predictive validity of the SCO scale dimensions.The disclosure dimension predicted both cross-sectionally and longitudinally the willingness to disseminate information that was critical of one's ingroup in the context of a conflict and portrayed it negatively.The self-censorship dimension predicted the willingness to conceal such information longitudinally (but not cross-sectionally).Both dimensions predicted their respective outcomes above and beyond all other variables assessed at Stage 1.Hence, the SCO scale may be a useful in predicting individuals' intensions to conceal or reveal information that may have negative implications for their group, particularly in the context of an intergroup conflict.Future research may extend the investigation of the scale's predictive validity by testing whether it can predict actual behavior of concealing or revealing information.
It should be noted, however, that the readiness to conceal or reveal information in any particular situation may be a function not only of one's general disposition, but also of various situational factors, including, for example, the nature of the information, the role and accountability of the person possessing the information, and the role(s) of the person(s) to whom the information might be revealed (Shahar et al., 2018).Future research may expand our understanding of self-censorship and disclosure of information by investigating the conditions under which each is more or less likely to occur (for further discussion see Sharvit, 2017).

Figure 1 .
Figure 1.Confirmatory factor analysis of the self-censorship orientation scale: A two-factor solution.

Table 2
Descriptive Statistics of Items in the SCO Scale and Results of an Exploratory Factor Analysis

Table 3
Fit Indices of Four Alternative Models of the Structure of the SCO Scale

Table 4
Bivariate Correlations Between the Measures at Stage 1

Table 5
Regression Analyses Predicting the SCO Factors by Other Measures at Stage 1

Appendix 4: Detailed Information About Confirmatory Factor Analysis Models
Item numbers correspond to Table 1 in the main text. Note: