Diane Boerma University of Amsterdam
Bachelor Thesis Social Psychology Student number: 10755071
Supervisor: Jonas Dalege Date: 02-06-2017
Word count (Abstract): 170 Word count (Thesis): 4998
Abstract
This study is the first in linking the Causal Attitude Network (CAN) model to the catastrophe model of attitudes, to explain attitude polarization. In the CAN model, attitudes are conceptualized as networks of causally connected evaluative reactions (e.g., beliefs, feelings, and behavior), while catastrophe theory deals with phenomena where abrupt change takes place. Here, we tested the prediction, derived from these two models, that interacting with an attitude object increases attitude polarization, even in the case of one’s initial attitude being neutral (instead of positive/negative). Using a sample of Dutch undergraduates (n = 32), we provided participants on five successive days with information regarding an attitude object. We expected that, as the amount of information increases, (initially neutral) attitudes toward this object would become stronger, resulting in polarization. Results showed that receiving information about an attitude object did not lead to attitude polarization. However, the present research did have some drawbacks. Therefore, results should be taken into account cautiously.
Keywords: Causal Attitude Network model, catastrophe theory, attitudes, attitude polarization.
The Influence of Information on Attitude Polarization
Today, political debates are becoming more and more polarized. For example, the strong social divergence that accompanied the last American presidential election between Donald Trump and Hillary Clinton was probably unparalleled by any other American presidential election in the last decades. A society based on Democratic values however, is built on compromise and it is therefore important to investigate causes of polarization of (political) attitudes. Based on the Causal Attitude Network (CAN; Dalege et al., 2016) model and the catastrophe model of attitudes (e.g., van der Maas, Kolstein, & van der Pligt, 2003), the prediction can be derived that interacting in any form with an attitude object increases attitude polarization. The aim of this study will be to test this prediction in an ecologically valid way.
An attitude is an overall evaluation of an attitude object, which can basically be any aspect of the (social) world: people, things, places, actions etc. (McGuire, 1985; Zanna & Rempel, 1988, as cited in Smith, Mackie, & Claypool, 2015). Attitudes differ from each other on two important dimensions. First, attitudes differ in valence (positive vs. negative). Second, attitudes differ in strength (while one may really hate the game of soccer, another one judges the game somewhat milder, Maio & Haddock, 2015). Several attributes related to attitude strength, such as certainty and accessibility, have been identified. One of those attributes, extremity, refers to the extent to which an individual’s attitude deviates from the midpoint of the favorable-unfavorable dimension (Krosnick, Boninger, Chuang, Berent, & Carnot, 1993). Attitude extremification corresponds to the notion of attitude polarization (Liu & Latané, 1998), which is discussed next.
Attitude Polarization
When attitudes become more extreme (i.e., initial positive/negative attitudes become even more positive/negative), attitude polarization is said to have occurred. As people are asked to think about a given attitude object (e.g., a person, place, issue) their attitudes towards this attitude object tend to polarize (Tesser, 1978), a phenomenon called ‘self-generated attitude change’ (Gladding, 2008). In line with this is the finding that more knowledge about a particular attitude object, results in greater polarization (e.g., Millar & Tesser, 1986, as cited in Monroe & Read, 2008). This, because the more knowledge, the more information there is to think about (Monroe & Read, 2008). The Causal Attitude Network (CAN) model
recognizes this. Moreover, this model claims that the mere interaction with an attitude object leads to stronger attitudes toward this object, resulting in attitude polarization.
CAN Model
The Causal Attitude Network (CAN) model conceptualizes attitudes as networks consisting of evaluative reactions (like beliefs, feelings, and behavior) and causal influences between these reactions. To reduce energy expenditure, attitude networks aim to have an optimized consistent attitudinal representation. In order to attain such a state, evaluative reactions of the same valence are connected through an excitatory edge, while evaluative reactions of different valence are connected through an inhibitory edge (Dalege et al., 2016).
The global connectivity of an attitude network can be seen as the strength of an attitude (Dalege et al., 2016). When all nodes (i.e., evaluative reactions) in a network are closely connected, this network is perceived as having a high global connectivity,
representing strong attitudes. The CAN model proposes that connections between evaluative reactions self-organize when the individual interacts with the attitude object and that the global connectivity of his or her attitude network increases as a result of this (Dalege et al., 2016b). So, the prediction can be derived that interacting in any form with an attitude object (e.g., thinking about or perceiving the attitude object) strengthens one’s attitude, leading to attitude polarization.
To illustrate how something like this might present itself in the real world, take a look at the following example. Consider Mary, a 22 years old American who is holding a negative attitude towards Donald Trump. His conflicting statements made Mary question Trump’s credibility, which in turn led her to judge him as dishonest and an immoral person. Now, a small network is already formed (see Figure 1). According to the CAN model, different evaluative reactions hold each other in check (Dalege et al., 2016). For Mary’s network, this means that while at first, she judged Trump as dishonest because of her judgment of Trump not being credible, now judging Trump as dishonest also serves as an indication of Trump’s credibility. Hence, none of Mary’s judgments can readily change without inflicting some change on the other judgments (Dalege et al., 2016). So, when Mary receives information about Trump that strengthens her judgment of him as being dishonest, her judgment of Trump not being credible will also strengthen to some extent and by now attitude polarization has occurred.
Figure 1. Attitude network of Mary. Nodes represent evaluative reactions, while edges represent causal influence between these reactions.
In line with the above, it is found that attitude change in weakly connected networks takes place on a continuum, while in strongly connected networks such a change occurs more in an all-or-none fashion (Dalege et al., 2016b). This finding links the CAN model to the catastrophe model of attitudes (Latané and Nowak, 1994; van der Maas, Kolstein, & van der Pligt, 2003; Zeeman, 1976), which is the topic of the next section.
Catastrophe Model of Attitudes
Catastrophe theory deals with phenomena where change is not continuous, but abrupt (Zeeman, 1976; Flay, 1978; Latané and Nowak, 1994). The theory has been introduced to the field of psychology by Zeeman (1976), who suggested catastrophe theory to be a sound explanation of attitude change (Latané and Nowak, 1994; van der Maas et al., 2003). Zeeman (1976) hypothesized that smooth changes in independent variables are able to induce sudden, discontinuous changes in attitudes (van der Maas et al., 2003). In order to interpret this hypothesis correctly, some understanding of catastrophe theory, is required.
This catastrophe model of attitudes includes two independent variables and one dependent variable. The independent variables are known as the normal factor (α) and the splitting factor (β) (Zeeman, 1976; Flay, 1978; Latané and Nowak, 1994; van der Maas et al., 2003). The normal factor, α, is called this way because when the splitting factor, β, is at minimum, the dependent variable is only affected by the magnitude of α and increases
linearly with it. The independent variable β is called the splitting factor because an increase in its magnitude causes the bimodality of the dependent variable to become more extreme. This is called divergence (Flay, 1978; van der Maas et al., 2003).
In the case of the favorability of one’s attitude being the dependent variable,
Nowak, 1994; van der Maas et al., 2003). The information variable can be perceived as a position on a dimension of positivity and is build up out of factors such as self-interest, environment, and previous experience. It features as the normal factor; it represents the
attitude as a function of the net positivity/negativity of the information one has about the topic (Latané and Nowak, 1994). Involvement, or issue importance, is the splitting factor; the more involved individuals are in an issue, the more their attitudes are bimodal and extreme (Flay, 1978; Latané and Nowak, 1994). See Figure 2.
Figure 2. Favorability of attitude as a function of involvement. A catastrophe theory perspective.
As an example, let us revisit the case of Mary. Imagine her as someone who used to be not really interested in politics, since she rather puts her energy in partying and having fun. You could say that she considered politics not very important and so her involvement in this issue is rather low (rear plan of Figure 2). Now, her previous experience with Donald Trump is him being a successful businessman and a few friends of her actually met him once and told Mary he was very friendly. Mary’s net positivity of the information she has towards Trump is therefore considered relatively high and since one’s attitude is a direct, linear function of this positivity of information, Mary’s attitude towards Trump is also expected to be quite positive.
A few months have passed and Donald Trump is now the president of the United States of America. Mary is currently in college and partying is not her number one priority anymore. She is now focused on her future and considers politics a very important issue (front plane of Figure 2). The net valence of the information she has about Trump has changed as well; due to Trump’s conflicting and sexist statements and people in her environment who voted for Hilary Clinton it is now almost down to the point of intermediate positivity, but has not reached it yet! Her attitude is still positive, though barely. Today however, Mary has come to know of a dishonorable act Trump performed. Now, her net valence of information has reached intermediate positivity, called the region of inaccessibility when involvement is high. With Mary’s consideration of politics being an important issue and thus being highly
involved, an intermediate or neutral attitude is now not possible. A sudden attitude change takes place and Mary’s positive attitude has now become negative.
Up until now, it has been assumed that attitude polarization only occurs in the
situation of people’s initial attitude being positive or negative. So, the arising question here is the following: what happens when one’s initial attitude is neither positive nor negative, but quite neutral instead? We believe that together, the CAN model and the catastrophe model of attitudes are able the answer this question, as explained below.
Involvement/importance is also believed to be one of the underlying features of
attitude strength (Krosnick et al., 1993). Because of this, bearing in mind that according to the CAN model strong attitudes are represented by a high global connectivity, we could replace involvement with connectivity of the attitudinal network as the splitting factor. In fact, since it is supposed to strengthen connectivity, interaction (with the attitude object) can be perceived as the splitting factor in the present study. In the current attitude model this means that two persons with initially almost the same neutral attitude can diverge into two (strong) opposites the more interaction with the attitude object takes place (van der Maas et al., 2003). This corresponds to the notion of polarization (Latané and Nowak, 1994; Latané, Nowak, & Liu, 1994; Tesser, 1976). This is exactly what we think is going to happen in the current study.
Present study
The current study investigates the prediction based on the CAN model that the mere interaction with an attitude object leads initial attitudes towards this particular object become stronger, resulting in attitude polarization. Because of this, we believe that providing people with neutral (instead of positive/negative) information regarding the attitude object will be sufficient to contribute to this effect.
From a catastrophe theory perspective, it is assumed that attitude polarization can even occur in the situation of an individual holding an initial neutral attitude. Therefore, attitude objects of which we expect participants to have no prior knowledge of are used. As a result, although we do expect attitude polarization to occur, no predictions about the direction of this polarization can be made; two initially proximate attitudes may bifurcate into (strong)
opposite attitudes.
Not only corresponds attitude extremification to the notion of attitude polarization (Liu & Latané, 1998). Also, extremity follows directly from high (attitudinal network) connectivity: in highly connected attitude networks it is virtually impossible that the sum score of the evaluative reactions takes a moderate value (Dalege et al., 2016). Because of this, attitude extremity features as the dependent variable in the present study.
The catastrophe model of attitudes predicts attitudes to become more extreme when importance increases. It is for this reason that we expect attitude extremity to be positively correlated with attitude importance. A measure of attitude importance is therefore included.
Finally, an exploratory analysis will be conducted, with attitude importance as the dependent variable.
Method Participants
The present study is conducted among 32 undergraduate students of the University of Amsterdam (UvA) (13 men and 19 women). Participant’s age ranged from 18 to 24 and the mean age was 20.15 (SD = 1.57). They had the possibility to participate by signing up for our project at the website of the UvA-lab (www.lab.uva.nl) or by contacting us directly via e-mail. In exchange for their participation they received 1.5 participation credits.
Materials
Materials included (news)articles about genetic modification and mortgage interest deduction1, a global measure of attitude extremity and attitude importance towards these issues, and content questions regarding the article participants just have read.
Articles. Short, neutral (news)articles regarding genetic modification and mortgage interest deduction were used in this study. Some of them were neutral in the sense of them providing as many advantages as disadvantages regarding the topic of the article, whilst
1
A pilot study has been conducted, prior to the present study, to make sure we would use attitude objects of/in which people have little prior knowledge and/or interest. It appeared that two issues (genetic modification and mortgage interest deduction) were the most suitable for our research. Among participants of this pilot study, which included undergraduate students, €20 had been raffled.
others were judged as neutral by providing information without taking a stand.
Attitude Extremity. Three items concerning the extremity of one’s attitude towards genetic modification and mortgage interest deduction were included (e.g., “How
positively/negatively do you feel about genetic modification?”). Items were based on the 2000 American National Election Studies (ANES) (Malhotra & Krosnick, 2007). These questions were answered on a continuous slider (with only the endpoints labeled with words; e.g., very negatively vs. very positively). Since extremity refers to the deviation from the midpoint, the absolute difference between the scores and the midpoint of the slider (50) was measured. A reliability analysis revealed a relatively high reliability for all items, across condition and time (Cronbach’s α ranged from .68 to .94).
Attitude Importance. Three items regarding the importance of one’s attitude, concerning genetic modification and mortgage interest deduction, were included (e.g., “How important is mortgage interest deduction for you?”). These questions were answered on a continuous slider (with only the endpoints labeled with words; e.g., very important vs. not important). A reliability analysis revealed a high reliability for all items, across condition and time (Cronbach’s α ranged from .86 to .96).
Content. Three content questions about the article that just has been read were
included. Each question was followed by three possible answers (e.g., “For at most how many years is the mortgage interest deductible; 10, 20, or 30 years?”). One question was excluded (i.e., “On the label of products in which genetic modification took place, this should be stated”), since a clearly correct answer to it was not among the possible answers given and, as a result, participant’s answers were strongly divided. A minimum score of 10 correctly
answered questions was required. Design and Procedure
This study takes the form of a 6 (time of measurement) x 2 (condition) x 2 (topic) mixed design, with time of measurement and topic being the within-subject factors and condition the between-subject factor. Participants were randomly assigned to one of the two conditions: genetic modification (n = 16) and mortgage interest deduction (n = 16).
On the first day, participants showed up at the university where they received
instructions and were asked to sign an informed consent. Next, they filled out a questionnaire regarding their attitudes towards both genetic modification and mortgage interest deduction, regardless off the condition they were in. After this, they read a short (news)article about the issue matching the condition they had been assigned to. Subsequently, to check if the
be answered. Finally, participants again filled out a questionnaire regarding their attitudes concerning the two issues.
On each of the next four days, participants received an email with a link to a Qualtrics page. Here they were asked to read another article (again regarding the issue corresponding to the condition they were in) and answer three content questions.
To motivate thorough reading of the articles, €20 was promised to be allocated to the five participants who answered the content questions best.
Results
Two participants were removed from further analysis, since they each answered seven content questions correctly, whereas the cut-off score was set at a minimum score of 10 (out of 15).
Correlation
Pearson’s correlations were conducted to assess the relation between attitude
extremity and attitude importance. Table 1 shows the correlations found. All correlations were found to be positive. This fits our hypothesis. However, most correlations turned out not to be significant. Since the correlation coefficient is an effect size in its own right, the significant value is not that important, but non-significant correlations found are indeed quite small.
Table 1
Correlations Between Attitude Extremity Scores and Attitude Importance Scores per Day for Both Topics
Note. *Correlations are significant at the 0.05 level (one-tailed). On the first day, two measurements took place (therefore called Day 1.1 and Day 1.2, respectively).
Topic Time of Measurement GMO MID
1. Day 1.1 .22 .18 2. Day 1.2 .34* .18 3. Day 2 .12 .12 4. Day 3 .07 .32* 5. Day 4 .17 .21 6. Day 5 .23 .32*
Attitude Extremity
A three-way mixed ANOVA was conducted, with condition as a between-subject factor, time of measurement and topic as within-subject factor, and extremity scores as dependent variable. Mauchly’s test indicated that the assumption of sphericity had been violated for the interaction effect between topic and time measurement, χ2
(14) = 27.36, p < .05. Therefore, Greenhouse-Geisser corrected tests are reported (ε = .73).
There was no significant effect of the interaction between topic, time of measurement, and condition, F(3.63, 101.65) = .91, p = .453, partial η2
= .03.This indicates that the interaction between topic and time of measurement was the same for participants of both conditions. This is not in line with our hypothesis.
To break down this interaction we split the data based on the condition participants were in and performed a two-way repeated measures ANOVA, with time of measurement and topic as within-subject factors.
The two-way interaction between topic and time of measurement was neither significant for the GMO condition, F(5, 70) = .45, p = .809, partial η2
= .03, nor was it for the MID condition, F(5, 70) = 2.03, p = .085, partial η2
= .13. This means that for both conditions, the effect of time of measurement on attitude extremity did not differ across topics. Since participants received information about just one topic, while the other topic just served as a control subject, this is not what we expected to find. See Figure 3 for a graphical representation.
However, concerning the MID condition this interaction came close to being significant and the effect size (partial η2
= .13) was of medium to large size. Looking at the interaction graph (Figure 3) we can see that attitude extremity regarding the topic MID is increasing over time, which is in line with our hypothesis. With respect to GMO as a topic however, we also see an increase over time, which is even greater than it is for MID. Recalling that GMO served as a control topic in this condition, we did not expect this.
Figure 3. Graphical representation of the interaction between time measurement and topic for the GMO condition and the MID condition.
To further break down the interaction, two one-way repeated measures ANOVA’s were conducted, with time measurement as the within-subject factor. One for topic being GMO and one for topic being MID. Results are reported according to condition.
Genetic Modification Condition. With regard to the GMO condition, Mauchly’s test indicated that the assumption of sphericity had been violated regarding the main effect of time measurement for both the GMO topic, χ2(14) = 24.58, p < .05 and the MID topic, χ2
(14) = 36.71, p < .001. Therefore, Greenhouse-Geisser corrected tests are reported (ε = .64 and ε = .61, respectively).
No significant effect of time measurement was found when the topic was GMO, F(3.18, 44.49) = 1.20, p = .322, partial η2
= .08, which indicates that the attitude extremity of the attitude towards GMO did not differ over time. This is not in line with our hypothesis.
No significant effect of time measurement was found when the topic was MID, F(3.06, 42.79) = 1.71, p = .178, partial η2
= .11, which indicates that the attitude extremity of the attitude towards MID did not differ over time. As MID served as the control subject in this condition, we did not expect to find an effect of time.
Mortgage Interest Condition. When we take a look at the MID condition, Mauchly’s test indicated that the assumption of sphericity had been violated regarding the main effect of time measurement for the GMO topic, χ2
(14) = 33.25, p < .05. Therefore, Greenhouse-Geisser corrected tests are reported (ε = .49)
No significant effect of time measurement was found when the topic was GMO, F(2.46, 34.40) = 1.71, p = .190, partial η2
the attitude towards MID did not differ over time. Since GMO featured as the control subject in this condition, we did not expect to find an effect of time.
No significant effect of time measurement was found when the topic was MID, F(5, 70) = 1.78, p = .127, partial η2
= .11, which indicates that the attitude extremity of the attitude towards MID did not differ over time. This is not in line with our hypothesis. Attitude Importance (exploratory analysis)
A three-way mixed ANOVA was conducted, with condition as a between-subject factor, time of measurement and topic as within-subject factors, and attitude importance as dependent variable. Mauchly’s test turned out to be significant for the interaction effect between topic and time of measurement, χ2
(14) = 46.65, p < .001, indicating a violation of the assumption of sphericity. Therefore, degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity (ε = .54).
There was no significant effect of the interaction between topic, time of measurement, and condition, F(2.72, 76.08) = 1.63, p = .193, partial η2
= .06. This indicates that the interaction between topic and time of measurement was the same for participants of both conditions.
To break down this interaction we split the data based on the condition participants were in and performed a two-way repeated measures ANOVA, with time of measurement and topic as within-subject variables, and attitude importance as dependent variable.
Mauchly’s test indicated that the assumption of sphericity had been violated for the interaction effect between topic and time measurement, for both the GMO condition, χ2
(14) = 39.48, p < .001 and the MID condition, χ2
(14) = 35.50, p < .001. Therefore, Greenhouse-Geisser corrected tests are reported (ε = .57 and (ε = .46, respectively).
The two-way interaction between topic and time of measurement was neither significant for the GMO condition, F(2.87, 40.19) = 1.13, p = .346, partial η2
= .08, nor was it for the MID condition, F(2.30, 32.15) = 1.30, p = .290, partial η2
= .09. This means that for both conditions, the effect of time of measurement on attitude importance did not differ across topics. Since participants received information about just one topic and the other topic just served as a control subject, this is not what we expected to find. See Figure 4 for a graphical depiction.
Figure 4. Graphical representation of the interaction between time measurement and topic, for the GMO condition and the MID condition.
To further break down the interaction, two one-way repeated measures ANOVA’s were conducted, with time measurement as the within-subject factor. One for topic being GMO and one for topic being MID. Results are reported according to condition.
Genetic Modification Condition. For the GMO condition, Mauchly’s test revealed a violation of the assumption of sphericity regarding the main effect of time of measurement for the MID topic, χ2
(14) = 40.22, p < .001. Therefore, Greenhouse-Geisser corrected tests are reported (ε = .38).
No significant effect of time measurement was found when the topic was GMO, F(5, 70) = 1.77, p = .130, partial η2
= .11, which indicates that the importance of the attitude towards GMO did not differ over time.
No significant effect of time measurement was found when the topic was MID, F(1.88, 26.33) = .89, p = .417, partial η2
= .06, which indicates that the importance of the attitude towards MID did not differ over time. Since MID served as the control subject in this condition, we did not expect to find an effect of time.
Mortgage Interest Deduction Condition. For the MID condition, Mauchly’s test revealed a violation of the assumption of sphericity regarding the main effect of time of measurement for both the GMO topic, χ2(14) = 45.28, p < .001 and the MID topic, χ2
(14) = 24.75, p < .05. Therefore, Greenhouse-Geisser corrected tests are reported (ε = .46 and ε = .50, respectively).
No significant effect of time measurement was found when the topic was GMO, F(2.30, 32.13) = .73, p = .508, partial η2
attitude towards MID did not differ over time. Since GMO served as the control subject in this condition, we did not expect to find an effect of time.
No significant effect of time measurement was found when the topic was MID, F(2.52, 35.25) = 2.09, p = .129, partial η2
= .13, which indicates that the importance of the attitude towards MID did not differ over time.
Discussion
In the present study, we examined the influence of information on attitude
polarization. Based on the CAN model and catastrophe theory, we hypothesized that as the amount of information about an attitude object increases, one’s initial neutral attitude toward this object would strengthen, resulting in attitude polarization. This hypothesis was not supported. From a catastrophe theory perspective, we expected attitude extremity to correlate positively with attitude importance. This hypothesis was supported, although not strongly. Results from exploratory analyses regarding attitude importance turned out to be quite similar to those regarding attitude extremity.
Limitations of the Present Study and Directions for Future Research
A lack of sufficient participants in our sample may have been the underlying cause of the unexpected results found. When a sample is quite small, the power to find a significant effect is fairly low. Therefore, it is possible that in fact an effect of receiving information on attitude polarization does exist, but we simply did not find it due to a small sample. Indeed, the number of participants that took part was not to our satisfaction. Due to a lack of entry’s and time constraints however, a larger number could not be achieved. Future research should make use of a larger sample.
Another limitation is that the two topics seem to differ from each other in the sense that attitude scores on the first time of measurement were higher for the GMO topic compared to the MID topic (see Figure 3 and Figure 4). Different attitude scores across the topics at the beginning of the study can be expected to yield different attitude scores at the end of the study. However, no polarization was found for either topic, so it does not seem to have been of much influence in the present study. Future research should make sure though, that baseline scores are similar across topics.
Furthermore, it is possible that merely providing participants with information may not be enough to induce polarization. The CAN model proposes that for polarization to occur, interaction with an attitude object is required. Maybe simply providing people with
information about a particular attitude object does not correspond to the concept of interaction. Maybe the ‘presence’ of this information limits participants’ imagination and
leads to a reduce in thought-induced polarization. To be sure, a condition in which
participants interact more intensely with the material by, for example, listing their thoughts about the attitude object, could be included in a future study.
Also, the articles we used in the study may have been of influence. Each day
participants read another article (regarding the same topic though). However, the articles used did not differ much with respect to their content, so participants did not get to read much ‘new’ information. So, the actual amount of information did not increase as much as intended. Maybe interaction did not actually increase over time as a result of this, which could be why participants’ attitudes did not strengthen enough to result in polarization. To be sure, it is recommended that articles with a more distinguishable content are used in future research. Moreover, it is of course possible that receiving neutral information is in fact not sufficient to induce attitude polarization. Therefore, a condition with participants receiving
positive/negative information could be included as well. Conclusions
The present study was the first in linking the CAN model to the catastrophe theory in order to explain attitude polarization. Albeit the results did not confirm our main hypothesis, this does not inevitably mean that we should toss the theory of the CAN model and/or catastrophe theory as an explanation of attitude polarization aside.
First of all, although not strongly, attitude extremity and attitude importance were found to be positively related. Second, the exploratory analysis of attitude importance revealed results similar to those of the main analysis of attitude extremity. Together, these findings imply that catastrophe theory is indeed a sound explanation of attitude polarization.
Our results did not provide support for the combination of the CAN model and the catastrophe theory as predictors of attitude polarization. However, as a considerable number of limitations were found, we cannot conclude that an explanation of attitude polarization based on these models is out of the question. For the moment though, we do have to conclude that in this study receiving information about an attitude object did not lead to attitude
References
Dalege, J., Borsboom, D., van Harreveld, F., van den Berg, H., Conner, M., & van der Maas, H. L. J., (2016). Toward a formalized account of attitudes: The Causal Attitude Network (CAN) model. Psychological Review, 123, 2-22.
Dalege, J., Borsboom, D., van Harreveld, F., van der Maas, H. L. J., (2016b). A network perspective on political attitudes: Testing the connectivity hypothesis. Manuscript submitted for publication.
Flay, B. R. (1978). Catastrophe theory in social psychology: Some applications to attitudes and social behavior. Behavioral Science, 23, 335-350.
Gladding, R. (2008). Structure on attitude polarization. The Osprey Journal of Ideas and Inquiry, 137.
http://digitalcommons.unf.edu/ojii_volumes/137
Krosnick, J. A., & Petty, R. E. (1995). Attitude Strength: An Overview. In R. E. Petty & J. A. Krosnick (Eds.), Attitude Strength: Antecedents and consequences (pp. 1-24).
Hillsdale, NJ: Erlbaum.
Krosnick, J. A., Boninger, D. S., Chuang, Y. C., Berent, M. K., & Carnot, C. G. (1993). Attitude strength: One construct or many related constructs? Journal of Personality and Social Psychology, 65, 1132-1151.
Latané, B., & Nowak, A. (1994). Attitudes as catastrophes: From dimensions to categories with increasing involvement. In R.R. Vallacher & A. Nowak (Eds.), Dynamical systems in social psychology) pp. 219-249). San Diego, CA: Academic Press.
Latané, B., Nowak, A., Liu, J. H. (1994). Measuring emergent social phenomena: Dynamism, polarization, and clustering as order parameters of social systems. Behavioral Science, 39, 1-24.
Liu, J. H., & Latané, B. (1998). Extremitization of attitudes: Does thought- and discussion-induced polarization cumulate? Basic and Applied Social Psychology, 20, 103-110. Maio, G. R., & Haddock, G. (2015). The Psychology of Attitudes and Attitude Change.
London: Sage.
Malhotra, N., & Krosnick, J. A. (2007). The effect of survey mode and sampling on
inferences about political attitudes and behavior: Comparing the 2000 and 2004 ANES to Internet surveys with nonprobability samples. Political Analysis, 15, 286-323.
Monroe, B. M., & Read, S. J. (2008). A general connectionist model of attitude structure and change: The ACS (Attitudes as Constraint Satisfaction) model. Psychological
Review, 115, 733-759.
Smith, E. R., Mackie, D. M., & Claypool, H. M. (2015). Social Psychology. New York: NY: Psychology Press.
Tesser, A. (1976). Attitude polarization as a function of thought and reality constraints. Journal of Research in Personality, 10, 183-194.
Tesser, A. (1978). Self-generated attitude change. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 11, pp. 289-338). San Diego, CA: Academic Press.
van der Maas, H. L. J., Kolstein R., & van der Pligt, J. (2003). Sudden transitions in attitudes. Sociological Methods and Research, 2, 125-152.
