Fear appeals in health communication.

(1)

Fear appeals in health communication.

What is the role of nationality, ethnicity and cultural

orientation in receivers’ responses to a fear appeal

mes-sage? An empirical study in the Netherlands, China and

South Africa.

15th of June, 2017 University of Groningen Faculty of Arts

Communication and Information Sciences Master thesis

Geke van der Kroef Student 2146924

g.van.der.kroef.1@student.rug.nl Supervisor and first assessor Prof. dr. C.J.M. Jansen

University of Groningen, the Netherlands Second assessor

Prof. dr. L.G. de Stadler

(2)

This study investigated the influence of nationality, ethnicity and cultural orientation on reactions to fear appeal messages (narrative fear appeals and short textual and partly visual fear appeals). 52 Dutch, 50 Chinese and 166 South African participants first read a self-targeted narrative fear appeal which focused on a girl suffering from the STD chlamydia, or participants read a family-targeted nar-rative fear appeal which focused on the misery of her parents. After this, participants were exposed to two short textual and partly visual fear appeals that show the negative effects of smoking. One of these short textual and partly visual fear appeals is rather implicit and shows a curved cigarette, the other short textual and partly visual fear appeal is even more explicit and shows a man with a throat tumour. Cultural orientation was measured using three instruments: the CVSCALE (Yoo, Donthu, & Lenar-towicz, 2011) and the VSM (Hofstede, 1994) were used to measure six cultural dimensions as intro-duced by Hofstede (power distance; uncertainty avoidance; individualism versus collectivism; long versus short term orientation; masculinity versus femininity; indulgence versus restraint), and the Familism Scale (Steidel & Contreras, 2003) was used for measuring familism. Participants were asked to read the self-targeted or the family-targeted narrative, were then exposed to either the implicit fear appeal first and then the explicit fear appeal or vice versa and after that were asked to fill out a ques-tionnaire. The results showed that nationality and cultural orientation were only partly related. Fur-thermore, it was found that both nationality, ethnicity and cultural orientation hardly had any influence on the effects of the different fear appeals that were used. Based on these outcomes, the generalisabil-ity of claims from earlies studies that nationalgeneralisabil-ity, ethnicgeneralisabil-ity and cultural orientation should be taken into account when developing a fear appeal message should be questioned.

(3)

1. This study

1.1 Background of the study

Global health matters. Health matters since health of the individual and that of the community contrib-utes to a socially and professionally more productive life (Harman, 2012). Additionally, health matters to individuals because they see being healthy as one of the main determinants of happiness (Gerdtham & Johannesson, 2001; Senik, 2014) and certainly, being healthy equates to a better quality of life (Lichtenberg, 2005; Philipson et al., 2012). On the other hand, health matters because health care is expensive and can put the consumer in financial distress (Christensen, Bohmer, & Kenagy, 2000). For instance, more than 150 million people per day suffer financially due to out-of-pocket expenditure on health services and these costs are huge for individuals as well as for governments (World Health Or-ganization, 2016). Between 2010 and 2012, there was a growth rate of 70% in health care costs for the US government (Powers, 2015) and research conducted in ten countries showed a rise in government health care expenditures between 1970 and 2002 (Hagist & Kotlikoff, 2005).

Promoting a healthy lifestyle can help with achieving these positive living conditions and can, for instance, prevent many cases of the estimated fatal and neonatal deaths caused by sexually trans-mitted diseases, as well as reduce the 6 million deaths a year caused by smoking (Koplan & Eriksen, 2015). There are various ways in which people can be informed about a healthy lifestyle or about the adverse effects of unhealthy behaviour: by means of cryptic messages, narratives or fear appeals (Jan-sen, 2017). This study focused on fear appeal messages.

1.2 What is a fear appeal message?

Fear appeals are messages that try to scare people into healthier behaviour. Fear appeals describe and/or visualize the negative effects of a behaviour to convince people that they should act differently in the future (Witte, 1992a). Harmful behaviours that are being addressed in fear appeals are for in-stance smoking, drinking and having unsafe sex (Hastings, Stead, & Webb, 2004; Smith & Stutts, 2003; Timmers & van der Wijst, 2007). Figure 1 and Figure 2 show fear appeal messages that try to stop people from smoking. These are pictures from cigarette packages and show what can happen if the recommended behaviour is not executed.

Figure 1 Implicit short textual and partly visual fear appeal. Figure 2 Explicit short textual and partly visual fear appeal.

(6)

the recipients the same fear appeal message may evoke different reactions in different people, which may affect subsequent results (Witte & Morrison, 2000; Ruiter, Verplanken, De Cremer, & Kok, 2004).

Figure 3 The Extended Parallel Process Model (EPPM) from Witte (1992a; 1998)

1.3 Fear appeals and culture

Several authors claim that it is important to take individual differences into account when presenting receivers with such a fear appeal message (see for instance, Chung & Ahn, 2013; Lee & Park, 2012; Murray-Johnson et al., 2001; Terblanche-Smit & Terblanche, 2011). Most of these authors investigat-ed the effects of messages focusinvestigat-ed on ‘self’ or on ‘family’ (Murray-Johnson et al., 2001; Jansen, van Baal, & Bouwmans, 2007; Jansen & Verstappen, 2014). For instance, Murray-Johnson et al. (2001) carried out two experiments. The first experiment compared Afro-Americans and Mexican immigrants from the US with each other. The effects of two fear appeals were measured, both about a young girl with AIDS who died from the consequences of this disease. One fear appeal focused on the conse-quences of AIDS for the young girl herself, the other one described the conseconse-quences for her family.

In this first experiment, Murray-Johnson et al. (2001) assumed that the participants varied in cultural orientation. To the researchers’ regret, however, this assumption was not tested which led them to carry out a second experiment. This second experiment had participants from the US and Tai-wan, and used the same fear appeals, with a threat focused on either the individual or the family. This time, Murray-Johnson et al. (2001) measured the cultural orientation of the participants using the INDCOL-scale (Hui, 1988). In doing this, they expected that the participants from the US would be mostly individualistic, while the Taiwanese participants were expected to be mostly collectivistic. To the researchers’ surprise, however, this was the other way round. According to Murray-Johnson et al. (2001) these results could be caused by shortcomings in the INDCOL-scale. The authors therefore plead for using more advanced measurement methods for cultural orientation in future studies.

(7)

different results on the cultural orientation of their groups of participants than were to be expected. The questionnaire for measuring horizontal and vertical dimensions of individualism and collectivism proved to generate unreliable outcomes and therefore could not be used in further analyses.

Furthermore, a manipulation check showed that, according to the participants, the texts being used did not significantly differ from each other in the direction of the threat (self or family). In their discussion section, Jansen et al. (2007) advocate for more cross-cultural research into fear appeals, for texts that are better designed and for more advanced measurement instruments for cultural orientation. In Jansen and Verstappen (2014), these suggestions were followed. They first of all created new fear appeals. Again, two versions were made, one focusing on the suffering of an individual (self-targeted fear appeal) and one on the suffering of a family (family-targeted fear appeal). Instead of AIDS, the subject was now chlamydia and in these new versions the girl did not die from the consequences of this sexual transmitted disease. The participants were from the Netherlands and from Spain. Jansen and Verstappen (2014) used two measurement scales for cultural orientation that were found to be reliable and valid in earlier studies: the Auckland Individualism and Collectivism Scale (AICS) from Shulruf, Hattie and Dixon (2007) and the Familism Scale from Steidel and Contreras (2003).

The AICS from Shulruf et al. (2007) consists of the five most frequent dimensions found in Oyserman, Coon and Kemmelmeier (2002). Shulruf et al. (2007) decided to exclude familial connec-tions from their instrument because in their view the extent to which someone focuses on the family is not an attribute of individualism nor collectivism. Just as Fischer (2000), Shulruf et al. (2007) see ‘familism’ as a separate domain, defining the concept as prioritizing one’s family over oneself (Schwartz, 2007). Assuming that familial connections could be of importance in their research, Jansen and Verstappen (2014) chose to measure focus on the family with eighteen questions from the Fami-lism Scale from Steidel and Contreras (2003). In Jansen and Verstappen (2014)’s study, both meas-urement instruments proved to generate reliable results. The outcomes of the AICS and the Familism Scale confirmed that individualism, collectivism and familism are indeed different aspects of a per-son’s cultural orientation (Jansen & Verstappen, 2014, p. 362).

Although Jansen and Verstappen (2014) used new fear appeals and new measuring scales, just like Murray-Johnson et al. (2001) and Jansen et al. (2007), Jansen and Verstappen (2014) found that participants were hardly influenced by the variables of cultural orientation (Jansen & Verstappen, 2014, p. 364). Cultural orientation was not found to have an impact on the target of the threat of the fear appeals (self-targeted or family-targeted). In conclusion, in this study, the suggestion was not supported that it is important to take the nationality or cultural orientation of the recipient into account when creating a fear appeal (see for instance, Chung & Ahn, 2013). An explanation for the results found in Jansen and Verstappen (2014) could perhaps be that Spain and the Netherlands are too simi-lar despite their different rankings in the Hofstede index on individualism and collectivism. Another explanation could be that in contrast to fear appeal messages found in most health promotion cam-paigns, just as in other studies into the influence of culture in the effects of fear appeals, Jansen and Verstappen (2014) did not use any visual material supporting the threats. In their discussion section, Jansen and Verstappen (2014, p. 364) remark that including visual material could result in different effects.

1.4 Context of this study

In this study, the recommendations made by the above mentioned studies (Murray-Johnson et al., 2001; Jansen et al., 2007; Jansen & Verstappen, 2014) were implemented. Data was collected in Gro-ningen (The Netherlands), Shanghai (China) and Stellenbosch (South Africa). The possible influence of both nationality, ethnicity and cultural orientation on the effects of fear appeal messages varying in the target of the threat (self or family) in a narrative fear appeal message or in the level of explicitness (high or low) of the dangers depicted in a short textual and partly visual fear appeal message (see Fig-ure 1, hereafter called implicit and FigFig-ure 2, hereafter called explicit) was investigated.

(8)

Lalwani, Shavitt & Johnson, 2006). Murray-Johnson et al. (2001), for example, found differences in the effects of the fear appeals from members of different ethnical groups in the US (p. 344-346). Addi-tionally, Huang and Shen (2016) found significant effects of ethnicity on persuasion when investigat-ing the effects of cultural tailorinvestigat-ing on cancer communication in their meta-analysis. Specifically possi-ble differences between different ethnical groups in South Africa were studied.

Nevertheless, literature reveals that there is much criticism on the concept of national culture. The values of the cultural dimensions distinguished by Hofstede (1980; 2001; 2011) are changing over time due to communication, tourism and immigration (Kelley, MacNab & Worthley, 2006) but their scores have not been revised ever since. Additionally, many definitions of culture include individual behaviour (for example: Parsons & Shils, 1951; Smith, Dugan & Trompenaars, 1996) and state that culture is not (only) a national concept, but can include smaller societies that share the same values (Atkinson, Morten & Sue, 1998). Therefore, a few earlier studies suggest that cultural differences should be measured on an individual level instead of on a national level (see for instance, Callow & Shiffman, 2002; Hornikx & Le Pair, 2017; Singelis & Brown, 1995; Van de Vijver & Leung, 1997; Zhang & Gelb, 1996). Hornikx and Le Pair (2017), for example, mention that measuring culture on the individual level may help with finding effects of complex visual images, like the short textual and partly visual fear appeals used in this study. Following these suggestions, it was decided to not only categorize participants in this study by their nationality and their ethnicity, but also according to their cultural orientation.

Figure 4 Expectations for cultural orientation per nationality, based on Hofstede (1984; 1991) and The Chinese Culture Connection (1987)

Note: scores on the cultural dimensions assessed by Hofstede are between 0 and 100.

Based on the literature stated above, the research question of this study was as follows: What is the

role of nationality, ethnicity and cultural orientation in Dutch, Chinese and South African receivers’ responses to different fear appeal messages? To answer this research question, the following eight

hypotheses were tested:

Hypothesis 1: Nationality is related to cultural orientation. Hypothesis 2: Ethnicity is related to cultural orientation.

Hypothesis 3: Nationality influences the effects of target of the threat in a narrative fear appeal mes-sage.

Hypothesis 4: Nationality influences the effects of level of explicitness in a short textual and partly visual fear appeal message.

Hypothesis 5: Receivers’ ethnicity influences the effects of target of the threat in a narrative fear ap-peal

Hypothesis 6: Receivers’ ethnicity influences the effects of level of explicitness in a short textual and partly visual fear appeal message.

Hypothesis 7: Receivers’ cultural orientation influences the effects of target of the threat in a narrative fear appeal message.

Hypothesis 8: Receivers’ cultural orientation influences the effects of level of explicitness in a short textual and partly visual fear appeal message.

0 10 20 30 40 50 60 70 80 90 100

Power distance Uncertainty avoidance

Individualism Long term orientation

Masculinity Indulgence

(9)

2. Method

2.1 Design

In a 2 (direction of the threat: self-target versus group-target) x 3 (nationality: Dutch, Chinese or South African) between-subjects experimental design, participants were randomly divided over two condi-tions. Participants first read a self-targeted narrative fear appeal, or participants read a family-targeted narrative fear appeal. The narratives (see Appendix 1.1 and Appendix 1.2) and the subsequent corre-sponding questions (Appendix 2.1) were copied verbatim from Jansen and Verstappen (2014).

Secondly, in a within-subjects experimental design, participants were exposed to two pictures from cigarette packages that show the negative effects of smoking and are meant to scare the receiver. Both visuals were supported by a textual warning. One of these short textual and partly visual fear appeals was rather implicit and shows a picture of a curved cigarette (see Figure 1) the other picture was more explicit and shows a man with a throat tumour (see Figure 2). After having seen each fear appeal, participants were asked to fill out the relevant part of the questionnaire (see Appendix 2.3).

Cultural orientation of the participants was assessed in the questionnaire using three instru-ments (see Appendix 2.2) that were found to be reliable and valid in earlier studies (see below). Final-ly, personal questions were asked to collect personal information from the participants (see Appendix 2.4). To secure that the ordering of the questionnaire had no influence on the reactions of the partici-pants, eight different questionnaires were developed for this study (see Appendix 3).

2.2 Materials

In this study, fear appeals in a narrative form and fear appeals containing a short text and a picture were used. The narrative fear appeals, derived from Jansen and Verstappen (2014), told the story of a girl called Maria who was infected with chlamydia and now could not have children. The self-targeted version of the narrative focused on the suffering of Maria herself, using sentences as: “The difficult situation Maria is in has brought about a great deal of sadness and misery in her life.” The family-targeted version accentuated the misery of her family, using sentences as: “The difficult situation Ma-ria’s family is in has led to a great deal of sadness and misery.” The short textual and partly visual fear appeals that depict the negative consequences of smoking were derived from the European Union. The implicit fear appeal showed a picture of a curved cigarette with the corresponding text: “Smoking may reduce the blood flow and causes impotence”. The explicit fear appeal showed a picture of a man with a tumour in his throat with the corresponding text: “Smoking can cause a slow and painful death”.

In order to approach the participants in their own language, three versions of all fear appeals were created that differ in language: Dutch, Mandarin and English. Since South Africa has eleven official languages but English is being used as lingua franca, it was decided to approach South African participants in English. The narratives were originally written in Dutch and thereafter translated into English by Jansen and Verstappen (2014). Subsequently, both versions were translated from English into Mandarin, the official language of China, and back translated to make sure that the translations were correct. Since Chinese women keep their last name when married, the Chinese version uses Ma-ria’s family name to safeguard the credibility of the story. The short textual and partly visual fear ap-peals were originally written in English. Subsequently, the same procedures as for the narratives were followed and the fear appeals were translated into Dutch for the Dutch participants and into Mandarin for the Chinese participants. For the complete message versions, translated into Dutch, Mandarin and English (see Appendix 1).

2.3 Participants and procedure

(10)

2.3.1 Dutch participants

Participants from the Netherlands (N = 52; 32 women; 20 men) were students from the University of Groningen (85%), the Hanze Hogeschool, the Leiden University and Amsterdam University. These students were approached personally and via social media. Participants filled out the questionnaire via an electronic device, such as smartphones, tablets and computers. 73% of the participants were be-tween 21 and 24 years old. In total, 7 out of 52 Dutch participants were born in another country than the Netherlands. However, all participants considered the Netherlands as the country where they come from and they considered Dutch their mother tongue.

2.3.2 Chinese participants

Via snowball sampling, Chinese participants (N = 50; 29 women; 21 men) from different universities in Shanghai and surroundings were approached. Next to students, 19 young workers from the compa-ny Xindao completed the questionnaire. The Chinese participants were approached via their email address and filled out the questionnaire digitally. The majority of these (graduated) students from Chi-nese universities (48%) were between 21 and 24 years old, another 30% between 25 and 30 years old. All Chinese participants were born in China and they considered China as the country they come from. The mother tongue of all Chinese participants was Chinese (Mandarin).

2.3.3 South African participants

In total, 103 women and 63 men (N = 166; M = 20.6; SD = 1.73) from Stellenbosch University were personally approached by the researcher and they completed the questionnaire on paper. As might be expected from the inhabitants of South Africa, the participants had different mother tongues, some participants were bilingual or spoke more than two languages, and there were participants from many different countries that considered South Africa as the country that they come from (for instance: Le-sotho, Zimbabwe, Namibia). Therefore, it was decided to create two groups based on ethnicity. The majority (N = 100, white) was white and spoke Afrikaans, English or both languages. The other group of participants (N = 66, non-white) also spoke one or more of the other official languages of South Africa.

2.4 Measures

2.4.1 Manipulation checks

The manipulation check questions were copied verbatim from Jansen and Verstappen (2014). Three questions assessed how much Maria suffered from having chlamydia (According to this text, how bad

is it for Maria that she has chlamydia?; According to this text, how much does Maria suffer from hav-ing chlamydia?; Accordhav-ing to this text, how awful does Maria find it that she has chlamydia?). Three

other questions assessed how much the parents of Maria suffer from the situation of their daughter (According to this text, how bad is it for the parents of Maria that Maria has chlamydia?; According

to this text, how much do the parents of Maria suffer from Maria having chlamydia?; According to this text, how awful do the parents of Maria find it that Maria has chlamydia?). All items were

fol-lowed by a seven-point Likert scale with 1. Not at all . . . 7. A lot. Reliability was found to be satisfac-tory for both the perceived suffering of the girl (Cronbach’s α = .80) and the perceived suffering of the parents (Cronbach’s α = .95).

2.4.2 Assessing cultural orientation

According to Hofstede (2011), it is important to differentiate the personality of the individual from the culture this person comes from (see for instance, Hofstede, 2001, p. 2): “When you describe yourself,

you talk about your personality; if you want to know about your culture, you describe where you feel at home.” Therefore, data were collected using the CVSCALE (Yoo, Donthu, & Lenartowicz, 2011),

(11)

avoidance; individualism versus collectivism; long versus short term orientation; masculinity versus femininity; indulgence versus restraint).

The first five cultural dimensions of Hofstede were measured with the CVSCALE since vari-ous studies have shown that the CVSCALE is reliable and valid to use for cultural value assessment on an individual level (Prasongsukarn, 2009; Mazanec et al., 2015). The CVSCALE consists of 26 questions (one is missing in this study, for the cultural dimension long versus short term orientation), with five items for power distance (People in higher positions should make most decisions without

consulting people in lower position.; People in higher positions should not ask the opinions of people in lower positions too frequently.; People in higher positions should avoid social interaction with peo-ple in lower positions.; Peopeo-ple in lower positions should not disagree with decisions by peopeo-ple in higher positions.; People in higher positions should not delegate important tasks people in lower posi-tions.; Cronbach’s α = .73), five items for uncertainty avoidance (It is important to have instructions spelled out in detail so that I always know what I’m expected to do.; It is important to closely follow instructions and procedures.; Rules and regulations are important because they inform me of what is expected of me.; Standardized work procedures are helpful. Instructions for operations are im-portant.; Cronbach’s α = .87), six items for individualism versus collectivism (Individuals should sac-rifice self-interest for the group.; Individuals should stick with the group even through difficulties.; Group welfare is more important than individual rewards.; Group success is more important than individual success.; Individuals should only peruse their goals after considering the welfare of the group.; Group loyalty should be encouraged even if individual goals suffer.; Cronbach’s α = .82), five

items for long versus short term orientation (How important is careful management of money for you?;

How important is going on resolutely in spite of opposition for you?; How important is personal steadiness and stability for you?; I would give up today’s fun for success in the future.; I would work hard for success in the future. Cronbach’s α = .55) and four items for masculinity versus femininity (It is more important for men to have a professional career than it is for women.; Men usually solve problems with logical analysis; women usually solve problems with intuition.; Solving difficult prob-lems usually requires an active, forcible approach, which is typical for men.; There are some jobs that a man can always do better than a women.; Cronbach’s α = .80). All items were followed by a

seven-point scale with 1. Don’t agree at all . . . 7. Fully agree. In order to reach satisfactory scores for Cronbach’s α, it was decided to remove 2 questions from the long versus short term orientation varia-ble, i.e. question 43 (How important is going on resolutely in spite of opposition for you?) and ques-tion 45 (I would give up today’s fun for success in the future.). Similar Cronbach α’s were found when inspecting the Dutch, Chinese and South African participants (see Table 2).

Table 2 Cronbach α’s for the cultural orientation variables and the Familism Scale. Items within the

(12)

The sixth cultural dimension of Hofstede (2001), indulgence versus restraint, refers to the extent to which people focus on having a good time and act impulsively or think about the consequences and repress these feelings because they appreciate strict social norm. Since this dimension may be of im-portance for this study and is not included in the CVSCALE, the corresponding items from the Values Survey Module (VSM) of Hofstede (1994) were added to the questionnaire (m11: In my private life,

keeping time free for fun is important to me.; m12: In my private life, moderation (having few desired) is important to me.; m16: I am a happy person). The items were followed by a 7-point scale to remain

in line with the questions of the CVSCALE, with 1. Of very little or no importance . . . 7. Of utmost

importance. The data were converted to a 5-point scale since the VSM measures 1. Of utmost im-portance . . . 5. Of very little or no imim-portance where the values have a .66 distance from each other

instead of 1.00 (1: 3/3 = 1; 2: 5/3 = 1.66; 3: 7/3 = 2.33; 4: 9/3 = 3; 5: 11/3 = 3.66; 6: 13/3 = 4.33; 7: 15/3 = 5).

Next, the index formula form Hofstede was applied: 35 (m12 - m11) + 40 (m17 - m16) + C(ir), where m11, m12, m16 and m17 are the means of the questions. Unfortunately, m17 was not measured in this study. However, since m17 tries to measure restraint, just as question m12 (question-naire item 52), it was decided to use the m12 scores for the missing m17 scores (see Table 3).

Table 3 Means and values for the variable indulgence versus restraint.

M11 = item 51 M12 = item 52 M16 = item 54 M17 = value missing: equal to m12

IVR value

All participants M = 1.73 M = 2.54 M = 1.89 M = 2.54 54.74 Dutch participants M = 1.65 M = 2.77 M = 2.01 M = 2.77 69.3 Chinese participants M = 1.74 M = 2.89 M = 1.70 M = 2.89 87.6 South African participants M = 1.75 M = 2.35 M = 1.90 M = 2.35 40.28

The orientation towards one’s family, from here familism, was measured with the Familism Scale since familism is related to, but distinct from the cultural orientation variable collectivism (Steidel & Contreras, 2003). The Familism Scale was first introduced by Steidel and Contreras (2003) and was based on a factor analysis performed with data from 124 Latino adults who lived in the US. The scale consists of 18 items which can be measured in two different ways; either with self-referenced items (for instance: I think that aging parents should live with their relatives.) or with culture-referenced items (for instance: In the country that I come from, aging parents should live with their relatives.), all followed by a seven-point scale with 1. Don’t agree at all . . . 7. Fully agree.

(13)

Also, the self-referenced and culture-referenced questions from the Familism Scale proved to be relat-ed but not interchangeable r = .74; p < .001 (see Table 4).

Table 4 Bivariate correlations between cultural orientation variables (all participants; Dutch participants; Chinese participants; South African participants.

PDI UAI INDCOL LSTO MASFEM INDRES SRF CRF

All participants PDI 1 UAI -.04 1 INDCOL .21** .27** 1 LSTO -.13* .49** .22** 1 MASFEM .45** .07 .26** .05 1 INDRES -.04 -.17** -.28** -.21** -.12 1 SRF .02 .28** .18** .23** .13* -.11 1 CRF -.09 .29** .20** .26** .06 -.11 .74** 1 Dutch participants PDI 1 UAI .16 1 INDCOL .48** .37** 1 LSTO .12 .50** .34** 1 MASFEM .38** .60** .60** .42** 1 INDRES -.38** -.36** -.62** -.25 -.56** 1 SRF .21 .31** .23 .02 .35** -.20 1 CRF .20 .15 .27 .07 .27 -.15 .31* 1 Chinese participants PDI 1 UAI -.03 1 INDCOL .21 .54** 1 LSTO -.02 .39** .52** 1 MASFEM .41** .12 .43** .23 1 INDRES -.14 -.24 -.54** -.39** -.27 1 SRF .02 .01 .29* .23 .21 -.28* 1 CRF -.12 .07 .19 .31* .10 -.24 .66** 1

South African participants

PDI 1 UAI .03 1 INDCOL .21** .14 1 LSTO -.11 .43** .04 1 MASFEM .36** .06 .21** .08 1 INDRES .05 .03 -.06 -.03 -.02 1 SRF .13 .30** .13 .23** .26** -.01 1 CRF .06 .37** .19* .22** .33** -.03 .77** 1

Note. PDI = power distance; UAI = uncertainty avoidance; INDCOL = individualism versus collectivism; LSTO = long versus short term

orientation; MASFEM = masculinity versus femininity; INDRES = indulgence versus restraint; SRF = self-referenced familism; CRF = culture-referenced familism.

(14)

2.4.3 Categorizing participants after cultural orientation

In order to determine what the impact of cultural orientation of the participants is, the cultural orienta-tion variables were converted into dichotomous variables. When the score for the cultural orientaorienta-tion variable was higher than the median, this resulted in a high score for the dichotomous variable. When the score was lower than the median score, this resulted in a low score for the dichotomous variable (see Table 5).

Table 5 Distribution of participants over the dichotomous variables (low/high). Median Grouped into: (low or high) All partici-pants Dutch par-ticipants Chinese participants South African participants

Power distance 2.00 Low 128 13 16 99

High 140 39 34 67

Uncertainty avoidance 5.80 Low 133 31 35 67

High 135 21 15 99

Individualism 4.33 Low 123 25 26 72

High 145 27 24 94

Long term orientation 6.33 Low 126 34 35 57

High 142 18 15 109

Masculinity 3.00 Low 128 13 12 103

High 140 39 38 63

Indulgence 46.55 Low 132 24 18 90

High 136 28 32 79

Self-referenced familism 4.67 Low 139 46 23 70

High 129 6 27 96

Culture-referenced familism 4.67 Low 140 52 21 67

High 128 0 29 99

2.4.4 Dependent variables

The following EPPM variables served as dependent variables for the self-targeted and family-targeted narrative, with three items for fear (Reading this text made me afraid.; Reading this text made me

tensed.; Reading this text made me anxious.; Cronbach’s α = .92), three items for severity (I think that chlamydia is a serious health problem.; I think that chlamydia is a serious illness.; I think that chla-mydia is a disease that can have a lot of influence on a person’s life.; Cronbach’s α = .84), three items

for susceptibility (I am at risk of becoming infected with chlamydia.; It is likely that I will be infected

with chlamydia.; It is possible that I will be infected with chlamydia.; Cronbach’s α = .83), two items

for response efficacy (Condoms are an effective means to prevent chlamydia.; I think using a condom

is effective in preventing chlamydia.; Cronbach’s α = .82) and three items for message minimization

(This text was misleading.; This text was manipulative.; This text was exaggerated.; Cronbach’s α = .81). Danger control was measured with one item (I intend to change my behaviour such that I won’t

get infected with chlamydia.). Due to a difference in data collection, only African participants filled

out two questions for self-efficacy (Using a condom is easy;. My sexual partner and I are capable of

using condoms in order to prevent chlamydia.) who were measured individually. Additionally, South

African participants filed out one question for fear control (I don’t want to think about chlamydia.). The dependent variables used were all measured on a seven-point scale with 1. Don't agree at all . . . 7. Fully agree. Similar Cronbach α’s were found when inspecting the Dutch, Chinese and South Afri-can participants (see Table 6).

The two short textual and partly visual fear appeals, implicit and explicit, were measured by the same fear arousal items as the narratives (Cronbach’s α implicit = .95; explicit = .93), severity (3 items: I think that smoking is a serious health problem.; I think that smoking is a serious problem.; I

think that smoking is a problem that can have a lot of influence on a person’s life.; Cronbach’s α

im-plicit = .87; exim-plicit = .92), susceptibility (3 items: I am at risk of becoming addicted to smoking.; It is

likely that I will be addicted to smoking.; I is possible that I will be addicted to smoking.; Cronbach’s

α implicit = .96; explicit = .95) and the same message minimization items as the narratives (Cronbach’s α implicit = .93; explicit = .91). Due to a difference in data collection, only African par-ticipants filled out one question for self-efficacy (I am capable of quitting smoking.)

Response efficacy was measured differently in the implicit fear appeal (2 items: Quitting

smoking is an effective way to prevent importance and trouble with the blood circulation.; Cronbach’s

(15)

painful death.; I think that quitting smoking can prevent a slow and painful death.; Cronbach’s α .85).

The dependent variables used for the implicit and explicit fear appeals were all measured on a seven-point scale with 1. Don't agree at all . . . 7. Fully agree. Similar Cronbach α’s were found when in-specting the Dutch, Chinese and South African data (see Table 6).

Table 6 Reliability of the EPPM values for the narratives, the implicit fear appeal and the explicit fear appeal: Cronbach’s αs. Items within the

questionnaire Cronbach’s α All participants Cronbach’s α Dutch partici-pants Cronbach’s α Chinese partici-pants Cronbach’s α South African participants Fear narrative 7, 8, 10 .92 .966 .943 .897 Severity narrative 11, 12, 13 .84 .849 .844 .83 Susceptibility narrative 14, 15, 16 .833 .873 .895 .691 Response efficacy narrative 17, 18 .817 .773 .782 .837 Message minimization narrative 23, 24, 25 .805 .727 .893 .789 Fear implicit 73, 75, 76 .950 .965 .951 .947 Severity implicit 77, 78, 79 .874 .730 .862 .913 Susceptibility implicit 80, 81, 82 .961 .947 .968 .959 Response efficacy implicit 83, 84 .878 .897 .977 .847 Message minimization implicit 87, 88, 89 .928 .901 .938 .93 Fear explicit 91, 93, 94 .929 .91 .918 .928 Severity explicit 95, 96, 97 .922 .92 .862 .934 Susceptibility explicit 98, 99, 100 .951 .938 .947 .949 Response efficacy explicit 101, 102 .854 .745 .928 .849 Message minimization

explicit

105, 106, 107 .914 .93 .896 .914

Note. ‘Self-efficacy narrative’ was measured using two items for the narratives. The Cronbach’s α was not sufficient (Cronbach’s α = .53). It

(16)

3. Results

First, possible effects of nationality on the different cultural orientation variables will be discussed. Next, the possible effects of nationality and cultural orientation on message version will be addressed. The results for the narratives (self- or family-targeted) will be presented first and thereafter the results for the short textual and partly visual fear appeals will follow.

3.1 Nationality, ethnicity and cultural orientation

To test the first hypothesis, the effects of nationality on the different cultural orientation variables were inspected (see Table 7 for mean scores and SD’s). A significant main effect of nationality was found for power distance (F [2, 265] = 16.21; p < .001; partial η2

= .109). Further inspection revealed that South African participants scored lower on power distance than Dutch participants (p < .001) and Chi-nese participants (p = .001). No significant difference was found between Dutch and ChiChi-nese partici-pants.

A significant main effect was found for uncertainty avoidance (F [2, 265] = 12.20; p < .001; partial η2 = .084). Further inspection revealed that South African participants scored higher on uncer-tainty avoidance than Dutch participants (p = .001) and Chinese participants (p < .001). No significant difference was found between Dutch and Chinese participants.

No significant effect of nationality on individualism versus collectivism was found.

Furthermore, a significant main effect was found on long term orientation (F [2, 265] = 14.03; p < .001; partial η2

= .096). All three nationalities score high on this dimension (see Table 7). Howev-er, South African participants scored significantly higher on long term orientation than Dutch partici-pants (p < .001) and Chinese participartici-pants (p < .001). No significant difference was found between Dutch and Chinese participants.

Another significant main effect was found for masculinity (F [2, 265] = 25.89; p < .001; par-tial η2

= .163). Further inspection revealed that South African participants scored significantly lower on masculinity than Dutch participants (p < .001) and Chinese participants (p < .001). No significant difference was found between Dutch and Chinese participants.

Additionally, a significant main effect was found for indulgence (F [2, 265] = 5.78; p < .005; partial η2

= .042). Further inspection revealed that South African participants scored significantly low-er on indulgence than Dutch participants (p = .05) and Chinese participants (p < .005). Again, no sig-nificant difference was found between Dutch and Chinese participants.

A significant main effect was found on self-referenced familism (F [2, 265] = 13.08; p < .001; partial η2

= .12). Further inspection revealed that Dutch participants scored significantly lower on self-referenced familism than Chinese participants (p < .001) and South African participants (p < .001). No significant difference was found between Chinese and South African participants.

Lastly, a significant main effect was found on culture-referenced familism (F [2, 265] = 53.52; p < .001; partial η2 = .288). Further inspection revealed that Dutch participants scored significantly lower on culture-referenced familism than Chinese participants (p < .001) and South African partici-pants (p < .001). No significant difference was found between Chinese and South African participartici-pants.

(17)

Table 7 Effects of nationality on cultural orientation variables (minimum score 1, maximum score 7*) using a univariate analysis. Dutch

participants

Chinese participants

All SA participants N = 166 Whites N = 110 Non-Whites N = 56 Power distance M = 2.68 SD = 1.19 M = 2.48 SD = 1.12 M = 1.85 SD = .96 M = 1.87 SD = 1.04 M = 1.82 SD = .78 Uncertainty avoidance M = 5.25 SD = 1.19 M = 5.06 SD = .98 M = 5.81 SD = 1.05 M = 5.71 SD = 1.06 M = 6.01 SD = 1.02 Individualism versus collectivism M = 4.13

SD = 1.12 M = 4.16 SD = 1.24 M = 4.41 SD = 1.26 M = 4.33 SD = 1.26 M = 4.58 SD = 1.24 Long- versus short term orientation M = 5.83

SD = .91 M = 5.79 SD = .82 M = 6.33 SD = .73 M = 6.29 SD = .71 M = 6.43 SD = .78 Masculinity versus femininity M = 4.09

SD = 1.73 M = 3.85 SD = 1.46 M = 2.64 SD = 1.41 M = 2.68 SD = 1.43 M = 2.54 SD = 1.37 Indulgence versus restraint M = 69.3

SD = 110.79 M = 87.6 SD = 97.38 M = 40.28 SD = 85.14 M = 54.74 SD = 82.14 M = 11.89 SD = 84.45 Familism Scale self-referenced items M = 4.02

SD = .66 M = 4.82 SD = .77 M = 4.81 SD = .92 M = 4.73 SD = .94 M = 4.96 SD = .88 Familism Scale culture-referenced items M = 3.47

(18)

3.2 The self-targeted and family-targeted narratives

First, the results of the manipulation check for the narratives (self- or family-targeted) will be dis-cussed. Next, the effects of nationality of the participants will be presented for message version (self- or family-targeted) on the variables of the EPPM (fear; severity; susceptibility; response efficacy; self-efficacy; danger control; fear control). After that, the effects of the same dependent variables for white and non-white South African will be presented. Finally, the effects of low or high scores on the vari-ous cultural dimensions on message version will be presented.

3.2.1 Manipulation checks

Two one-way analyses of variance were performed, both with message version as the only factor: one with ‘perceived suffering of Maria’ and the other with ‘perceived suffering of the parents’ as depend-ent variable. The manipulation proved to have been successful. When combining the data for all par-ticipants, mean scores for perceived suffering of the girl were significantly higher (F [1, 266] = 61.29; p < .001; partial η2

= .022) for those who read the self-targeted version (M = 6.07; SD = 1.15) than for those who read the family-targeted version (M = 5.87; SD = 1.33). Mean scores for the perceived suf-fering of the family were significantly higher (F [1, 266] = 77.49; p < .001; partial η2 = .226) for those who read the family-targeted version (M = 5.87; SD = 1.33) than for those who read the self-targeted version (M = 3.97; SD = 2.11). When inspecting the manipulation checks data for the Dutch, Chinese and South African participants, similar effects were found (see Table 8 for mean scores and SD’s). Table 8 Mean and SD’s for Manipulation Check Variables (All Participants; Dutch Participants; Chinese participants; South African partici-pants).

Self-targeted narrative Family-targeted narrative Effect All participants

Perceived suffering of Maria

Perceived suffering of the parents

M = 6.07 SD = 1.15 M = 3.97 SD = 2.11 M = 5.7 SD = 1.29 M = 5.87 SD = 1.33 p < .05; partial η2 = .022 p < .001; partial η2 = .226 Dutch participants

M = 6.51 SD = .89 M = 2.97 SD = 1.70 M = 4.76 SD = 1.35 M = 6.18 SD = 1.41 p < .001; partial η2 = .382 p < .001; partial η2 = .523 Chinese participants

M = 6.69 SD = .45 M = 4.83 SD = 1.58 M = 5.15 SD = 1.1 M = 6.71 SD = .47 p < .001; partial η2 = .459 p < .001; partial η2 = .412 South African participants

Perceived suffering of the parents

(19)

3.2.2 Nationality, ethnicity and message version

To test the second hypothesis, univariate analyses were performed with nationality, ethnicity and mes-sage version as independent variables. The EPPM variables served as dependent variables: fear arous-al, perceived severity, perceived susceptibility, perceived response efficacy, danger control and in case of the South African participants perceived self-efficacy and fear control. The means scores and SD’s of all participants can be found in Table 9.

A significant main effect of nationality was found for severity (F [2, 262] = 3.15; p < .05; par-tial η2

= .023). Further inspection revealed that Dutch participants scored significantly lower on severi-ty than South African participants. No significant main difference of nationaliseveri-ty was found between Dutch participants and Chinese participants or between Chinese participants and South African partic-ipants. A significant main effect of message version for severity was found (F [1, 262] = 4.86; p < .03; partial η2

= .018). Participants scored higher on severity after reading the family-targeted narrative than after reading the self-targeted narrative. A significant univariate interaction effect of nationality and message version was found for severity (F [2, 262] = 6.00 p = .003; partial η2 = .044). Further inspection revealed that Chinese participants who read the family-targeted narrative scored higher on severity than those who read the self-targeted narrative. No significant differences in scores for the two narratives were found for Dutch or South African participants. No significant univariate interac-tion or main effects of ethnicity and message version were found for severity.

Second, a significant main effect of nationality for susceptibility was found (F [2, 262] = 24.56; p < .001; partial η2 = .158). Further inspection revealed that South African participants scored higher on susceptibility than the Dutch (p < .001) and Chinese participants (p < .001). No significant difference was found between Dutch and Chinese participants. A significant main effect of message version on susceptibility was found (F [1, 262] = 30.98; p < .001; partial η2

= .106). Participants scored significantly higher on susceptibility after reading the self-targeted narrative than after reading the family-targeted narrative. Next, a significant univariate interaction effect of nationality and mes-sage version was found for susceptibility (F [2, 262] = 15.25; p < .001; partial η2

= .104). Both Chi-nese (p < .01) and Dutch participants (p < .001) who read the self-targeted narrative scored higher (p < .001) than those who read the family-targeted narrative. No significant effect of nationality and mes-sage version could be reported for the South African participants. No significant univariate interaction or main effects of ethnicity and message version were found for susceptibility.

A significant main effect of message version on response efficacy was found (F [1, 262] = 4.49; p < .05; partial η2

= .017). Further inspection revealed that participants scored higher on response efficacy after reading the self-targeted narrative than after reading the family-targeted narrative. No significant univariate interaction or main effects of ethnicity and message version were found for re-sponse efficacy. No significant univariate interaction effect or significant main effect of nationality and message version could be found for response efficacy.

Lastly, a significant main effect of nationality for danger control was found (F [2, 262] = 4.42; p = .013; partial η2

= .033). Further inspection revealed that Dutch participants scored lower on danger control than South African participants. No significant differences were found between Dutch and Chinese participants or between Chinese and South African participants. A significant univariate main effect of message version was found for danger control (F [2, 262] = 8.43; p < .005; partial η2

= .031). Further inspection revealed that participants scored significantly higher on danger control after reading the family-targeted narrative than after reading the self-targeted narrative. A significant univariate interaction effect of nationality and message version was found for danger control (F [2, 262] = 3.82; p < .05; partial η2

= .028). Further inspection revealed that Dutch participants (p < .05) and Chinese participants (p < .01) scored lower on danger control after reading the self-targeted narrative than after reading the family-targeted narrative. No such effect was found for South African participants. No significant univariate interaction or main effects of ethnicity and message version were found for dan-ger control.

(20)

Table 9 Means and standard deviations of the EPPM scores for the (1) self-targeted narrative (N = 134), and the (2) family-targeted narrative (N = 134). Nar ra-tive All partici-pants Dutch participants Chinese participants

All partici-pants N=166 Whites N=110 Non-Whites N=56 Fear 1 M = 3.44 SD = 1.72 M = 3.65 SD = 1.58 M = 3.47 SD = 1.69 M = 3.36 SD =1.78 M = 3.49 SD = 1.73 M = 3.12 SD = 1.89 2 M = 3.51 SD = 1.83 M = 3.7 SD = 2.1 M = 3.74 SD = 1.93 M = 3.38 SD = 1.71 M = 3.18 SD = 1.6 M = 3.76 SD = 1.88 Severity 1 M = 5.9 SD = 1.09 M = 5.47 SD = .96 M = 5.36 SD = 1.12 M = 6.18 SD = 1.02 M = 6.15 SD = 1.11 M = 6.25 SD = .84 2 M = 5.99 SD = 1.21 M = 5.8 SD = 1.44 M = 6.35 SD = 1.26 M = 5.94 SD = 1.11 M = 5.84 SD = 1.25 M = 6.14 SD = .77 Susceptibility 1 M = 2.41 SD = 1.5 M = 3.51 SD = 1.73 M = 3.65 SD = 1.22 M = 1.72 SD = 1 M = 1.81 SD = 1.12 M = 1.55 SD = .7 2 M = 1.94 SD = 1.23 M = 1.72 SD = 1.1 M = 2.43 SD = 1.75 M = 1.86 SD = 1.04 M = 1.81 SD = 1.02 M = 1.94 SD = 1.08 Response efficacy 1 M = 5.86 SD = 1.24 M = 6.3 SD = .83 M = 5.9 SD = 1.33 M = 5.71 SD = 1.3 M = 5.64 SD = 1.16 M = 5.86 SD = 1.54 2 M = 5.46 SD = 1.5 M = 5.63 SD = 1.57 M = 5.7 SD = 1.2 M = 5.33 SD = 1.56 M = 5.33 SD = 1.63 M = 5.34 SD = 1.44 Self-efficacy 1 (Question 19, SA par-ticipants only) 1 * * * M = 6.05 SD = 1.36 M = 6.11 SD = 1.14 M = 5.93 SD = 1.74 2 * * * M = 5.68 SD = 1.41 M = 5.67 SD = 1.37 M = 5.71 SD = 1.51 Self-efficacy 2 (Question 20, SA par-ticipants only) 1 * * * M = 6.07 SD = 1.6 M = 5.91 SD = 1.73 M = 6.39 SD = 1.29 2 * * * M = 5.78 SD = 1.77 M = 5.63 SD = 1.8 M = 6.07 SD = 1.7 Danger control 1 M = 4.56 SD = 2.14 M = 3.46 SD = 1.61 M = 3.88 SD = 1.54 M = 5.10 SD = 2.26 M = 5.05 SD = 2.18 M = 5.18 SD = 2.45 2 M = 4.99 SD = 2.10 M = 4.73 SD = 2.01 M = 5.27 SD = 1.89 M = 4.99 SD = 2.19 M = 4.85 SD = 2.20 M = 5.25 SD = 2.19 Fear control 1 * * * M = 4.54 SD = 2.19 M = 4.66 SD = 2.22 M = 4.74 SD = 2.16 2 * * * M = 4.67 SD = 2.10 M = 4.29 SD = 2.14 M = 4.54 SD = 2.01 * No scores were available for this variable due to the differences in the questionnaires distributed among the three countries.

3.2.3 Cultural orientation and message version

To test the third hypothesis, univariate analyses were performed with cultural orientation (participants scoring low/high) as independent variable. The EPPM variables served as dependent variables: fear arousal, perceived severity, perceived susceptibility, perceived response efficacy, danger control, fear control and in case of the South African participants perceived self-efficacy. See Table 10 for mean scores and SD’s.

Power distance

Significant main effects of power distance (low/high) were found for fear (F [1, 264] = 11.42; p = .001; partial η2 _{= .041), severity (F [1,264] = 9.01; p < .05; partial η}2

= .033) and self-efficacy 2 (1, 162) = 6.71; p = .01; partial η2

= .040). Further inspection revealed that participants scoring high on power distance scored higher on fear than participants scoring low. Participants scoring low on power distance scored higher on severity than participants scoring high on power distance. Participants scor-ing low on power distance scored higher on danger control than participants scorscor-ing high on power distance. Significant main effects of message version were found for susceptibility (F [1, 264] = 7.88; p = .005; partial η2

= .029) and response efficacy (F [1, 264] = 5.45; p = .02; partial η2 = .020). Further inspection revealed that participants scored higher on susceptibility and response efficacy after reading the self-targeted narrative than after reading the family-targeted narrative. These main effects, howev-er, were qualified by a significant univariate interaction effect of power distance (low/high) and mes-sage version for severity (F [1, 264] = 5.57; p < .05; partial η2

(21)

Uncertainty avoidance

Significant main effects of the cultural orientation variable uncertainty avoidance (low/high) were found for severity (F [1. 264] = 27.47; p < .001; partial η2 = .094), susceptibility (F [1, 264] =6.87; p < .01; partial η2 _{= .025) and danger control (F [1, 264] = 2.09; p < .05; partial η}2

= .024). Further inspec-tion revealed that participants scoring low on uncertainty avoidance scored lower on severity than participants scoring high on uncertainty avoidance. Participants scoring low on uncertainty avoidance scored higher on susceptibility than participants scoring high on uncertainty avoidance. Participants scoring low on uncertainty avoidance scored lower on danger control than participants scoring high on uncertainty avoidance. Significant main effects of message version were found for susceptibility (F [1, 264] = 6.90; p < .01; partial η2

= .025) and response efficacy (F [1, 264] = 6.39; p = .012; partial η2 = .024). Further inspection revealed that participants scored higher on susceptibility and response effica-cy after reading the self-targeted narrative than after reading the family-targeted narrative. These main effects, however, were qualified by a significant univariate interaction effect of uncertainty avoidance (low/high) and message version for susceptibility (F [1, 264] = 6.77; p = .10; partial η2

= .025). Further inspection revealed that when scoring low on uncertainty avoidance, participants scored higher on susceptibility after reading the self-targeted narrative than when scoring high on uncertainty avoid-ance.

Individualism

Significant main effects of individualism (low/high) were found on message version for susceptibility F [1, 264] = 7.57; p = .006; partial η2

= .028) and response efficacy (F [1, 264] = 6.20; p = .013; partial η2

= .023). Further inspection revealed that participants scored higher on susceptibility and response efficacy after reading the self-targeted narrative than after reading the family-targeted narrative. No significant univariate interaction effects of individualism (low/high) and message version or main effects of individualism (low/high) were found.

Long term orientation

Significant main effects of the cultural orientation variable long term orientation (low/high) were found for severity (F [1, 264] = 31.20; p < .001; partial η2

= .106), susceptibility (F [1, 264] = 15.53; p < .001; partial η2

= .056), response efficacy (F [1, 264] = 4.84; p < .05; partial η2 = .018) and danger control (F [1, 264] = 7.31; p < .01; partial η2

= .027). Further inspection revealed that participants scor-ing low on long term orientation scored lower on severity than participants scorscor-ing high on long term orientation. Participants scoring low on long term orientation scored higher on susceptibility than par-ticipants scoring high. Parpar-ticipants scoring low on long term orientation scored lower on response efficacy than participants scoring high on long term orientation. Participants scoring low on long term orientation scored lower on danger control than participants scoring high on long term orientation. Significant main effects of message version were found for susceptibility (F [1, 264] = 8.81; p = .003; partial η2 _{= .032) and response efficacy (F [1, 264] = 7.18; p = .008; partial η}2

= .026). Further inspec-tion revealed that participants scored higher on susceptibility and response efficacy after reading the self-targeted narrative than after reading the family-targeted narrative. These main effects, however, were qualified by a significant univariate interaction effect of long term orientation (low/high) and message version on fear (F [1, 264] = 4.22; p < .05; partial η2

= .016) and susceptibility (F [1, 264] = 27.56; p < .001; partial η2

= .095). Further inspection revealed that participants scoring low on long term orientation, scored higher on fear and susceptibility after reading the self-targeted narrative than after reading the family-targeted narrative. No significant difference was found between the self- and family-targeted narratives for participants scoring high on long term orientation.

Masculinity

Significant main effects of the cultural orientation variable masculinity (low/high) were found for se-verity (F [1, 264] = 4.14; p < .05; partial η2

= .015) and susceptibility (F [1, 264] = 12.21; p = .011; partial η2

= .024). Further inspection revealed that participants scoring low on masculinity scored higher on severity than participants scoring high on masculinity. Participants scoring low on masculin-ity scored higher on susceptibilmasculin-ity than participants scoring high on masculinmasculin-ity. Significant main ef-fects of message version were found for susceptibility (F [1, 264] = 10.34; p = .001; partial η2 = .038) and response efficacy (F [1, 264] = 5.31; p = .022; partial η2

(22)

participants scored higher on susceptibility and response efficacy after reading the self-targeted narra-tive than after reading the family-targeted narranarra-tive. These main effects, however, were qualified by a significant univariate interaction effect of masculinity and message version for self-efficacy 1 (F [1, 264] = 5.05; p < .05; partial η2

= .030). Further inspection revealed that when scoring low on masculin-ity, participants scored higher on self-efficacy 1 after reading the self-targeted narrative than after reading the family-targeted narrative. No significant difference was found between the self- and fami-ly-targeted narratives for participants scoring high on masculinity.

Indulgence

Significant main effects of message version were found for susceptibility (F [1, 264] = 5.74; p = .017; partial η2 _{= .021) and response efficacy (F [1, 264] = 4.16; p = .042; partial η}2

= .016). Further inspec-tion revealed that participants scored higher on susceptibility and response efficacy after reading the self-targeted narrative than after reading the family-targeted narrative. No significant univariate inter-action effects of indulgence and message version or main effects of indulgence (low/high) were found.

Self-referenced familism

Significant univariate main effects of self-referenced familism (low/high) were found for severity (F [1, 264] = 12.83; p < .001; partial η2

= .046), susceptibility (F [1, 264] = 4.36; p < .05; partial η2 = .016) and danger control (F [1, 264] = 9.22; p < .005; partial η2

= .034). Further inspection revealed that participants scoring low on self-referenced familism scored lower on severity than participants scoring high on self-referenced familism. Participants scoring low on self-referenced familism scored higher on susceptibility than participants scoring high on self-referenced familism. Participants scoring low on referenced familism scored lower on danger control than participants scoring high on self-referenced familism. Significant main effects of message version were found for susceptibility (F [1, 264] = 7.31; p = .007; partial η2

= .027) and response efficacy (F [1, 264] = 5.35; p = .021; partial η2 = .020). Further inspection revealed that participants scored higher on susceptibility and response effica-cy after reading the self-targeted narrative than after reading the family-targeted narrative. No signifi-cant univariate interaction effects of self-referenced familism and message version were found.

Culture-referenced familism

Significant main effects of the cultural orientation variable culture-referenced familism (low/high) were found for fear (F [1, 264] = 18.77; p < .05; partial η2

= .022), severity (F [1, 264) = 14.29; p = .001; partial η2 _{= .041), response efficacy (F [1, 264] = 5.21; p < .05; partial η}2

= .019), self-efficacy 1 (F [1, 264] = 5.21; p < .05; partial η2

= .019) and danger control (F [1, 264] = 16.82; p < .001; partial η2

= .060). Further inspection revealed that participants scoring high on culture-referenced familism scored higher on fear than participants scoring low on culture-referenced familism. Participants scor-ing high on culture-referenced familism, scored higher on severity than participants scorscor-ing low on culture-referenced familism. Participants scoring high on culture-referenced familism scored higher on response efficacy than participants scoring low on culture-referenced familism. Participants scoring high on culture-referenced familism scored higher on self-efficacy 1 than participants scoring low on culture-referenced familism. Participants scoring low on culture-referenced familism scored lower on danger control than participants scoring high on culture-referenced familism. Significant main effects of message version were found on susceptibility (F [1, 264] = 7.50; p = .007; partial η2

= .028), re-sponse efficacy (F [1, 264] = 5.21; p = .023; partial η2

= .019) and self-efficacy 1 (F [1, 264] = 3.93; p = .049; partial η2 = .024). Further inspection revealed that participants scored higher on susceptibility, response efficacy and self-efficacy 1 after reading the self-targeted narrative than after reading the family-targeted narrative. These main effects, however, were qualified by a significant univariate in-teraction effect of culture-referenced familism (low/high) and message version for susceptibility (F [1, 264] = 6.29; p < .05; partial η2

(23)

Table 10 Means and standard deviations of the EPPM scores for the (1) self-targeted narrative, and the (2) family-targeted narrative per cultural orientation variable (low/high).

Low/ high score n a rr a ti v e

Fear Severity Suscep-tibility Re-sponse efficacy Self-efficacy 1 (SA partici-pants only) Self-efficacy 2 (SA partici-pants only) Danger control Fear control

(24)

3.3 The short textual and partly visual fear appeals

First, the effects of nationality of the participants will be presented for message version (implicit fear appeal or explicit fear appeal) on the variables of the EPPM (fear; severity; susceptibility; response efficacy; self-efficacy; danger control; fear control). After that, the effects of the same dependent vari-ables for white and non-white South African will be presented. Next, the scores for low/high scoring participants on the various cultural dimensions will be presented in relation to message version. Final-ly, the effects of cultural orientation (low/high) on the dependent variables will be discussed per fear appeal.

3.3.1 Nationality, ethnicity and message version

To test the second hypothesis, repeated measures analyses were performed with nationality and ethnic-ity as between-subjects factors and the scores on the EPPM variables per fear appeal as dependent variables. The means scores and SD’s of all participants on the EPPM variables can be found in Table 11 (see below).

A significant main effect of nationality was found for fear (F [2, 265] = 25.34; p = .012; partial η2

= .033). Further inspection revealed that Chinese participants scored significantly higher on fear (p < .05) than South African participants. No significant main difference of fear was found between Dutch participants and Chinese participants or between Dutch participants and South African partici-pants. Last, a significant main effect of message version for fear (F [1, 265] = 189.68; p < .001; partial η2

= .417) was found. Participants scored higher on fear after being exposed to the explicit fear appeal than after being exposed to the implicit fear appeal. These main effects were qualified, however, by a significant interaction effect of nationality and message version for fear (F [2, 265] = 13.13; p < .001; partial η2

= .090). Further inspection revealed that Dutch participants (p < .001), Chinese participants (p < .001) and South African participants (p < .001) experienced more fear after seeing the explicit fear appeal than after seeing the implicit fear appeal. No significant univariate interaction or main effects of ethnicity were found for fear.

A significant main effect of nationality was found for severity (F [2, 265] = 10.42; p < .001; partial η2

= .073). Further inspection revealed that Dutch participants scored lower on severity (p < .001) than Chinese participants. Chinese participants scored lower on severity (p < .03) than South African participants. A significant main effect of message version for severity was found (F [1, 265] = 25.17; p < .001; partial η2

= .087). Further inspection revealed that participants scored higher on sever-ity after seeing the explicit fear appeal than after seeing the implicit fear appeal. No significant uni-variate interaction or main effects of ethnicity were found for severity. No significant interaction effect was found for severity.

A significant main effect of nationality was found for susceptibility (F [2, 265] = 13.29; p < .001; partial η2

= .091). Further inspection revealed that Dutch participants (p < .01) and Chinese par-ticipants (p < .001) scored lower on susceptibility than South African parpar-ticipants. No significant main effect of nationality for susceptibility could be reported between Dutch and Chinese participants. Ad-ditionally, a significant main effect of message version could be reported for susceptibility (F [1, 165] = 9.85; p = .002; partial η2

= .036). Participants scored significantly higher on susceptibility after be-ing exposed to the explicit fear appeal than after bebe-ing exposed to the implicit fear appeal. These main effects were qualified by a significant interaction effect of nationality and message version for suscep-tibility (F [2, 265] = 5.10; p < .01; partial η2

= .037). Further inspection revealed that Dutch partici-pants (p < .05) and Chinese participartici-pants (p < .05) scored higher on susceptibility after being exposed to the explicit fear appeal than after being exposed to implicit fear appeal. No significant differences could be reported for South African participants. No significant univariate interaction or main effects of ethnicity were found for susceptibility.

Fear appeals in health communication.