Social norms and child marriage in Cameroon: An application of the theory of normative spectrum

ABSTRACT This paper reports on a qualitative study on social norms and child marriage in rural Cameroon, a country with high prevalence of child marriage but largely ignored in the literature. Study participants (n = 80) were men and women from four different ethnic groups living in four rural villages (two in the Far-North, two in the East). With the assistance of four local interviewers, we conducted 16 semi-structured focus groups to understand how existing social norms contributed to child marriage in participants’ communities. We found great variety in the influence of social norms on people's health-related practices: across these four communities, social norms made compliance with the child marriage practice (respectively) possible, tolerated, appropriate, and obligatory. Effective health promotion interventions should be grounded within sound theoretical understandings of the varying influence of social norms. Using data on child marriage, this paper offers a case study of how that understanding can be developed.

Despite the high proportion of Cameroonian women married before 18 (30%) (UNICEF, 2018), not much is known about the causes and dynamics of CM in Cameroon. The existing literature, however, suggests that an ecology of material, institutional, individual and social factors intersect in sustaining CM. Risk factors are similar to those that occur globally including poverty, low levels of education, and living in rural areas (Delprato, Akyeampong, Sabates, & Hernandez-Fernandez, 2015; shake hands) its influence is stronger. The fourth is the interdependence of the practice: when noncompliance by any one individual threatens the success of the practice within the greater community (norms against overuse of community resources are examples of this). The theory hypothesizes that these four characteristics intersect, giving origin to four levels of influence that norms can exert as mentioned above (obligatory, appropriate, tolerated, and possible). In obligatory practices, actors are under an extreme amount of social pressure to conform; the practice is the only way to achieve a collective outcome, to the point that actors reciprocally police themselves into compliance. Driving is a classic example: no single actor can deviate from the norm without disrupting everyone else. Appropriate practices allow for deviations from the norm, but such deviations are not advisable if actors want to achieve a specific individual outcome. For instance, an adolescent might smoke to gain social status, but his or her peers are not affected by this adolescent's choice. Practices that are allowed (or tolerated) usually do not carry either strong rewards or punishments. Actors anticipate that others around them will accept their actions, but the approval of other is not their primary motivation. Finally, at the weakest level of influence (possible), norms can offer a model of what people are doing, making it cognitively accessible to them as a form of social learning. An example is that on learning that many people are trying a new diet and perceiving a potential benefit, an individual decides to try the new diet as well. In this paper, we investigated how the influence of a child marriage norm varied across four ethnic groups in Cameroon, analysing qualitative evidence collected as detailed in the following section.

Methods
Qualitative research is appropriate when researchers are trying to uncover patterns in participants' beliefs, including those that rise to the level of social norms (Berg, 2001;Creswell, 2007). In this study, we use a specific set of data from a larger qualitative data corpus. The corpus includes data collected using Social Network Analysis, Focus Groups, Individual Interviews, Ethnographic Observations, and Cultural Models strategies (Shakya, Mackie, Nkwi, & Cislaghi, 2018). In this paper, we specifically analysed data on social norms in the focus group dataset.

Focus group participants
Participants in the focus groups (n = 80) were men and women from four rural villages in Cameroon, two located in the Far-North region, and two in the Eastern region. Each village was fairly homogeneous per ethnic group, offering the opportunity to investigate norms as they emerged in these groups. Respectively, the four ethnic groups were: Mafa (Far/North, Christian), Maka (East, Christian), Mbororo (East, Muslim), and Musgum (Far-north, some Muslim and some Christian). All participants gave informed consent. Ethical approval was obtained by the University of California San Diego and by the Cameroonian Ministry of Health.

Data collection
Eight interviewers conducted 4 focus groups discussions (FGDs) in each village, for a total of 16 focus groups with 5 people each. FGDs were chosen for their potential in exploring shared beliefs within groups, including social norms (Cislaghi & Heise, 2017). FGDs offer participants a space where they can discuss the cultural beliefs that surround them, as well as the sources of contention and change surrounding those beliefs. We used a factorial design, an approach that involves holding discussions with separate groups, with common control characteristics, but different break characteristics (Morgan, 1996). For instance, in this study the sample was split by age (18-35, 36+) and sex (male and female) obtaining four groups: older men and younger men, older women and younger women (See Table 1).
Using common interview questions across the focus groups, we explored how cultural beliefs diverged or converged across gender, geography, and generation.
The focus group guide included questions about existing marriage practices and ceremonies, as well as a series of vignettes. The vignettes presented participants with realistic stories of child marriage (created with the help of local researcher). After each vignette, we asked participants to respond to a series of questions on the vignette scenario (see Table 2).
All FGDs and interviews were audiotaped. During data collection, the authors and the qualitative researchers debriefed on a daily basis to record notes on contextual details and impressions that could be helpful later during the analysis.
In addition to the focus group discussions, two researchers collected ethnographic data. Coauthor and Cameroonian anthropologist Dr. Paul Njochi Nkwi and his colleague, Dr. Deli Tize Teri (University of Yaoundé) (who had both conducted ethnographic work for decades in these regions) visited the four study settings, spending approximately one month in each, aiming to gather information on non-discursive practices relevant to our study. They held unstructured ethnographic conversations with several key informants (about 30 in total, although the nature of the conversations makes it impossible to give a precise number). These key informants included, in each village: 1 village chief, 1 religious leader, 1 leader of women's group, 1 traditional midwife, and 2-4 members To do so, I would like for us to use our imagination to create three little stories. This is the first story. It has two protagonists: Jane and John (change to relevant names). I will be asking a few questions about them, and I would like for you to pretend John and Jane are a typical couple from this village. John and Jane will soon get married.
a. This is John's first marriage. How old is he? Does he live alone or with his parents? b. Jane has also never been married. How old is she? Does she live alone or with her parents?
[…] f. How would John's family decide whether Jane is suitable for John to marry? What qualities would they be looking for in Jane? i. of the council of elders, however other community members also talked to them about topics related to the research. The two ethnographers took fieldnotes that were used as both sources of information on the historical and cultural context (see the first part of the result section) and means to triangulate the focus group data. Both observations and discussions were guided by four themes: 1. Presence of married girls under 18; 2. Presence of teenage mothers; 3. Cultural acceptability of adolescent and adult marriages; and 4. Cultural meanings attached to marriage practices and ceremonies. Both ethnographers contributed to the interpretation of the findings and their fieldnotes were taken into account during the analysis of the focus group data.

Data analysis
Two local interpreters transcribed all audio tapes, translated them into English, and removed all identifying information. Co-author PN as well as field researcher Damien Pereyra conducted quality control on the transcriptions. For translation quality control, each interpreter also back-translated a sample of the other interpreters' work. We used theoretical thematic analysis strategies (Braun & Clarke, 2006) informed by social norms theory, but we also inductively let the data drive some of the analysis when they didn't fit into our understanding of social norms. We first familiarised ourselves with the data, reading all the transcripts several times. We then generated initial codes. In an effort to counteract possible biases, we independently analysed the English transcripts; three research assistants also independently coded the data with us. Next, we met in person (at the University of San Diego) to discuss our preliminary observations. There, we identified some assumptions in our coding and uncovered possible themes and patterns of interpretation. After the meeting, we independently reviewed and refined the themes, looking for internal coherence and external heterogeneity across the themes (Braun & Clarke, 2006;Henwood & Pidgeon, 2003). In a series of conference calls, we discussed them until there was an overall agreement among us. Through this process, we eventually reached both coding and meaning saturation. Code saturation refers to the point in the data analysis process when further coding is no longer possible and the codebook has stabilised. Meaning saturation refers to the point in the process when no new information from the data can be obtained and no further nuances are found (Fusch & Ness, 2015;Hennink, Kaiser, & Marconi, 2017). In line with what the literature suggests, we reached code saturation relative early in our process (after having code little more than four focus groups), and meaning saturation soon after (after having analysed approximately 80% of the full dataset) (Hennink, Kaiser, & Weber, 2019). Examples of themes and codes are provided below in Table 3.
In a further series of conference calls, we discussed the interpretation (the 'so what') of the themes that were emerging in the data, until we had preliminary findings to present to cultural insiders for feedback. Interpretation of data was largely semantic, but the ethnographic work of the Cameroonian co-author and his colleagues helped us go beyond the surface of the data. Preliminary findings were presented in Cameroon to local researchers and people familiar with the cultural context who provided feedback on their interpretation. The co-authors then continued the data analysis iteratively (independently and then with conference calls), taking that feedback into account. In a second workshop, led by Cameroonian co-author PN, the final set of results was again presented to a team of local researchers and practitioners who reviewed and approved them. In the results section, we present our themes related to the child marriage norm. Unless explicitly stated, quotes in each sub-group come from different participants.

Mafa
The study village where Mafa participants lived is located in a mountainous region, was well supplied with water and electricity and included a kindergarten, a primary school, and a high school. The Mafa Table 3. Examples of themes and codes emerged in the focus group data.

Themes
Codes Example from the dataset Girls' age at marriage Typical age at marriage for girls 'In our community, girls get married when they are between 12 and 15. There are girls who grow up more quickly, and at 13 reach puberty, and others who wait until they're 15.' Opinions on appropriate marriage age Family negotiations 'Now, the man's family leaves for the woman's village, carrying gifts and money. There, the two families meet and discuss the dowry. The two families sit facing each other, all the presents in the middle. Next, the girl is asked to enter the room; she must choose one of the presents and give it to the head of her family if she wants this marriage.' Dowry 'Among us the Musgum, marriage is expensive but, at least, the wife cannot easily ask for a divorce. If she does, her family will have to pay the dowry back. So, she's like a precious property that you buy and look after with care'. Place of wedding ceremony 'In our culture, the groom's family travels for the ceremony to the bride's village' Wedding cost 'A marriage can cost 800 000 FCFA … the parents of the bride are the ones who spend more.'

Respectable women marry early
Refusing an early marriage 'People will think that, if a girl doesn't get married when she is 15, she is a hypocrite, that she is not honest. People will think that she wants to prostitute.' Social consequences for late marriage for the girl 'A woman who is not married by age 15 is seen as a disobedient girl. She lost her chances to get married. It will make her grow old without being married.' Social consequences for late marriage for the family 'If a father refuses to marry his daughter off at the appropriate time [14], then people will say it's his fault if she doesn't get married in the future. Those are the consequences, his daughter stayed at home, he searched for it.' Commonality of late marriage 'We have never had such a situation here.' Respectable women do not have premarital sex Unmarried girl sees a boy 'If a girl hangs out with a boy at school, her father will be ashamed and feel humiliated by his daughter.' Sanctions for premarital sex 'If a girl has sex before wedding it will be a humiliation to the girl's family. This is the reason why parents send their daughters for marriage at the age of 14, 15 and 16.' Marriage to avoid premarital sex 'When a girl is 12, she is already quite old for marriage; we do not allow girls here to remain unmarried at such an old age, because if she's not careful she will get pregnant, and that will be shameful.'

Girls' voice and parents' role
Love marriage (Girl and boy choose each other) 'A boy and a girl meet one another. They love each other, that's all. The parents of the children will respect their choice and arrange the procedures of marriage'. Negotiated marriage (Parents propose, girl can refuse) '[When a father proposes a suitor] If the girl likes the boy, she will accept the father's decision, but if she does not like him, she will refuse.' Forced marriage (Parents choose for girl) 'At times what happens is that the man wants to marry a girl who doesn't want him. But, if the man's family finds an agreement with the girl's family, they force the girl to marry him.' Note: Here are included some of the themes and codes that were used in this paper. The full codebook included approximately 25 themes and 45 codes.
ethnic group is the majority in the village, but in the surrounding mountains lived other ethnic groups: Kapsiki, Hina, Moufou, Daba, and the Bana. The main economic activities are agriculture, livestock breeding, petty trading, and foraging. Descent in this village is patrilineal, with rights and privileges derived from membership in a given patrilineage. Although monogamous marriage is not infrequent, the Mafa tend to be polygynous. When ready to marry, a young man will declare his intentions to marry to his parents, who negotiate the marriage contract with the girl's family. In the past, childhood betrothal was commong in the village, but now the choice of spouse tended to be consensual. Traditionally and currently, Mafa girls in this community married at around age 20.

Maka
Situated in the East Region of Cameroon, the study village was originally composed of five main patrilineages. During the early settlement of the five lineages the principle of endogamy (the necessity to marry inside of one's own clan or community) was very much in force and marriage between the five lineages was not possible. With the reduction of tensions among the different clans and lineages in more settled village life, exogamy has become more common. At the time of our field work, it was normal for women and men to take marriage partners from outside their lineage and village of residence. Usually, marriage is considered valid only when the bridewealth is paid. This consists of required gifts from the groom's family which are in turn accepted by the bride's family. Ethnographers observed several pregnant adolescent girls. These girls, often not married, were between the ages of 14 and 15 years old. Getting pregnant out of wedlock is considered acceptable within their communities.

Mbororo
The village has a population of around 10,000, predominantly Mbororo, with the surrounding area populated by members of the Bamileke and Bamoums ethnic groups. Marriage is agreed upon when the required bridewealth is paid. This usually consists of a mat, wine, and the promise of labour, with the groom required to work on the farms of his parents-in-law. Once the marriage has been celebrated and the groom confirms that his bride is a virgin, he is required to plant a spear in front of his father-in-law's house as to symbolise that he was the first to have a sexual intercourse with his daughter. In case of divorce, the husband could request the return of the bride wealth. Generally, marriage is both endogamous and exogamous, reflected by the increasing intermixing of young people from different families. Parents are often involved in the choice of wives for their sons, but young men have a say in the matter. If the girl was chosen very early in life, the boy's family would be responsible for the care of the child-bride until she is of age (13-15 year) or has had her first menses.

Musgum
The Musgum village where this study was conducted was close to a rich aquatic resource, prone to frequent overflow. Main economic activities are agriculture, fisheries, and livestock. The population is ethnically homogeneous, with a Muslim majority and a Christian minority coexisting peacefully in the village. The two groups attach similar meaning to marriage, as a union between the two families. The Muslim Musgum traditionally marry their girls relatively soon after menarche, while the Christian community members are said to wait longer, until girls are 17 or more. Our ethnographers said they were told there were no unmarried Muslim girls older than 15 (nor they could observe any). Both valued girls' virginity at marriage, although as a condition for marriage it was more strongly valued among the Muslim community. Pregnancies before marriage are rare in the village, and occasions for girls to meet boys outside the home are infrequent. For a girl to move around the village she must be accompanied by a male family member.

Results
Several norms emerged that contribute to child marriage in these communities as, for instance, the norms 'respectable girls do not have premarital sex' and 'respectable parents are always obeyed by their children'. In this paper, we chose to conduct a deeper examination of the norm that 'respectable girls marry soon after they reach puberty' (the child marriage norm) as a model of how the TNS can help understand how the same norm affects practices in different ways across different settings. We decided to focus on this one norm for two reasons. By limiting our analysis to one norm it allows us to focus on that norm in depth, while providing a straightforward example, uncomplicated by the presence of multiple norms, of the way in which the TNS (likely to be unfamiliar to most readers) can be used. The second reason is that the relationship of this norm with child marriage is the most direct among all the norms we found. A direct relation suggests this norm might be the one with the strongest influence on CM (Cislaghi & Heise, 2018a), increasing the importance of its analysis for practitioners in the field.

Mafa -No norm (CM inconceivable)
Amongst the Mafa we found no evidence of a norm in favour of CM. Mafa participants mentioned that a good age for a girl to marry is when she is 17 or 18 years old. Shared beliefs that no girls in their village would marry before 17 or 18 made it impossible for Mafa participants to conceive that a girl could, indeed, get married at age 15. Asked explicitly if they thought 15 to be a good age for marriage, Mafa participants remarked that 15 would be too early. One young man said: 'A 15-year-old girl is still small and doesn't know anything about keeping a household'. And one older woman responded: 'No, for she is still a child'. However, while Mafa participants were united in their belief that marrying before 15 was too young, they also reported that getting married after 18 would reduce the girl's chances to marry, what Mafa participants called, the man of 'her dreams'. In our vignette we asked participants whether a 22-year girl could get find a husband. One young man said that: 'it is easy for her to find a husband [now that she's 22], but she will not likely find the type of husband she ever dreamed of when she was 17 or 18 years old. She can get married to a polygamous man and be his second or third wife, or even to a widower'. And, in another focus group, a young woman also said: 'It will be difficult for her to find a husband as young as herself [now that she is 22]. If she finds a husband, he will probably be an old man. This has happened before.' And another woman, in the group of the female elder participants, said: 'She may have a chance to marry a man of her own age, but usually the girls like that marry the old'. Mafa participants for the most part did not comment on what people would think about a girl who wasn't married by the age of 22. Two older women, however, did speak about that: 'People will say that she is cursed or mystically attached', and: 'People will say that this is her parents' fault, and others will say that the girl was guilty of serious misconduct'. Only one elder man was of the opposite opinion: 'It's easier for a 22-ear-old woman to get married, because now she has the good age for that'. However, an elder woman in another focus group said that in practice that doesn't happen: 'Take for instance my daughter: when she was young all men were asking her hand. I told them my daughter should go to school and that she was still young … [Some time passed and now] no man wants to marry her.' Even though Mafa participants seemed to favour marriage at 18 or more, they also anticipated negative sanctions for women who waited too long to get married.

Maka -Weakest (CM tolerated)
Amongst the Maka participants, we found a descriptive norm of CM (most girls get married before 18) but a weak injunctive norm that made it tolerated in the community. Maka participants reported, on average, an age of marriage of 13 for girls in their village, in spite of late marriage also taking place in the community. Despite the belief that many girls in their community married before age 18, most Maka participants, when asked whether a girl who refused to marry at age 15 would have chances to marry later, thought that girls would have no problems marrying as adults. One young woman, for example, remarked: 'Of course [she will get married one day]; she is still young and has her future ahead of her, especially if she is educated'; and: 'Is marriage a race? It is like having a child: it comes at its time, we don't force it. She will get married one day'. Two young men also expressed a similar opinion when they said: 'Yes; she will be old enough and with enough experience [to get married]' and: 'She is well prepared for marriage, so she can have a job and take good care of her family'. Some participants also acknowledged that there were some women who, indeed, married later. A younger man, for instance, offered a concrete example from his family: 'Yes [it does happen that women marry later in life after having refused a first marriage]. Take my younger sister: she graduated at an advanced level and just after that she had a child with someone. They are in a stable relation now and she is preparing for public professional exams'.
Maka participants' attitudes towards child marriage were mostly negative: 'I would refuse [to marry off a 15-year-old daughter], because that's not good', said an older woman; and, '[Children] should just be patient [and don't rush into marriage]' said an older man. Only one male leader said to support child marriage because, he suggested, a 'woman is made for marriage'. Despite participants' negative attitudes towards child marriage, however, the practice continued together with late marriage also taking place. That is, the negative attitudes towards CM were too weak to rise to the level of a strong injunctive norms, making it plausible for community members who did want to engage in CM that it was acceptable or tolerated for them to do so.

Musgum -Weakest (CM possible) and stronger (CM appropriate)
Among Musgum participants, we found evidence suggesting that norms and attitudes towards marriage formation differed between participants identifying as Muslims and Christians. Muslim participants said that CM was common and approved within their community, while Christian participants mostly disapproved of CM and said it didn't happen among them, although they were aware of it taking place among Muslim community members. For instance, asked in the focus groups, four different participants answered as follows: We found, in other words, indication that, among the Musgum participants, there existed two streams of thought with regard to CM, one favouring it, the other mostly against it With the assistance of the ethnographic notes and what participants said with regard to their religious identity, we advanced the hypothesis (also confirmed by other data in the larger corpus) that these two views might be associated, respectively, with the two groups of Musgum participants who identified respectively as Muslims or Christians. Take, for instance, the following exchange in the male elder focus group that highlights the two contrasting positions: Today, and according to the Koran, girls have to be sent for marriage at the age of 12 and at times 13. But, we know that at such an age the girl is still young. So, we wait until she is 15, 16 or even 20 before sending her for marriage.
To us Christians, we give our girls for marriage when they are at least 18 years old if not we refuse. So actually, to us Christians, some girls get married at about 25 years old.
The ethnographic fieldnotes confirm the hypothesis around the religious split. During the ethnographic conversations, the representative of the local youth group, for instance, explained: 'In our community here, things are regulated by religious rules. Among Christians, a girl is sent into marriage at 17 years. But the Muslims say that the age should be earlier, a girl has to go into marriage at 13 years.' Musgum participants who approved of child marriage stressed the importance of bodily preparedness for deciding whether the girls is ready for marriage. One older man, for instance, said: 'It is when a girl already has breasts that she can be sent for marriage'. And another older man confirmed: 'Yes, she is 15 years old. So: she can get married normally. But if she has a small body and without breasts, she can delay [but note this man says 'delay', not 'cancel'] the wedding'. One young man explained: 'Among the Musgum, what matters is when the girl has her first period. It may happen that the girl sees her first menstruation when she's 13; then she can get married; age depends on the body of the girl.'

Mbororo -Strongest (Obligatory)
Mbororo participants showed very little variation in their responses: they all agreed that girls should marry early and disapproved of girls marrying later. For this group, CM was obligatory, and was supported by the strongest norms between all four groups in our study. During the focus groups, some Mbororo participants were concerned they might get in trouble for openly supporting child marriage. When the interviewer asked older female participants the ideal age of marriage, one whispered to others: 'don't say below 18 years'. Ironically, the whisper unwittingly has great significance in normative terms. The whisper is evidence of injunctive norms in favour of child marriage: the woman anticipated others in the group approving of child marriage and warned them not to do so in front of the interviewer.
Even though the focus group began with this participant warning others, later in the focus group participants were explicitly arguing against later marriage; for instance one woman in this group said: 'To me, it will be very difficult for a girl to get married [when she's 20]. At that age, all her sisters would be in their marital homes. Up to today, when she's older than 20, she is still in her father's compound, she sleeps with her mother, and this is not good'. Not only did older women think other people approved of child marriage (an injunctive norm), they also thought that older marriage was rare (a descriptive norm). Two, for instance, asked whether they knew of a girl marrying a later age, said 'l do not know a single case like the one you described', and: 'In this village, l have not known of a similar case and, since the day I have been here, I have never witnessed a situation of that sort'.
Older men, asked whether it was possible for a girl to marry late, did not even discuss that possibility as they didn't think that it would happen. They instead responded to the vignette as if it asked what happens when a girl refuses to get married to one particular man, and then gets married to another one. One man, for instance, said: 'Another person can marry her. At times a woman has many suitors and she can choose any one of her choice'. Note how this man doesn't consider the possibility that the 14-year-old girl doesn't get married at all at that age; he only mentions how she can choose another suitor. The same was true in the group of younger men. Asked to provide examples of girls marrying later, they could only give cases of girls divorcing and then remarrying at an older age: 'There are girls who get married and then divorce later, then after some months, having fulfilled the necessary conditions, they get married, especially if the fault comes from the husband' and: 'There are girls who divorce and never get married again'. Nobody, except one, considered a woman marrying at a later age. The one who did, referred to polygamy when he said: 'Within our Muslim culture, it's easy [for a 22-year old woman to get married] because she loses the chance of having a young husband, but she will end up with someone who has three wives already.' Younger women also provided evidence that a norm existed in their village against later marriage. One, for instance, said: 'She refused to get married at the age of 14 and now she is 22 years old. She will find it difficult to get married. She is old.' And another onesimilar to the discussion in the group of the older men described abovesuggested that if a girl refuses her first suitor 'Provided that she is of good character, she will soon get married'. Note how this woman says 'soon', almost implying that, if the marriage doesn't happen soon, or doesn't happen at all, the girl might not be of good character. The presence of a norm here becomes evident: girls who don't get married early after puberty are not respectable, and not of good character. And again, asked whether they have ever witnessed a girl refusing a marriage or one getting married late, young female participants couldn't cite any. One reflected on the differences between this village and her village of origin by saying: 'Here, I do not know [of a girl marrying later], but in my family [in another village and possibly ethnic group] there are such cases'. The interviewer asked the same of other participants, but they couldn't give any example either, the typical response being: 'I do not know of similar event'.

Discussion
Recall that the TNS suggests that norms can have varying strength of influence. This influence can be nonexistent (when there is no norm), weakest (when the norm makes the practice possible), weaker (when the norm makes the practice tolerated), stronger (when the norm makes the practice appropriate), or strongest (when the norms make the practice obligatory with few realistic options for people to not comply). In this paper, we looked at the strength of influence of the norms 'respectable girls marry soon after they reached puberty.' We found the TNS helpful in understanding the strength of influence that the child marriage norm had on participant practices. Table 4 summarises our findings across the four ethnic groups informed by this theory. We further looked at the in-group differences by age and gender [see Table 5]. As we mention later in greater detail, this analysis was limited by the small number of in-group participants. The analysis was however helpful for data triangulation purposes: we found a consistent homogeneity of responses across the sub-groups, which suggested a good reliability of participants' accounts.
Data collected among the Mafa participants suggest that there existed no norm of child marriage. Recall how they agreed among generations and genders that all girls in their village get married around 18, and that a 15-year-old getting married sounded preposterous to them. We did however find a strong belief that, once they reach age of marriage, girls should marry quickly or they might miss their window of opportunity.
For Maka participants CM was happening in spite of the fact that most participants disapproved of it. Many reported that girls would have no problem marrying later but were adamant in their responses that child marriage did happen in their community and was something that they found We found no evidence of a norm of child marriage among Mafa participants. However, we found evidence of a norm that girls should marry quickly when right age of marriage comes.

Musgum (Christian) Weakest (Possible)
Musgum Christian participants said they did not practice child marriage and did not share beliefs that girls should get married before 18. However, they knew that many Musgum Muslim practicsd it and that child marriage was possible.

Maka
Weak (Tolerated) Maka participants reported that, albeit there was no injunctive norm that girls should consummate marriage at puberty, it could happen (albeit rarely) that girls would marry as children (without consummation), and that people tolerated that.

Musgum (Muslim)
Stronger Musgum Muslim participants said it was in the girl's interest to get married soon after puberty. That is, that it was appropriate for her to do so. It was possible for her to marry later on, but the consequences would be negative.

Mbororo
Strongest Mbororo participants said that it was impossible for a girl to get married later in life. An unmarried woman and her entire family would be ostracised and stigmatised by the entire community.
regrettable. In other words, child marriage was not considered appropriate, but it was tolerated. Elsewhere (Cislaghi, Nkwi, Mackie, & Shakya, Under Review) we analysed in greater detail the Maka data, exploring extensively the relation between their opposing attitudes and the presence of CM, suggesting that several conditions (including girls' expanding capacity to take CM-related decisions) weakened the injunctive norm, making the practice tolerated in the community. Musgum participants were split in how they described child marriage. Participants who identified as Christians said that child marriage was not happening. It was possible for a girl to get married at a young age (15-18), but it was extremely uncommon and unpreferable. The group of Musgum participants who identified as Muslim had the opposite view: that most girls married soon after puberty. Later marriage did happen, but they were a second choice for girls, who might then become someone's third wife. We inferred that the split among participants might represent a split within their village of two communities with different sets of cultural norms. The norm of CM seemed to be strong among the community of Musgum Muslim participants, where it was considered appropriate, while it was very weak among the community of Musgum Christian participants, where CM was something they knew was possible (as they witnessed it among the Muslim community) but that they considered inappropriate.
Data collected with the Mbororo participants suggest the presence of a very strong norm that made compliance with the practice obligatory: no options were left to girls, as soon as they reached puberty, they had to get married. Social punishments for not doing so, participants reported, were too heavy for anyone to bear. Mbororo participants did not know of any girl who married at a later age in their village and, when asked whether it would be even possible for one to do so, replied that, at most, she might refuse a suitor as long as she is ready to marry another one soon after.
Normative differences across groups have important implications for effective practice in the abandonment of child marriage in Cameroon. In the case of the Mafa participants, where the CM norm was nonexistent in spite of belief in support of quick marriage when the appropriate age has been reached, a community-based intervention might facilitate a discussion on the reasons why people do not practice child marriage, eventually stretching the window of opportunity so that girls could wait even longer if beneficial to them. An example of similar community conversations comes from the SASA! project, that changed norms of domestic violence in East Africa by helping community activist unite and voice out their concerns publicly (Starmann et al., 2018). Among the Christian Musgum, where the norm made the practice only possible, an intervention could strengthen the existing descriptive norm that CM is extremely uncommon, making public the small percentage of girls who are currently getting married as children in their community. Similar social marketing strategies have been used, for instance, to publicise the non-drinking norm in university campuses (Halim, Hasking, & Allen, 2012). In communities similar to the Maka village in our study, where the practice is continuing despite people's disapproval, an intervention might publicise the extent to which people disapprove of CM, strengthening the injunctive norm proscribing it. In the case of the community of Muslim Musgum participants, who thought child marriage to be appropriate, an intervention could help community members recognise positive deviant cases and look at the advantages for the girls' lives and discuss how other girls could benefit from the same advantages. An example is found in the 'ring the bell' campaign that gave visibility to role models who intervened in case of domestic violence (Michau, Horn, Bank, Dutt, & Zimmerman, 2015). Finally, in places where the norm makes CM obligatory, as in the case of the community where the Mbororo participants lived, an intervention would need to facilitate conversations among community members on the system of beliefs sustaining the practice, first helping a core group of participants generate new attitudes towards CM, and then equipping them with skills and knowledge to reach out to others, including traditional and religious leaders. The NGO Tostan offers a model of such an intervention, with proven evidence of its capacity to reduce child marriage (Cislaghi, Gillespie, & Mackie, 2016). While the application of the TNS can be directly applied to programme strategies, it can also be an important tool for informing quantitative work. Programmatic efforts, especially at scale, require proper measurement, and while qualitative research provides deep insights, it does not provide the ability to quantify the issues at hand. One challenge is that currently most quantitative surveys lack questions on social norms, which are particularly crucial to capture for complex issues of gender inequity like CM (Raj, McDougal, & Trivedi, 2017). However qualitative research is an important component of proper survey development (Jose, Bhan, & Raj, 2017a, 2017b. This qualitative work demonstrates the utility of the TNS in understanding important distinctions in the relationship between social norms and CM, and provides an important foundation for quantitative measurement development as well as for direct intervention design strategies.

Limitations
Our study has two key limitations. The first is that we didn't design it from the start as a test of the TNS. Rather, as we analysed the data, we discovered they could offer an empirical indication of its possible validity as a heuristic model. We do believe that these findings provide enough evidence that the TNS is, at this point, ripe for testing but are adamant that other methods might be more appropriate for conducting a proper test. For instance, a quantitative survey in these same four ethnic groups might help measure possible misalignments between the prevalence of a normative belief and its strength, by, for instance, asking participants both whether they think child marriage is acceptable, and what social consequences they anticipate for complying or not with it. The second limitation refers to the sampling. Even though we reached saturation within the ethnic group, the presence of only one subgroup of each sex/age intersection limited the extent to which we could provide definitive conclusions on the trend within those subgroups. We carefully tried to present the results as suggestive of possible in-group trends, and tried to address this limitation with ethnographic observations, but we acknowledge that a greater number of focus groups (beyond our resource availability) would have offered greater capacity for sub-group patterns. Even taking into account those limitations, however, this paper provides evidence to look further at the validity of the Theory of Normative Spectrum for applied research and practice on social norms and harmful practices.

Conclusion
In this paper, we reported a qualitative study of child marriage among four ethnic groups living in rural Cameroon. We conducted focus groups with men and women residing in four rural villages in the East and Far-North regions of the country. Through the use of vignettes, we invited them to comment on hypothetical scenarios of a girl given in marriage at age 15, and another who got married at a later age. We used the Theory of Normative Spectrum to interpret their responses. With the exception of the Musgum participants, we found homogeneity in the system of norms existing in each community. We also found that the child marriage norm ('girls should get married soon after puberty') had varying strength across the four groups: from nonexistent (Mafa), to possible (Christian Musgum), tolerated (Maka), appropriate (Muslim Musgum), and obligatory (Mbororo).
Understanding how social norms are sustaining health-related practices is important for the design of effective cross-cultural health promotion interventions. We suggest that the work of practitioners and researchers alike would benefit from using frameworks that differentiate among weaker and stronger norm, as this could both help both represent the complexity of the socio-cultural niche in which harmful practice take place and choose strategically the most appropriate set of activities for each context. Our findings suggest that the TNS can serve as such a framework, as the cross-cultural understanding of how social norms affect health-related practices increases.