Are Children Interviewees Just Like Any Others?

Reading, Writing, Understanding: Academic Requirements for the Questionnaire

Surveying primary school students using a self-administered questionnaire immediately raises the question of whether they are able to answer it, i.e. read the questions, understand them and formulate appropriate responses. Written communication is a skill that is mastered progressively over the course of primary school, with different students progressing at different rhythms, particularly depending on their social background, which remains the primary explanatory variable for differences in academic achievement in France (Peugny, 2013). Moreover, questionnaires are more likely to be equated with an academic task – especially when it is administered at school. The risk that certain children do not have the capabilities required to answer it is then associated with the risk that their desire to respond (diligently) is determined by their proximity to academic culture.

To get around these difficulties, one possibility consists in administering the questionnaire face to face – but this supposes either having substantial means or accepting a limited number of participants. The most simple solution is to have the questionnaire administered during class time (which is therefore also the easiest for teachers to accept); it is also the most efficient method, if the goal is to collect enough responses for statistical analysis, in a limited amount of time. To ensure that the children understand the questionnaire, certain researchers assist them by reading the questions and providing explanations if necessary (Lignier and Pagis, 2017). This technique allows non-readers to participate in self-administered questionnaires too. The research team involved with the project Learn to Live in Democracy in Germany thus designed a questionnaire based on pictograms that children had to circle after each question had been read and the meaning of the different pictures had been explained (Van Deth et al., 2011). A final possibility, which is the one I adopted, consists in allowing the children to respond autonomously, while the researcher moves around the classroom to answer any comprehension questions that might arise. This choice, however, supposes that the children have a reasonable reading level, which is why I conducted the research with pupils in the last three years of primary school. Consultation with the teachers led me to believe that most students at this level had sufficiently mastered reading and writing. However, it is important to note that the academic levels of the students are highly variable, as can be seen in the differences in the time required to respond to the questionnaire (anywhere between 10 minutes and 40 minutes). Moreover, a certain number of children – primarily students in CE2 or those recently arrived in France – did have difficulty filling in the questionnaire. Around 20 respondents had to be helped by the teacher, their learning support assistant ¹ , or myself, to read the questions, and sometimes also to write the answers. Some of these questionnaires were not included in the analysis, when the adult appeared to have overly influenced the responses, which introduces a slight bias in the sample by excluding the responses of the least academically successful students. These problems associated with reading level must not be underestimated; they would make it impossible to use this kind of protocol in the lower classes, for example. The ability to respond depends on the degree to which academic skills have been acquired, which is correlated with age, and grade level; but it is not exclusive to children. Indeed, difficulty reading and understanding questions is also a problem that can be encountered in surveys conducted with adults (see for example Beauchemin et al., 2016).

Most respondents did manage to answer the question more or less independently. However, this does not mean that they understood all the questions nor that they followed the expected logics in answering them. Although I was careful to formulate questions in ways that were clear and as short as possible, using a vocabulary accessible to the children, this was not always sufficient to prevent misunderstandings, in spite of the tests conducted with around 10 children. For example, the question “do you have a religion?” seemed to be relatively simple and yet proved to be problematic. Many children did not know the word “religion”, including those who actively practiced one (in hindsight, a question like “do you believe in God?” would have been clearer). Allowing children to ask questions during the questionnaire also made it possible to identify the prompts that were problematic and limit the risk of misinterpreting results. More anecdotally, there was another question that also gave rise to inappropriate answers, due to the unintentional ambiguity of the question:

Do your parents read a paper □ Yes □ No   If yes, which one……………

Apart from the few items mentioned above, the questionnaire was generally well understood by the children. However, the fact that they understand the question does not mean that they answer “seriously”, or in other words, with the intention of providing the best answer. Experiments conducted with adults (Krosnick, 1991) have shown that the respondents’ efforts to provide an answer vary according to their social backgrounds, their interest in the subject, and the difficulty that the study represents for them. As far as children are concerned, we can suppose that answering a 30-minute questionnaire on politics requires a substantial effort. This is particularly true for the test of political knowledge, not only because it appears particularly similar to a school test, but also because it was included at the end of the questionnaire and therefore may suffer from the tiredness or impatience of the respondents. The possibility that the respondents’ diligence in their answers depends on their attitudes toward school is a problem for the validity of the results. In order to verify whether this test did in fact measure political competency, and was not subject to a bias linked with the ways of responding, I introduced a “placebo” question, designed to verify the validity of results:

What is the capital of France?  □ Tokyo  □ Paris  □ New York

This test question therefore allowed us to invalidate 13 questionnaires (out of 538) and thus to reduce the bias linked to individual variations in the ways the questions were answered. Above all, it enabled us to demonstrate that the vast majority of students answered carefully: 96% of respondents correctly answered this question, which was among the last items on the questionnaire. The proportion of students who gave up, or answered at random, is therefore particularly low – and probably less than in most adult surveys (Caron, 2006). Any doubts about the children’s diligence, or the obstacle posed by their reading and writing skills, have therefore been laid to rest, at least for this study.

Do Children Speak the Truth?

There is a second challenge to the validity of the results that stems from the reliability of children’s comments. Indeed, they are often suspected of making things up, exaggerating, and even lying. These widespread visions of childhood are often present in the methodological texts designed to serve as “guides” on questioning children in legal and/or medical studies, for example. In a summary text on the subject, Poole and Lamb (1998) state that children rarely deliberately “lie”, but that they are highly impressionable. In particular, they may believe that a “false” event occurred when it is told to them, the border between fiction and reality being only partially established. The reliability of children’s responses has also been questioned by a study in political science. In 1973, Pauline Vaillancourt interviewed children three times in a single year and observed significant instability in their answers, both in terms of their opinions and also for factual questions such as their religion (stability rate: 72%), or their father’s profession (66%) – which led her to conclude that it is inappropriate to use questionnaires for research with children. And yet longitudinal studies with adults also demonstrate a certain percentage of incoherent responses from one round to the other (for example concerning the number of children they have, Régnier-Loilier, 2014), which does not lead the authors to question the reliability of their respondents, but rather the formulation of their questions. We can thus hypothesize that the instability that Vaillancourt measured is largely due to problems with the formulation of the questions; we have already seen that the term “religion” gave rise to certain misunderstandings. For the question on the father’s profession, Vaillancourt used a closed question (although she does not specify the exact response categories) – and therefore the hypothesis according to which the categories were not clear for some of the children is all the more probable given many of them don’t know their parents’ exact professions. Vaillancourt’s reservations about the reliability of children’s responses are therefore not valid as far as my own research is concerned. The latter was conducted over several years, with a small number of children (N=77) participating twice, which allows us to test the stability of answers over time, on a limited scale. ³ The comparison of a simple question repeated in the same terms in the different rounds of the study – “what are your parents’ professions?”- demonstrates a substantial stability in the children’s responses. Only three of them gave different responses for their parents’ professions between years, but the reported change is sociologically probable (for example housewife declared the first year, waitress declared the second) so it is possible that parents may have actually changed profession. More commonly, the children alter their formulation from one year to the next, generally gaining in precision (for example: “he makes houses/ is a builder”; “organizes performances/ is artistic director of a music festival”). In a few cases, the children did not wish to answer the question in the first round and gave an answer in the second (or vice versa). None of the 77 children therefore seemed to “make up” this information (unless they gave the same invented answer twice in a row, which seems unlikely). The doubts expressed about the reliability of children’s answers therefore seems unfounded here.

Yet Vaillancourt’s study does demonstrate that children are likely to answer questions even when they do not really understand them, which does constitute a substantial methodological difficulty. Once again, this difficulty is not specific to children, but there are reasons to believe that they are particularly subject to it. Hughes and Grieve’s (1980) study also shows that children are liable to answer absurd questions (for example: “Is red heavier than yellow?” Or “Two flies are on a wall. Which one gets there first, the one on the right or the one on the left?”) and that they are able to justify their answers afterwards (for example: red is heavier, because it has “more colour”). This issue was later pursued further by Blades, Spencer and Waterman (2001), who showed that the tendency to answer absurd questions above all depended on the format of the question: closed questions like those used by Hughes and Grieve incited children to answer, whereas they were more likely to refuse to do so in an open question. Moreover, the authors show that children are aware that these questions are absurd, but that they feel obliged to respect the implicit rules of conversation that a question requires an answer, especially when the question is posed by an adult, and even more so when that adult is a stranger to them. These studies show that the children’s answers must be interpreted carefully, particularly when the questions are removed from their everyday experience. However, the analysis of certain results in the questionnaire used for my thesis demonstrates that the risk of children answering without understanding the questions is ultimately quite low. The interviewees were invited to situate themselves according to a series of statements, by ticking “I agree”, “I disagree” or “I do not know” (Table 1 below).

OPEN IN VIEWER

Table 1.

Reactions to Different Statements (N=538)

	“I agree”	“I disagree”	“I don’t know”
“More things need to be done to protect the planet”	89%	2%	9%
“It is good that women participate in politics”	57%	6%	37%
“Foreigners who live in France should be allowed to vote in elections”	61%	10%	29%
“There are too many foreigners in France”	25%	37%	39%
“It is good that gay people can get married”	39%	29%	32%
“Palestine should become a country”	29%	6%	65%

Although most of these questions are controversial, the statement “more things need to be done to protect the planet”, which it is objectively difficult to disagree with, only attracted 2% of negative responses from the children – which demonstrates that they did in fact understand the question and were not responding at random. Moreover, the particularly high “I do not know” response for the item on Palestine, which stems from the children’s lack of knowledge on this issue, also indicates that most of them preferred to not respond when they felt unable to do so – which is itself a gauge of the validity of the other items.

Another issue relating to the validity of the responses is the “acquiescence bias” which, according to certain authors, is particularly pronounced among children. Bruck, Ceci and Hembrooke (2002) demonstrate that young children are highly susceptible to false testimony, particularly because they tend to confirm what they believe to be the expectations of the investigator: they tend to agree for example with closed questions, even when they are absurd (for example the tendency to answer “yes” when asked whether the nurse they saw before the interview licked their knee). This tendency to answer yes is accompanied by a reluctance to give negative opinions in open-ended questions. Kolson and Green (1970) showed that children tend to give positive answers to opinion questions, including when they relate to figures they do not know. Kolson and Green conclude that the supposed “valorisation” of the political world by children (presented by behaviourists as a key stage in political socialisation) is a simple effect of methodological bias. The data from my questionnaire invalidate this hypothesis. Although there can be no doubt that behaviourists overestimated children’s positive image of the political world (particularly for reasons to do with the formulation of the questions ⁴ ), the interviewees who participated in my research did not appear particularly reluctant to give negative opinions. I asked them about the three contemporary figures that they were most familiar with, François Hollande, Nicolas Sarkozy and Marine Le Pen. For each of them, they were encouraged to tick “I love him/her”, “I like him/her” (positive opinions), “I don’t like him/her”, “I hate him/her” (negative opinions), “I don’t know”, or “I don’t know him/her” (lack of opinion). The Table 2 records the positive and negative opinion given by the children, in relation to the total number of opinions they provided (Table 2 below).

OPEN IN VIEWER

Table 2.

Opinions Given by the Children about François Hollande, Nicolas Sarkozy and Marine Le Pen (N=514)

	3 opinions	2 opinions	1 opinion	0 opinion	Total
All positive	21	26	30	–	77
All negative	47	51	56	–	154
Two positives, one negative	33	–	–	–	33
Two negatives, one positive	90	–	–	–	90
One negative, one positive	–	63	–	–	63
Total	191	140	86	97	514

There is only a small minority of children who give exclusively positive opinions (77 overall) and only 21 children positively respond to all the three figures. Inversely, many more of them provide only negative opinions (154 total) or give more negative opinions than positive ones. If there is a general trend in the table, therefore, it is rather that children tend to express negative opinions on political figures – which is in striking contrast to the results presented by behaviourists for American children in the 1960s. This is not the place to comment in more detail on this result, but it is worth noting that the discrepancy is probably due to the cultural and social context as well as the behaviourists’ methods. At any rate, the assumption that children are overwhelmingly acquiescent in their responses appears unfounded.

These different results demonstrate that children generally respond seriously to the questionnaire. The only challenge concerning the reliability of the results is due to the fact that they are sometimes inclined to reply to questions without fully understanding them – which demonstrates more an excessive zeal than a lack of diligence. From this point of view, the methodological difficulties resulting from the children’s age can be resolved by adapting the survey protocol: like in any research, it is necessary to ensure that the questions are clearly understood by the respondents. When there is a large social and cultural distance between the interviewer and the interviewees, even more caution must be taken, not only to avoid risks of misunderstanding and misinterpretation of results, but also because feelings of insecurity or illegitimacy can influence the way respondents answer. For example, children may be reluctant to tell an adult that their question is absurd. The imbalance in the research relationship must therefore be understood as a second major issue to be taken into account in research with children.

Social Status, Age, Class, Race: Decisive Factors in the Distance between Interviewer and Interviewee

The researchers belong to another world; they are an outsider to the school, and a stranger to the children. As an adult, they are in a situation of authority, which is accentuated by the fact that they are asking questions in a school setting. Children are indeed subject to domination, both tacit and naturalized, by the adults that surround them. This domination plays out on economic, legal, political, but also symbolic levels, through an injunction to obedience that they often have to submit to, and the superior social status of adults. Children must listen to adults, obey them, and avoid contradicting them or challenging their authority. The forms of domination that children are subject to vary significantly however, depending on parental educational models and social conditions, like gender, socio-economic backgrounds, age (being a “big” or “little” child), but also academic results and physical strength. These different factors often encourage domination between children, but also with certain adults (a child may be in a situation of superiority in relation to a domestic worker employed by their parents). The twofold domination of the adult researcher over the child participant can also be accentuated or tempered by other factors that determine the extent of the distance between the researcher and the participant. The research relationship is likely to be particularly destabilizing for the children from working class backgrounds, for whom the distance associated with age and status are accentuated by class and/or race relations. The following exchange, with a working class child (whose father is a builder and whose mother is a housewife) who described himself as “from a Moroccan background”, enrolled in a school on disadvantaged urban fringes, provides a good illustration of the different aspects this distance can take:

The remark “You can see you come from Paris” can refer to a social distance – which can be seen through language (“you speak Parisian”), which is probably associated with an ethno-racial categorization (“you can see it”) – in any event, it is clearly a mark of otherness. Like this child, several interviewees had a tendency to equate me with a teacher, which is particularly visible in their use of the formal “vous”, their reflex of raising their hand when they want to speak (even though the interviews are conducted with only two children at a time) and sometimes, like in this extract, a tendency to call me “teacher”. These signs that I was categorized by the children as a “school adult” were particularly frequent with students from working-class backgrounds, probably because the students from upper classes were more likely to be familiar with adults outside the school with whom they could equate me. My efforts to blend into the school environment were to no avail – as this child said, “it’s the same”. As an adult within the school system it was not really possible to build a relationship as an equal with the students – not without engaging in long-term participant observation as some researchers have, in which case it is possible to progressively change status in the eyes of the participants (Lignier, 2008).

The fact that I was equated with school personnel did not necessarily mean, however, that the children did not differentiate between the interview situation and the classroom. The children’s experiences at school are vastly different – the adult-child relationships in particular vary depending on the status of the adult and the different “moments” of the school day (lessons, activities, playtime, canteen, excursions etc.) and cannot be reduced to a purely vertical relationship between the student and the teacher. Patrick Rayou for example demonstrates that children’s perceptions of actors in the school environment are heavily dependent on their status (1999: 62). The children seem to be conscious that the research relationship is distinct from a typical school relationships, and many of them allowed themselves to step outside the rules that dominate there, for example by getting up during the interview, joking, using swearwords, and sometimes being generally rowdy. However, overall the interviews were generally in keeping with the school environment: the students had to stay seated and answer the questions, all of which was previously planned by the adult who supervised the questionnaire. The survey relationship that was established was therefore clearly one between a child and an adult. Is this relationship less “authentic” than one between a child and a child? We can consider that research relationships are like any other “ordinary” social relationship (Mauger, 1991; Merklé and Octobre, 2015) and are only problematic when the researcher is not aware of the potential effects of this interaction on the results obtained and these are not taken into account.

The status of the researcher is therefore neither more nor less an obstacle for interviewing children than it is with other populations whose characteristics (for example, socio-professional or cultural etc.), distance them from the researchers. Although this distance interviewer-interviewee is not a problem in itself, it raises an issue on an ethical level, particularly as far as consent is concerned (Throssell, 2018; Rouyer et al., this issue). Children can feel that they are tacitly obliged to participate (Denscombe and Aubrook, 1992), in particular when the activities take place in the school environment where participation is generally compulsory. Much like unemployed or incarcerated respondents, or employees interviewed in the workplace, children have reasons to fear repercussions if they refuse to participate. The fact that the institution is the link between interviewees and interviewers introduces an effect on the way the former consider the research. Anglophone research has emphasized the importance of obtaining “enlightened consent” from child participants (David, 2001; Hill, 2005), in other words ensuring that they understand the meaning of the researcher’s approach and that they do not feel constrained to participate. They also suggest children be asked regularly in the interview if they wish to continue. In the context of my research, these precautions above all revealed that the children were enthusiastic about participating. Out of more than 80 students who were asked, only three did not wish to participate. ⁷ At the end of the interview it was not rare that the children asked to keep going, or to start again. Of course, the children’s enthusiasm can be partly explained by the fact that their study allowed them to get out of lessons for a while. Even when it is spontaneous, consent can therefore result from logics other than an interest in the study itself. In any event, this does not seem to constitute an unpleasant experience for the participants, particularly because the interview protocol was designed to be adapted to them.

The Interview: Child’s Play?

The exploratory interviews conducted as part of my Masters research convinced me that it was necessary to establish an innovative interview protocol. During these interviews (conducted individually with working class children), I used a “traditional” semi-directive interview technique, beginning with questions on family and habits and then progressively moving on to political questions. These interviews were not particularly successful: the children were intimidated, answered with very short sentences, made no comments or discussion, and generally seemed anxious for the interview to be over. Shyness and intimidation, as a general rule, is a problem that is frequently associated with interviews with children. Naturally, it varies depending on the personality traits of the individual, which are also correlated with gender and social background (girls from working class backgrounds are particularly likely to be shy). In any event, this is a result of the adult’s status as the interviewer. “Getting kids to talk” about politics through semi-directive interviews, in other words asking questions that are not very precise so that they “launch” into their own reflections, enabling the interviewer to prompt them when relevant, proved therefore very difficult.

From this perspective, children are very different interviewees than adults. They do not easily partake in free discussion, and do not seem to be “good candidates” for semi-directive interviews – particularly very young children or those from working class backgrounds. It therefore seems necessary to adapt interview techniques for children, not because the latter are “unable” to have a serious conversation, but rather because their discursive habits are different from those required in a directive interview. When they talk amongst themselves children use language that is metaphorical and evocative (Montmasson-Michel, 2016), changing subjects rapidly and relying on many comic interludes (Garitte, 2000). Adult-child language, by contrast, tend to use questions not as a conversational device but for educational or informative purposes (Poole and Lamb, 1998). When an adult asks a question, children expect that they already know the answer (Graue and Walsh, 1998), and that they will provide feedback (validation or correction; Westcott and Littleton, 2005). These discursive schemas are visible in the ways in which children respond to questions during interviews. The interviews seem more like interrogation and less like discussion, with the respondents often looking for the validation of their responses and struggling to present their discourse as a personal opinion. Reversing the perspective, it seems that semi directive interviews are broadly modelled on the conversational style of adults in dominant classes (Mauger and Pouly, 2019), which explains why children are not comfortable with it, particularly when they come from working class backgrounds.

Among researchers who conduct interviews with children it is widely recognised that survey techniques must be specifically adapted. Interviews are often conducted at home, sometimes in the children’s bedroom, and participants are often interviewed several times (Throssell, 2015; Green and Hill, 2005). Moreover, a wide range of techniques is used to avoid “head-on” discussion: interviewers asked children to draw, react to an anecdote, video, or image (for example Hill et al., 1996; Throssell, 2015). Some asked children to tell stories (Engel, 2005), react to hypothetical situations (Rayou, 1999) or “rank” cards (Zarca, 1999; Lignier and Pagis, 2017). One of the other solutions to make the children feel comfortable, used in particular by Lignier and Pagis (2017) was interviewing the children in pairs. This approach, which I used in most of my interviews, stems from a desire to limit the interviewer’s domination in the interaction, by establishing a more balanced situation. Interviewees are less intimidated when they are in pairs (based on the children’s friendships) and enjoy the interview situation more. The impacts of this choice on interaction are, of course, very important: the children’s responses are at least partly determined by those of their partner. However, this method does allow the researcher to take more of a backseat and observe the interaction between the children. It therefore produces an interaction dynamic that is very different from individual interviews, but which has other advantages that are just as relevant for analysis – which have been described several times in relation to adults (Duchesne and Haegel 2004; Braconnier, 2010). Interviewing children in pairs is not sufficient, however, to produce fluid interactions similar to those we would expect from semi-directive interviews. In general terms, the questions children are asked directly can be uncomfortable, particularly when they are to do with politics. For this reason, I adopted a specific interview protocol. I created a “game” in the form of laminated cards of different colours, on which questions, keywords, or images were printed. The children were asked to choose one at random after they threw the dice which indicated which colour they should choose. In turn, they had to answer the question, read the keyword, or describe the image and then tell me what they thought about it: I only intervened to ask for additional details, or to repeat the instructions. In this respect, this was a kind of directive interview in which the questions were self-administered.

Yacine throws the dice and lands on a five, which corresponds to the image-cards. “Yes!” he yells with satisfaction and takes a card. Before he reveals the image, his partner Karim tries to guess what is on it. “Marine Le Pen!” (The two previous images were photographs of François Hollande and Barack Obama). In fact, it is a picture of a primary school teacher.

Using a “game” to ask the questions is thus a kind of trick to be able to ask very specific questions without giving children the impression they are being interrogated. This use of this game meant that the interview was organized into different questions that were covered bit by bit. As a result, the interviews are quite disjointed and do not follow a discursive logic as is generally the case in a semi-directive interview. But this method is satisfactory for several reasons. Firstly, the fact that it was fun, although very basic, meant that it appealed to the respondents. One of the signs of their enthusiasm was that they scrupulously respected the alternation between partners and awarded particular value to the image-cards (as we can see in the extract above), which were bigger than the others. It also seems that the self-administration of the questions makes them less intimidating and allows the children to take charge of the rhythm of the interview and decide when they move onto the next question. In certain cases, they answered the question very quickly in order to move onto the next one, their enthusiasm for the game being in fact a sign of their curiosity to discover the questions. Re-prompting to have more detailed answers was therefore sometimes difficult. But this relative lack of power for the interviewer in terms of how the interview unfolded was broadly compensated for by the general dynamic energy of the exchanges, rhythmed by the shift from one question to another. More fundamentally, it seems that the forms of dialogue produced by these interviews are adapted to children’s conversational forms. Indeed, with this survey method, the children tended to change subject very rapidly and their discussions often moved forward in stops and starts, rather than according to a structured progression. It is important to note, however, that the linguistic practices of children are heavily dependent on their social spheres (Bernstein, 1975) and that some of the respondents would probably have been quite comfortable in a more traditional interview situation. The advantage of the game, from this perspective, is that the children could appropriate it as they wished. Moreover, games are a way of getting around the difficulties raised by group interviews (Duchesne and Haegel, 2004), by guaranteeing a genuine alternation between respondents. Finally, this method also has the advantage of producing uniformly structured interviews (although the questions were asked in a different order each time), and questions that were worded identically, which enables a detailed comparison between the different interviews. Although conversational habits between children and the distance between the interviewer and the interviewees resulting from age make innovative interview techniques essential, these methods have a range of advantages and need not be reserved for children. These games, as well as projective materials, and other activities sometimes used in interviews with adults provide access to respondents’ representations in contexts where semi-directive interviews are not ideal, whether for reasons linked to the interviewees’ profile, the subject, the configuration of the interview, or its objectives.