Development of odour awareness in pre-schoolers: A longitudinal study

Physiology & Behavior 204 (2019) 224–233

Contents lists available at ScienceDirect

Physiology & Behavior journal homepage: www.elsevier.com/locate/physbeh

Development of odour awareness in pre-schoolers: A longitudinal study a,b,⁎

Lenka Martinec Nováková a b c

b,c

, Jan Havlíček

T

Department of Anthropology, Faculty of Humanities, Charles University, U Kříže 8, 158 00 Prague 5 - Jinonice, Czech Republic National Institute of Mental Health, Topolová 748, 250 67 Klecany, Czech Republic Department of Zoology, Faculty of Science, Charles University, Viničná 7, 128 44 Prague 2, Czech Republic

ARTICLE INFO

ABSTRACT

Keywords: COBEL Children Metacognition Olfaction Reactivity Smell

Conducting interviews about children's olfactory behaviours offers a feasible way of learning about the earliest perceptions and knowledge of one's odour world. However, little is known about the stability and development of such self-reports. Here we present the first longitudinal study to repeatedly test children's odour awareness using the Children's Olfactory Behavior in Everyday Life (COBEL) questionnaire in five waves over a two-year period. We expected that higher scores would be attained by girls relative to boys and by older children compared to younger ones and that the scores would increase further into the study. We found a linear time-related increase in the total COBEL scores and in the food and environmental components, whereas awareness of social odours decreased over time. Girls were more olfaction-oriented in the context of social and environmental, but not food, odours. All the reported effects were small. The age at which the children entered the study did not affect their scores. We suggest that the unexpected findings regarding social odours warrant replication in particular and extension in longitudinal studies carried out over a broader time span.

1. Introduction Humans exhibit a high degree of inter-individual variability in how they perceive and interact with their olfactory environment, and in the significance they ascribe to their sense of smell [1–4]. This aspect of olfactory perception has been termed “odour awareness” [3,4] or “olfactory reactivity and awareness” [5,6], and can be assessed by means of various olfaction-related metacognitive measures [2,4,5,7]. These measures lend insight into people's olfactory behaviours that are not directly observable or reproducible in the laboratory. For instance, while some people are readily aware of ambient odours and tend to comment on their presence, others will only notice them after they have been brought to their attention by someone else. Some people let ambient smells affect their attitudes and behaviour, whereas others do not. Also, some people seek pleasant olfactory stimuli and avoid unwanted, potentially disturbing odours more actively than others. In children, the most established tool for investigations of odour awareness is a questionnaire entitled “Children's Olfactory Behaviors in Everyday Life” (COBEL) [5,8]. Designed to evaluate children's self-reported awareness of odours present in their everyday environment, their active seeking of olfactory stimuli, and affective responses towards them, it comprises three components that assess the food-, social-, and environment-related olfactory contexts separately. For instance, to

⁎

learn about a child's food-related odour awareness, participants are asked what they will do when presented with an unfamiliar dish, whether they ever wonder what will be for dinner based on cooking smells, or what their food dislikes are (and after they respond it is assessed whether the dislikes have an olfactory basis). The social context is evaluated by asking a child if they ever try to smell parts of their body or articles of clothing, whether they have ever noticed that their relatives have a specific body odour and how they feel about it, or whether they ever register other people's personal odours. Finally, awareness of environmental odours is operationalised, for instance, as a tendency to seek comfort in sadness by smelling odours, as memory for odours encountered on the previous day, or as an attitude towards tobacco odours. COBEL can be administered either as a questionnaire to literate children, or, with pre-literate children, a structured interview can be conducted based on the questionnaire. Several cross-sectional studies show that between-subject variability in COBEL scores is apparent from early in ontogeny and that it relates to the proxies of age and gender in healthy [5,6,8–10] and visually-impaired [11] children. Specifically, based on their self-reports, girls of all ages were found to be more olfaction-oriented than boys [5,6,10,12] and older children scored higher than younger ones [5,6,8]. Thus, most of our knowledge about the development of odour awareness in childhood comes from studies which employed one-off assessments

Corresponding author. E-mail addresses: [email protected], [email protected] (L. Martinec Nováková).

https://doi.org/10.1016/j.physbeh.2019.02.035 Received 5 August 2018; Received in revised form 1 January 2019; Accepted 23 February 2019 Available online 26 February 2019 0031-9384/ © 2019 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).

Physiology & Behavior 204 (2019) 224–233

L. Martinec Nováková and J. Havlíček

only. The only two longitudinal studies to date in which children were interviewed using the COBEL questionnaire indicated that children's odour awareness at the end of the study varied as a function of scores achieved at its beginning [13] and that their awareness was positively affected by exposure to more diverse food-related odours [12]. A research question which still remains largely unanswered is whether olfactory-related behaviours remain consistent over time and multiple repeated assessments or whether they undergo development even in the short-term. Here we present the first longitudinal study to investigate the development of odour awareness in children over a two-year period in five waves of data collection. This cohort was of special interest because of the potential effects of experiential factors related to school entry. Every six months, children were interviewed about their olfactory behaviours and their perceptions of their everyday odour environment. We hypothesised that higher scores would be attained by girls relative to boys and by older children compared to younger ones. Also, we expected that the scores would increase further into the study.

children entered the study at different times (Wave 1 vs. Wave 2) and the pattern of skipping the sessions was highly individual, experiential effects on the children's reports could not be simply operationalised in terms of waves. For example, of the 39 children recruited to the study during Wave 1 who participated in Wave 5, which would have been, in theory, their fifth testing occasion, in reality only two children had actually attended all four of the previous sessions, while 19 children had been tested only three times and 16 only twice, and two had thus far attended only a single session. Therefore, the level of individual experience with the interview within the specific wave was expressed as a cumulative total of sessions attended by the given child thus far. Only a single session was attended by 31 children (19.7%). The number of children who attended 2, 3, or 4 sessions was N = 33 (21.0%), 51 (32.5%), and 40 (25.5%), respectively. Additionally, the time interval since any previous testing was taken into consideration. For instance, for children recruited to the study during Wave 1 who participated in Wave 5, this time interval ranged from 5 to 25 months. The number of children (boys, girls, and the total sample) participating in each wave is given in Table 1, along with descriptive data on their age, experience level, and COBEL scores. Binomial tests showed that the proportion of boys and girls within each group participating in each wave was not significantly different from the 50:50 ratio: p = .30 (Wave 1), p = .39 (Wave 2), p = 1.00 (Wave 3), p = .18 (Wave 4), and p = .65 (Wave 5). The proportion of boys and girls did not differ across the waves, p = .31. Children participating in each wave were significantly older than those who attended the preceding one (one-tailed independent-sample t-test): p = .001 (mean age 70.03 months in Wave 1 vs. 73.41 in Wave 2), p < .001 (mean age in Wave 3 was 77.75 months), and p < .001 (77.61 months in Wave 4 vs. 91.10 in Wave 5). Children participating in Waves 3 and 4 were of the same age, 77.75 months vs. 77.61, p = .46. Other comparisons (i.e., Wave 1 vs. Waves 3, 4, 5; Wave 2 vs. Waves 4 and 5, and Wave 3 vs. Wave 5) were also statistically significant (p < .001), with mean age increasing in later waves. This means that the fact of different children participating across the individual waves did not result in a drop in mean age. This also held true for boys and girls as well as the recruitment groups analysed separately. All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2008 (5). The study was approved by the Institutional Review Board (IRB) of the Faculty of Science, Charles University. The children's parents provided written informed consent and the children themselves gave oral informed consent in the presence of a teacher employed by the school. The child-parent pairs each received 300 CZK (approx. 12 EUR) in compensation.

2. Materials and methods 2.1. Participants The participants were 157 children of Czech origin: 79 boys, mean age at study commencement 5.87 ± 0.71 years (70.41 ± 8.62 months), range 3.58–7.08 years (43–85 months), and 78 girls, mean age at study commencement 5.75 ± 0.57 years (69.13 ± 6.71 months), range 4.42–7.00 years (53–84 months). At the start of the study, the children attended one of six public mixed-sex kindergartens in various districts of Prague. The kindergartens were attended by children from varied social backgrounds. Kindergarten principals were contacted via telephone, e-mail, and in person to inform them about the planned study. Those who provided permission to perform the study on the kindergarten's premises were asked to pass the information on to the teachers, who distributed leaflets to the children's parents. We kept the e-mail addresses and phone numbers of the parents in order to contact them later with an invitation for their children to take part in subsequent sessions. After a given child started school, the session took place either on the school premises or in our laboratory. The sessions were scheduled in approximately six-month intervals. Namely, data was collected in the late autumn and early winter of 2010 (Wave 1), late spring and early summer of 2011 (Wave 2), late autumn and early winter of 2011 (Wave 3), late spring and early summer of 2012 (Wave 4), and late autumn and early winter of 2012 (Wave 5). Half the children were recruited and first tested during Wave 1 (48.41%, N = 76, 33 boys), while the other half entered the study and were first tested during Wave 2 (51.59%, N = 81, 47 boys). Because, on the first testing occasion, these two groups did not differ in age, t (153.72) = 0.45, p = .65 (see Table 1 for details), or baseline total COBEL scores, t(148) = −0.91, p = .36, group membership was disregarded. All the children were tested for the final time during Wave 5, meaning that there was data from the maximum of five testing occasions available for those who entered the study during Wave 1, but a maximum of four for those who were recruited during Wave 2. Because of the longitudinal nature of the study, missing data was a concern. For the most part, data gaps were caused by participant absence on the day of data collection rather than attrition from the study. The numbers of COBEL interviews conducted were 73 (100% preschoolers) in Wave 1, 128 (100% pre-schoolers) in Wave 2, 88 (50% pre-schoolers) in Wave 3, 35 (94.3% pre-schoolers) in Wave 4, and 81 (100% school age) in Wave 5. For logistical reasons, we were unable to collect data for school children in Wave 4, resulting in a lower N, a greater proportion of pre-schoolers, and a drop in the mean age within this wave (see Discussion for details). Participant-wise, complete data was available from 23 children. Thus, 85.35% of participants had some missing data due to skipping at least one testing occasion. Because the

2.2. Children's Olfactory Behaviors in Everyday Life questionnaire (COBEL) Odour awareness was assessed by means of an interview based on the Children's Olfactory Behaviours in Everyday Life (COBEL) questionnaire [5]. Originally developed for 6- to 10-year-olds, it consists of 16 questions designed to evaluate self-reported awareness of odours in salient everyday contexts, i.e., those related to food (e.g., “When you smell a food odour, do you try to guess for fun what it is (response format: never/sometimes/often)?”), social perception (e.g., “Do you happen to smell parts of your body (never/sometimes/often)?”), and environment (e.g., “Imagine someone is smoking next to you. Do you care/like or not like/love or hate this odour?”). Each item was scored on a 3-point scale, rating the child as poorly (0), moderately (0.5), or highly (1) olfaction-oriented in the given situation. Frequencies of scores for the individual COBEL items achieved by children across testing occasions (i.e., with progressively greater experience with the interview) regardless of wave are shown in Supplementary Material 1. Although this tool had already been used successfully in a previous 225

226

COBEL environmental

COBEL social

COBEL food

Sessions attended Interval since last testing COBEL total

Wave 3 Age (months)

COBEL environmental

COBEL social

COBEL food

Sessions attended Interval since last testing COBEL total

Wave 2 Age (months)

COBEL environmental

COBEL social

COBEL food

COBEL total

Wave 1 Age (months)

86.33 ± 6.49 (76.00–96.00), 13 2 6.79 ± 0.64 (6.00–8.00), 13 7.08 ± 1.73 (4.00–10.50), 13 2.15 ± 0.32 (1.50–2.50), 13 1.23 ± 0.99 (0–3.00), 13 3.69 ± 1.25 (2.00–5.50), 13

79.60 ± 5.86 (69.00–90.00), 27 1 6.82 ± 0.64 (6.00–8.00), 27 6.20 ± 2.10 (2.50–9.50), 25 1.66 ± 0.76 (0–2.50), 25 1.16 ± 0.80 (0–2.50), 25 3.38 ± 1.20 (1.50–5.50), 25

70.59 ± 8.40 (43.00–82.00), 33 6.45 ± 2.25 (2.50–11.50), 31 1.32 ± 0.74 (0–2.50), 31 1.74 ± 1.12 (0–4.00), 31 3.39 ± 1.03 (1.50–5.50), 31

Boys

84.58 ± 5.47 (74.00–94.00), 19 2 8.06 ± 2.40 (6.00–14.00), 19 7.74 ± 2.42 (3.50–12.00), 19 1.82 ± 0.61 (0.50–2.50), 19 1.82 ± 1.16 (0–4.00), 19 4.11 ± 1.17 (2.50–7.00), 19

78.06 ± 5.33 (66.00–87.00), 30 1 6.86 ± 0.67 (6.00–8.00), 30 7.77 ± 2.66 (2.50–12.50), 26 1.65 ± 0.83 (0–2.50), 26 2.04 ± 1.25 (0–4.00), 26 4.08 ± 1.10 (2.00–6.50), 26

69.61 ± 6.07 (53.00–80.00), 43 6.70 ± 1.94 (3.00–11.00), 42 1.38 ± 0.71 (0–2.50), 42 1.93 ± 1.12 (0–4.00), 42 3.39 ± 1.00 (1.00–5.50), 42

Girls

Group 1

85.29 ± 5.87 (74.00–96.00), 32 2 7.55 ± 1.98 (6.00–14.00), 32 7.47 ± 2.16 (3.50–12.00), 32 1.95 ± 0.53 (0.50–2.50), 32 1.58 ± 1.12 (0–4.00), 32 3.94 ± 1.20 (2.00–7.00), 32

78.79 ± 5.59 (66.00–90.00), 57 1 6.84 ± 0.65 (6.00–8.00), 57 7.00 ± 2.51 (2.50–12.50), 51 1.66 ± 0.79 (0–2.50), 51 1.61 ± 1.13 (0–4.00), 51 3.74 ± 1.19 (1.50–6.50), 51

70.03 ± 7.14 (43.00–82.00), 76 6.60 ± 2.07 (2.50–11.50), 73 1.36 ± 0.72 (0–2.50), 73 1.85 ± 1.12 (0–4.00), 73 3.39 ± 1.00 (1.00–5.50), 73

Total

72.74 ± 7.77 (60.00–88.00), 26 1 6.42 ± 1.40 (5.00–9.00), 25 6.89 ± 2.16 (2.50–13.00), 26 1.60 ± 0.68 (0.50–2.50), 26 1.73 ± 0.88 (0.50–3.50), 22 3.83 ± 1.17 (2.00–7.00), 26

7.27 ± 2.16 (3.00–11.00), 31 1.73 ± 0.67 (0.50–2.50), 30 1.92 ± 1.00 (0.50–3.50), 31 3.68 ± 1.28 (1.00–6.00), 31

6.67 ± 2.24 (1.50–11.00), 46 1.63 ± 0.61 (0.50–2.50), 44 1.80 ± 0.91 (0.50–3.50), 37 3.67 ± 1.14 (1.00–5.50), 46 74.38 ± 9.00 (60.00–91.00), 33 1 6.74 ± 1.50 (5.00–10.00), 33 6.03 ± 1.88 (3.00–9.50), 30 1.66 ± 0.62 (0.50–2.50), 28 1.54 ± 0.86 (0.50–3.50), 26 3.15 ± 0.95 (1.50–5.00), 30

68.41 ± 7.71 (54.00–84.00), 33 0 NA

NA

NA

NA

NA

NA

Girls

70.39 ± 8.76 (54.00–85.00), 47 0 NA

NA

NA

NA

NA

NA

Boys

Group 2

73.66 ± 8.45 (60.00–91.00), 59 1 6.60 ± 1.46 (5.00–10.00), 58 6.43 - 2.04 (2.50–13.00), 56 1.63 ± 0.65 (0.50–2.50), 54 1.63 ± 0.87 (0.50–3.50), 48 3.46 ± 1.10 (1.50–7.00), 56

6.92 ± 2.21 1.50–11.00), 77 1.67 ± 0.63 (0.50–2.50), 74 1.85 ± 0.95 (0.5–3.50), 68 3.68 ± 1.19 (1.00–6.00), 77

69.57 ± 8.35 (54.00–85.00), 80 0 NA

NA

NA

NA

NA

NA

Total

77.76 ± 9.92 (60.00–96.00), 46 1 6.75 ± 1.31 (5.00–10.00), 46 6.35 ± 1.88 (3.00–10.50), 43 1.82 ± 0.59 (0.50–2.50), 41 1.44 ± 0.90 (0–3.50), 39 3.31 ± 1.06 (1.50–5.50), 43

73.75 ± 8.98 (54.00–90.00), 74 0 6.82 ± 0.64 (6.00–8.00), 27 6.51 ± 2.19 (1.50–11.00), 71 1.64 ± 0.66 (0–2.50), 69 1.54 ± 0.92 (0–3.50), 62 3.57 ± 1.16 (1.00–5.50), 71

70.59 ± 8.40 (43.00–82.00), 33 6.45 ± 2.25 (2.50–11.50), 31 1.32 ± 0.74 (0–2.50), 31 1.74 ± 1.12 (0–4.00), 31 3.39 ± 1.03 (1.50–5.50), 31

Boys

77.74 ± 9.03 (60.00–94.00), 45 1 7.13 ± 2.05 5.00–14.00), 44 7.24 ± 2.29 (2.50–13.00), 45 1.69 ± 0.65 (0.50–2.50), 45 1.77 ± 1.01 (0–4.00), 41 3.94 ± 1.16 (2.00–7.00), 45

73.01 ± 8.22 (54.00–87.00), 63 1 6.86 ± 0.67 (6.00–8.00), 30 7.50 ± 2.39 (2.50–12.50), 57 1.70 ± 0.74 (0–2.50), 56 1.97 ± 1.11 (0–4.00), 57 3.86 ± 1.21 (1.00–6.50), 57

69.61 ± 6.07 (53.00–80.00), 43 6.70 ± 1.94 (3.00–11.00), 42 1.38 ± 0.71 (0–2.50), 42 1.93 ± 1.12 (0–4.00), 42 3.39 ± 1.00 (1.00–5.50), 42

Girls

Total

–2.65 (86)

–1.55 (78)

.05

–2.00 (86) 0.95 (84)

.01

.13

.34

.30

.99

.17

.02

.64

.02

.80

.62

.98

.48

.73

.61

.55

p

–1.05 (88)

0.01 (89)

–2.45 (126) –0.47 (123) –2.31 (108.79) –1.38 (126)

–0.25 (55)

0.50 (135)

–0.02 (71)

–0.70 (71)

-0.34 (71)

–0.51 (71)

0.59 (74)

t(df)

Gender difference

(continued on next page)

77.75 ± 9.44 (60.00–96.00), 91 1 6.94 ± 1.71 (5.00–14.00, 90 6.81 ± 2.14 (2.50–13.00), 88 1.75 ± 0.62 (0.50–2.50), 86 1.61 ± 0.97 (0–4.00), 80 3.64 ± 1.15 (1.50–7.00), 88

73.41 ± 8.62 (54.00–90.00), 137 0 6.84 ± 0.65 (6.00–8.00), 57 6.95 ± 2.32 (1.50–12.50), 128 1.66 ± 0.70 (0–2.50), 125 1.75 ± 1.03 (0–4.00), 119 3.70 ± 1.18 (1.00–6.50), 128

70.03 ± 7.14 (43.00–82.00), 76 6.60 ± 2.07 (2.50–11.50), 73 1.36 ± 0.72 (0–2.50), 73 1.85 ± 1.12 (0–4.00), 73 3.39 ± 1.00 (1.00–5.50), 73

Total

Table 1 Mean ± SD (range) and valid N of the children's age, interval in months since any previous testing occasion, COBEL scores (total, food, social, and environmental) and the median of sessions already attended across the individual waves, subdivided according to gender and group.

L. Martinec Nováková and J. Havlíček

Physiology & Behavior 204 (2019) 224–233

2.00, 1

5.00, 1

COBEL social

COBEL environmental

227

Group 1

95.60 ± 5.21 (83.00–105.00), 23 2 13.82 ± 5.34 (5.00–25.00), 23 7.46 ± 1.78 (4.00–11.00), 23 1.67 ± 0.78 (0–2.50), 23 1.80 ± 0.91 (0–3.50), 23 3.98 ± 1.33 (2.50–6.50), 23

80.35 ± 4.03 (78.00–83.00), 2 2 4.32 ± 0.87 (4.00–5.00), 2 7.75 ± 3.89 (5.00–10.50), 2 1.75 ± 1.06 (1.00–2.50), 2 2.00 ± 1.41 (1.00–3.00), 2 4.00 ± 1.41 (3.00–5.00), 2

Girls

Significant findings are highlighted in bold.

COBEL environmental

COBEL social

COBEL food

Sessions attended Interval since last testing COBEL total

96.64 ± 5.61 (87.00–106.00), 16 2 12.64 ± 4.05 (5.00–18.00), 16 6.34 ± 1.65 (3.50–9.00), 16 1.56 ± 0.54 (1.00–2.50), 16 1.06 ± 0.89 (0–2.50), 16 3.72 ± 1.03 (2.00–5.50), 16

2.00, 1

COBEL food

Wave 5 Age (months)

9.00, 1

2 4.93, 1

82.23, 1

Boys

Sessions attended Interval since last testing COBEL total

Wave 4 Age (months)

Table 1 (continued)

96.03 ± 5.33 (83.00–106.00), 39 2 13.34 ± 4.83 (5.00–25.00), 39 7.00 ± 1.80 (3.50–11.00), 39 1.63 ± 0.69 (0–2.50), 39 1.50 ± 0.97 (0–3.00), 39 3.87 ± 1.21 (2.00–6.50), 39

80.98 ± 3.05 (78.00–83.00), 3 2 4.52 ± 0.71 (4.00–5.00), 3 8.17 ± 2.84 (5.00–10.50), 3 1.83 ± 0.76 (1.00–2.50), 3 2.00 ± 1.00 (1.00–3.00), 3 4.33 ± 1.15 (3.00–5.00), 3

Total

87.52 ± 8.89 (73.00–103.00), 22 2 9.75 ± 4.62 (5.00–19.00), 21 6.05 ± 2.35 (2.50–10.00), 22 1.76 ± 0.59 (1.00–2.50), 19 1.64 ± 0.82 (0.50–3.00), 14 3.48 ± 1.07 (1.50–5.50), 22

78.48 ± 7.20 (67.00–94.00), 21 2 7.74 ± 1.69 (5.00–12.00), 21 6.50 ± 1.84 (3.50–11.00), 21 1.76 ± 0.68 (0.50–2.50), 21 1.59 ± 1.08 (0.50–3.50), 16 3.52 ± 0.90 (1.50–5.00), 21

Boys

85.44 ± 6.79 (76.00–100.00), 20 2 9.15 ± 3.76 (5.00–19.00), 20 6.38 ± 2.49 (2.50–11.50), 20 1.56 ± 0.75 (0.50–2.50), 17 1.29 ± 0.77 (0.50–3.00), 17 3.95 ± 1.25 (2.00–6.50), 20

75.03 ± 3.09 (68.00–79.00), 11 2 7.45 ± 1.80 (5.00–12.00), 11 6.36 ± 1.76 (2.50–8.50), 11 1.50 ± 0.58 (1.00–2.50), 10 1.50 ± 0.83 (0.50–3.00), 9 3.77 ± 1.31 (1.50–6.00), 11

Girls

Group 2

86.53 ± 7.93 (73.00–103), 42 2 9.46 ± 4.18 (5.00–19.00), 41 6.20 ± 2.39 (2.50–11.50), 42 1.67 ± 0.67 (0.50–2.50), 36 1.45 ± 0.80 (0.50–3.00), 31 3.70 ± 1.17 (1.50–6.50), 42

77.29 ± 6.27 (67.00–94.00), 32 2 7.64 ± 1.71 (5.00–12.00), 32 6.45 ± 1.78 (2.50–11.00), 32 1.68 ± 0.65 (0.50–2.50), 31 1.56 ± 0.98 (0.50–3.50), 25 3.61 ± 1.05 (1.50–6.00), 32

Total

91.36 ± 8.86 (73.00–106.00), 38 2 11.00 ± 4.56 (5.00–19.00), 37 6.17 ± 2.07 (2.50–10.00), 38 1.67 ± 0.57 (1.00–2.50), 35 1.33 ± 0.89 (0–3.00), 30 3.58 ± 1.05 (1.50–5.50), 38

78.65 ± 7.07 (67.00–94.00), 22 2 7.61 ± 1.76 (5.00–12.00), 22 6.61 ± 1.87 (3.50–11.00), 22 1.77 ± 0.67 (0.50–2.50), 22 1.62 ± 1.05 (0.50–3.50), 17 3.59 ± 0.93 (1.50–5.00), 22

Boys

90.87 ± 7.83 (76.00–105.00), 43 2 11.65 ± 5.18 (5.00–25.00), 43 6.95 ± 2.18 (2.50–11.50), 43 1.63 ± 0.76 (0–2.50), 40 1.59 ± 0.88 (0–3.00), 40 3.97 ± 1.27 (2.00–6.50), 43

75.85 ± 3.65 (68.00–83.00), 13 2 6.96 ± 2.04 (4.00–12.00), 13 6.58 ± 2.03 (2.50–10.50), 13 1.54 ± 0.62 (1.00–2.50), 12 1.59 ± 0.89 (0.50–3.00), 11 3.81 ± 1.27 (1.50–6.00), 13

Girls

Total

91.10 ± 8.28 (73.00–106.00), 81 2 11.35 ± 4.88 (5.00–25.00), 80 6.59 ± 2.15 (2.50–11.50), 81 1.65 ± 0.67 (0–2.50), 75 1.48 ± 0.89 (0–3.00), 70 3.78 ± 1.18 (1.50–6.50), 81

77.61 ± 6.12 (67.00–94.00), 35 2 7.37 ± 1.86 (4.00–12.00), 35 6.60 ± 1.90 (2.50–11.00), 35 1.69 ± 0.65 (0.50–2.50), 34 1.61 ± 0.98 (0.50–3.50), 28 3.67 ± 1.06 (1.50–6.00), 35

Total

–1.48 (79)

–1.19 (68)

0.30 (73)

–1.65 (79)

–0.59 (78)

0.26 (79)

–0.58 (33)

0.07 (26)

0.99 (32)

0.05 (33)

0.99 (33)

1.54 (32.60)

t(df)

Gender difference

.14

.24

.77

.10

.56

.79

.57

.95

.33

.96

.33

.13

p

L. Martinec Nováková and J. Havlíček

Physiology & Behavior 204 (2019) 224–233

Physiology & Behavior 204 (2019) 224–233

L. Martinec Nováková and J. Havlíček

study with slightly older Czech children [6], in the present study it transpired that children did not fully understand the rating format of Item 3, termed “Senses in nature” [9,10]. Prompted by the question “When you walk in nature, what do you prefer?”, children were asked to rank the four sensory modalities of touching, smelling, watching, and listening in order of preference. Specifically, most children tended to disregard the items to be ranked and offered their own alternative, which mostly involved “playing” or “running around”. Therefore, this item was excluded from the interview. Thus, the total COBEL score, computed as the sum of the 15 items, ranged from 0 to 15. In addition to the total score, component scores for food (items 1, 2, and 16), social (items 11, 12, 13, and 14), and environmental odours (items 4, 5, 6, 7, 8, 9, 10, and 15) were computed after Ferdenzi and colleagues [5,8]. The theoretical ranges for food, social, and environmental odours were thus 0–3, 0–4, and 0–8, respectively. The actual ranges seen in the data are given in Table 1. The amended version of the questionnaire used in this study has been previously published by Martinec Nováková and colleagues [12].

in Wave 1 and were first tested during Wave 2, for these participants the starting date was dummy-coded. The code entered did not affect the results. The continuous variables of the initial age (that is, the age at the commencement of the study) and the interval since any previous testing were treated by a grand mean centering method (i.e., by subtracting the mean), which is generally recommended in order to simplify the interpretation of the results [19]. Next, following the strategy suggested by Singer and Willett [18], several models were fitted and then compared by means of the −2 log likelihood (i.e., the likelihood ratio test/deviance test) and the Akaike Information Criterion (AIC, “smaller is better”) in order to select the best model. Namely, to compare the models, we calculated delta AIC (Δi) as follows: AICi – AICmin, where AICi is the AIC value for model i, and AICmin is the AIC value of the “best” model. Then we followed the rule of thumb, whereby a ∆i < 2 indicates substantial evidence for the given model, values between 3 and 7 suggest that the model has considerably less support, whilst a ∆i > 10 says that the model is very unlikely [20]. Firstly, two unconditional models were generated to examine mean differences in the outcome variable (i.e., total or component COBEL scores) across individuals and to compare the fit of models estimated by means of the restricted maximum likelihood method (REML; default option) and the maximum likelihood (ML) method. As a better model fit was achieved by means of the latter method, all subsequent models were estimated using ML. Secondly, an unconditional growth model was tested, which served as a baseline model to explore whether the growth curves were linear or curvilinear. Thirdly, three higher-order polynomial models (quadratic, cubic, and exponential respectively) were fitted to investigate whether the rate of change accelerated or decelerated across time. In this way we found that only the linear growth effect was significant. Therefore, only the linear term was retained in the subsequent models, meaning that a linear slope model was always hypothesised. In other words, we modelled the individual change of COBEL scores as following a straight line rather than a curve. Fourthly, a conditional model was formed to determine whether the variables of gender, initial age, the cumulative total of sessions actually attended by a given child thus far, the time interval since any previous testing, and kindergarten/primary school attendance were related to the growth parameters (the initial status, known as the intercept, and the linear growth, known as the slope). It transpired that based on AIC, the best model fit was obtained when along with time, only gender (coded as -1 = girl, 1 = boy) and initial age were retained in the model. Thus, the number of sessions, the time interval since any previous testing, and kindergarten/primary school attendance were dropped from the subsequent models. Finally, several covariance structure models were explored to assess the error covariance structure of the longitudinal data, whereby we determined that the unstructured one tended to yield the best model fit. We allowed the intercept and the linear slope to vary across individuals. Missing data was handled through pairwise/listwise deletion. This procedure was followed to model the effects on each of the four COBEL scores (the total score and the food, social, and environmental components).

2.3. Procedure The testing sessions took place between 9 a.m. and 3 p.m. in a secluded, well-ventilated room without strong ambient odours. At the beginning of each session, children were briefly familiarised with the tasks, which were presented as a game, and ensured that they could stop or quit at any time. Since this study was part of a larger project (see [12,13] for other studies on this sample), the children also took various olfactory tests. The order of the olfactory tests and the interview based on COBEL was randomised across children. The sheer number of the tasks presented a cognitive load that could only be alleviated by splitting them over two sessions. Therefore, each child was tested on two consecutive days or within a week at the very latest. Each session took circa 30 min. Parents or teachers were never present in the room during the testing session. 2.4. Statistical analysis 2.4.1. Data exploration Analyses were carried out with SPSS 24.0 [14]. The normality of the raw data was checked, firstly, by visually examining the individual histograms of all relevant variables, secondly, by producing skewness and kurtosis values and their respective standard errors, from which zscores were computed and compared to the value of 1.96, as suggested by Field [15], and, thirdly, with multiple Shapiro-Wilk's W tests. 2.4.2. Linear mixed modelling Two approaches to analysis of longitudinal data were adopted to ascertain the robustness of the results. Both utilised linear mixed models, were run using the SPSS syntax MIXED command, and yielded very similar results. 2.4.2.1. Individual growth curve models. The first data analytic strategy consisted in fitting individual growth curve (IGC) models. In so doing, we followed the procedure recommended by Shek and Ma [16]. One of the advantages of IGC models is that they retain all of the information and variability in the data when examining the rate of change in the dependent variables (DVs) [17]. This information is of great importance to researchers in developmental psychology because children tend to vary not only in their initial status (i.e., the model's intercept), but also the rate of change (i.e., the slope). Also, in contrast to a repeatedmeasures approach, the IGC models allow the irregularity of the number and spacing of waves by means of a time-structured predictor (“time”) [18]. Thus, at Wave 1, the values of time were set at 0, and the number of months from the date of data collection within Wave 1 was calculated for each subsequent wave (i.e., Waves 2–5). In order to be able to perform these calculations for the children who did not take part

2.4.2.2. Repeated-measures analysis with time-dependent covariates. The other data analytic strategy involved a repeated-measures analysis with time-dependent (time-varying) covariates. Waves represented the repeated-measures effect, gender was treated as a fixed factor, and a child's age on a specific testing occasion, the cumulative total of sessions attended by the given child thus far, the time interval since any previous testing, and kindergarten/primary school attendance were handled as individual-level covariates that were measured repeatedly across the waves. However, the best model fit was also obtained after the number of sessions, the time interval since any previous testing, and kindergarten/primary school attendance were dropped from the models. Along with the main effect of gender, the gender*age interaction was also included in the model. The residual covariance matrix structure was diagonal with heterogeneous variance, which is 228

Physiology & Behavior 204 (2019) 224–233

L. Martinec Nováková and J. Havlíček

Table 2a F-statistics, p-values, estimates, 95% confidence intervals (95% CIs), and Cohen’s f2 effect sizes global (i.e., across waves) and local (i.e., within waves) for the effects of time, age at initial testing, and gender on COBEL measures in non-imputed data. Time Non-imputed data COBEL total F (df1, df2) p estimate 95% CI Cohen’s f2 Global Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 COBEL food F (df1, df2) p estimate 95% CI Cohen’s f2 Global Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 COBEL social F (df1, df2) p estimate 95% CI Cohen’s f2 Global Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 COBEL environmental F (df1, df2) p estimate 95% CI Cohen’s f2 Global Wave 1 Wave 2 Wave 3 Wave 4 Wave 5

Initial age

Gender

3.964 (1, 259.023) 0.048 0.020 [0, 0.041]

0.072 (1, 151.457) 0.789 -0.005 [-0.038, 0.029]

3.728 (1, 147.491) 0.055 0.507 [-0.012, 1.026]

< < < < < <

< 0.02 0.02 < 0.02 < 0.02 < 0.02 < 0.02

< 0.02 < 0.02 0.04 0.05 < 0.02 0.04

7.465 (1, 330.434) 0.007 0.010 [0.003, 0.017]

0.986 (1, 186.876) 0.322 0.005 [-0.005, 0.014]

0.108 (1, 184.389) 0.743 -0.025 [-0.173, 0.123]

0.02 < 0.02 < 0.02 < 0.02 0.02 0.02

< 0.02 < 0.02 < 0.02 < 0.02 0.02 0.02

< 0.02 < 0.02 < 0.02 < 0.02 0.04 < 0.02

6.123 (1, 106.343) 0.015 -0.014 [-0.025, -0.003]

0.332 (1, 139.837) 0.565 -0.005 [-0.020, 0.011]

6.531 (1, 134.082) 0.012 0.313 [0.071, 0.556]

< 0.02 < 0.02 < 0.02 < 0.02 < 0.02 0.14

< 0.02 0.05 < 0.02 < 0.02 0.02 < 0.02

< 0.02 < 0.02 0.05 0.03 < 0.02 0.08

7.223 (1, 90.639) 0.009 0.016 [0.004, 0.028]

0.402 (1, 145.236) 0.527 0.006 [-0.012, 0.023]

4.574 (1, 139.178) 0.034 0.294 [0.022, 0.566]

< 0.02 < 0.02 < 0.02 < 0.02 < 0.02 0.02

< 0.02 < 0.02 < 0.02 < 0.02 < 0.02 0.02

< 0.02 < 0.02 0.02 0.10 < 0.02 0.04

0.02 0.02 0.02 0.02 0.02 0.02

Table 2b F-statistics, p-values, estimates, 95% confidence intervals (95% CIs), and Cohen’s f2 effect sizes global (i.e., across waves) and local (i.e., within waves) for the effects of time, age at initial testing, and gender on COBEL measures in imputed data.

Imputed data COBEL total F (df1, df2) p estimate 95% CI Cohen’s f2 Global Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 COBEL food F (df1, df2) p estimate 95% CI Cohen’s f2 Global Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 COBEL social F (df1, df2) p estimate 95% CI Cohen’s f2 Global Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 COBEL environmental F (df1, df2) p estimate 95% CI Cohen’s f2 Global Wave 1 Wave 2 Wave 3 Wave 4 Wave 5

the default covariance structure for repeated effects. The model was run separately for the total COBEL scores and the three component scores (food, social, and environmental). The assumptions of the test were met since the DVs were linearly related to the independent variables (IVs) over at least some of the waves and they were normally distributed. Since the results matched those obtained with the first analytic strategy both in terms of statistical significance and effect size, they are not reported in the paper.

Time

Initial age

Gender

13.542 (1, 348.292) < 0.001 0.032 [0.015, 0.050]

0.157 (1, 145.027) 0.693 -0.006 [-0.034, 0.023]

1.179 (1, 148.715) 0.279 0.252 [-0.207, 0.712]

0.03 < 0.02 < 0.02 < 0.02 0.02 < 0.02

< 0.02 < 0.02 < 0.02 < 0.02 0.02 0.02

< 0.02 0.02 < 0.02 < 0.02 0.05 < 0.02

1.755 (1, 4081.798) 0.185 0.011 [-0.005, 0.027]

1.983 (1, 9.811) 0.190 -0.007 [-0.017, 0.004]

1.352 (1, 9.897) 0.272 -0.089 [-0.261, 0.082]

< 0.02 < 0.02 < 0.02 < 0.02 < 0.02 0.12

< 0.02 < 0.02 < 0.02 < 0.02 0.02 < 0.02

< 0.02 < 0.02 0.03 0.03 < 0.02 0.05

0.328 (1, 402.390) 0.567 -0.003 [-0.011, 0.006]

0.273 (1, 164.735) 0.602 -0.003 [-0.015, 0.009]

4.575 (1, 170.263) 0.034 0.212 [0.016, 0.408]

0.04 < 0.02 < 0.02 < 0.02 < 0.02 0.02

< 0.02 0.06 < 0.02 < 0.02 < 0.02 < 0.02

< 0.02 0.07 0.01 0.10 < 0.02 0.03

24.532 (1, 43.367) < 0.001 0.027 [0.016, 0.039]

2.895 (1, 14.151) 0.111 0.010 [-0.003, 0.022]

12.931 (1, 14.865) 0.003 0.339 [0.138, 0.540]

0.02 < 0.02 < 0.02 < 0.02 < 0.02 < 0.02

< 0.02 < 0.02 < 0.02 < 0.02 < 0.02 0.02

< 0.02 < 0.02 < 0.02 0.05 < 0.02 0.03

expressed in units of standard deviations. Hence, when d = 1, the means of two groups differ by one standard deviation. For mixed models, Cohen's f2 was computed using SAS PROC MIXED according to Selya [22]. Cohen's f2 for a given IV is a ratio of the proportion of the variance in the DV uniquely explained by the IV to the proportion of the variance in the DV unexplained by any variable in the model. In Tables 2a, b, global effect sizes across the waves (i.e., for the overall model) are reported as well as local ones within the individual waves. According to Cohen's [21] guidelines, f2 ≥ 0.02, f2 ≥ 0.15, and f2 ≥ 0.35 represent small, medium, and large effect sizes respectively. Cohen's f2 < 0.02 is below the recommended minimum effect size representing a “practically” significant effect for social science data [23], which is why the

2.4.3. Calculation of effect size For t-tests, a standardised measure of effect size, Cohen's d, was calculated after Cohen [21]. It is the difference between two means 229

Physiology & Behavior 204 (2019) 224–233

L. Martinec Nováková and J. Havlíček

exact values are not reported. 95% confidence intervals (95% CIs, given in square brackets) for the estimates were taken from the SPSS output. CIs can be interpreted in various ways [24]. Here we favoured the one stating that a 95% CI is an 83% prediction interval for the effect size estimate of a replication experiment [25]. 2.4.4. Data imputation To model the relationships on a larger dataset, we then re-ran the analyses on imputed data. There were 136 cases for which there were IVs available but the DVs were missing. This data was imputed in the R [26] with the missForest package [27], available from the Comprehensive R Archive Network (CRAN). Recommended particularly for conducting multiple imputation of mixed data (numeric and factor variables in one data frame) [28], it has been compared to other imputation methods and found to have the least imputation error for both continuous and categorical variables and the smallest prediction difference (error) [29]. The default settings were used [27,30]. 3. Results 3.1. Non-imputed data IGC models were fitted to assess the effect of the time-structured predictor (“time”), age at study entry, and gender on children's reports. The cumulative total of sessions attended by a given child thus far, the time interval since any previous testing, and kindergarten/primary school attendance were omitted, as explained above. As shown in Tables 2a, b all the effects were small or barely qualified as such (Cohen's f2 < 0.02), regardless of statistical significance. The only statistically significant effect on the total COBEL score was the linear, timerelated increase, F(1, 259.02) = 3.96, p = .048, Cohen's f2 < 0.02. To be specific, the longer a child was in the study, the higher their score. Food-related odour awareness also showed linear growth, F(1, 330.43) = 7.47, p = .007, Cohen's f2 < 0.02, with children achieving higher scores further into the study. In contrast, scores on the social component of COBEL decreased with passing time, F(1, 106.34) = 6.12, p = .015, Cohen's f2 = 0.02. Girls also scored higher than boys on the social component, F(1, 134.08) = 6.53, p = .012, Cohen's f2 < 0.02. Finally, children's awareness of environmental odours also increased further into the study, F(1, 90.64) = 7.22, p = .009, Cohen's f2 = 0.02, and girls scored higher than boys on this component, F(1, 139.18) = 4.57, p = .034, Cohen's f2 < 0.02. Fig. 1 shows component residual plots for the regression of total and component COBEL scores on the residuals of time in non-imputed data. Comparisons of COBEL scores in girls and boys are given in Fig. 2. 3.2. Imputed data In imputed data (N = 498 for the food and environmental components and N = 451 for the social component and the total score), we also found only small effects. Namely, the total COBEL score linearly increased with time, F(1, 348.29) = 13.54, p < .001, Cohen's f2 < 0.02. Food-related odour awareness did not change with time, F (1, 408,18) = 1.76, p = .185, Cohen's f2 = 0.02, in contrast to the nonimputed data. In line with the original data, girls exhibited greater awareness of social odours, F(1, 170.26) = 4.58, p = .034, Cohen's f2 < 0.02. There were, however, no changes in the social component with time, F(1, 402.39) = 0.33, p = .567, Cohen's f2 < 0.02. Finally, in line with what we reported for the non-imputed data, the environmental component also increased further into the study, F(1, 43.37) = 24.53, p < .001, Cohen's f2 = 0.04, and was higher in girls, F (1, 14.87) = 12.93, p = .003, Cohen's f2 < 0.02.

Fig. 1. Component residual plots for the regression of total and component COBEL scores on the residuals of time in non-imputed data. The fitted black dashed lines represent a linear relationship, loess fit lines are shown in grey. Deviations of the loess fit lines from the linear ones indicate non-linearity.

development of children's odour awareness across five waves of data collected over a two-year period. We investigated the effect of two proxies routinely invoked to explain inter-individual differences in

4. Discussion The present study represents the first longitudinal examination of 230

Physiology & Behavior 204 (2019) 224–233

L. Martinec Nováková and J. Havlíček

substantial cross-cultural differences, Saxton and colleagues [6] reported very similar results in Czech and Namibian children: girls turned out to be more acutely aware of social and environmental, but not food, odours. An alternative explanation for the effect of gender would be that the girls' higher COBEL scores were but a marginal expression of more general phenomena, such as better verbal skills. Nevertheless, in the original study on COBEL [5], girls showed greater verbal fluency than boys, but their reports of higher attention and reactivity to some odours (social, affective, and partly environmental ones, but not food ones) were not driven solely by their better verbal skills. Since COBEL is a metacognitive measure, another alternative could be that girls' greater metacognitive skills underlie their higher odour awareness scores. Although there is preliminary evidence that girls and boys do not differ in metacognition [38], there is clearly a knowledge gap that needs to be bridged by future studies. A time-related decrease in the scores reflecting awareness of social odours is surprising because the exact opposite was expected based on the previous studies. For instance, Ferdenzi and colleagues [5] noted a rise in children's awareness of certain personal odours. In an item-byitem analysis it transpired that the proportion of participants who were aware of the odour of their relatives increased between 6 and 10 years of age and that more and more children sought social odours when they were feeling sad [5]. A possible explanation of the present contrasting findings would be that the growing awareness of social odours may not necessarily become verbalised. On the contrary, children may become less explicit about their perceptions as they grow up. Scrutiny of the individual items comprising the social component suggested that at the beginning of the study, children did not hesitate to offer frank, sometimes even deliberately provocative, views on their own personal odours, those of their relatives, and peers. For example, at the commencement of the study, they would not feel embarrassed to state that their parents or siblings tended to have bad breath in the morning, break wind, or leave unpleasant smells behind after using the bathroom. Such statements would occur rather rarely towards the end of the study, leading to a drop in the scores since they were not replaced by more “socially appropriate” comments or were substituted with a simple “yes”. Even though these statements themselves were not scored, they would provide the context needed for the assessment of what was said next. For instance, at the start of the study, asking a question “Do you find that your parents/siblings smell of something?” would elicit vivid and detailed accounts of offensive odours emitted by various family members. When further prompted (“Imagine they did not smell of anything anymore: would you not care/would it bother you/would it suit you – a little/a lot? Why?”), a child's response would be guided by their preceding answer. Hence, a child reporting, for example, that someone tended to break wind would proceed to declare that “it is impolite and no one should do it” and that a change in the person's habits would have suited them a lot, scoring maximum points. Towards the end of the study, the same child might respond to the same question in the affirmative without going into detail. The lack of overt elaboration might have toned down their next statement and the degree of liking or disliking, resulting in lower scores. However, the absence of a detailed response may in fact reflect a child's growing, not diminishing, awareness of personal odours, combined with greater appreciation of social norms and their integration. It is possible that with experience, children become more acutely aware of the moralisation of body odours [39] and that they therefore begin to deliberately skirt around certain topics, recognising them as taboo. When spontaneous reports of offensive odours outnumber those of pleasant ones, this may, paradoxically, lead to lower scores. Statistical analysis of these ideas was not possible because for some children, only the scored parts, not the opening remarks providing the context for further interviewing, were available. Thus, we present these points more as suggestions for future research. One of the limitations of the questionnaire is that it lists several

Fig. 2. Violin plots for total and component COBEL scores in boys and girls. Box indicates the interquartile range (IQR), solid line represents the median, and dotted line indicates the mean. The upper whisker extends from the hinge to the highest value that is within 1.5 * IQR of the hinge and the lower whisker extends from the hinge to the lowest value within 1.5 * IQR of the hinge. The area around the box plot represents kernel probability density of the data at different values.

olfaction (i.e., gender and age), but also tested for within-subject changes over time. In general, we found a significant within-subject effect of individual maturation over time and a gender difference, but the initial age did not play a role. In imputed data, the statistical significance of some of the effects disappeared but, more importantly, the effect sizes remained small. The present findings indicate, firstly, that the effect of a child growing up (i.e., that of the “time” variable) on odour awareness in general seems rather small. Greater odour awareness in older children has been observed in several previous studies in participants aged 6–10 and 7–11 years [5,6,8], even though not consistently so [10,12]. A similar conclusion can be drawn from the study of Martinec Nováková and Vojtušová Mrzílková [10], who report a Kendall τ correlation of 0.061 between age and the total COBEL score, which gives a Cohen's d of 0.19 [31]. This, too, represents a small effect [21]. Effect sizes of similar magnitude can be extracted from the study on the original use of the questionnaire [5]. This is further in line with the finding of Saxton and colleagues [6], who, in the total sample of older children (between 8 and 15 years of age), reported that “COBEL scores increased with age (F(1, 169) = 5.51, p = 0.020)”. At N = 174, assuming equal sample sizes, this gives a Cohen's d of roughly 0.36 [32] for the total score, which again falls in the small effect range. They also analysed the individual components of COBEL and arrived at similar results, with a d of 0.33 and 0.42 for the food and the environmental contexts, respectively, but < 0.10 for the social one. The present study provides corroborating evidence. Secondly, another small effect was that of gender. One possibility is that the reports indeed reflect girls' greater reliance on their sense of smell compared to that of boys'. This would be in line with previous studies in children [6,8] and adults [3,33–37]. For example, in a crosscultural study of Finnish and French children [8], girls reported greater attention and reactivity to everyday odours than boys. Specifically, girls scored significantly higher on six items, of which four predominantly related to social odours (body odour of relatives, tendency to smell parts of own body and clothes, and odours sought when sad, which most of the time were odours of self or of significant others). Despite 231

Physiology & Behavior 204 (2019) 224–233

L. Martinec Nováková and J. Havlíček

items which may imply the desired direction of a response, such as “Imagine your parents present you a dish you do not know: will you do something before putting it in your mouth?”, “When you smell a food odour, do you try to guess for fun what it is?”, or “Do you find that people smell of something, even without perfume or deodorant?” These items are just as problematic in Czech as they are in English. These socalled leading questions are phrased in such a way as to suggest the desired answer. Their inherent biases can operate as “suggestions”, encouraging children to answer in the affirmative. Young children are particularly susceptible to the effects of “leading” information [40]. Pre-schoolers are more susceptible to suggestion in general than older children and adults [41,42]. A child's responses may be biased especially when the interviewer is perceived as a person in authority [43] and when the interview is conducted at a kindergarten or school [44]. The impact of such biases decreases as the social distance between an adult interviewer and a respondent shortens, i.e., as the child grows up or the level of formality decreases [45]. To counteract such effects, it is recommended to exclude a child's first answer as it may represent a response that they believe is expected, desirable, or appropriate [46]. The research protocol in the present study did not allow for such an approach. Further, children tend to perform poorly when they feel that they are expected to give an answer, no matter how confident they are [47]. Allowing for an “I don't know” response in future studies may reduce children's susceptibility to such biases in interviewing [48]. A limitation of the present study was that it did not delve into the actual factors underlying the age-related variation in self-reports of odour awareness. This tends to be attributed to a combination of general cognitive maturation (including memory, verbal skills, and the ability to interpret and describe the environment), together with the cumulative effects of odour exposure [5,8]. For instance, we have already shown that the effects of verbal fluency and odour exposure, respectively, were greater than the effect of age [12]. Besides controlling for verbal fluency, researchers might also find it useful to employ an independent measure of metacognitive skills, as discussed above, such as the train-track task [38], which requires a child to build a traintrack from wooden pieces to match a shape in a plan. Another limitation of the study was the amount of missing data and related issues, such as the higher proportion of pre-schoolers in Wave 4 compared to Wave 3, the similar mean age of the children participating in these two waves, and the low N in Wave 4. The data collection within Wave 4 took place in May and June 2012, which was towards the busy end of the school year. As school principals, teachers, and parents each time provided a one-time permission only and had to be asked again within the next wave, sometimes they responded in the negative to our request. Kindergarten teachers and parents of pre-schoolers were generally more relaxed about the testing and more interested or even enthusiastic about cooperating. Besides, pre-schoolers tended to have similar schedules and could be approached within the kindergarten almost on any given day, as opposed to school children. When the first participants started school during Wave 3, they turned out to be difficult to reach because of their busy curricula and after-school activities. Parents of the children who could not be tested at school were then invited to visit our laboratory, but they often found it logistically inconvenient. Future studies could overcome many of these issues by recruiting children from institutions with more relaxed curricula, such as outdoor kindergartens and elementary schools. This approach might also allow ecologically (externally) valid observation of olfactory behaviours.

higher than boys, with the single exception of the food component. All the effects were small in both the original (i.e., non-imputed) and imputed data. The age at which children entered the study did not affect their scores. We suggest that particularly the unexpected findings regarding social odours warrant replication and extension in longitudinal studies carried over a broader time span. Declarations of interest None. Author contributions Conceived and designed the study: LMN and JH. Performed the study: LMN. Analysed the data: LMN. Wrote the paper: LMN and JH. Acknowledgements The authors would like to express their gratitude to Jitka Fialová and Markéta Sobotková for their help with data collection and Lydie Kubicová for her assistance in maintaining the participant database. A special word of thanks goes to David Le Sage for proofreading. We are grateful to the children and their parents for participation, and to school principals and teachers for allowing us to perform the study on the premises of their schools. This study is a result of research funded by the project LO1611 with financial support from the Ministry of Education, Youth, and Sports (MEYS) under the NPU I program. It has also been supported by the Charles University Research Centre (UNCE) program UNCE/HUM/025 (204056). Further, LMN was supported by the Czech Science Foundation (GA17-14534S), PROGRES program Q22 “Antropologická bádání v rámci přírodních, humanitních a historických věd” at the Faculty of Humanities, Charles University within the Institutional Support for Long-Term Development of Research Organizations from MEYS, and by the Specific Academic Research project (Specifický vysokoškolský výzkum, SVV) number 260 469 (“Adaptivní mechanismy v lidské psychice”) carried out at the Faculty of Humanities, Charles University. The funding sources were not involved in the study design, in the collection, analysis, and interpretation of the data, in the writing of the report, or in the decision to submit the article for publication. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.physbeh.2019.02.035. References [1] L. Nováková, J.V. Valentova, J. Havlíček, Engagement in olfaction-related activities is associated with the ability of odor identification and odor awareness, Chemosens. Percept. 7 (2014) 56–67. [2] I. Croy, D. Buschhuter, H.S. Seo, S. Negoias, T. Hummel, Individual significance of olfaction: development of a questionnaire, Eur. Arch. Otorhinolaryngol. 267 (2010) 67–71. [3] M.L. Dematte, I. Endrizzi, F. Biasioli, M.L. Corollaro, M. Zampini, F. Gasperi, Individual variability in the awareness of odors: demographic parameters and odor identification ability, Chemosens. Percept. 4 (2011) 175–185. [4] M.A.M. Smeets, H.N.J. Schifferstein, S.R. Boelema, G. Lensvelt-Mulders, The odor awareness scale: a new scale for measuring positive and negative odor awareness, Chem. Senses 33 (2008) 725–734. [5] C. Ferdenzi, G. Coureaud, V. Camos, B. Schaal, Human awareness and uses of odor cues in everyday life: results from a questionnaire study in children, Int. J. Behav. Dev. 32 (2008) 422–431. [6] T.K. Saxton, L. Martinec Nováková, R. Jash, A. Šandová, D. Plotěná, J. Havlíček, Sex differences in olfactory behavior in Namibian and Czech children, Chemosens. Percept. 7 (2014) 117–125. [7] A. Wrzesniewski, C. McCauley, P. Rozin, Odor and affect: individual differences in the impact of odor on liking for places, things and people, Chem. Senses 24 (1999)

5. Conclusion To the best of our knowledge, this paper presents the first longitudinal study to repeatedly test children's odour awareness across five waves over a two-year period. We found a time-related linear increase in the total COBEL scores and the food and environmental components, whereas awareness of social odours decreased over time. Girls scored 232

Physiology & Behavior 204 (2019) 224–233

L. Martinec Nováková and J. Havlíček 713–721. [8] C. Ferdenzi, S. Mustonen, H. Tuorila, B. Schaal, Children's awareness and uses of odor cues in everyday life: a Finland-France comparison, Chemosens. Percept. 1 (2008) 190–198. [9] L. Martinec Nováková, R. Vojtušová Mrzílková, A. Kernerová, Gender differences in influences of temperament on olfactory reactivity and awareness, Sci. Rep. 7 (2017) 8920. [10] L. Martinec Nováková, R. Vojtušová Mrzílková, Children's exposure to odors in everyday contexts predicts their odor awareness, Chemosens. Percept. 9 (2016) 56–68. [11] C. Ferdenzi, G. Coureaud, V. Camos, B. Schaal, Attitudes toward everyday odors for children with visual impairments: a pilot study, J. Vis. Impair. Blind. 104 (2010) 55–59. [12] L. Martinec Nováková, J. Fialová, J. Havlíček, Effects of diversity in olfactory environment on children's sense of smell, Sci. Rep. 8 (2018) 2937. [13] L. Martinec Nováková, J. Fialová, J. Havlíček, Development of children's olfactory abilities and odor awareness is not predicted by temperament: a longitudinal study, Chemosens. Percept. 11 (2018) 59–71. [14] IBM Corp, IBM SPSS Statistics for Windows, 24th ed., IBM Corp, Armonk, NY, 2016. [15] A. Field, Discovering Statistics Using SPSS, SAGE Publications, London, 2005. [16] D.T.L. Shek, C.M.S. Ma, Longitudinal data analyses using linear mixed models in SPSS: concepts, procedures and illustrations, Sci. World J. 11 (2011) 42–76. [17] J.L. Miner, K.A. Clarke-Stewart, Trajectories of externalizing behavior from age 2 to age 9: relations with gender, temperament, ethnicity, parenting, and rater, Dev. Psychol. 44 (2008) 771–786. [18] J.D. Singer, J.B. Willett, Applied Longitudinal Data Analysis, Oxford Press, New York, 2003. [19] J.J. Hox, Multilevel Analysis: Techniques and Applications, Erlbaum, Hillsdale, NJ, 2002. [20] K.P. Burnham, D.R. Anderson, Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach, 2nd ed. Springer-Verlag, New York, 2002. [21] J.E. Cohen, Statistical Power Analysis for the Behavioral Sciences, Lawrence Erlbaum Associates, Inc, Hillsdale, NJ, 1988. [22] A.S. Selya, J.S. Rose, L.C. Dierker, D. Hedeker, R.J. Mermelstein, A practical guide to calculating Cohen's f(2), a measure of local effect size, from PROC MIXED, Front. Psychol. 3 (2012) 111. [23] C.J. Ferguson, An effect size primer: a guide for clinicians and researchers, Prof. Psychol. Res. Pract. 40 (2009) 532–538. [24] G. Cumming, The new statistics: why and how, Psychol. Sci. 25 (2014) 7–29. [25] G. Cumming, R. Maillardet, Confidence intervals and replication: where will the next mean fall? Psychol. Methods 11 (2006) 217–227. [26] R. Development Core, Team. A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, 2008. [27] D.J. Stekhoven, missForest: Nonparametric Missing Value Imputation Using Random Forest, 1.4 ed., R Foundation for Statistical Computing, Vienna, Austria, 2013. [28] J. Starkweather, A new recommended way of dealing with multiple missing values:

[29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48]

233

using missForest for all your imputation needs, Benchmarks RSS Matters 2014 (July 2014). A.K. Waljee, A. Mukherjee, A.G. Singal, Y. Zhang, J. Warren, U. Balis, et al., Comparison of imputation methods for missing laboratory data in medicine, BMJ Open 3 (2013) e002847. D.J. Stekhoven, Package ‘missForest’: Nonparametric Missing Value Imputation using Random Forest, Swiss Federal Institute of Technology, Zürich, Switzerland, 2013. D.A. Walker, JMASM9: converting Kendall's tau for correlational or meta-analytic analyses, J. Mod. Appl. Stat. Methods 2 (2003) 525–530. M.W. Lipsey, D.B. Wilson, Practical Meta-Analysis, Sage Publications, Thousand Oaks, 2001. R.S. Herz, E.D. Cahill, Differential use of sensory information in sexual behavior as a function of gender, Hum. Nat. 8 (1997) 275–286. J. Havlíček, T.K. Saxton, S.C. Roberts, E. Jozífková, S. Lhota, J. Valentová, et al., He sees, she smells? Male and female reports of sensory reliance in mate choice and non-mate choice contexts, Personal. Individ. Differ. 45 (2008) 565–570. R.S. Herz, M. Inzlicht, Sex differences in response to physical and social factors involved in human mate selection - the importance of smell for women, Evol. Hum. Behav. 23 (2002) 359–364. K. Sucker, H. Berresheim, H. Ramcke-Krüll, P. Schulze, T. Brüning, J. Bünger, Approach to characterize a sub-group susceptible to odour annoyance, Chem. Eng. Trans. 23 (2010) 99–104. E. Buron, A. Bulbena, G. Pailhez, A.B. Cabre, The Spanish version of two olfactory scales: reliability and validity, Rev. Psiquiatr. Salud Ment. 4 (2011) 187–194. D. Bryce, D. Whitebread, The development of metacognitive skills: evidence from observational analysis of young children's behavior during problem-solving, Metacogn. Learn. 7 (2012) 197–217. M.L.M. Soo, R.J. Stevenson, The moralisation of body odor, Mank. Q. 47 (2007) 25–56. S. Moston, How children interpret and respond to questions: situational sources of suggestibility in eyewitness interviews, Soc. Behav. 5 (1990) 155–167. G.S. Goodman, C. Aman, Children's use of anatomically detailed dolls to recount an event, Child Dev. 61 (1990) 1859–1871. G.S. Goodman, R.S. Reed, Age differences in eyewitness testimony, Law Hum. Behav. 10 (1986) 317–332. S.J. Ceci, D.F. Ross, M.P. Toglia, Suggestibility of children's memory: psycholegal implications, J. Exp. Psychol. Gen. 116 (1987) 38–49. F. Leach, Researching gender violence in schools: methodological and ethical considerations, World Dev. 34 (2006) 1129–1147. M.S. Kwock, G.A. Winer, Overcoming leading questions: effects of psychosocial task variables, J. Educ. Psychol. 78 (1986) 289–293. S. Walker, K. Irving, D. Berthelsen, Gender influences on preschool children's social problem-solving strategies, J. Genet. Psychol. 163 (2002) 197–209. H.R. Dent, G.M. Stephenson, An experimental study of the effectiveness of different techniques of questioning child witnesses, Br. J. Clin. Psychol. 18 (1979) 41–51. N. Schwarz, F. Strack, H.-J. Hippler, G. Bishop, The impact of administration mode on response effects in survey measurement, Appl. Cogn. Psychol. 5 (1991) 193–212.