Edited by:
Reviewed by:
*Correspondence:
This article was submitted to Language Sciences, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Emotions play an important role in human communication, and the daily-life interactions of young children often include situations that require the verbalization of emotional states with verbal means, e.g., with emotion terms. Through them, one can express own emotional states and those of others. Thus, the acquisition of emotion terms allows children to participate more intensively in social contexts – a basic requirement for learning new words and for elaborating socio-emotional skills. However, little is known about how children acquire and process this specific word category, which is positioned between concrete and abstract words. In particular, the influence of valence on emotion word processing during childhood has not been sufficiently investigated. Previous research points to an advantage of positive words over negative and neutral words in word processing. While previous studies found valence effects to be influenced by factors such as arousal, frequency, concreteness, and task, it is still unclear if and how valence effects are also modified by age. The present study compares the performance of children aged from 5 to 12 years and adults in two experimental tasks: lexical decision (word or pseudoword) and emotional categorization (positive or negative). Stimuli consisted of 48 German emotion terms (24 positive and 24 negative) matched for arousal, concreteness, age of acquisition, word class, word length, morphological complexity, frequency, and neighborhood density. Results from both tasks reveal two developmental trends: First, with increasing age children responded faster and more correctly, suggesting that emotion vocabulary gradually becomes more stable and differentiated during middle childhood. Second, the influence of valence varied with age: younger children (5- and 6-year-olds) showed significantly higher performance levels for positive emotion terms compared to negative emotion terms, whereas older children and adults did not. This age-related valence effect in emotion word processing will be discussed with respect to linguistic and methodological aspects.
The ability to verbalize emotional states is a crucial stepping stone not only in language acquisition, but also for a child’s social-emotional development, since it enables children to participate in social contexts which form an essential learning environment. In their third year of life children begin to produce their first words to express emotional states (such as
Valence, the perceived value of a stimulus’ pleasantness, was shown to be one of two basic dimensions along which humans classify emotional content from their environment (
The present study aims to shed light on the question of if and how valence influences children’s processing of emotion terms. Can an enhanced processing of positive or negative emotional content in children be observed in language, and do changes occur in the course of development? Answering these questions will contribute to a comprehensive understanding of the development of emotion processing skills during childhood. As stated above, emotion terms are characterized by carrying a specific value of valence. Moreover, their acquisition starts in early childhood (
When investigating emotion and valence effects researchers often use the well-known psycholinguistic lexical decision paradigm (lexical decision task, LDT) in which visually or verbally presented letter or phoneme strings are to be judged as a word or a pseudoword (
It is generally agreed that affective features of a word’s meaning (such as emotionality, characterized by a certain value of valence and arousal) influence performance levels in word processing. A large number of studies using LDT point toward a processing advantage of emotionally toned words over neutral ones (for English:
Research with adult participants shows heterogeneous results regarding the question of whether negative or positive words are processed more efficiently. A negativity bias, i.e., an improved perception of negative words in the form of faster responses and/or higher accuracy rates, was found in several studies using VDTs (e.g.,
However, the majority of studies with adults using VDT, LDT, and other experimental approaches such as memory, attention, reading, or naming tasks (e.g.,
As mentioned above, all previously reported studies investigating the effect of emotionality and valence were conducted with adult participants using written word stimuli. Studies on affective and emotion word processing in children are sparse. One study by
Another more indirect task to explore the valence effect was used by
On the other hand, a positivity advantage has not been consistently found in children.
Several factors have been shown to modulate the role of valence in word processing: word concreteness, arousal, word frequency, and the task itself. To begin with concreteness,
In summary, many studies with adults have consistently reported an enhanced processing of emotional over neutral word stimuli, while evidence for valence effects is more inconsistent. Although the majority of study results point to a positivity advantage (shown in many studies with adults and in two studies with children), some studies either found no differences between the two valence groups or even the opposite pattern (a negativity advantage). Hence, the direction of the valence effect still remains unresolved. In addition, the factors leading to an enhanced processing (e.g., faster responses or higher accuracy rates) of positive or of negative word stimuli (such as concreteness, arousal, frequency, and task) have not always been sufficiently controlled for in previous research. Most importantly, there is a considerable imbalance with respect to the participant groups investigated so far: the majority of findings is based on adults’ behavior, whereas only three studies investigated the effect of valence in children. To our knowledge, children’s and adults’ processing of affective words or emotion terms have never been compared so far. Thus, it remains unclear whether and how valence effects are modified by age, as developmental studies on emotion and valence effects are extremely sparse.
In light of the reported lack of developmental studies, we employed two experimental reaction time tasks in order to investigate how accurately and how quickly children (from 5 to 12 years of age) and adults process verbally presented German emotion terms: (1) a LDT with neutral words as well as positive and negative emotion terms and (2) a VDT with positive and negative emotion terms. We used emotion terms exclusively, in order to avoid a mixture of semantically different words (emotion terms vs. affective words). Since all participants performed both tasks the experimental design additionally allows for conclusions about the expected valence effect with respect to the task relevancy factor, i.e., whether the stimulus’ valence is important for the optimal solution of the task or not. For the LDT mentally stored lexical units need to be accessed and compared with a presented string of phonemes for checking its existence as a proper word. Meanwhile, for the VDT, semantic information regarding the word’s valence must be accessed from the mental representation of the emotion term. Thus, in VDT valence is clearly task-relevant, whereas in LDT it is not. The order of the two experiments was kept constant for all participants, to maintain their level of motivation. The longer (and potentially more complex) LDT was always conducted first. Error rates and response times for LDT and VDT were analyzed to detect effects of valence and age. Based on the reported findings from children and adults, we put forward the following questions and hypotheses:
Developmental changes in task performance:
In line with previous word processing studies with children (e.g.,
Valence effect:
Can a valence effect be observed in LDT and VDT? Given that the majority of findings are in favor of a positivity advantage, valence effects were expected in both tasks, with positive emotion terms being processed faster and more accurately than negative emotion terms. Based on the task relevancy factor, we expected valence effects in VDT to be stronger compared to the LDT.
Does age influence the appearance and direction of this effect? Since developmental studies comparing adults’ and children’s performance in the processing of affective words and emotion terms are as yet unavailable, we cannot derive task-specific directional hypotheses concerning the question of whether, and if so, how, age impacts the expected valence effects.
Participants were 96 typically developing children of four age groups recruited from local daycare services and schools in the cities of Gießen, Marburg, and surrounding areas (Hesse, Germany): 5-year-olds (
Twenty-four adult participants (students from the University of Gießen) served as a control group (
All participants grew up as monolingual native speakers of German. Gender was balanced both within and between the different age groups (12 female and 12 male participants per group).
The stimuli consisted of 24 positive (e.g.,
Descriptive statistics for selected neutral words, as well as positive, and negative emotion terms controlled for arousal (on a scale from 1 = low-arousing to 5 = high arousal), valence (on a scale from 1 = very negative, over 4 = neutral, to 7 = very positive), concreteness (on a scale from 1 = very abstract, over 4 = neutral, to 7 = very concrete).
Variables | Valence Children rating by Bahn et al. (under review) | Valence Adults emotion terms: rating by Bahn et al. (under review) neutral words: BAWL-R | Arousal Children rating by Bahn et al. (under review) | Arousal Adults emotion terms: rating by Bahn et al. (under review) neutral words: BAWL-R | Concreteness online-rating with adults, unpublished data, obtained by Bahn and colleagues |
---|---|---|---|---|---|
(1) Positive emotion terms | 5.87 | 5.65 | 2.88 | 3.15 | 3.72 |
(2) Negative emotion terms | 2.43 | 2.40 | 2.99 | 3.36 | 3.78 |
(3) Neutral words | Children rated only positive (1) and negative (2) words | 3.89 | Children rated only positive (1) and negative (2) words | 2.06 | 5.42 |
One-way ANOVA on each factor with |
Descriptive statistics for selected neutral words, as well as positive, and negative emotion terms controlled for age of acquisition, frequency, number of phonemes and morphemes, neighborhood density, mean duration and mean pitch of recorded word stimuli.
Variables | Age of acquisition (age; months) Online-rating with adults, unpublished data, obtained by Bahn and colleagues | Absolute Frequency 1/Mio ChildLex | Number of phonemes BAWL-R | Number of morphemes BAWL-R | Neighborhood density BAWL-R | Mean duration of recorded word stimuli in seconds | Mean pitch of recorded word stimuli in Hz |
---|---|---|---|---|---|---|---|
(1) Positive emotion terms | 4;7 | 62.49 | 5.38 | 1.71 | 7.54 | 0.76 | 153 |
(2) Negative emotion terms | 5;1 | 40.74 | 5.25 | 1.71 | 7.21 | 0.74 | 149 |
(3) Neutral words | 4;5 | 36.71 | 5.21 | 1.71 | 10.79 | 0.72 | 148 |
One-way ANOVA on each factor |
The stimuli were originally selected from the BAWL-R (
Since it seemed questionable whether adult norms of valence and arousal are appropriate for experiments with children, we then conducted two additional rating studies with 60 typically developing 9-year-old children and 60 adults (Bahn et al., under review). Participants judged the value of valence and arousal of each of the 48 preselected emotion terms. In contrast to the BAWL-R ratings, items in our rating studies were presented audibly in order to make the task equally feasible for children. Results showed that children’s and adults’ values were very similar (see
To obtain age of acquisition norms (AoA), which were not available for the full word list, two rating surveys were conducted online with 96 employees and students of two German universities (Marburg and Gießen) for the emotion terms, and 202 employees and students for the neutral words. Participants estimated on a 7-Point Likert Scale at what particular age (from the age of 2 to 8 years and older) a child most probably knows the meaning of the words. AoA values were derived from the means of all responses. Norms of concreteness for the three word sets were collected from 411 participants (employees and students) in another online rating study at the University of Marburg. Using a 7-Point Likert Scale from 1 (very abstract) to 7 (very concrete), participants were asked to assign a specific value of concreteness to each of the emotion terms and neutral words. Again, norms were derived by averaging values for each item across all participants. Values of absolute frequency (1/mil) were taken from the ChildLex corpus (
One-way ANOVAs confirmed that: (1) the three word sets (positive, negative, and neutral) differed significantly with respect to valence, regardless of whether adult norms or children’s norms were considered. (2) Positive and negative words did not differ in their mean value of arousal, regardless of whether adult norms or children’s norms were considered. (3) Neutral words and emotion terms significantly differed with respect to arousal. This is inevitable, since emotion words always involve a higher degree of emotional activation than neutral words. (4) Neutral words and emotion terms significantly differed in their average value of concreteness: Emotion terms were less concrete than neutral words. (5) Positive and negative emotion terms did not show a difference in the mean value of concreteness. (6) The three word sets did not differ with respect to Age of Acquisition, frequency (neither for ChildLex norms, nor for CELEX norms, based on adult text corpora), morphological complexity, word length, or neighborhood density (see
Next, for each word stimulus an appropriate pseudoword (
Finally, all word and pseudoword stimuli were recorded in a soundproofed booth, spoken by one female and one male trained native speaker of standard German using neutral prosody for all three word categories and pseudowords. As shown in
First, all participants were informed about the study. After receiving all information and any remaining questions were answered, informed consent forms needed to be signed by all adult participants and parents. Children needed to verbally agree to their participation. Parents additionally filled in a developmental questionnaire in order to exclude any delays or disorders regarding their child’s language or cognitive development.
With 5-, 6-, and 9-year-old children, the whole procedure comprised two sessions with the following order of tasks: First session: vocabulary test and CPM, second session: LDT and VDT, whereas 12-year-olds and adults performed all tasks during one session. One break between both tasks of each session served to keep the participants attentive. Children were either tested in the laboratory, at school or kindergarten. Adults always participated in the laboratory. Children were given a small gift, adults were either monetarily rewarded or received student credit. The study was approved by the local Ethics Committee of the University of Gießen.
For the LDT, participants were seated in a quiet room with a laptop with a 15.4 inch LCD screen in front of them. Items were presented verbally via headphones. OpenSesame (
Average accuracy and response times were calculated for each participant and item. For the analysis of accuracy, reactions from 15 5-year-olds, 21 6-year-olds, 24 9-year-olds, 24 12-year-olds, and 24 adults could be considered, which corresponded to 90% of all collected single reactions (120 participants∗192 words). For reaction times, from these analysable reactions outliers were excluded in a stepwise procedure: (1) only correct responses to words and pseudowords were considered for analysis (11% excluded, 89% remaining of analysable reactions). (2) Exclusion of the lowest and highest 5% of reactions that were seen as extreme outliers, e.g., due to equipment error or distraction (10% excluded, 79% remaining of all analysable reactions). (3) Exclusion of single reactions that were at the same time atypical both for a particular participant and for a particular item, i.e., above or below 2 standard deviations from the participant’s AND the item’s mean (1% excluded, 78% remaining of all analysable reactions). (4) Exclusion of participants that showed an atypical mean response time with respect to their age-mates, i.e., more than 2 standard deviations above or below the group mean (corresponds to three participants, one each for the 9-year-olds, 12-year-olds and adults, 2% reactions excluded). Thus, 76% of all analysable reactions remained for the analysis (68% of the data from the 5-year-olds, 71% from the 6-year-olds, 77% from the 9-year-olds, 78% from the 12-year-olds, and 80% of the adult data).
Using the cleaned up raw trial-by-trial data for accuracy and response time, we carried out linear (LMER) and generalized linear (GLMER) mixed effect regression analyses in the R programming environment (
For the analysis of accuracy a generalized linear mixed-effects regression modeling approach was chosen (link = logit, fitted by Laplace approximation). The model included a random intercept for each participant and a random slope of valence for each participant. Further, we included an intercept for each word-item in the analyses. In
Estimated means of age and valence for LDT.
Estimated mean accuracy and mean response time with (SE) | 5 years | 6 years | 9 years | 12 years | Adults |
---|---|---|---|---|---|
Accuracy (positive, negative, and neutral) | 77% (4.8%) | 91% (2.1%) | 96% (0.9%) | 98% (0.4%) | 99% (0.3%) |
Accuracy (positive) | 92% (3%) | 96% (1%) | 98% (0.8%) | 99% (0.3%) | 99% (0.3%) |
Accuracy (negative) | 67% (9%) | 89% (4%) | 96% (1%) | 99% (0.5%) | 100% (0.1%) |
Accuracy (neutral) | 73% (6%) | 87% (3%) | 95% (1%) | 97% (0.9%) | 98% (0.6%) |
Response time (positive, negative, and neutral) | 1049 ms (35 ms) | 844 ms (30 ms) | 699 ms (28 ms) | 605 ms (28 ms) | 319 ms (28 ms) |
Response time (positive) | 1015 ms (43 ms) | 867 ms (36 ms) | 697 ms (34 ms) | 579 ms (34 ms) | 300 ms (34 ms) |
Response time (negative) | 1005 ms (41 ms) | 793 ms (35 ms) | 664 ms (34 ms) | 995 ms (34 ms) | 301 ms (34 ms) |
Response time (neutral) | 1126 ms (36 ms) | 871 ms (31 ms) | 735 ms (29 ms) | 641 ms (29 ms) | 356 ms (29 ms) |
Accuracy age contrast | β-value | Response time age contrast | β-value | ||||||
---|---|---|---|---|---|---|---|---|---|
5–6 | -13% | 5% | -2.69 | 0.055 | 5-6 | 205 ms | 45 ms | 4.61 | 0.000 |
5–9 | -19% | 5% | -3.93 | 0.001 | 5-9 | 350 ms | 44 ms | 8.01 | 0.000 |
5–12 | -21% | 5% | -4.39 | 0.000 | 5-12 | 444 ms | 44 ms | 10.16 | 0.000 |
5–adults | -22% | 5% | -4.51 | 0.000 | 5-adults | 730 ms | 44 ms | 16.71 | 0.000 |
6–9 | -5% | 2% | -2.53 | 0.084 | 6-9 | 145 ms | 39 ms | 3.68 | 0.003 |
6–12 | -8% | 2% | -3.72 | 0.002 | 6-12 | 239 ms | 39 ms | 6.06 | 0.000 |
6–adults | -8% | 2% | -4.01 | 0.001 | 6-adults | 525 ms | 39 ms | 13.33 | 0.000 |
9–12 | -2% | 0.9% | -2.45 | 0.102 | 9-12 | 94 ms | 38 ms | 2.43 | 0.115 |
9–adults | -3% | 0.9% | -3.21 | 0.012 | 9-adults | 380 ms | 38 ms | 9.88 | 0.000 |
12–adults | -0.7% | 0.4% | -1.46 | 0.588 | 12-adults | 286 ms | 38 ms | 7.45 | 0.000 |
5 years positive-negative | 25% | 8% | 3.05 | 0.006 | 5 years positive-negative | 9 ms | 34 ms | 0.29 | 0.955 |
5 years positive-neutral | 18% | 6% | 3.27 | 0.003 | 5 years positive-neutral | -111 ms | 34 ms | -3.32 | 0.003 |
5 years negative-neutral | 6% | 8% | -0.78 | 0.714 | 5 years negative-neutral | -121 ms | 34 ms | -3.58 | 0.001 |
6 years positive-negative | 7% | 3% | 2.01 | 0.110 | 6 years positive-negative | 73 ms | 29 ms | -2.57 | 0.030 |
6 years positive-neutral | 8% | 3% | 2.78 | 0.015 | 6 years positive-neutral | -5 ms | 29 ms | -0.17 | 0.985 |
6 years negative-neutral | 2% | 4% | 0.52 | 0.864 | 6 years negative-neutral | -78 ms | 30 ms | -2.58 | 0.029 |
9 years positive-negative | 2% | 1% | 1.29 | 0.402 | 9 years positive-negative | 33 ms | 28 ms | 1.20 | 0.454 |
9 years positive-neutral | 3% | 1% | 2.12 | 0.087 | 9 years positive-neutral | -37 ms | 28 ms | -1.32 | 0.384 |
9 years negative-neutral | 1% | 2% | 0.81 | 0.698 | 9 years negative-neutral | -71 ms | 30 ms | -2.38 | 0.049 |
12 years positive-negative | 0.5% | 0.5% | 1.02 | 0.566 | 12 years positive-negative | -16 ms | 27 ms | -0.60 | 0.821 |
12 years positive-neutral | 2% | 0.9% | 2.64 | 0.023 | 12 years positive-neutral | -63 ms | 28 ms | -2.23 | 0.069 |
12 years negative-neutral | 2% | 0.9% | 3.03 | 0.105 | 12 years negative-neutral | -46 ms | 30 ms | -1.57 | 0.264 |
Adults positive-negative | -0.4% | 0.3% | -1.44 | 0.321 | Adults positive-negative | -1 ms | 27 ms | -0.04 | 0.999 |
Adults positive-neutral | 1% | 0.6% | 2.13 | 0.083 | Adults positive-neutral | -56 ms | 28 ms | -2.01 | 0.113 |
Adults negative-neutral | 2% | 0.6% | 2.93 | 0.010 | Adults negative-neutral | -55 ms | 29 ms | -1.87 | 0.150 |
Percentages of correct responses for different word types (neutral words and positive and negative emotion terms) in LDT across age groups (error bars indicate SE and ∗indicate significance).
Turning to the analysis of response times, we implemented a linear mixed-effects regression approach (fitted by Restricted Maximum Likelihood, with Satterthwaite approximations to degrees of freedom). Here, we estimated a random intercept for each participant and a random slope of valence for each participant. Further, we included a random intercept for each word-item. Age and valence were included as fixed predictors as described above. Since neutral words and emotion terms (positive vs. negative) significantly differed with respect to their mean concreteness value, concreteness was added as a continuous covariate (grand-mean centered) to the model in order to control for its potential influence. Although a main effect of concreteness could be observed [
Average response times for different word types (neutral words and positive and negative emotion terms) in LDT across age groups (error bars indicate SE and ∗indicate significance).
As expected, an age-dependent improvement in LDT performance (higher accuracy and faster responses) was observed, suggesting that in general, children become more mature in accessing information from the mental lexicon as they grow older. Regarding developmental trajectories, patterns for accuracy and response times turned out to be somewhat different: With respect to accuracy, improvement was most pronounced between the two youngest groups (from 77% in 5-year-olds to 91% in 6-year-olds). After this early developmental boost, there were no other significant differences between two adjacent age groups, which points to a continuous, i.e., a slow but steady growth in accuracy over a long time period beginning at age 6. For reaction time, no such early developmental boost was observed. The results rather point to a continuous acceleration of processing speed over middle childhood with a developmental plateau between 9 and 12 years of age.
With respect to valence, we found a positivity advantage in accuracy which is in line with previous findings (
Experiment 2 was performed by the same children and adults who already participated in experiment 1 (see section “Participants” of Experiment 1 for a detailed description of participants). Three percent of the participants (two 5-year-olds and two 6-year-olds) needed to be excluded due to achieving a mean accuracy of less than 60%.
In the VDT, we used a subset of the stimuli from Experiment 1: 24 positive and 24 negative German emotion terms. See Section “Stimuli” of Experiment 1 for a detailed description of selection criteria and the process of stimulus construction.
The procedure of the VDT was identical to the LDT except for the task-relevant instruction. As in the LDT, participants were informed that they will hear a tone followed by a word at which point they had to indicate whether that word carried a positive or negative meaning as quickly and accurately as possible. This time, a sun symbol indicated a positive word meaning, and a raincloud indicated a negative one. The meaning of “positive” and “negative” was further explained to all children using phrases such as “positive means something good, nice, or pleasant. For example, the word
As for the LDT, average accuracy and response times were calculated for each participant and item. Twenty-two participants in each of the 5- and 6-year-old group as well as 24 participants in each of the 9-year-old, 12-year-old, and adult group were left for the analysis of accuracy, which corresponds to 97% of all collected single reactions (120 participants∗48 items). Compared to the LDT, one additional exclusion criterion (1.1) was added to the data reduction process: The VDT responses for words for which a child gave an incorrect response in the LDT were excluded from the data analysis, assuming that one cannot adequately determine a word’s valence without first knowing that it is a word at all. Based on five steps (the same as for LDT except for 1.1) the following proportions of single reactions were excluded from the response time data: (1) only correct responses from the VDT were considered for analysis (11% excluded, 89% remaining of all analysable reactions). (1.1) After exclusion of words with incorrect responses in the LDT from the VDT data (9%) 80% of all analysable reactions remained for the analysis. (2) Exclusion of the lowest and highest 5% of reactions that were seen as extreme outliers, e.g., due to equipment error or distraction (9% excluded, 71% remaining of all analysable reactions). (3) Exclusion of single reactions that were at the same time atypical for both a particular participant AND for a particular item, i.e., above or below 2 standard deviations from the participant’s and item’s mean (2% excluded, 69% remaining of all analysable reactions). (4) Exclusion of participants that showed an atypical mean response time with respect to their age-mates again using the rule of 2 standard deviations from the mean (corresponds to five participants, one each for every age group, 3% reactions excluded). In total, 66% of all analysable reactions could be considered for reaction time analysis (42% of the data from the 5-year-olds, 61% from the 6-year-olds, 73% from the 9-year-olds, 76% from the 12-year-olds, and 78% of the adult data).
We carried out LMER and GLMER mixed effect regression analyses in the R programming environment (
For the analysis of accuracy, we once again used a generalized linear mixed-effects regression modeling approach (link = logit, fitted by Laplace approximation). The model included a random intercept for each participant and a random intercept for each word. Age and valence were included as fixed predictors as described above.
Estimated means of age and valence for VDT.
Estimated mean accuracy and mean response time with (SE) | 5 years | 6 years | 9 years | 12 years | Adults |
---|---|---|---|---|---|
Accuracy (positive and negative) | 79% (3%) | 87% (2%) | 96% (1%) | 97% (1%) | 98% (0.4%) |
Accuracy (positive) | 87% (3%) | 94% (1%) | 97% (1%) | 97% (1%) | 99% (0.2%) |
Accuracy (negative) | 71% (5%) | 80% (4%) | 94% (1%) | 96% (1%) | 97% (1%) |
Response time (positive and negative) | 1033 ms | 1003 ms | 890 ms | 770 ms | 325 ms |
Response time (positive) | 944 ms | 927 ms | 864 ms | 753 ms | 322 ms |
Response time (negative) | 1122 ms | 1078 ms | 917 ms | 788 ms | 329 ms |
Accuracy age contrast | β-value | SE | Response time age contrast | β-value | |||||
---|---|---|---|---|---|---|---|---|---|
5–6 | -8% | 3% | -2.45 | 0.102 | 5–6 | 30 ms | 69 ms | 0.44 | 0.992 |
5–9 | -17% | 3% | -5.38 | 0.000 | 5–9 | 143 ms | 67 ms | 2.12 | 0.218 |
5–12 | -18% | 3% | -5.82 | 0.000 | 5–12 | 263 ms | 67 ms | 3.92 | 0.002 |
5–adults | -19% | 3% | -6.19 | 0.000 | 5–adults | 708 ms | 67 ms | 10.55 | 0.000 |
6–9 | -8% | 2% | -3.94 | 0.001 | 6–9 | 112 ms | 66 ms | 1.72 | 0.430 |
6–12 | -10% | 2% | -4.63 | 0.000 | 6–12 | 232 ms | 65 ms | 3.55 | 0.005 |
6–adults | -11% | 2% | -5.22 | 0.000 | 6–adults | 678 ms | 65 ms | 10.36 | 0.000 |
9–12 | -1% | 1% | -1.31 | 0.684 | 9–12 | 120 ms | 64 ms | 1.90 | 0.330 |
9–adults | -2% | 0.9% | -2.63 | 0.065 | 9–adults | 565 ms | 64 ms | 8.90 | 0.000 |
12–adults | -1% | 0.7% | -1.50 | 0.560 | 12–adults | 445 ms | 63 ms | 7.02 | 0.000 |
5 years positive–negative | -16% | 4% | -3.50 | 0.001 | 5 years positive–negative | 178 ms | 54 ms | 3.31 | 0.001 |
6 years positive–negative | -14% | 3% | -4.29 | 0.000 | 6 years positive–negative | 151 ms | 47 ms | 3.21 | 0.002 |
9 years positive–negative | -3% | 1% | -1.86 | 0.064 | 9 years positive–negative | 53 ms | 44 ms | 1.21 | 0.232 |
12 years positive–negative | -1% | 1% | -1.05 | 0.292 | 12 years positive–negative | 36 ms | 43 ms | 0.82 | 0.414 |
Adults positive–negative | -2% | 0.8% | -3.19 | 0.001 | Adults positive–negative | 7 ms | 43 ms | 0.17 | 0.868 |
Percentages of correct responses for positive and negative emotion terms in VDT across age groups (error bars indicate SE and ∗indicate significance).
Results with respect to response times revealed a comparable pattern. Here, we implemented a linear mixed-effects regression approach (fitted by Restricted Maximum Likelihood, with Satterthwaite approximations to degrees of freedom) estimating a random intercept for each participant, a random slope of valence for each participant and a random intercept for each word-item. Age and valence were included as fixed predictors as described above. All estimated average response times are shown in
Average response times for positive and negative emotion terms in VDT across age groups (error bars indicate SE and ∗indicate significance).
In accordance with our hypothesis, accuracy rates and processing speed in VDT increased with the age of the participants. Furthermore, the developmental pattern of this age-related improvement in emotion term processing seems to differ with respect to the measure of performance. Regarding accuracy, the biggest developmental change seemed to occur between the ages of 6 and 9. During this time period children become proficient in correctly categorizing emotion terms along their valence, which finally results in an adult-like accuracy at the age of 9. However, response times of children remain slower compared to those of adults throughout the entire range of children’s ages that were tested. Furthermore, the increase of processing speed occurs steadily during a long developmental phase from 5 to 12 years of age.
Both measures of performance showed the predicted modulation by the stimulus’ valence and the participants’ age: results point to an early advantage (faster and more accurate processing in 5- and 6-year olds) of positive words compared to negative words. For accuracy, we additionally found a positivity bias in adults and a trend toward the same for 9-year-old children. However, given that adults made only a few errors in VDT overall, one more single wrong response to a negative word (compared to reactions to positive words) might have influenced the appearance of the valence effect more strongly than would have been the case for children. For this reason, our results point toward an overall continuous decrease up to an absence of the positivity bias in older children and adults. This, however, does not fully converge with the results of
The aim of the present study was to detect developmental changes in emotion word processing. Performance in two psycholinguistic tasks (LDT and VDT) was compared and analyzed regarding possible influences of the two factors of age (children aged 5, 6, 9, 12 and adults) and valence (neutral words vs. positive emotion terms vs. negative emotion terms). To our knowledge, this was the first study that aimed at detecting similarities or differences in emotion word processing between children of different age groups as well as between children and adults. Furthermore, the presented sets of emotion terms and neutral words were carefully controlled for valence, arousal, age of acquisition, concreteness, frequency and a number of linguistic variables to ensure that valence and age effects should not be weakened by confounding emotional or linguistic factors.
Briefly, our results clearly demonstrate a general improvement of word processing with increasing age. In addition to this expected general improvement with age, we were able to uncover developmental trajectories. Depending on task and outcome measure, we found characteristic patterns of development over time. Most importantly, the present study demonstrated a shift in the processing of positive and negative words in the course of development: While young children showed a better performance for positive words, this preference disappeared with increasing age. Possible explanations for the age- and valence-related findings will be discussed below.
Age effects were stable across outcome measures and task. As expected, performance in both tasks became significantly better (higher accuracy and faster responses) with increasing age, suggesting that mental representations of emotion terms become better accessible with age. Furthermore, age-related trajectories were characterized by both a continuous improvement of processing abilities and by developmental boosts followed by plateaus, depending on the outcome measure (accuracy or response time) as well as on the task. For LDT and VDT, children reached an adult-like performance level in accuracy, whereas response times continued to improve until adulthood. Five- and six-year-old children showed severe difficulties in the LDT, where they had to discriminate between words and pseudowords. Their performance in this task should therefore be interpreted with caution. In 5-year-olds 38%, and in 6-year-olds 13% of the children had to be excluded because of accuracy scores below 60%. Their lower performance levels might have appeared for different reasons: (1) Although the selected positive, negative, and neutral words had an average age of acquisition value of less than six years of age (see
Children at the age of 5 and 6 showed a stable positivity advantage in three of the four different outcome measures (LDT: positivity advantage in accuracy, VDT: positivity advantage in accuracy and response times). Responses to positive words in these two age groups were faster and more accurate than to negative words or neutral words (in LDT). This result matches the findings by
Considering the large body of studies on non-verbal emotion processing with infants and children that strongly points toward an enhanced processing of negative cues compared to positive ones (e.g.,
As mentioned in the introduction,
Although the results of the present study confirmed the expected valence effect (in the form of a positivity bias), as well as a modulation of this effect by the factor age, the (almost complete) absence of the positivity bias in older children and adults contrasts with a large number of the previously reported findings (e.g.,
It is also important to mention that the comparability of the results of previous studies and those of the present investigation is reduced for several reasons: (1) we controlled the stimuli for arousal, while other studies that found valence effects in adults did not report doing so (e.g.,
The present study showed that children’s processing of emotion terms, as investigated by two word processing tasks (LDT and VDT), improves during childhood. While both tasks were difficult for young children (age 5 and 6), children at the age of 9 and 12 had acquired well-specified semantic representations of emotion terms, as reflected by almost adult-like error rates in both tasks. Regarding processing speed, development continued until adulthood. The focus of the present study was on the effect of valence in emotion term processing. The results demonstrated a clear positivity advantage that turned out to be age- as well as task-dependent: First, preferential processing of positive over negative terms was characteristic for young children, but decreased with age. Second, the valence effect was more pronounced in the emotional categorization task, where access to a word’s valence is strongly task-relevant.
Our findings demonstrate a positivity bias in children for emotion terms exclusively. Future studies should investigate children’s word processing using matched sets of emotion terms, and of affective words with different levels of concreteness. In addition, words were presented in isolation in the present study, and not embedded in a linguistic context. Therefore, no conclusions can be drawn about children’s perception of emotion terms in natural communication. Recent studies point toward a strong influence of contextual information on emotion processing (
This study was carried out in accordance with the recommendations of “Lokale Ethikkommission des Fachbereichs 06 Justus-Liebig-Universität Gießen” with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the “Lokale Ethikkommission des Fachbereichs 06 Justus-Liebig-Universität Gießen.
Acquisition, analysis, and interpretation of data for the work: DB, MV, JG, GS, and CK. Drafted the work and revised it critically for important intellectual content: DB, MV, JG, GS, and CK. Final approval of the version to be published: DB, MV, JG, GS, and CK. Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved: DB, MV, JG, GS, and CK.
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.
We like to thank all families and students for their participation, Arne Nagels for his fruitful comments, Franziska Degé for her support in the project as well as Johanna Sommer, Mareike Neumann, Cecilia Sweitzer, Isabell Debus, and Stefanie Tuerk for their help in data collection.