The acquisition of emotion-laden words from childhood to adolescence

Studies investigating how children acquire emotional vocabularies have mainly focused on words that describe feelings or affective states (emotion-label words, e.g., joy) trough subjective assessments of the children’s lexicon reported by their parents or teachers. In the current cross-sectional study, we objectively examined the age of acquisition of words that relate to emotions without explicitly referring to affective states (emotion-laden words, e.g., cake, tomb, rainbow) using a picture naming task. Three hundred and sixty participants belonging to 18 age groups from preschool to adolescence overtly named line drawings corresponding to positive, negative, and neutral concrete nouns. The results of regression and mixed model analyses indicated that positive emotion-laden words are learnt earlier in life. This effect was independent of the contribution of other lexical and semantic factors (familiarity, word frequency, concreteness, word length). It is proposed that the prioritized acquisition of positive emotion-laden words might be the consequence of the communicative style and contextual factors associated with the interaction between children and caregivers. We also discuss the implications of our findings for proposals that highlight the role of language in emotion perception and understanding.


Introduction
A critical motivation for infants to acquire language is to share and interpret thoughts and feelings about the objects, persons and events for which they care about.Children's affective experiences are closely related to word learning since they are central to the relationship with the personal and physical world (Beck et al., 2012;Bloom, 1998).At around 18 months, infants already have the ability to form associative links between novel labels and human facial configurations expressing emotions (Ruba et al., 2021).Immediately after, they begin to use emotion-descriptive language (Bretherton & Beeghly, 1982;Ridgeway et al., 1985).
The set of known words referring to affective concepts is referred to as emotional lexicon.A distinction is relevant between emotional words that directly denote affective states or describe feelings (emotion-label words, e.g.sad or happy; henceforth EM) and words with emotional connotations that do not convey specific emotional states (emotion-laden words, e.g., candy or gun; henceforth EL) (Pavlenko, 2008).This classification has received empirical support from behavioral and neurobiological studies (Altarriba & Basnight-Brown, 2011;Knickerbocker & Altarriba, 2013;Wang et al., 2019;Zhang et al., 2017Zhang et al., , 2019; see Wu & Zhang, 2020, for a review).In this vein, the recognition of EM words is faster compared to the identification of EL words (Kazanas & Altarriba, 2015).EM words are also associated to enhanced priming (Kazanas & Altarriba, 2015) and repetition blindness (Knickerbocker & Altarriba, 2013) effects than EL words.Finally, evidence from event-related potential studies have shown differences in short-latency components (e.g., N170, P2, Early Posterior Negativity) indicating processing differences between EM and EL words at early stages of word processing (Liu et al., 2022;Wang et al., 2019;Zhang et al., 2017).
Our knowledge about the development of the emotional lexicon is scarce.The acquisition of EM words has been examined by relying on subjective reports provided by parents (Baron-Cohen et al., 2010;Bretherton & Beeghly, 1982;Li & Yu, 2015;Ridgeway et al., 1985) or on children's assessment of their own knowledge (Baron-Cohen et al., 2010).A different approach was to ask children to list as many emotion words as they knew (Beck et al., 2012) or to define EM words (Nook et al., 2020).Finally, more objective estimations of the size and characteristics of the EM vocabulary were provided by testing children's ability to label faces, stories or vignettes describing feelings (Grosse et al., 2021;Streubel et al., 2020;Widen & Russell, 2003, 2008), listing as many emotion words as they knew (Beck et al., 2012) or asking definitions for EM words (Nook et al., 2020).Overall, these studies show that most children between 28 and 36 months have already learnt words describing feelings such as funny, love or scared (Bretherton & Beeghly, 1982;Nook et al., 2017;Ridgeway et al., 1985;Widen & Russell, 2003, 2008), and that EM vocabularies grow quickly at ages 4-11 (Baron-Cohen et al., 2010;Grosse et al., 2021;Li & Yu, 2015;Nook et al., 2020;Streubel et al., 2020).Also, EM words denoting positive affective states are learnt earlier in life compared to those expressing negative feelings (Baron-Cohen et al., 2010;Li & Yu, 2015).The advantage in the acquisition of positive relative to negative EM words might be associated with some features of the way in which mothers, fathers, and caregivers speak with infants and young children.In this sense, infant-directed speech typically conveys positive emotions (Golinkoff et al., 2015;Singh et al., 2002).Also, there is evidence showing that caregivers mainly use positive vocalizations and prosody to communicate with children (Dave et al., 2018;Kitamura & Lam, 2009;Saint-Georges et al., 2013), and that infants display a preference for melodic sound contours and vocalizations with positive emotional valence (e.g., approving) relative to those with negative emotional valence (e.g., disapproving) when listening to parental and adult speech (Fernald, 1993;Mumme & Fernald, 2003;Papoušek et al., 1990).
So far, however, little attention has been given to the acquisition of emotional-laden words that do not directly label affective states or feelings.Also, research on children's early lexical development has focused on questions such as the acquisition of words belonging to different grammatical categories (nouns and verbs; e.g., Imai et al., 2008) or the lexical and syntactic factors associated with vocabulary growth (e.g., Garlock et al., 2001), although these studies have neglected the affective properties of words.One exception are the studies by Ponari and colleagues who focused exclusively on abstract words.These authors reported that 7-9 years old children know more (Ponari et al., 2018) and learn better (Ponari et al., 2020) abstract positive EL words compared to neutral abstract words.
To go deeper in our understanding of the acquisition of EL words, in the current study we tested the ability of children from 32 months to 15 years of age to name line drawings denoting negative, neutral and positive Spanish EL concrete nouns. 1 Based on a dimensional approach to emotions in which valence (ranging from unpleasant/negative to pleasant/ positive) and arousal (ranging from quiet to activated) are the main affective dimensions (Lang, 1995;Russell, 2003), we examined their role on the acquisition of EL words.In line with prior research showing that adults mainly communicate positive concepts and affect in their linguistic interactions with infants (Ponari et al., 2018), we expect that positive EL words are learnt earlier than both negative EL and neutral words.This finding would resemble that observed for EM words (Baron-Cohen et al., 2010;Li & Yu, 2015).Nonetheless, our hypothesis is tentative since there are no prior studies that systematically investigated this issue.Of note, the existence of processing differences between EM and EL words (Knickerbocker & Altarriba, 2013;Wang et al., 2019) suggests a distinct conceptual representation of these emotional vocabularies in the lexicon, and highlights the need to further investigate whether they share a similar pattern of acquisition in infancy and childhood.We hypothesize a lower contribution of arousal since there is evidence indicating that children's representation of emotions mainly relies on a positive-negative dichotomy whereas they place reduced emphasis in other affective dimensions (Nook et al., 2017).Additionally, we examined if the acquisition of EL words is mediated by lexical and semantic factors closely associated with word learning or the representation of emotional words such as word length, word frequency, familiarity, or concreteness (Alario et al., 2004;Hinojosa et al., 2020;Lotto et al., 2010).

Participants
Data were collected from 360 Spanish preschool, school and high school children and adolescents (180 girls 1 Although prior studies suggest that the rate of growth of EM vocabulary levels off between 12-and 16-years-old, we included participants up to 15 years since this is the first systematic approach to the acquisition of EL words.and 180 boys) from eight schools in the Comunidad de Madrid and one school from Galicia, Spain.Their ages ranged from 32 months (2 years 8 months) to 181 months (15 years 1 month).Children from ages of 32 months (2 years 8 months) to 97 months (8 years 1 month) were tested in 6-months age bands, whereas older children and adolescents were tested in 12-month age bands.Each of the 18 age bands included 20 children (10 girls and 10 boys).The sample size was established according to previous studies using picture naming tasks to investigate word acquisition (e.g., Álvarez & Cuetos, 2007;Lotto et al., 2010).Nonetheless, statistical power values were calculated post hoc for each analysis using G*Power (Faul et al., 2007), yielding values greater than 0.80 in all cases (assuming a medium effect size).Information about the 18 age bands is shown in Table 1.All participants were native Spanish speakers, had normal or corrected to normal vision and no history of developmental disorders, nor special educational needs.Children and adolescents lived in urban or suburban areas.Informed consent was obtained from parents.The study was performed in accordance with the ethical standards in the Declaration of Helsinki and approved by a local ethics committee at the Instituto Pluridisciplinar.

Stimuli
First, we selected positive, negative and neutral concrete EL words from several normative studies with adults reporting affective norms for Spanish words (Ferré et al., 2012;Guasch et al., 2016;Hinojosa et al., 2016;Redondo et al., 2007;Stadthagen-Gonzalezet al., 2017).These words could be easily and unequivocally represented by simple line drawings.We excluded words denoting descriptions of feelings and affective states, abstract words that were difficult to represent by simple line drawings (e.g., peace or violence), and words denoting emotional content that were judged as inappropriate for young children.After carefully scrutinizing these datasets, a total of 201 EL words were selected (67 negative: spider, bullet; 62 neutral: comb, gloves; and 72 positive: dolphin, gift; all nouns).Valence scores for negative words were below 4.03, scores for neutral words ranged from 4.03 to 6, and scores for positive words were above 6 (in a 9-points Likert scale ranging from 1 negative to 9 positive).In addition, we gathered adult ratings for arousal, concreteness (i.e., the degree to which a word denotes a perceptible entity), and familiarity (i.e., subjective exposure to a word) from Ferré et al. (2012), Guasch et al. (2016), Hinojosa et al. (2016), Redondo et al. (2007) and Stadthagen-Gonzalez et al. (2017), and log frequency (i.e., the number of times a particular word is encountered in a representative sample of texts or speech; here we used the logarithm to base 10 of word frequency) from EsPal (Duchon et al., 2013).The descriptive statistics of these variables are shown in Table 2.
Subsequently, we collected from the internet a set of line drawings depicting those words to be used in the picture naming task.Several pictures were modified to keep the visual complexity similar for all stimuli by removing some parts that displayed objects or entities apart from those denoted by the target word, or by equating image size.Each drawing was presented in black and white on an independent DIN-A4 size sheet.Figure 1 shows some stimulus examples.The complete set of stimuli will be provided upon request.

Procedure
We followed a procedure used in prior picture naming studies investigating word acquisition (Álvarez & Cuetos,  (Chalard et al., 2003;Morrison et al., 1997), a phonetic cue (e.g., the initial phoneme) was provided whenever children produced a synonym or if the name was on 'the tip of the tongue' to help them focus on the intended word related to a particular drawing.Prior research has shown high significant correlations for the age of acquisition of words between studies providing cues and those in which cues were not provided to children (Cannard & Kandel, 2008).Therefore, responses following cues were considered correct if children generated the appropriate word.Once the participant named the pictures, the experimenter coded the answers in an Excel spreadsheet (1 = correct answer; 0 = incorrect answer/unknown word).Children were regularly encouraged throughout the picture naming task.Younger children were allowed to rest whenever they were tired.
As in prior research (e.g., Álvarez & Cuetos, 2007), we started testing the 158 (13 years 2 months)-169 months (14 years 1 month) age group so that the youngest children had to name a smaller set of line drawings.We assumed that younger children would not be able to name those stimuli that older children could not name.The order of presentation of stimuli was randomized for each participant.Pictures were divided into two blocks with a 5-min rest break between blocks.Additional rest breaks were allowed upon request by children.
The age of acquisition for each word was determined using the 75% rule adopted in prior studies (Álvarez & Cuetos, 2007;Morrison et al., 1997).Once all stimuli were shown to each participant in an age group, correct and incorrect responses for each drawing were collated.
Subsequently, we calculated the mean age of children in that group, as well as the average naming agreement for each picture.The age of acquisition of a particular word was considered to be the age of the youngest group in which at least 75% of the children could name the concept represented by the line drawing.

Results
Data is available at https:// osf.io/ u863v/.As shown in Fig. 2 the histogram of the age of acquisition suggests high positive skewness (skewness = 1.69,SE = 0.17), indicating that most words included in our study were acquired rather early in life (i.e., 75% of words have been learnt under 65.5 months).Table 3 shows a frequency distribution of the age of acquisition of words.We examined the relationship between the age of acquisition of words and the affective dimensions of valence and arousal through a multiple regression analysis.We included several lexical and semantic variables such as concreteness, familiarity, log frequency and word length (number of letters), which exert different effects on word processing (e.g., Stadthagen-Gonzalez & Davis, 2006).This analysis was performed on a subset of 157 words for which values in all variables examined in this study were available.The age of acquisition of words was entered as a criterion, while valence, arousal, concreteness, familiarity, log frequency and number of letters were tested as predictors.We used the logarithm of the age of acquisition of words instead of the raw value since the original distribution of the values was positively skewed (see Fig. 2).Variables were introduced using the stepwise method.No multicollinearity was observed (tolerance > 0.66).The resulting regression model significantly predicted the age of acquisition of words, F(4,152) = 16.71,p < 0.001, explaining 31% of the total variance (R 2 = 0.31).The model included familiarity, number of letters, valence, and log frequency (see Table 4).The two variables with the highest predictive power were familiarity (R 2 change = 0.21) and number of letters (R 2 change = 0.04), adding a total of 25% of explained variance.Moreover, log frequency accounted for a 3% of the variance and showed a negative relationship with the age of acquisition of words.These results indicate an earlier acquisition of those words that for adults are more familiar, frequently used and shorter words (see Figs. 3, 4 and 5).
Importantly, valence explained a 3% of the variance and showed a negative relationship with the age of acquisition of words (see Fig. 6).This result indicates that words showing higher valence scores (positive words) for adults are acquired earlier in life.Of note, this relationship cannot be attributed to the effect of any other lexical or semantic variable examined in this study (e.g., word frequency, familiarity, number of letters, or concreteness).Supplementary materials show the results from additional analyses to further examine the consistency of valence effects on word acquisition, as well as the validity of our data with valence, arousal and word frequency norms from children aged 9-and 11-(Martínez & García, 2004;Sabater et al., 2020).

Discussion
The present study aimed to shed some light on children's gain of emotional vocabularies.Besides the contribution of lexical and semantic variables such as word frequency or word length (Alario et al., 2004;Morrison et al., 1997), we observed that familiarity had the highest predictive power of the age of acquisition of words.Our data indicates that words that are encountered more often in the linguistic environment are learnt earlier in life.This result is in agreement with previous developmental studies showing that children's experience with words plays a key role in the acquisition of language.In this sense, novel objects or events are more likely to be linked to familiar word forms to build new lexical representation (Bannard & Matthews, 2008;Bortfeld et al.,  2013).Nonetheless, the main and novel finding of this study was the effect of valence in word acquisition, which is independent of the contribution of familiarity, word frequency, or word-length.In particular, words with a positive connotation    that do not explicitly express affective states were acquired earlier than negative or neutral words.In contrast, we failed to report effects of arousal to the acquisition of EL words, which is in line with evidence showing that emotional representations in young children mainly rely on the valence dimension (Grosse et al., 2021;Nook et al., 2017).
Prior studies examining the acquisition of EM words based on parents' or teachers' reports found that children learn words denoting positive feelings earlier that those describing negative or neutral affective states (Baron-Cohen et al., 2010;Li & Yu, 2015).Here, we extend prior knowledge by showing a similar acquisition advantage for positive EL words.In line with prior claims, this finding points to a role of EL words on the acquisition and representation of emotional concepts that describe affective states (Lindquist et al., 2016;Satpute & Lindquist, 2021).Besides, current results highlight the need to consider word affective features when answering questions such as the acquisition of noun and verb lexical categories in children (Bates et al., 1994;Imai et al., 2008).Finally, most research on emotion vocabulary development has focused on children who are English native speakers.Here we show that the advantage in the acquisition of positive words extends to Spanish vocabulary.This finding is in line with the results from multilingual analyses that reported similar trajectories in vocabulary development in languages such as English, French, Korean, Hebrew, or Spanish (Bornstein et al., 2004).Nonetheless, since these studies did not assess the acquisition of affective words, potential cross-cultural differences in the development of English and Spanish emotion vocabulary should be explicitly tested in multilingual comparative studies.In this sense, there are data indicating that the ability to recognize and understand others' emotions develops later in Spanishspeaking pre-schoolers relative to English-speaking peers (Downs et al., 2007).
Regarding the possible reasons for a positive advantage in the acquisition of EL words, there is evidence indicating that adults make extensive use of words denoting positive concepts when they speak to children.In this line, Ponari and collaborators (2018) examined the subcorpus of childdirected speech (MacWhinney, 2000) from the Language Data Exchange System (CHILDES) to show that more than half of the most frequent words in English were positive and none were negative.Of note, infant-directed speech reflects positive emotions through the use of specific intonation contours and certain parameters of the spectral composition (e.g., high mean fundamental frequency or pitch, proportion of high frequency energy), or increases in the speech rate (Dave et al., 2018;Fernald & Kuhl, 1987;Singh et al., 2002), as well as through exaggerated facial expressions (Tamis-LeMonda et al., 2014).It seems likely that caregivers convey EL words with positive referents (e.g., toy or hug) in ways to capture children's attention by emphasizing these features when uttering within the course of linguistic interactions with children.Since infants display an early and persistent preference towards these paralinguistic cues (Golinkoff et al., 2015;Mumme & Fernald, 2003;Singh et al., 2002) and for affectively positive stimuli (Singh et al., 2004), increased attention would facilitate children's mapping of EL words to their positive-valence referents and, in turn, growth in positive EL vocabulary.Also, within the embodied cognition framework (Pulvermüller, 2005;Simmons et al., 2005), children's positive experiences associated with words' referents could also improve the mapping between objects or events and their lexical labels in positive EL words.In sum, we speculate that a more frequent exposure to EL positive words with idiosyncratic prosodic features allows children to observe adult's use of positive EL words to label objects, persons, or events 2 (Hoemann et al., 2019).These interactions might play a 'bootstrapping' role that attracts children's interest and provides with a communicative context for the prioritized acquisition of EL words.
Current data have theoretical implications for those views arguing for a role of language in the development of emotion understanding (Beck et al., 2012;Hoemann et al., 2019;Ornaghi & Grazzani, 2013).In particular, theories of constructed emotions assume that conceptual knowledge associated with emotional words plays a key role in shaping the way people perceive emotions in themselves and others (Barrett et al., 2007;Lindquist et al., 2015;Shablack & Lindquist, 2019).In agreement with this view, several studies have reported a relationship between children's vocabulary size of words to label affective states and both their ability to recognize emotions and their knowledge of emotion regulation strategies (Grosse et al., 2021;Streubel et al., 2020).Accordingly, children's prioritized acquisition of positive EL and positive EM words might be related to the earlier recognition of positive emotions in several domains.In this sense, Widen (2013) has shown that by the third year of life children make a preferential use of the term happy to label facial expressions.Additionally, Denham and Couchoud (1990) reported that 2-4-years-old children identified more easily positive situations relative to negative situations in vignettes showing puppets expressing feelings.Finally, children were better at naming and pointing to happy facial expressions compared to expressions of sadness, fear, or anger (Camras & Allison, 1985;Vesker et al., 2018). 2 We indirectly test this possibility by examining the relationship between word frequency and valence scores in 7-, 9-, and 11-yearold children (Martínez & García, 2004;Sabater et al., 2020).We observed a positive correlation in all age-groups (7-years-old, r = .249,p = .003;9-years-old, r = .384,p < .001;11-years-old, r = .226,p = .007),which suggest that positive words for children are those more frequently used by adults.
One limitation of this study is that the selection of the stimuli was based on affective scores from adults, which might not necessarily reflect how children perceived the emotional features of words.Although the question of when emotional vocabularies are learnt (e.g., Grosse et al., 2021;Li & Yu, 2015;Nook et al., 2020) is slightly different from the question of how children assess the affective features of words (e.g., Monnier & Syssau, 2017;Nook et al., 2017;Sabater et al., 2020), this is an intrinsic limitation of this line of research since it would be nearly impossible to collect valence and arousal ratings from infants and young children who do not understand these dimensions in the same way as adults do (Nook et al., 2017).There might be a similar concern regarding scores for some of the subjective lexico-semantic variables analysed in the current study (e.g., familiarity, concreteness).However, the results of our supplementary analyses conducted with valence and word frequency norms from 9-and 11-yearsold children are informative here.Indeed, these analyses replicated positive valence effects on the age of acquisition of words independent of word frequency, suggesting that our findings are rather robust.Also, recent evidence suggests that word emotional valence in adults can be predicted by children´s representations of this affective dimension, and that valence judgements by children and adults are highly correlated (Martínez-Huertas et al., 2021).Nonetheless, current findings might also be relevant by at least showing that children learn earlier those EL words that for adults show positive connotations.
In conclusion, we observed that positive EL words were acquired earlier that both negative EL and neutral words, which might arise from prior exposure to words to label objects, animals, or persons with positive connotations during adults' interactions with children.New research may shed some light on the potential mediators that influence the prioritized acquisition of EL positive words such as verbal knowledge or general cognitive skills (Grosse et al., 2021;Nook et al., 2017).Also, further studies are needed to disentangle potential confounding variables including differences in the availability of visual referents for positive and negative EL words in child's environment or differences in word features that facilitate their acquisition such as iconicity (Nielsen & Dingemanse, 2021).Current results have also some methodological implications.In this sense, we provide a data source that might be useful for a broad community of researchers examining emotional language processing in both children and adults.Just to give a few examples, the age of acquisition of EL words reported here might be valuable for investigating aspects such as the acquisition of emotional words in second languages or the interaction between lexico-semantic variables and word emotional properties in language comprehension and production studies.

Fig. 3
Fig. 3 Partial regression plot of familiarity.A partial regression plotshows the relationship between two variables included in a regression (the criterion and a predictor) after controlling for the effect of the other variables.Specifically, it shows the residuals of the criterion (on the Y-axis) and the residuals of the predictor (on the X-axis) when both variables are regressed on the rest of the predictors.Each dot represents a word, and the solid line shows the linear regression fit

Fig. 5
Fig. 5 Partial regression plot of log frequency.Each dot represents a word, and the solid line shows the linear regression fit

Fig. 6
Fig. 6 Partial regression plot of valence.Each dot represents a word, and the solid line shows the linear regression fit

Table 1
Age Bands and Mean Age.Each age band included a total of 20 participants (10 girls and 10 boys)

Table 2
Descriptive statistics of the variables examined in the study Lotto et al., 2010;Morrison et al., 1997)1997).Children were tested individually in an empty classroom.They were told that the experimenter was going to show them several pictures.Thereafter, children were instructed to look carefully at each line drawing and to answer the following question: "How would you name what you see in the picture?".Following a standard procedure

Table 3
Frequency distribution of the age of acquisition of words

Table 4
Coefficients of the multiple linear regression model