Spanish affective normative data for 1,406 words rated by children and adolescents (SANDchild)

Abstract

Most research on the relationship between emotion and language in children relies on the use of words whose affective properties have been assessed by adults. To overcome this limitation, in the current study we introduce SANDchild, the Spanish affective database for children. This dataset reports ratings in the valence and the arousal dimensions for a large corpus of 1406 Spanish words rated by a large sample of 1276 children and adolescents from four different age groups (7, 9, 11 and 13 years old). We observed high inter-rater reliabilities for both valence and arousal in the four age groups. However, some age differences were found. In this sense, ratings for both valence and arousal decreased with age. Furthermore, the youngest children consider more words to be positive than adolescents. We also found sex differences in valence scores since boys gave higher valence ratings than girls, while girls considered more words to be negative than boys. The norms provided in this database will allow us to further extend our knowledge on the acquisition, development and processing of emotional language from childhood to adolescence. The complete database can be downloaded from https://psico.fcep.urv.cat/exp/files/SANDchild.xlsx.

Age-Related Differences in Affective Norms for Chinese Words (AANC)

“…The available evidence suggested that young children's ratings of valence were more extreme than those of adolescents and adults (Monnier and Syssau, 2017;Vesker et al, 2018). Furthermore, youngest children considered more words to be positive than adolescents (Sabater et al, 2020). These findings have confirmed that the affective evaluations of information vary with age.…”

“…In addition to language differences, affective meanings of words seem to vary with age. Age-related differences in emotional context have been evidenced when words were rated by children and adults (Monnier and Syssau, 2017;Vesker et al, 2018Vesker et al, , 2020Morningstar et al, 2019;Sabater et al, 2020). The available evidence suggested that young children's ratings of valence were more extreme than those of adolescents and adults (Monnier and Syssau, 2017;Vesker et al, 2018).…”

“…Familiarity ratings have been interpreted as a measure on the frequency of exposure to a word (Stadthagen-Gonzalez and Davis, 2006;Eilola and Havelka, 2010). These might be related to the exposure effect (Zajonc, 1968) or familiarity effect (Warriner et al, 2013;Sabater et al, 2020). Repeated exposure of people to these stimuli (familiarity) enhances their positive attitude toward them, such as preferences, positivity bias, and control.…”

“…Dominance refers to the degree of control exerted by a stimulus (ranging from weak/submissive to strong/dominant). Based on Bradley and Lang's (1999) procedure, affective norms of words are available in a number of languages, including English (Stadthagen-Gonzalez and Davis, 2006;Eilola and Havelka, 2010;Warriner et al, 2013;Scott et al, 2019), French (Gilet et al, 2012;Monnier and Syssau, 2017), German (Grühn and Smith, 2008;Kanske and Kotz, 2011;Schmidtke et al, 2014), Spanish (Ferré et al, 2017;Stadthagen-Gonzalez et al, 2017;Sabater et al, 2020), Portuguese (Soares et al, 2012), Dutch (Moors et al, 2013), Polish (Imbir, 2015), Italian (Montefinese et al, 2014), and Chinese (Wang et al, 2008;Yu et al, 2016;Yao et al, 2017;Liu et al, 2018). Although most affective norms are rated by young adults, these works have clearly demonstrated that affective meanings of words vary with languages and cultures.…”

Emotional Valence Precedes Semantic Maturation of Words: A Longitudinal Computational Study of Early Verbal Emotional Anchoring

“…(2) Different univariate neural network models were trained to predict the emotional valence (output of neural networks) from amodal LSA vector representations (input of neural networks) only in the child developmental stage. In this case, neural networks were trained and validated predicting the child's emotional judgments of the SANDchild data set (Sabater, Guasch, Ferré, Fraga, & Hinojosa, 2020) using the vector Note: Neural network models that were trained and validated in the child semantic space were also tested in the adult semantic space representations of the child semantic space. As it will be described below, not all the words in the child semantic space took part in the training and validation set of the neural network as we only used words from the SANDchild data set (908 words for training and 302 words for validation data sets).…”