Emotions are a fundamental part of the human experience. They can be triggered by different kinds of stimuli in everyday life and have a deep impact on cognition and behavior (Hinojosa et al., 2019). From the dimensional perspective of emotions, proposed by Wundt (1896) and revived by Osgood et al. (1957), emotions have three dimensions: valence (characterized as a spectrum that ranges from pleasantness to unpleasantness), arousal (the degree of physiological activation that ranges from calmness to excitement), and dominance (being in control or controlled by a situation). Over the years, several studies have elicited emotion in an experimental setting using large datasets of emotionally loaded stimuli to explore the interactions between these dimensions (e.g., Imbir et al., 2019; Kensinger & Corkin, 2003; Lu et al., 2017; Russell & Mehrabian, 1977). Consistently, these studies found that emotional response variability is better explained by the interaction between valence and arousal. These two dimensions form a bidimensional space in which highly arousing stimuli can be found on either end of the valence spectrum; that is, the most pleasant and unpleasant stimuli show high levels of activation while neutral stimuli do not (Monnier & Syssau, 2014; Montefinese et al., 2013). On the other hand, dominance only explains a small proportion of variance and has been linked to societal expectations and beliefs (e.g., Angelini, 2007; Irrazabal et al., 2015).

Since triggering emotions remains a challenge, different resources have been developed to fulfill this need. The Center for the Study of Emotion and Attention (CSEA) has provided various affective datasets for this purpose, such as the International Affective System Pictures (IAPS; Lang et al., 1999), the Affective Norms of English Words (ANEW; Bradley & Lang, 1999a), the Affective Norms for English Texts (ANET; Bradley & Lang, 2007), and the International Affective Digitized Sounds (IADS; Bradley & Lang, 1999b). Each of these datasets has been adapted to different languages and countries (e.g., Fernández-Abascal et al., 2008; Irrazabal et al., 2015; Kristensen et al., 2011; Montefinese et al., 2013; Soares et al., 2012). The IAPS, for instance, comprises different sets of pictures rated along the three emotional dimensions described above. Its wide use relies on its cross-cultural appeal because images can be understood by most Western populations and pictures can generate strong emotional reactions (Kensinger & Schacter, 2006; Uhrig et al., 2016).

Some of these picture sets have been adapted to an Argentinian sample (Irrazabal et al., 2015; Irrazabal & Tonini, 2020). The results replicated the existence of a typical boomerang-like dispersion along the bidimensional emotional space, in which unpleasant and pleasant images were also more arousing. According to this model, emotions represent a disposition towards action guided by two motive systems, appetitive/pleasant and aversive/unpleasant (i.e., the valence spectrum). Arousal, on the other hand, represents the variations in metabolic or neural activation of both systems (Lang, 1995). For the IAPS adaptation, the authors also found some differences by gender, since women showed greater activation than men for pleasant pictures related to children, romantic love, and food, and unpleasant pictures that depicted physical aggression and gender-related violence (such as rape and assault). On the other hand, men showed grater activation for pleasant pictures that depicted sexual or erotic scenes, and sporting events; and for unpleasant pictures that depicted war, suicide, and tumors (Irrazabal & Tonini, 2020). Women also showed overall less dominance than men (Irrazabal et al., 2015).

Words are a powerful tool for the expression and modulation of emotion, either one’s own emotions or those of others, but rely deeply on language knowledge. In working with words, we can control a series of variables that are not possible to control with images and can influence the processing of the stimuli, such as concreteness, age of acquisition, and frequency (Monnier & Syssau, 2014). The emotional connotation of words vary from one culture to another (Eilola & Havelka, 2010; Fumagalli et al., 2015; Montefinese et al., 2013) and validation of the norms in each country is relevant to research, since cultural differences can be found in different languages (Spanish vs. English, e.g., Redondo et al., 2007), and within the same language but with different sociocultural backgrounds, such as Brazilian vs. European Portuguese (Kristensen et al., 2011; Soares et al., 2012), British vs. American English (Eilola & Havelka, 2010), or European vs. Latin-American Spanish (Haensch, 2002; Kilgarriff & Renau, 2013). Even though these language varieties share both lexical and grammatical information, some differences in naming, spelling, and word familiarity can be found. These differences, along with differences derived from use and custom, could be transferred to the emotional assessment of words, therefore having adaptations for each country would be extremely useful for research. Some European Spanish words are not used in other Spanish-speaking countries, rendering them hard to identify or unfamiliar. For example, in Rioplatense Spanish, a regional variation of Spanish spoken in South America, words such as flemón (toothache) are not used, while words like ordenador (computer), móvil (cellphone), and maleta (suitcase) have alternative naming conventions: computadora, celular, and valija, respectively (Haensch, 2002; Kilgarriff & Renau, 2013; Manoiloff et al., 2010). Consequently, ratings of the same word may vary from one culture to another, due to the connotation the word may have in each specific country. For highly emotional words, such as insults and taboo words, differences in perceived offensiveness have been found between speakers of British English and American English (Dewaele, 2016).

In addition, cross-language adaptations pose different challenges for semantic equivalence, because some words may not have a single-word translation into other languages (Ogarkova, 2016). In Spanish, for instance, the English word toothbrush is translated as cepillo de dientes, that is, three separate words. Therefore, finding equivalent meanings across languages or countries is often difficult (Fumagalli et al., 2015). In view of these differences, the ratings provided by one population cannot be extrapolated to another without taking into account their sociodemographic features. The adaptation of databases to other languages and their regional variations enables a more accurate comparison of empirical findings on emotion processing across different cultures.

The role of valence and arousal on memory is controversial. Extensive evidence suggests a processing advantage for emotional words over neutral ones on memory (e.g., Diaz Abrahan et al., 2020; Kensinger & Corkin, 2003, for reviews on the topic Ack Baraly et al., 2017; Talmi, 2013) and word recognition (e.g., Imbir et al., 2019; Zhao et al., 2018), but discrepant evidence has also been found. Other authors have shown that valence is more relevant than arousal, and that valence has a graded effect: reaction times are faster for positive than neutral or negative words (Madan et al., 2012; Talmi & Moscovitch, 2004; Rodriguez-Ferreiro & Davies, 2019). Overall, most researchers agree that emotion impacts how we process words (its acquisition, storage, and retrieval); therefore, large databases with different subjective and objective measurements are essential tools for emotional word research and psycholinguistics. However, other variables could also play a role in emotion modulation (Guasch et al., 2016; Hinojosa et al., 2016). While affective dimensions (valence and arousal) and other psycholinguistic variables (concreteness, imageability, and familiarity) tend to be studied separately, in recent years this trend has started to revert, and both types of variables and their interactions have begun to be studied as an interconnected system (Monnier & Syssau, 2014; Montefinese et al., 2013; Yao et al., 20172017).

Different studies have shown that concreteness influences the processing of emotional words (e. g., Palazova et al., 2013; Vigliocco et al., 2014; Yao et al., 2016). While concrete words denote concepts that are easily perceived in space and time, and can be acted upon (e.g., table), abstract words refer to concepts that are internally represented, more related to emotional states (Altarriba & Bauer, 2004), and usually acquired later in life (e.g., apple vs. truth; Kousta et al., 2011). Some studies have shown the existence of a concreteness effect, in which high concreteness levels facilitate word processing (e.g., Kanske & Kotz, 2007) and recall (e.g., Tse & Altarriba, 2009). Other studies have reported a reverse concreteness effect, whereby abstract words have a processing advantage over concrete words (e.g., Kousta et al., 2011). This reverse concreteness effect appears when other lexico-semantic variables are considered, reflecting the need for a dataset that groups these variables together.

Two other psycholinguistic variables need to be considered. Firstly, imageability refers to the degree to which a concept enables a mental image or a sensory experience (Rofes et al., 2018). Some studies have shown that words with high or low imageability scores are less arousing (Monnier & Syssau, 2014; Montefinese et al., 2013; Warriner et al., 2013) than the words in-between (Montefinese et al., 2013). One study indicated that pleasant words are more imageable than unpleasant ones (Warriner et al., 2013); however, Guasch et al. (20162016) found that there is a negative correlation between emotionality and imageability.

Secondly, familiarity indicates the perceived frequency of occurrence of a given word; highly familiar words are read or heard more often, while unfamiliar words are deemed rarer (Dampuré et al., 2011). Similarly, positive words are more familiar (Yao et al., 2017) and more frequent (Monnier & Syssau, 2014; Montefinese et al., 2013; Warriner et al., 2013) than negative ones. Words that had the highest score on dominance were the most familiar (Montefinese et al., 2013). The growing body of studies that have analyzed these relationships between lexico-semantic and affective variables in word processing indicate the need for further development of linguistic resources for experimental manipulation.

The first goal of our work was to provide affective measures for 1034 words translated from the ANEW to Rioplatense Spanish (RS). The second goal was to incorporate psycholinguistic ratings and objective measures for the translations and to study their possible relations. To this end, we created a general database suitable for experiments that investigate the effects of emotions as well as other semantic variables on word processing. The third goal was to compare our dataset with normative data from the United States and European Spanish (ES). Finally, given that the Argentinian IAPS sample showed gender differences and considering that other ANEW adaptations also found similar differences when compared with the original sample (Redondo et al., 2007; Soares et al., 2012), the fourth goal was to analyze the information by gender. In this way, we proposed to assess whether these differences are replicated with different kind of stimuli. The adaptation of this database to other languages or cultures enables the comparison of empirical findings on emotion processing (Schmidtke et al., 2014).

Method

Subjects

The sample included 483 volunteers (338 women and 135 men; 320 university studentsFootnote 1 and 163 non-university studentsFootnote 2) who rated the six variables (valence, arousal, dominance, familiarity, imageability, and concreteness). They were recruited from several educational centers in central (Córdoba, Buenos Aires) and southern (Rio Negro) provinces of Argentina. All participants were required to comply with the following inclusion criteria: (a) be over 18 years old, (b) be Argentinian born, 1) (c) be a native speaker of Rioplatense Spanish, and (d) do not consume mood stabilizers. All the participants were naïve as to the purpose of the study and were randomly allocated to the different wordlists. Besides, 90.6% were right-handed. Twelve participants were excluded because they did not meet the inclusion criteria. The final sample included 471 subjects, a total of 302 subjects performed the valence, arousal and dominance ratings (210 women, 92 men; mean age = 24.46 years old, SD = 6.87, range 18–52; mean education = 15.65 years, SD = 3.17) and 169 subjects performed the familiarity, imageability, and concreteness ratings (125 women and 43 men, and one volunteer self-identified as non-binary; mean age = 25.54 years old, SD = 8.72, range 18–66; mean education = 16 years, SD = 2.95). The study was conducted in accordance with the Declaration of Helsinki guidelines and the procedure was approved by the Ethics Committee of Instituto de Investigaciones Médicas “Alfredo Lanari”, Protocol #283.

Adaptation procedure

The data set contained 1034 Rioplatense Spanish (RS) words translated from the ANEW (Bradley & Lang, 1999a). The set contains 712 nouns, 243 adjectives, 65 verbs, and 13 words that could be interpreted as either nouns or adjectives. Furthermore, only one word could be interpreted as both a noun and a verb (poder [power]). The 1034 words (100%) were independently translated from English to RS by two bilingual judges, who were asked to provide a primary translation and alternative translations for each word. Interjudge agreement reached 69.34% for the primary translations. For the remaining 30.66%, a third bilingual judge revised the alternative translation provided by the two judges and the Redondo et al. (2007) translations and along with the other judges decided the final word to be included. For example, the ES translation for the word noose was soga [rope], a highly unusual translation in RS. The primary translation provided by one of the judges was horca, which coincided with the alternative translation provided by the second judge. Therefore, the word horca was ultimately selected. Due to cultural and lexical differences between ES and RS (Haensch, 2002), 202 words required different translations (e.g. hairpin translates as horquilla in ES, which is not used in RS, so it was replaced by hebilla). In addition, two words did not have a one-word equivalent in Spanish (rollercoaster, skijump). Thus, alternative words were provided to preserve the emotionality of the original US sample: Rollercoaster translates as montaña rusa in Spanish; to preserve its high valence and high arousal, the word adrenalina [adrenaline] was chosen. Following Bradley and Lang (1999b), once all words were translated, a pen-and-paper procedure was developed.

All 1034 words were pseudo-randomly divided into six lists of 172 or 173 words each. To collect ratings of the six variables, we constructed two versions of the questionnaire with the same words. One version was aimed to collect ratings of valence, arousal, and dominance, and the other one was intended to collect ratings of familiarity, imageability, and concreteness (FIC), resulting in 12 different booklets exposing the subjects to ten words or fewer per sheet. Each version contained 172–173 words, began and ended with either a positive or a neutral word from Bradley and Lang’s (1999a) ratings, and was completed by at least 46 participants for the affective ratings and 24 participants for the psycholinguistic ratings. To reduce bias, participants were not informed about the order of the sheets on each booklet, so they could start from whichever end they preferred.

All six measures were rated on two nine-point scales. For the three affective ratings, the Self-Assessment Manikin (SAM), a non-verbal self-report measure designed by Bradley and Lang (1994), was employed. This measure has the advantage that it can be used across cultures because it is easy to understand and does not need language scaffolding due to its pictographic nature. Response scales ranged from very unpleasant (1) to very pleasant (9) for valence; from very calm (1) to very excited (9) for arousal; and from very controlled (1) to in-control (9) for dominance. The other three scales (FIC) ranged from 1 (unfamiliar, unimaginable, and abstract) to 9 (highly familiar, imaginable, and concrete). We used the SAM scale in order to maintain methodological consistency with other affective norms (Bradley & Lang, 1999a; Imbir, 2015; Kristensen et al., 2011; Monnier & Syssau, 2014; Montefinese et al., 2013; Redondo et al., 2007; Schmidtke et al., 2014; Soares et al., 2012).

Rating procedure

The participants signed their informed consent and then completed a demographic questionnaire in which the following information was assessed: age, gender, laterality, nationality, place of birth, years of formal education, primary occupation, mental health history, and current use of medication.

The rating tasks employed a paper-and-pencil format. Each volunteer received, randomly, a list printed on an A4-size rating-booklet. The front page included the written instructions in Spanish for each dimension (valence, arousal, and dominance or familiarity, imageability, and concreteness). The rest of the pages contained a maximum of ten words each, printed on the left side of the page in Arial 12-point font. The SAM scale was printed on the right side of each word. Figure 1a shows the format employed. For the FIC rating task, nine-point scales were designed but, unlike the SAM scale, they did not provide pictorial aid. Following Montefinese et al. (2013), the dimension of familiarity was shown first to avoid bias. Figure 1b shows the format employed.

Fig. 1
figure 1

Rating format employed for affective variables (a): valence, arousal, and dominance respectively; and for psycholinguistic variables (b): familiarity, concreteness, and imageability, respectively

Full size image

Participants completed the assessments in group sessions in quiet rooms of their corresponding universities, institutes, or homes. Each participant rated one questionnaire in a single session lasting about 45 min. We did not give participants explicit instructions about ambiguous words; instead, they were encouraged to select the first meaning that came to mind and to answer each point as fast as possible. Unknown words were either left blank or rated according to their first impression. However, participants could also use a glossary provided upon request.

At the beginning of the test session, participants were told they would read a series of words for us to assess their first impressions; the volunteer nature of their participation, as well as its confidentiality, was emphasized. Afterward, the rating system for the SAM or FIC scale was explained orally. The groups completed only one scale, either SAM or FIC. To ensure that the participants understood the task, four words not included in the ANEW set were used as practice items (mate, a typical infusion from South America, creencia [belief], bacteria, and lira [lyre]).

The instructions for the SAM scale were taken from the original ANEW database (Bradley & Lang, 1999a) and adapted into RS (for the original version, see Appendix). The instructions were as follows: For the valence scale, indicate how much a given word refers to something pleasant or unpleasant. If it makes you feel completely pleased, satisfied, happy, or hopeful, you should choose the rating 9 (very pleasant). On the other hand, if it makes you feel angry, bored, hopeless, or annoyed, you should choose the rating 1 (very unpleasant). For the arousal scale, you should indicate how calm or active you feel when you read each word. If it makes you feel fully active, stimulated, and awake, you should mark the rating 9 (very excited). By contrast, if it makes you feel completely bored, calm, or lazy, you should select the rating 1 (very calm). For the dominance scale, you should indicate how dominant or submissive a given word makes you feel. If it makes you feel autonomous, important, and in-control you should choose the rating 9 (very in-control). By contrast, if it makes you feel influenced, submissive, or guided, you should indicate the rating 1 (very controlled).

Besides, instructions for the FIC scale were taken from Montefinese et al. (2013) and presented as follows: For the Familiarity scale, we ask you to indicate how often a given word appears in your everyday life, whether spoken or written. If you read or hear the word every day, several times a day, you should choose the rating 9 (very familiar). If, however, you rarely hear or read the word, you should choose the rating 1 (very unfamiliar). For the concreteness scale, we ask you to indicate how much a given word triggers something that can be perceived directly through the senses. If a given word can be easily perceived through them, you should choose the rating 9 (concrete), while if a given word is abstract (that is, it cannot be perceived through the senses), you should choose the rating 1 (abstract). For the imageability scale, we ask you to indicate how easy it is to create a mental image of a given word, like a sound, a picture, or any other sensory experience. If the mental image comes to mind quickly, you should choose the rating 9 (very imageable). On the other hand, if the word is very difficult to imagine, you should choose the rating 1 (very unimageable).

Results and discussion

Description of database

The database, available as supplementary material, contains 1034 RS words with normative affective ratings (valence, arousal, dominance) and complementary subjective (familiarity, imageability, concreteness) and objective (e.g., number of letters, syllables, regional word frequency, grammatical class, etc.) psycholinguistic ratings are reported (see Appendix for more information). For regional word frequency, indexes were retrieved between the months of April–July 2020 from CORPES XXI (RAE), a corpus comprised of oral and written texts from Spain, America, the Philippines, and Equatorial Guinea. It provides the number of occurrences for a given target word for every million words and, crucially, it allows users to search for occurrences in a specified region (Rio de la Plata, for instance).

First, we present the descriptive results obtained in the adaptation of the ANEW dataset for RS considering the three emotional dimensions, and then we report the analyses run by dividing the sample by gender. The next step was to compare the RS dataset with the United States and the Spanish European version of the ANEW. Finally, we analyzed the relationship between the affective, subjective, and objective psycholinguistic indexes. Since Freq-CORPUS-Reg and Freq-CORPUS-Gen had a correlation of .997, p < .0001, only the Freq-CORPUS-Reg was employed in the analysis. The statistical analysis was performed with the SPSS 19 package.

Reliability of the measures

To ensure rating consistency in estimating inter-rater reliability, multiple iterations (total iterations: 100) of two groups of participants were randomly formed. Then a Pearson correlation was run for the three affective variables on each iteration. Afterwards, the mean correlation was calculated for each variable. The three correlations were high and positive: Valence r = .972, p < .0001, Arousal r = .818, p < .0001, and Dominance r = .821, p < .0001. These results indicated a high reliability between measures, and in agreement with previous findings (Monnier & Syssau, 2014), valence was the most consistent variable, with greater variability for arousal and dominance. Furthermore, Cronbach’s alpha was used to assess internal consistency for the three affective variables across the six booklets. Responses had a high degree of internal consistency for Valence (alpha = .81), Arousal (alpha = .93), and Dominance (alpha = .93).

The same analysis was run for the subjective psycholinguistic variables. The three correlations were high and positive: Familiarity r = .813, p < .01, Imageability r = .861, p < .01, and Concreteness r = .917, p < .01. Internal consistency was also assessed for the three psycholinguistic variables across the six booklets. Responses were found to have a high degree of internal consistency for Familiarity (alpha = .96), Imageability (alpha = .98), and Concreteness (alpha = .98).

Affective variables: Valence, arousal, and dominance

Figure 2 shows the distribution of the RS word ratings in the bidimensional space of Valence x Arousal for both male and female respondents. The distribution adjusted to the boomerang shape previously reported (Bradley & Lang, 1999a; Guasch et al., 2016; Redondo et al., 2007; Soares et al., 2012). This shape indicated that words rated as highly pleasant or highly unpleasant were also rated as more arousing which indicated that the relationship is best described as quadratic. Valence was considered an independent factor and Arousal a dependent variable for the quadratic regression analysis y = 8.51x2 – 1.14x + .096, which was highly significant, explaining 24% of the variance, R = .488 F(2,1031) = 161.51, p < .0001. The linear regression indicated a r = .353 F(1,1032) = 146.64, p < .0001, which, although significant, only explained 12% of the variance. The Pearson correlation for Valence and Arousal indicated a negative correlation between the variables (Guasch et al., 2016; Imbir, 2015; Schmidtke et al., 2014; Soares et al., 2012).

Fig. 2
figure 2

Distribution of mean values (by type of word) for the 1034 words of the Rioplatense Spanish adaptation of the ANEW in the valence and arousal affective dimensions

Full size image

To further investigate the relation between Valence and Arousal, we classified the words into three Word Types using the criteria employed by Ferré et al. (2012), Monnier and Syssau (2014), and Yao et al. (2017). Negative items: words between 1 and 4 points on the Valence scale; neutral items: words between 4.01 and 6; and positive items: words between 6.01 and 9. From this classification, we obtained 353 negative words (34% of the sample, range 1.07–3.98, M = 2.62, SD = 0.6); 217 neutral words (21% of the sample, range 4.02–6.06, M = 5.16, SD = 0.57); and 464 positive words (45% of the sample, range 6.02–8.73, M = 7.3, SD = 0.72). Then, we calculated correlations between Valence and Arousal within each category. There was a negative correlation between Valence and Arousal for negative words r = – .383, p < .0001, and a negative correlation also for neutral words r = – .284, p < .0001; the lower the value of the word, the more arousing. For positive words, there was a positive correlation between the variables r = .225, p < .0001; the higher the value of the word, the more arousing. If we divide Fig. 2 into four quadrants, while the negative words are concentrated in the upper left quadrant, indicating that these words were more arousing, the positive words are distributed mostly along the arousal dimension. This could indicate that for positive items, valence and arousal could be independent in comparison with negative items. Similar distributions were found by Guasch et al. (2016); Imbir (2015); Monnier and Syssau (2014); Schmidtke et al. (2014); Soares et al. (2012); Yao et al. (20172017). Nevertheless, the words of this corpus cover a wide range of scores in both the arousal (1.77–8.43) and valence dimensions (1.07–8.73), allowing researchers to select items combining both dimensions.

Dominance is the least studied variable, neglected as it is in several studies. Our results on this variable are depicted in Fig. 3a, b. Our analysis yielded a positive linear correlation between Dominance and Valence, which was high and significant, r = .839, p < .0001, indicating that the more pleasant the word, the more in-control the person felt about it (e.g., risa [laughter] and logro [achievement]; Imbir, 2015; Montefinese et al., 2013; Warriner et al., 2013). Regarding Dominance and Arousal, the analysis, although significant, indicated a lower linear correlation r = – .129, p < .0001. This showed that the more arousing the word, the less in-control the person felt (e.g., words such as peligro [danger], violación [rape], and pánico [panic]). Other authors have found that the quadratic analysis adjusted better to this relation (Montefinese et al., 2013; Warriner et al., 2013). This was not our case, since both analyses yielded similar results (quadratic regression for Dominance and Valence R = .705 F(2,1031) = 1230.02, p < .0001; quadratic regression for Dominance and Arousal R = .141 F(2,1031) = 84.48, p < .0001)

Fig. 3
figure 3

Distribution of mean values for the 1034 words of the Rioplatense Spanish adaptation of the ANEW divided by type of word in the Dominance and Arousal (a) and Dominance and Valence (b) affective dimensions

Full size image

When divided by Word Type, the analysis of Dominance x Valence indicated a positive correlation for the three Word Types (negative: r = .390, p < .001; neutral: r = .443, p < .001; positive: r = .567, p < .001), which indicates that the more unpleasant the word, the more controlled the person felt, while the more pleasant the word, the more in-control the person felt. Regarding Dominance x Arousal, there was a negative correlation for the negative and neutral words (negative: r = – .071, p < .01; neutral r = – .284, p < .001) but a positive correlation for the positive words (r = .513, p < .001). For negative and neutral words, the more arousing the word, the more controlled the person felt. On the other hand, for positive words, the more arousing the word, the more in-control people felt.

Gender differences

A previous study on an Argentinian sample found differences between female and male subjects in emotional appraisal for different kinds of stimuli (Irrazabal et al., 2015; Irrazabal & Tonini, 2020). Therefore, to analyze possible rating differences between genders, we performed was a MANOVA with Gender (male subjects vs. female subjects) and Word Type (negative, neutral, positive) as the between factors and Valence, Arousal, and Dominance as the dependent variables. This analysis yielded a main effect of Gender on Arousal F(1,2067) = 4.92, p = .026. The post hoc test indicated that female subjects rated words as more arousing than male subjects, in line with the results found by Irrazabal et al. (2015) with IAPS, and by Soares et al. (2012) in the European Portuguese adaptation of ANEW, as opposed to the results found on the American sample (Bradley & Lang, 1999a). Also, the MANOVA showed a main effect of Word Type on Valence F(2,2067) = 9149.52, p < .0001; Arousal F(2,2067) = 173.3, p < .0001 and Dominance F(2,2067) = 1105.04, p < .0001. The main effect of Word Type on Valence indicated that positive words were rated with higher valence than neutral words, and these with higher valence than negative words (Monnier & Syssau, 2014; Soares et al., 2012). Regarding Arousal, negative words were more activating than positive ones, and these were more arousing than neutral words (in agreement with Montefinese et al., 2013; Soares et al., 2012; in contrast to Monnier & Syssau, 2014). For Dominance, positive words were more linked to being in-control than neutral words, and the neutral words more linked to being in-control than negative words (Imbir, 2015; Montefinese et al., 2013; Soares et al., 2012).

The interactions of Gender x Word Type for Valence F(2,2067) = 22.90, p < .0001, Arousal F(2,2067) = 6.57, p < .0001 and Dominance F(2,2067) = 41.19, p = .001 were significant. Regarding Valence, female subjects rated positive words as more pleasant and negative words as more unpleasant than did the male participants, without differences for the neutral words. For example, amado [loved] had a mean rating of 8.60 for women and 7.13 for men, while cachetazo [slap] had a mean rating of 1.57 for women and 3.38 for men. This pattern of results is different from that found by Irrazabal et al. (2015) with the IAPS for Argentinian male subjects, since their male sample had a positive bias for positive images. In accordance with our results, their female sample also had a negative bias for unpleasant images.

Concerning Arousal, negative words were rated as more arousing by women than men. For example, violación [rape] (women 8.37; men 7.82), masacre [massacre] (women 8.05; men 6.94), and piojos [lice] (women 7; men 5.60). Finally, for Dominance, the post hoc analyses showed that men indicated more control for negative words and women for positive ones. Regarding Arousal and Dominance, the results are similar to the outcomes found by Irrazabal et al. (2015), Montefinese et al. (2013), and Soares et al. (2012), but in the Arousal domain they are different from the outcomes of Monnier and Syssau (2014), although it should be noted that not all the cited studies divided the list by Word Type.

Ratings of men and women were highly correlated for Arousal (r = .749, p < .0001) and Dominance (r = .720, p < .0001), but especially for Valence (r = .961, p < .0001), suggesting that Valence was the most stable variable between women and men living in the same country (Fig. 4).

Fig. 4
figure 4

Distribution of mean values (for female and male subjects) for the 1034 words of the Rioplatense Spanish adaptation of the ANEW in the Valence and Arousal affective dimensions

Full size image

Cultural differences

To analyze possible differences between adaptations, the relationship between the different affective variables were inspected. The correlations between Valence, Arousal, and Dominance for Argentina–Spain and Argentina–USA were highly significant. Argentina and Spain showed a positive correlation for Valence r = .933, p < .01, Arousal r = .756, p < .01, and Dominance r = .758, p < .01. For Argentina and the USA, there was a positive correlation for Valence r = .895 p < .01, Arousal r = .665, p < .01, and Dominance r = .794, p < .01. When the correlations were run divided by gender, they remained highly significant (see Table 1 for further details). Once again, Valence was the most stable variable (Eilola & Havelka, 2010; Guasch et al., 2016; Monnier & Syssau, 2014; Redondo et al., 2007; Soares et al., 2012).

Table 1 Linear correlations between Rioplatense Spanish, European Spanish (Spain), and American (USA) ANEW word ratings in the three affective dimensions for all subjects and for female and male samples separately
Full size table

To further analyze possible rating differences by country, we performed two MANOVAs with Country (Argentina vs. Spain and Argentina vs. USA), Gender (female vs. male subjects) and Word Type (negative, neutral, positive) as the between-factors and Valence, Arousal, and Dominance as dependent variables. Only the analysis involving the variable Country is explained here. Results can be seen in Fig. 5 and Table 1.

Fig. 5
figure 5

Distribution of mean values for the 1034 ANEW words in American (USA), European Spanish (Spain), and Rioplatense Spanish (Argentina) for Valence and Arousal affective dimensions

Full size image

In the comparison with Spain, there was a significant interaction of Country x Word Type for the three dependent variables, Valence: F(2,4124) = 7.37, p < .0001; Arousal: F(2,4124) = 34.77, p < .0001, and Dominance: F(2,4124) = 35.17, p < .0001. There was also a significant interaction between Country x Word Type x Gender for Valence F(2,4124) = 5.15, p = .006, and Dominance F(2,4124) = 6.72, p = .001. The post hoc test for the interaction of Country x Word Type for Arousal (other analyses are subjected to the Country x Word Type x Gender interaction) indicated that negative and neutral words were more arousing for the Argentinian sample, while positive words were more arousing for the Spanish sample. The post hoc analyses for the Country x Word Type x Gender interaction showed that, regarding Valence, negative words were rated with higher valence by Argentinian male and female samples than their Spanish counterparts. For neutral and positive words, only Argentinian women rated the items with higher valence than Spanish women. Finally, in the Dominance dimension, Argentinian men and women felt more in-control for positive and negative words than the Spanish sample.

In the comparison with the USA sample, there was a main effect of Country for Arousal F(1,4124) = 287.43, p < .0001 and Dominance F(1,4124) = 8.45, p = .004. The interaction of Word Type x Country was significant for Valence F(2,4124) = 6.78, p = .001; Arousal F(2,4124) = 65.24, p < .0001, and Dominance F(2,4124) = 12.38, p < .0001. The Gender x Country interaction was significant for Dominance F(1,4124) = 19.31, p < .0001. The post-hoc analyses for Dominance (since the other analyses are subjected to the interaction mentioned below) indicated that the Argentinian sample felt more in-control for the negative words than the USA sample, and Argentinian women felt more in-control than USA women. Finally, the interaction Word Type x Country x Gender achieved significance for Valence F(2,4124) = 3.71, p = .025 and Arousal F(2,4124) = 3.92, p = .02. In this interaction, the post-hoc analyses for Valence indicated that negative words were given higher ratings by the male USA sample, and the positive words were given higher ratings by the male Argentinian sample. Regarding Arousal, negative and neutral words were perceived to be more arousing by the Argentinian male and female samples than by the USA sample.

Relation between affective, subjective, and objective psycholinguistic variables

Table 2 and Figs. 6 and 7 show the results of Pearson correlations between affective and other subjective and objective psycholinguistic variables.

Table 2 Correlations between affective, psycholinguistic, and objective indexes for the Rioplatense Spanish words
Full size table
Fig. 6
figure 6

Scatterplot matrix with linear regressions of affective and subjective psycholinguistic variables

Full size image
Fig. 7
figure 7

Scatterplot matrix with linear regressions of subjective psycholinguistic variables

Full size image

Regarding the relationships between subjective psycholinguistic and affective variables, for Valence the correlations indicated that the words with higher valence were also more imaginable, concrete, and familiar. Particularly, for Valence and Concreteness, a quadratic regression analysis explained the relationship between variables R = .250, F(2,1033) = 34.48, p < .0001, in line with Yao et al. (2017). The three subjective psycholinguistic variables showed lower significant correlations with Arousal. The analysis indicated that the more arousing the word, the lower the ratings for Familiarity, Concreteness, and Imageability. Like previous adaptations (Guasch et al., 2016), the quadratic analysis explained the relationship between Arousal and Concreteness R = .133, F(2,1033) = 9.32, p < .0001. Likewise, Dominance only showed a significant moderate correlation with Familiarity, which indicated higher in-control ratings for familiar than unfamiliar words. Although the Pearson correlation did not yield significant differences for Dominance and Concreteness, the quadratic analysis did, R = .240 F(2,1033) = 31.65, p < .0001.

Regarding the subjective psycholinguistic variables themselves, as can be seen, the strongest correlation was between Concreteness and Imageability, indicating that it was easier to elicit a mental image for concrete words than for abstract ones (e.g., silla [chair] vs. unidad [unit]). Familiarity and Imageability also showed a moderately significant correlation, indicating that participants found it easier to form a mental image when reading a familiar than an unfamiliar word (e.g., hermano [brother] vs. pitón [python]). Another significant, but low, correlation was between Familiarity and Concreteness. Results showed that highly familiar words were also rated high in Concreteness (e.g., tormenta [storm] vs. resentirse [resent]). The results were like those found by Guasch et al. (2016), Warriner et al. (2013), and Yao et al. (2017). Likewise, a previous study carried out by Martinez-Cuitiño et al. (2015) on an Argentinian sample reported similar results for linguistic variables associated with images. Unlike our results, this study also found differences between Argentinian locations (Córdoba vs. Buenos Aires). However, our results about the relation between concreteness and imageability differ from the outcomes of Eilola and Havelka (2010), who found a negative correlation between the variables.

To further analyze the possible differences between psycholinguistic and affective variables, a one-way ANOVA was run with Word Type as the factor and Concreteness, Familiarity, and Imageability as the dependent variables. Regarding Concreteness, a significant effect emerged, F(2,1033) = 34.18, p < .0001. Post hoc analyses indicated that neutral words were more concrete than positive words, and these were more concrete than negative words. For Familiarity, the ANOVA yielded a significant effect F(2,1033) = 90.43, p < .0001, and post hoc analyses showed that positive words were more familiar than neutral ones, and these more familiar than negative words. Finally, a significant effect for Imageability was found, F(2,1033) = 19.54, p < .0001. Post hoc analyses indicated no differences between neutral and positive words, and these were more imageable than negative words.

Objective psycholinguistic indexes included number of Letters, Syllable Length, and Frequency. To analyze the relationship between Word Length and Frequency, Pearson correlations were carried out (Table 2). Results showed a negative but low correlation between syllables and frequency, which indicates a higher frequency for short words (i.e., words with fewer syllables).

The relationship between objective psycholinguistic indexes and affective variables was also examined. The analyses revealed positive but low correlations between Frequency, Valence, and Dominance, which indicated that the more frequent the word, the more pleasant feelings it evoked (Monnier & Syssau, 2014; Montefinese et al., 2013), and the more in control the person felt. There was also a positive low correlation between Syllables, Letters, and Arousal, indicating that higher word length was also more arousing (similar results for a romance language can be found in Monnier & Syssau, 2014). Finally, there was a negative low correlation between Valence and Letters, indicating that fewer letters corresponded to more pleasant words.

While significant, all correlations between objective and subjective psycholinguistic indexes were low. Results showed negative correlations between Syllables, Letters, and the psycholinguistic variables, indicating that more syllables and letters were associated with less familiar, concrete, and imageable words. Finally, there was an expected positive correlation between Frequency and Familiarity, which indicated that higher frequency was associated with higher familiarity ratings.

Our data show that affective ratings are at least partially dependent on other objective and subjective psycholinguistic features, like concreteness, frequency, and word length. Furthermore, regional variations of the same language (Spanish) have also been shown to elicit rating differences across affective dimensions.

Summary and conclusions

To precisely identify the affective characteristics of the words that affect cognitive performance, researchers need databases and normative sets that enable the control and manipulation of these variables (Monnier & Syssau, 2014). The advance of different emotional models depends, at least in part, on valid and reliable experimental paradigms. Therefore, it is essential to have well-grounded stimuli for the empirical investigation of emotions (Kristensen et al., 2011). In this sense, the ANEW is a very useful tool; however, it is necessary to consider that there are sociocultural differences that might modulate emotional response, since the affective assessment of emotional words seems to be language- and culture-specific. As norms for one language and country cannot extend to others, the existence of comprehensive normative ratings is of great importance. Therefore, our main goal was to provide RS norms for the ANEW dataset and to compare our results with the original USA dataset and the European Spanish version, the latter being the direct comparison of our work. Likewise, different genders tend to have diverse responses to affective stimuli (Angelini, 2007; Irrazabal et al., 2015; Irrazabal & Tonini, 2020; Soares et al., 2012; Spalek et al., 2015) and thus our second goal was to provide separate scores for male and female subjects. Finally, to supplement the affective ratings, other linguistic variables were considered. Therefore, our final goal was to provide subjective and objective psycholinguistic indexes along with affective ones for every word on the dataset.

Summarizing, our results are like those found in other ANEW adaptations (e.g., Bradley & Lang, 1999a; Guasch et al., 2016 ; Yao et al., 2017), validating our outcomes and making the database suitable for future studies and comparisons. Regarding the bidimensional affective space of valence and arousal, we replicated the boomerang shape, in which the most pleasant or unpleasant words were also the most arousing ones. In accordance with previous studies, we found a negative correlation between valence and arousal, and while the pleasant words were scattered along the arousal dimension, the unpleasant words were concentrated in the upper quadrant, which indicates that arousal and valence are independent variables for pleasant words (Guasch et al., 2016; Imbir, 2015; Monnier & Syssau, 2014; Schmidtke et al., 2014; Soares et al., 2012; Yao et al., 2017). Concerning dominance, pleasant words were associated with higher in-control ratings and unpleasant words were associated with lower ratings (Imbir, 2015; Montefinese et al., 2013).

When gender was considered, our results yielded differences between female and male subjects, as opposed to the European Spanish sample (Redondo et al., 2007). Female subjects rated unpleasant words with lower valence and pleasant words with higher valence than did male subjects. Also, female subjects found unpleasant words more arousing, and they felt more in-control when reading pleasant words in comparison to males, who felt more in-control when reading unpleasant words (Monnier & Syssau, 2014; Montefinese et al., 2013; Soares et al., 2012).

Correlations between Argentinian female and male subjects, and between countries (Argentina vs. Spain; Argentina vs. USA) were highly significant for arousal and dominance but especially for valence, suggesting that valence is the most stable variable not only among women and men of the same country but also across countries (Monnier & Syssau, 2014).

While the correlations between Argentina–USA and Argentina–Spain were high and significant, the MANOVAs performed showed significant differences between countries, which indicates cross-cultural differences between languages and regions. Although Argentina and Spain share the same language, historical and cultural backgrounds give rise to dialectical variations that influence the emotional appraisal. Our results show that it is necessary to validate the findings across countries. Furthermore, RS is spoken not only in Argentina but also in Uruguay, and thus further studies with different Latin American populations are needed to analyze these differences. A limitation of this set is that we did not evaluate it in every Argentinian province, and our sample is composed of young adults mostly from metropolitan areas. Therefore, future studies could expand on these results by adding more provinces and age-groups to the Argentinian sample.

Within the RS adaptation, our dataset showed that higher valence was associated with higher control, imageability, familiarity, and concreteness. Also, pleasant words were rated with higher frequency values. Arousal was associated with high control and unfamiliarity ratings as well as lower concreteness and imageability. These data show that some effects previously attributed to concreteness can also be explained by word emotionality (Guasch et al., 2016; Yao et al., 2017).

The concreteness effect could be explained by familiarity, imageability, or context availability, and our dataset enables researchers to control for familiarity and imageability. Future research could expand on these norms by providing ratings for other psycholinguistic variables such as context availability, age of acquisition, and orthographic neighborhood (Guasch et al., 2016).

The main strength of our study is that we provide affective, objective, and subjective psycholinguistic indexes for 1034 RS words and that the affective index can also be divided by gender. Additionally, it expands on different published word samples; for example, it shows only a 4.35% overlap with the norms published by Hinojosa et al. (2015) and Hinojosa et al. (2016). Since we found differences between men and women, future research should take these differences into account to make it possible to select the stimuli. Furthermore, since the evaluation was performed in several Argentinian provinces and the sample comprised people born across the country, these norms are suitable for generalization. We expect that this version of ANEW will prove useful for research on affective word processing, enabling reliable comparisons across languages and geographical locations.