Since very early on in research on emotion, valence and arousal have been central to the classification of affective experiences (Osgood, Suci, & Tannenbaum, 1957; Wundt, 1912/1924). Valence is a subjective assessment that describes how pleasant a stimulus is (ranging from pleasant to unpleasant), whereas arousal refers to the subjective level of activation or intensity that a stimulus elicits (ranging from quiet to active). Although there is no absolute consensus about the basic components of affect (e.g., Barrett & Russell, 1999; Fontaine, Scherer, Roesch, & Ellsworth, 2007; Lang, 1995; Larsen & Diener, 1992; Reisenzein, 1994; Russell, 1980; Thayer, 1989; Watson & Tellegen, 1985), most current models recognize valence and arousal as fundamental axes of that construct (Barrett, 1998; Kuppens, Tuerlinckx, Russell, & Barrett, 2012). The dimensional model of emotions originally proposed by Wundt includes, along with valence and arousal, a third variable: dominance, a subjective assessment of how “in control” a word makes you feel. However, this third variable has not been widely used in the literature (Montefinese, Ambrosini, Fairfield, & Mammarella, 2014), is not a strong predictor of the variance of affective judgments (Bradley & Lang, 1994; Lang, Bradley, & Cuthbert, 2008), and is strongly correlated with valence (Warriner, Kuperman, & Brysbaert, 2013).

In addition to their importance in the study of emotion itself, affective variables have been shown to play an important role in multiple areas of cognition; for example, they influence lexical processing (e.g., Kanske & Kotz, 2007; Kousta, Vinson, & Vigliocco, 2009; Kuchinke, Võ, Hofmann, & Jacobs, 2007; Kuperman, Estes, Brysbaert, & Warriner, 2014; Ortigue et al., 2004; Scott, O’Donnell, Leuthold, & Sereno, 2009; Syssau & Laxén, 2012), they affect recall in both short-term memory (Majerus & D’Argembeau, 2011; Mammarella, Borella, Carretti, Leonardi, & Fairfield, 2013; Monnier & Syssau, 2008) and long-term memory (Dewhurst & Parry, 2000; Kensinger & Corkin, 2003; Talmi & Moscovitch, 2004), and they modulate attention effects (Mathewson, Arnell, & Mansfield, 2008; Stormark, Nordby, & Hugdahl, 1995). The interest in the effects of emotional properties on language processing goes beyond single words, with several studies having looked into their effect on the morpho-syntactic processing of sentences. For example, Hinojosa et al. (2014) proposed that words with negative content can facilitate the processing of gender agreement violations (at least in the early stages of processing). Similarly, Martín-Loeches et al. (2012) found evidence that the valence of adjectives modulated how sentences are processed at both the syntactic and semantic levels, but Díaz-Lago, Fraga, and Acuña-Fariña (2015) did not find an interaction in the processing of grammatical and emotional properties of words embedded within sentences.

Valence and arousal are, by their very nature, subjective characteristics that reflect experiences and associations with particular objects, events, and words. Values for these variables are usually collected via subjective ratings: Participants are exposed to different stimuli (usually lists of words) and are asked to state how happy or sad, how active or quiet, each stimulus makes them feel. Probably the most influential set of affective norms is Bradley and Lang’s (1999) Affective Norms for English Words (ANEW). In ANEW, participants provided their ratings on the basis of the Self-Assessment Manikin (SAM), a nonverbal scale that depicts humanoid figures representing different values along each emotional dimension. Although this method has been widely used in several replications of ANEW, Warriner et al. (2013) showed that equivalent normed data can be obtained by using verbal anchors on a numerical scale.

Norms of the affective properties of words have been published in multiple languages, such as European Portuguese (Soares, Comesaña, Pinheiro, Simões, & Frade, 2012), Brazilian Portuguese (Kristensen, de Azevedo Gomes, Justo, & Vieira, 2011), French (Bonin et al., 2003; Gilet, Grühn, Studer, & Labouvie-Vief, 2012; Monnier & Syssau, 2013, among others), German (Kanske & Kotz, 2010; Lahl, Göritz, Pietrowsky, & Rosenberg, 2009; Võ et al., 2009; Võ, Jacobs, & Conrad, 2006), Polish (Imbir, 2015), Finnish (Eilola & Havelka, 2010; Söderholm, Häyry, Laine, & Karrasch, 2013), Italian (Montefinese, Ambrosini, Fairfield, & Mammarella, 2014), and Dutch (Moors et al., 2013). In Spanish, the largest set of emotional norms so far was published in Redondo, Fraga, Padrón, and Comesaña (2007), which includes valence, arousal, and dominance norms for the Spanish translation equivalents of the original 1,034 items from ANEW using a method very similar to that in Bradley and Lang (1999). Similarly, Ferré, Guasch, Moldovan, and Sánchez-Casas (2012) also used Bradley and Lang’s SAM to obtain valence and arousal norms for 380 words belonging to three semantic categories—namely, animals, people, and objects. Hinojosa et al.’s (2015) “Madrid Affective Database for Spanish” represents an interesting development, in that it includes not only norms for arousal and valence for 875 words, but also ratings for five discrete emotional categories—happiness, anger, sadness, fear, and disgust—thus allowing a more fine-grained study of the emotional properties of words. Hinojosa et al. also included a useful table listing the details of most previous sets of norms of affective characteristics in various languages.

The availability of compatible norms in several languages not only opens the possibility of conducting relevant research in that particular language, but also allows for cross-language comparison studies (Campos & Astorga, 1988; Harris, Ayçiçegi, & Gleason, 2003; Russell, 1991); such norms are useful in the study of how bilingualism affects processing, whether in terms of lexical variables or of the expression of emotion (e.g., Altarriba & Canary, 2004; Opitz & Degner, 2012). Research in these areas is very active and has yielded interesting results for both bilingualism and the study of emotions. For example, there is evidence that in bilinguals, the activation of emotion is modulated by their level of proficiency in each language (Degner, Doycheva, & Wentura, 2011; Robinson & Altarriba, 2015). Bilinguals often report that emotional words carry a higher level of intensity in their first (or dominant) language, particularly regarding taboo or swear words (e.g., Dewaele, 2004; Pavlenko, 2012), a difference also observed with measures of autonomic response (e.g., Colbeck & Bowers, 2012; Harris et al., 2003). On the other hand, there is evidence that when a bilingual is highly proficient in both languages, equivalent emotional activation is found in the two languages, even for late bilinguals (e.g., Eilola, Havelka, & Sharma, 2007). Also, some cross-linguistic differences may relate to emotion words. For example, some recent work (e.g., Kazanas & Altarriba, 2015) has investigated the distinction between emotion words (e.g., joy/anger) and emotion-laden words (e.g., puppy/coffin). The authors found that in English, emotion-word processing occurs more quickly than emotion-laden word processing, especially for positive words. However, this effect seems to be less robust in Spanish. Research on the relationship between emotional words and bilingualism also encompasses other areas of cognition, such as memory. For example, is has been shown that recall is better for emotional words than for neutral words in monolingual speakers (e.g., Kensinger & Corkin, 2003; LaBar & Phelps, 1998; MacKay & Ahmetzanov, 2005), but there is a debate as to whether that advantage is also present for the second language of a bilingual. Anooshian and Hertel (1994) did not find such advantage, whereas Ayçiçegi and Harris (2004) found that, under certain circumstances, the effect was even larger in the second language. As can be seen, the literature on bilingual processing of emotional words encompasses many different topics and is far from settled, so much remains to be done in that regard. What all these studies have in common is the need for information on the emotional characteristics of large sets of words in multiple languages. Large sets of emotional norms are necessary in order to create well-controlled item sets in experimental studies, as well as for input in regression and other types of analysis.

In most of the norming studies mentioned before, the numbers of items have ranged from a few hundred to just over a thousand. Recent years have seen the publication of larger sets of norms of different lexical characteristics, usually including several thousand words (e.g., Brysbaert, Stevens, De Deyne, Voorspoels, & Storms, 2014; Kuperman, Stadthagen-Gonzalez, & Brysbaert, 2012; Warriner et al., 2013). The availability of such large norm databases has greatly facilitated the creation of stimulus sets, as well as making it possible to include such variables in the automated analysis of text samples (Leveau, Jhean-Larose, Denhière, & Nguyen, 2012).

In addition, large data sets make it possible to study the relationship between the affective properties of words and other lexical or semantic variables. Warriner et al. (2013), for instance, reported that positive words have a higher frequency than negative words. They related this finding to the observation that the English language has a bias toward positivity or pro-social benevolent communication interactions (see also Warriner & Kuperman, 2015). The relationship between word frequency and arousal has tended to be negative (high-frequency words are rated, on average, as calmer), although this overall relationship hid the fact that some of the high-frequency words were very arousing (such as love, god, life, money, hell, kill, fuck, happy). Children seem to be taught positive, low-arousing words first, since there is a negative correlation between age of acquisition (AoA) and valence, together with a positive correlation between AoA and arousal. Finally, Warriner et al. reported that positive words tended to refer to concepts with high imageability.

In these large norming studies, data have tended to be collected in a modular fashion, dividing up the stimulus list into blocks that can be rated in a single session, and introducing interrater reliability measures to allow for the integration of the blocks into a single list. This approach allows for the collection of ratings for very large word lists. So far the biggest set of affective word norms for English is Warriner et al. (2013), which provides ratings for valence, arousal, and dominance for nearly 14,000 words. Our purpose in the present study was to generate a set of norms for a comparable number of Spanish words, using procedures very similar to those of Warriner et al. Our large database will also allow us to explore how valence and arousal ratings relate to each other and to other lexical variables—namely frequency, AoA, familiarity, imageability, and concreteness—that are available in Spanish for large sets of words. Warriner et al. found weak correlations between emotional variables and most lexical variables included in their analyses, but found interesting relationships between their emotional ratings and the above-mentioned lexical characteristics.

An additional objective was to evaluate whether affective properties such as valence and arousal are fully determined by the lemma of a word. A morphologically rich language such as Spanish allows us to evaluate whether, for example, verb conjugations modify the affective properties of the infinitive form. As was noted by Rivera, Bates, Orozco-Figueroa, and Wicha (2010), very few lexical databases include information for inflected verbs. Warriner et al.’s (2013) emotional norms included only infinitive verbs, on the assumption that emotional values generalize to inflected forms, but so far this assumption has not been empirically tested. Moreover, whereas English verbs undergo little inflection, in languages with morphologically rich inflections such as Spanish, infinitives can be quite different from their conjugated forms. We were interested in whether there are differences in emotional ratings with regard to person (first vs. third) and tense (present, future, and conditional). With regard to person, our interest arose in view of the literature on self-referential processing of emotional words (e.g., Fossati et al., 2003; Northoff et al., 2006), which points to differences in the processing of emotional words when they are framed as personal attributes of the participant versus of another person. With regard to verb tense, we were interested in whether differences in ratings would emerge between present (ongoing, immediate actions), future (certain but unfulfilled actions), and conditional (potential actions). It is also possible that such differences, if they exist, only emerge in the presence of larger differences in word form—that is, for irregular verbs—so we looked into that contrast as well. Determining whether different verb conjugations share rating values is important for two reasons: (1) to determine the extent to which emotional ratings are dependent on the word form or are shared by variations of a base word (a “lemma”), and more pragmatically, (2) to determine whether it is necessary to include conjugated forms in these types of ratings, or whether infinitives are enough.

Method

Participants

A total of 512 native speakers of Spanish took part in our study. All respondents were taking undergraduate psychology courses at the University of Murcia, Spain, and received extra credit for their participation. Participants had the choice to rate up to three “blocks” of words per variable, and to provide ratings for one or both variables (ratings for arousal were collected at least 11 weeks after those for valence). A total of 233 participants rated words for both valence and arousal, whereas the rest only did so for one or the other variable. Valence ratings were obtained from 350 raters (of which 294, or 80 %, were female); of those, 79 % completed three blocks of ratings, 9 % two blocks, and 12 % one block. The raters for valence had an average age of 22 years 3 months (range = 18 to 62 years, SD = 6 years 1 month). Arousal ratings were obtained from 395 raters (of which 327, or 82.7 %, were female); of those, 56 % completed three blocks of ratings, 21 % two blocks, and 23 % one block. The raters for arousal had an average age of 22 years 5 months (range = 18 to 62 years, SD = 5 years 8 months). The dominance of female participants is in line with current practice in psychological experiments and is acceptable, given the high correlation of the valence and arousal ratings provided by males and females (Montefinese et al., 2014; Moors et al., 2013). Still, it is fair to say that the present data may not be the most optimal to predict the responses of an all-male participant group.Footnote 1

Materials

The 14,037 words included in this set of norms were assembled from four sources (with some items being present in more than one source): (1) First, we used all 1,034 words from the adaptation of the ANEW norms to Spanish by Redondo, Fraga, Padrón, and Comesaña (2007). These words were used as “controls” for validation purposes. (2) In order to include words from a variety of semantic categories, we included 3,509 words from Marful, Díez, and Fernandez (2014), a Spanish adaptation of Battig and Montague (1969). In that study, participants were asked to write as many words as they could for 56 semantic categories. For our word list, we included all words with an availability of 0.7 or more and excluded proper nouns (male and female), geographical names (provinces, countries, and cities), and institutional names (colleges and brands). If both singular and plural were present in the list, only the singular form was included. If only the plural was present, the plural form was included. (3) To explore the extent to which conjugated forms share emotional ratings with their infinitive form, a sample of 104 verbs were added. The list included regular verbs with all three infinitive terminations (-ar, -er, and -ir), as well as irregular verbs. We took care not to include verbs with conjugations that could also be a different part of speech (e.g., juego is both the conjugated verb “I play” and the noun “game”). For each verb, we included its infinitive form as well as the following conjugations: present indicative first person, present indicative third person (which is also the first person imperative for regular verbs), conditional (first and third person share the same word form), and future first person, for a total of 520 words. (4) The rest of the words (8,974) were taken from EsPal (Duchon, Perea, Sebastián-Gallés, Martí, & Carreiras, 2013), a Spanish database that provides information about many lexical properties based on a very large corpus. In order to include content words with a variety of grammatical functions, we included words with at least one count per million in EsPal’s individual listings of verbs, nouns, adjectives, and adverbs (excluding adverbs ending in -mente, the equivalent of “-ly” in English). Many of the words included in this list can function as more than one part of speech (e.g., the same word form can be both a verb and a noun), but taking into account just the most frequent part-of-speech category for each word according to EsPal, our list had the following composition: nouns = 61.8 %, adjectives = 21.3 %, verbs = 15.8 %, adverbs = 0.7 %, determiners = 0.2 %, and pronouns = 0.1 %. An effort was made to eliminate foreign words, proper nouns, acronyms, and multiword utterances from the list. In terms of frequency, the average Zipf value (van Heuven, Mandera, Keuleers, & Brysbaert, 2014) of the set, according to EsPal, was 3.65 (SD = 0.73, range = 0.51 to 7.46, median = 3.58). This means that our stimuli are nicely centered in the frequency range (Zipf values of 0–3 indicate low-frequency words, and values of 4–7, high-frequency words).

Procedure

The stimuli were distributed across 43 blocks with between 351 and 354 words each. Each block consisted of nine calibrator words, 41 control words from Redondo et al.’s (2007) Spanish adaptation of ANEW, and a random selection of the other words in the list. The calibrators consisted of three each of verbs, adjectives, and nouns, spanning the full range of the scale in the Redondo et al. norms. Different sets of calibrators were chosen for valence and arousal and presented at the beginning of each block, in order to give participants a sense of the entire range of the stimuli that they would encounter. The calibrators for valence were (in ascending order of valence): sangriento (“bloody”), funeral (“funeral”), contaminar (“to pollute”), mendigo (“beggar), afectar (“to affect”), martillo (“hammer”), liso (“smooth”), mejorar (“to improve”), and alegre (“happy”). The calibrators for arousal were (in ascending order of valence): dormir (“to sleep”), silla (“chair”), nublado (“cloudy”), árbol (“tree”), ahorrar (“to save money”), insecto (“insect”), capaz (“able”), guapo (“handsome”), and ganar (“to win”). Control words were randomly interspersed with the other words on the list and used to assess the reliability of our ratings with regard to the Redondo et al. norms. The sets of control words could appear in more than one block, and they were similar but not identical for the valence and arousal blocks. All words in a block (except the calibrators) were randomized once and presented in the same order to each participant rating that block.

Each block was rated by 20 participants. Blocks with ratings that correlated with the mean ratings per items at less than .10 were removed and replaced with ratings from a different participant (2.7 % of the collected blocks; 26 blocks for valence, 20 blocks for arousal). These changes are already reflected in the participant information provided above.

Participants accessed the blocks online through Qualtrics. They first completed a brief demographic questionnaire and were then given written instructions for the relevant variable. The instructions given were similar to those of the ANEW norms; the exact wording in Spanish and an English translation are provided in the Appendix. Most participants rated each block in 30 min or less. Participants were asked to rate each word on a 9-point scale, as follows: valence from 1 = infeliz (“unhappy”) to 9 = feliz (“happy”), and arousal from 1 = tranquilo(a) (“quiet”) to 9 = exitado(a) (“excited”). There was a further option to indicate that they did not know the word (No conozco la palabra). Following Redondo et al. (2007), the direction of the scales was reversed from those used in ANEW, a change intended to make the rating process more intuitive to participants. Each screen displayed approximately 24 words, and the scale (1 to 9) and anchors (infeliz–feliz, tranquilo–excitado) were displayed at the top of each group of six words. Following Warriner et al. (2013), we decided not to include the SAMs that had been used in the ANEW study and other similar norms; in the “Results and discussion” section, we show that our numerical ratings correlated highly with the SAM ratings from ANEW.

Results and discussion

For valence, only 222 words were rated less than ten times, and 3.2 % of the responses were removed due to participants not recognizing the word. For arousal, only 183 words were rated less than ten times, and 3.5 % of the responses were removed due to participants not recognizing the word.

Both valence and arousal were positively skewed, with respective average ratings of 5.2 (SD = 1.27) and 5.3 (SD = 1.50) that exceeded the midpoint of the measurement scale (refer to Table 1 for the valence and arousal extremes). This finding is consistent with other rating studies (Warriner et al., 2013) and, for valence, supports the bias toward positive word types that is well-established in other languages (cf. Warriner & Kuperman, 2015). Histograms for the distributions of ratings for valence and arousal are shown in Fig. 1. The valence ratings had a larger range than did arousal; the average valence ratings ranged from 1.1 to 8.8, whereas the average arousal ratings ranged from 1.4 to 8.4. The ratings for both valence and arousal were relatively consistent across participants (SDs = 1.27 and 1.50, respectively) when compared to other norming studies of valence and arousal (SDs = 1.68 and 2.30, respectively; Warriner et al., 2013).

Table 1 Words with the highest and lowest valence and arousal ratings
Full size table
Fig. 1
figure 1

Distribution of ratings for valence and arousal

Full size image

The valence ratings were relatively stable across the valence spectrum (Fig. 2, left panel), whereas the variance in the arousal ratings was highest in the lowest-arousal words and decreased linearly as arousal increased (Fig. 2, right panel).

Fig. 2
figure 2

Average standard deviations (variance among responders) across the valence and arousal ranges

Full size image

Figure 3 shows the relationship between valence and arousal. As had been seen in previous studies (e.g., Warriner et al., 2013), the two variables had a U-shaped relationship; very positive and very negative words were highly arousing, whereas neutral words were less arousing (R 2 = .314, p < .001).

Fig. 3
figure 3

Average valence and arousal ratings per word, plotted with the quadratic best fit

Full size image

To test whether the correlations with frequency, AoA, and imageability reported by Warriner et al. (2013) for the English language also hold for Spanish, we made use of the word frequency data as well as the subjective ratings for familiarity, imageability, and concreteness reported by Duchon et al. (2013), and the AoA ratings reported by Alonso, Fernandez, and Diez (2015). In general, our data showed the same patterns as those reported by Warriner et al. (2013; see also Montefinese et al., 2014). As the frequency, familiarity, imageability, or concreteness of a word increased, the valence rating of that word increased, but the arousal rating decreased. One could speculate that this was due to the “familiarity” or “mere exposure” effect, which has been shown in multiple domains: Frequent exposure to a particular stimulus leads to more positive evaluations of it (Bornstein & Craver-Lemley, 2004; Murphy & Zajonc, 1993; Zajonc, 1968, 2001, among many others; see also Bornstein, 1989, for a meta-analysis). Along the same lines, frequent items could be taken as being more commonplace, and therefore less exciting or arousing. This trend was reversed when looking at AoA: Words that were learned earlier in life resulted in higher valence ratings and lower arousal ratings. Words learned later in life were more arousing and more negative (see Fig. 4 for plots, and Table 2 for correlations). An explanation could be that children are shielded from “unpleasant” words, such as taboo words and insults, which are only acquired later in life, or at least that people norming words for AoA tend to believe so when they provide their ratings.

Fig. 4
figure 4

Average frequency (n = 13,932), familiarity (n = 5,280), imageability (n = 5,118), concreteness (n = 5,296), and age of acquisition (n = 5,081) across the valence and arousal spectra. The word frequencies used to calculate the Zipf value, as well as the subjective ratings for familiarity, imageability, and concreteness, were obtained from EsPal (Duchon et al., 2013), whereas the age-of-acquisition ratings came from Alonso et al. (2015)

Full size image
Table 2 Correlations between word characteristics and their arousal and valence
Full size table

Reliability of our norms

In order to determine the reliability of our methods, we split the participants randomly into two groups and calculated their mean ratings for each word. After recording the correlations between these two groups of participants, we performed this task another 99 times to get a set of 99 correlations. The mean correlation coefficient provided us with the measure of split-half reliability, which for valence amounted to .88 (.87 and .89, 95 % confidence interval) and for arousal amounted to .75 (.73 and .76, 95 % confidence interval), suggesting that our methods were highly reliable.

Comparison with other norms

Our norms were highly correlated to other Spanish word ratings, including Hinojosa et al. (2015) (for 636 words in common: r = .97, p < .001, and r = .71, p < .001, respectively, for valence and arousal) and Redondo et al. (2007; for 1,031 words in common: r = .98, p < .001, and r = .75, p < .001, respectively, for valence and arousal). We translated the words of our list into EnglishFootnote 2 and found 9,403 one-word translation equivalents in common with Warriner et al.’s (2013) list. Previous studies (e.g., Eilola & Havelka, 2010; Montefinese et al., 2014; Redondo et al., 2007; Soares et al., 2012; Warriner et al., 2013) have shown that ratings for both valence and arousal are relatively consistent across languages, a finding confirmed by the strong correlation between our Spanish norms and Warriner et al.’s English norms (r = .79, p < .001, and r = .54, p < .001, respectively, for valence and arousal).

Conjugated verbs

In Spanish, syntactic information about a word is often expressed in inflectional morphemes added to the word stem (similarly to the suffix “-s” in the English verb “walks,” indicating the third person singular). We examined several word forms of Spanish verbs to test whether the syntactic characteristics of verbal tense, person, mode, or aspect systematically affect emotional responses to verbal word forms. As can be seen in Table 3, there were no differences between the first and third person ratings for either valence or arousal. This seems to indicate that the differences between first and second person found for self-referential processing of emotional words using other paradigms (e.g., Fossati et al., 2003; Northoff et al., 2006) are not captured by subjective ratings such as the ones presented here. Furthermore, despite the fact that most of the other differences between conjugated verb forms were significant, the size of such differences was in all cases very small (a maximum of 4.7 % of the 9-point scale for valence, and 4.0 % for arousal). Tables 4 and 5 show the words with the largest difference between the different conjugations for valence and arousal, respectively.

Table 3 Comparisons between each tense (reporting mean ± SE and paired two-sample t tests)
Full size table
Table 4 Words with the biggest difference in valence between conjugations
Full size table
Table 5 Words with the biggest difference in arousal between conjugations
Full size table

It is possible that differences between the emotional ratings of different verb conjugations would be more marked for irregular verbs, since the differences between word forms are larger. However, when we compared ratings for just the irregular verbs in our set (n = 58), we obtained results similar to those from before (see Table 6): Several differences were significant, but they were very small in magnitude (less than 6.6 % of the scale for valence, and less than 6.3 % for arousal). Taken together, and given the small variations between the different conjugations, these results indicate that it is generally safe to assume that emotional ratings for infinitive verbs can be applied to their inflected forms.

Table 6 Comparisons between each tense of irregular verbs (reporting mean ± SE and paired two-sample t tests)
Full size table

Availability of the norms

The full set of norms is available in a comma-separated (.csv) file as supplementary materials to this article. The words are organized alphabetically, and the headings for the table are as follows: ValenceMean = mean valence value for all valid responses, ValenceSD = standard deviation of the valence ratings, %ValenceRaters = number of participants who knew and rated the word for valence, ArousalMean = mean arousal value for all valid responses, ArousalSD = standard deviation of the arousal ratings, and %ArousalRaters = number of participants who knew and rated the word for arousal. %ValenceRaters and %ArousalRaters are included as measures of word prevalence, which has been shown to be a strong predictor of reaction times in lexical decision (Keuleers, Stevens, Mandera, & Brysbaert, 2015). We also indicate each word’s dominant part of speech according to EsPal (Duchon et al., 2013). Ninety-five of the words do not have an entry in EsPal, so their parts of speech were determined manually. Entries with less than ten raters in either or both variables are marked with an asterisk.

General discussion

In summary, the present article expands on the work of Redondo et al. (2007) by providing valence and arousal ratings for over 14,000 Spanish words, greatly increasing the availability of items normed on those variables. We were able to collect such a large set of norms by using a modular approach, and we have shown that our norms are reliable and compatible with previous, similar studies. We also showed that there is little variation in ratings for different conjugations of the same verb. The availability of such a large set of norms will not only be useful for opening up research opportunities in Spanish by helping in the creation of stimulus sets, correlational studies, and text analyses, but its compatibility with the Warriner et al. (2013) norms and others will also be of value for research on emotional aspects of bilingualism, translation studies, and cross-language research.