It is generally accepted that one's mother tongue is the language for expressing feelings and emotions, especially if the topic is personally significant to the speaker (Dewaele, Reference Dewaele2011; Marian & Kaushanskaya, Reference Marian and Kaushanskaya2004; Pavlenko, Reference Pavlenko2005). For instance, a large web questionnaire study by Dewaele (Reference Dewaele2004b) revealed that first language words lead to higher emotional arousal, and, if asked to read out a list of taboo words, people report feeling more anxious when the words are in their native language rather than a second language (Gonzalez-Regiosa, Reference Gonzalez-Regiosa, Spielberger and Diaz-Guerrero1976). Bilinguals also prefer to use their native language when asked to talk about positive events in their lives (Javier, Barroso, & Munoz, Reference Javier, Barroso and Munoz1993). In contrast, later-learned languages are often believed to offer emotional distance, which is illustrated by the finding that bilinguals often feel it is easier to discuss embarrassing topics in their second language rather than in their native language, in which they would be more fluent (Bond & Lai, Reference Bond and Lai1986). This phenomenon of code-switching to the nonnative language can distance the speaker from an uncomfortable situation (Javier & Marcos, Reference Javier and Marcos1989), which has frequently been noted in the psychoanalytic literature (Gonzalez-Regiosa, Reference Gonzalez-Regiosa, Spielberger and Diaz-Guerrero1976; Schrauf, Reference Schrauf2000). Anxiety-arousing words are also more easily expressed in the nonnative language (Dewaele, Reference Dewaele2004a, Reference Dewaele2004b).

Self-report measures can be highly distorted and biased, however. There have therefore been attempts to conduct controlled experiments to reveal the differential emotionality of languages. One interesting approach has been to study memory for different types of words in bilinguals. Research using monolingual samples has consistently found the so-called emotion-word memory advantage, which means that people's memory tends to be better for emotion (e.g., “happy,” “sad”) and emotion-laden (e.g., “cancer,” “spider”) words compared to emotionally neutral words (e.g., “chair,” “letter”; Altarriba & Bauer, Reference Altarriba and Bauer2004; Jay, Caldwell-Harris, & King, Reference Jay, Caldwell-Harris and King2008; Kensinger & Corkin, Reference Kensinger and Corkin2003; Rubin & Friendly, Reference Rubin and Friendly1986; Talmi & Moscovitch, Reference Talmi and Moscovitch2004). Negative and taboo words, in particular, benefit from superior recall, which is not surprising because these words are believed to involve more complex representations and deeper processing than other types of words because of their attention-grabbing properties (MacKay & Ahmetzanov, Reference MacKay and Ahmetzanov2005; Peeters & Czapinski, Reference Peeters and Czapinski1990). Taboo words, in particular, tend to generate very strong responses in people (Jay, Reference Jay2000). When this emotion-word memory advantage is combined with self-reports that the native language is perceived as more emotional, it can be hypothesized that the emotion-word memory advantage should be restricted to words presented in one's mother tongue and not seen in later-learned languages.

Anooshian and Hertel's (Reference Anooshian and Hertel1994) pioneering study showed that the emotion-word memory advantage applied only to the first language. They employed an unexpected free recall procedure, and their results were the same for both balanced English–Spanish and balanced Spanish–English bilinguals living in a bilingual community, and who had started to acquire the second language at the age of 8 years or later. This finding indicated that the emotionality effects did not depend on the languages themselves, but that they were visible only in the mother tongue. Anooshian and Hertel also found that when participants rated the words for emotional intensity, subsequent recall was higher than when the words were rated for ease of pronunciation, but this effect was only visible in the native language. These results seem to fit neatly with the qualitative evidence showing that the native language is perceived as more emotional. The authors proposed that first language words are more likely to be associated with emotional episodes and other emotional exchanges in infancy and early childhood than later-learned languages, and thus, only first language words tend to have emotional connotations.

However, later-learned languages might not be as unemotional as they are usually thought to be. Using unbalanced Turkish–English bilinguals living in America who had acquired their second language after the age of 12, Aycicegi and Harris (Reference Aycicegi and Harris2004) found that emotionality effects could also be seen in the second language and that the recall of taboo words and reprimands was actually better when they were presented in the nonnative language. Furthermore, positive words were recalled significantly better than neutral words but only in the second language. In the recognition test, more negative words were observed in the second language, which further suggested that the second language benefitted more from the emotionality of the words.

The finding that participants had better memory and recognition for negative words presented in the second language compared to the native language (Aycicegi & Harris, Reference Harris2004) could be explained by a so-called mood change. This means that processing anxiety-provoking stimuli, such as negative words, leads people to experience negative emotions. Negative words presented in the native language are thought to have more emotional associations because early language learning coincides with the socialization of emotions (Cervantes, Reference Cervantes2002) and concept development as well as emotional regulation (Bloom & Beckwith, Reference Bloom and Beckwith1989). In order not to feel anxious, people will then try to avoid the processing of the aversive stimuli (Foa & Kozak, Reference Foa and Kozak1986), in which case the stimulus, for instance, the negative word, will be less deeply processed and hence not so accurately recalled. By contrast, negative words in languages learned after early childhood do not possess similar kinds of emotional connotations, and therefore there is no need to minimize their processing. This could then lead to better memory for negative words in a nonnative language. The reason why Anooshian and Hertel's (Reference Anooshian and Hertel1994) study failed to find any emotionality effects in the second language might be that their word list was biased toward positive words, because 13 out of 18 emotional words were positive. Aycicegi and Harris (Reference Aycicegi and Harris2004), in contrast, had an equal number of words in each word type category. However, their stimuli were predominantly negative words because they employed taboo words and childhood reprimands as two additional categories in addition to negative words. Furthermore, the presentation of the words in their study was mixed, and this could have made the nonnative language words seem unusual and novel, which may also have facilitated memory for those words. That participants were living in the second language environment may also have enhanced memory for words in that language.

To explore the topic further, Aycicegi-Dinn and Caldwell-Harris (Reference Aycicegi-Dinn and Caldwell-Harris2009) recently ran a study with late Turkish–English unbalanced bilinguals to see whether there were any effects of different processing tasks on memory. Using the same stimuli as in Aycicegi and Harris (Reference Aycicegi and Harris2004), they compared participants’ recall performance after they had completed either a shallow or deep processing task. The results showed that, across the different processing tasks, emotionality effects were equally strong in the first and the second language. An exception to this was the reprimands, which were better recalled in the second language than in the first. This, the authors argued, was due to a novelty effect, because reprimands in the second language could have been interpreted as being more unusual. However, although the emotionality effects did not differ overall in the first and second language, the specific pattern of results did vary depending on the exact task. For instance, when the words were rated for their emotional intensity, which was a deep processing task, emotion-memory effects were visible only in the first language, providing support for the traditional view that only the native language can have emotional connotations. In contrast, in a shallow processing task involving letter counting, no language differences were observed, but taboo words had superior recall in both the first and the second language. These results suggest that emotionality effects may depend on the task that is used and how the word stimuli are processed. Furthermore, there was no emotion-memory effect for negative words over neutral words in any of the tasks employed, which was quite surprising. The authors argued that this was because the negative words used (e.g., “poison,” “crime”) were more likely to have a later age of acquisition than the positive words used (e.g., “mother,” “father”).

Another recent unexpected free recall study by Ferre, Garcia, Fraga, Sanchez-Casas, and Molero (Reference Ferre, Garcia, Fraga, Sanchez-Casas and Molero2010) tested both early and late proficient bilinguals in order to see whether age and context of second language acquisition had any effect on memory for emotional words. Perhaps it is surprising that both groups of bilinguals showed similar emotionality effects in both of their languages: a superior recall for positive and negative words over neutral words was observed for both the first and the second language regardless of the age or context of acquisition. The authors concluded that proficiency was the main factor influencing the emotion-memory effects, because the emotionality effects in the first and the second language were equally strong regardless of whether the second language was learned early in an immersion context or in a school setting. It must be noted, however, that the authors used filler words as well as a filler task prior to the unexpected memory test in order to reduce any primacy and recency effects, which was not done in previous free recall studies. Thus, these methodological differences may account for the rather unexpected results that age and context of acquisition had no effect on the emotion-memory advantage.

Overall, very few studies have looked at emotion-word memory advantage in bilinguals, and results are still somewhat contradictory as to the extent of emotionality effects in nonnative languages. The two most recent studies by Aycicegi-Dinn and Caldwell-Harris (Reference Aycicegi-Dinn and Caldwell-Harris2009) and Ferre et al. (Reference Ferre, Garcia, Fraga, Sanchez-Casas and Molero2010) do certainly suggest that emotionality effects can be observed in a later-learned language, and that they can be as strong as in the first language. Other quantitative measures have provided more evidence that emotionality effects are not restricted to the native language. For instance, studies using the emotional Stroop paradigm, which is perhaps the paradigm used most frequently to investigate emotional processing at the single word level, have shown that the level of interference from negative and taboo words is similar in native and nonnative languages in both late and early bilinguals (Eilola, Havelka, & Sharma, Reference Eilola, Havelka and Sharma2007; Sutton, Altarriba, Gianico, & Basnight-Brown, Reference Sutton, Altarriba, Gianico and Basnight-Brown2007). Similar levels of interference have been detected also in native and nonnative speakers of English (Eilola & Havelka, Reference Eilola and Havelka2011). These behavioral tasks imply that (automatic) emotional activation occurs in both the first and the second language, and that threatening words are processed similarly across languages. However, others have found that Stroop interference is stronger for the dominant language (Zied et al., Reference Zied, Phillipe, Pinon, Havet-Thomassin, Aubin and Roy2004), although it is visible in the second language as well. Note that Zied at al. (Reference Zied, Phillipe, Pinon, Havet-Thomassin, Aubin and Roy2004) employed the classic Stroop task rather than an emotional version of the task, which is thought to involve different cognitive mechanisms (Algom, Chajut, & Lev, Reference Algom, Chajut and Lev2004). Therefore, the findings from these two tasks may not be directly comparable. When looking at the emotional Stroop task, it is interesting that response times to positive words are usually not significantly different from neutral words (Eilola et al., Reference Eilola, Havelka and Sharma2007; McKenna & Sharma, Reference McKenna and Sharma1995), which further suggests that negatively valenced words are processed differently from emotionally positive and neutral words.

However, physiological evidence regarding the emotionality of languages tends to support the proposal that emotionality effects are confined to the mother tongue. A skin conductance study by Harris, Aycicegi and Gleason (Reference Harris, Aycicegi and Gleason2003) using unbalanced Turkish–English bilinguals living in the second language environment showed that the response to aurally presented taboo words, and aurally or visually presented childhood reprimands, was lower in the second language, which would imply that taboo words in the native language elicit stronger autonomic responses, at least when they are listened to. It was suggested that aurally presented words might result in more emotional activation, because speech is usually very interpersonal in its context and hence provides more emotional associations. However, the authors failed to find any significant differences between native and nonnative languages when the participants were either listening to or looking at printed negative and positive words, or when the taboo words were presented visually. Follow-up experiments (Harris, Reference Harris2004; Harris, Gleason & Aycicegi, Reference Harris, Gleason, Aycicegi and Pavlenko2006) showed that the differences in skin conductance activity were reduced in early bilinguals when they were presented with childhood reprimands or emotion words, which led them to propose that the emotionality of languages depends on the age at which they are acquired and how proficient the speaker is. Recently, Caldwell-Harris and Aycicegi-Dinn (Reference Caldwell-Harris and Aycicegi-Dinn2009) extended their earlier finding (Harris et al., Reference Harris, Aycicegi and Gleason2003) that skin conductance response to reprimands was greatest in the native language to late unbalanced bilinguals living in their native-language environment. It is interesting that Eilola and Havelka (Reference Eilola and Havelka2011) did not detect any differences in performance by native and nonnative English speakers during emotional and taboo Stroop tasks, but they did find significantly higher skin conductance responses to taboo and negative words in native speakers. These studies seem to suggest that although the meanings of the second language words may well be activated and accessed during these different tasks, they do not result in increases in autonomic activation as they seem to do in the native language.

Combining the above studies, it would seem that the issue of the emotionality of first and later languages is not as clearcut as has been thought traditionally. The age and context of acquisition as well as proficiency are likely to be among the most critical factors determining the degree of emotionality of the languages in speakers of more than one language. The acquisition age and context can also be seen as reflecting the amount and type of exposure to a language: the earlier a language is acquired, the more likely the acquisition is to happen in a naturalistic setting as opposed to at school. Proficiency would also be expected to be greater if a language was acquired early in life in multiple, social contexts. Dewaele's (Reference Dewaele2004b; see also Dewaele, Reference Dewaele2008) web questionnaire study with more than 1,000 participants showed that the most important factors affecting the perceived emotionality of languages were the age at which later-learned languages were acquired, and whether they were learned in purely educational settings, or by living in a foreign language environment, or both. Other significant factors were the perceived fluency and the frequency of use of the language. It is important to note that so far experimental studies on memory and the multilingual lexicon in general have looked at only two languages at a time, and often used participants living either in the second language environment or in a bilingual community. This can easily result in findings such as those of Aycicegi and Harris (Reference Aycicegi and Harris2004), in which the nonnative language has apparently the same, or even stronger, emotional connotations than the mother tongue. However, if emotionality effects depend on the amount and type of exposure to a particular language, one would expect to see a graded difference in the degree of the emotionality, with the effects being largest in the native and most fluent language and progressively smaller in later-learned languages, depending on the type and amount of exposure to these languages. This issue may be best studied using unbalanced speakers of more than two languages living in their native-language environment, as this would control for exposure to the nonnative languages.

The current study aimed to examine the possibility that there is a graded emotionality effect across languages. It was decided that the study would be conducted in Finland because of the country's interesting linguistic situation. To summarize, Finland was ruled by Sweden from approximately 1155 to 1809, when the country was invaded by Russia. Despite Finnish having always been used by the majority, Swedish was the language of prestige and the only official language until 1863 (Engman, Reference Engman and Tägil1995). Now, Swedish remains as one of the official languages in Finland. However, as of 2008, 90.9% of the population have Finnish as their native language, compared to only 5.4% who speak Swedish as their first language (Tilastokeskus, 2009). Nevertheless, Swedish is still a mandatory language in Finland, but it is generally regarded as a burden or a relic from earlier times. English is also mandatory, but it is seen as useful and universal, and it is usually learned earlier than Swedish, with more than 90% of school-aged children learning English as their first foreign language (Sajavaara, Reference Sajavaara2006). The aim of the current study was to use this situation in Finland to test the idea that emotionality effects are graded across the languages known by the speaker, rather than being an all-or-none phenomenon visible only in the native language.

There are at least three advantages to investigating the emotionality effects of languages using a sample from Finland. First, because a typical Finn is trilingual with Finnish as the first, English as the second, and Swedish as the third language, it is possible to examine the potentially graded nature of emotionality effects across these three languages. Second, as Finnish people generally do not start to learn foreign languages until the age of 9, most of them can be considered to be late learners of English and Swedish. This makes it possible to look at the effect of age on emotionality effects to see whether languages learned after early childhood will have emotional connotations similar to the native language. Third, there is a clear pattern of dominance in how fluent a typical Finn is in his or her languages, with Finnish being the native language, and English and Swedish learned only after early childhood, in that particular order. This makes it possible to predict that any emotionality effects will be strongest in Finnish, weaker in English, and weakest or possibly nonexistent in Swedish.

To summarize, the current study was designed to investigate the processing of emotion-laden words in multilinguals to see whether emotionality effects are restricted to the first language or also seen in later-learned languages. Because it has been well established that deeper processing of the stimuli leads to better recall (e.g., Craik & Lockhart, Reference Craik and Lockhart1972), the current study also employed either a deep or shallow processing task before the free recall in order to see whether the recall of words across languages and different word categories would be affected by the depth of processing of the stimuli. As shown in the study by Aycicegi-Dinn and Caldwell-Harris (Reference Aycicegi-Dinn and Caldwell-Harris2009), the nature of the processing task can have an effect on emotion-memory effects. Native speakers of Finnish with knowledge of English and Swedish were chosen as the target population. In a mixed design, young adults (N = 41) completed an orienting word-rating task (deep or shallow processing) on 36 words: 12 negative/taboo, 12 positive, and 12 neutral presented in the three languages. Participants were then given an unexpected free recall test. It was hypothesized that there would be a clear emotion-word memory advantage for words presented in the mother tongue Finnish and that it would also be visible in English, but to a lesser degree. For Swedish words, the emotion-word memory advantage was expected to be very small, or nonexistent. Because of the differential processing of negative information, the memory advantage was anticipated to apply especially to negative/taboo words. Additional self-report measures regarding the subjective emotionality of participants’ later-learned languages and the extent to which they used these languages in everyday life were also collected to see whether participants regarded English as more emotional and more widely used than Swedish.

METHOD

Participants

The total number of the participants was 41 (66% females), and their mean age was 20.17 (SD = 3.17, range = 17–25 years). Participants were required to be native Finnish speakers with at least a fair knowledge of English and Swedish, and they were recruited from eastern Finland. They were considered to be late, unbalanced multilinguals because they had learned their later languages after early childhood. Opportunistic sampling was employed, which meant that those who consented to participate perceived themselves to be at least fairly competent in English and Swedish. The vast majority (88%) had started to learn English before Swedish, and for 5%, Swedish was a fourth- or later-learned language. Thus, most of the participants had started to learn English at the age of 9 and Swedish at the age of 13. On average, the participants had been learning English for 11 years and Swedish for 7 years. Most of the participants (61%) had never been abroad in a situation where they would have had to use Swedish as the main means of communication. As for English, 27% stated that they had never used English as the only language of communication. All participants reported being more fluent in English than in Swedish.

Design and materials

A mixed design with three independent variables was used, and this study generally adopted an approach similar to that of Anooshian and Hertel (Reference Anooshian and Hertel1994). However, the present study was more quasiexperimental in nature as no attempt was made to control for participants’ language skills except for the requirement that they were native Finnish speakers and late learners of English and Swedish (i.e., learning of nonnative languages started after the age of 7). The independent variables were the language of presentation (Finnish, English, Swedish), the word category (negative/taboo, positive, neutral), and the rating task (deep processing, shallow processing). The dependent measure for the main analysis was the percentage of different word types correctly recalled across these three factors.

Participants were shown a list of words (one word at a time) and asked to complete an orienting word-rating task. This was then followed by an unexpected free recall test in which participants were asked to write down as many words as they could remember. The words used in the present study (Appendix A) were randomly selected from those used and validated by Eilola et al. (Reference Eilola, Havelka and Sharma2007), who investigated emotionality effects in English and Finnish using an emotional Stroop task. The selected words were translated into Swedish by a native Swedish speaker and then checked by another Swede. Normative ratings for Finnish and English words were obtained from Eilola and Havelka's (Reference Eilola and Havelka2010) recent study, which is currently the only collection of normative emotional ratings for Finnish words. No similar database for Swedish words exists to date. Therefore, it was not possible to examine if any differences existed between the word categories with regard to their familiarity, emotional valence, arousal, and offensiveness.

There were 108 words used in the study: 36 from each language (Finnish, English, Swedish) and 12 words in each word category (negative/taboo, positive, neutral). The negative/taboo category consisted of five taboo words and eight negative words. The word presentation was blocked, which meant that participants were always presented with words in one language followed by the other two languages, and in order to eliminate possible primacy and recency effects, six different word lists were constructed to account for each possible order combination (i.e., first Finnish, second English, third Swedish; first Finnish, second Swedish, third English; first English, second Finnish, third Swedish). Each participant was randomly assigned to one of the six word lists. Each word list contained 12 different words presented in each language, and these words consisted of four words in each of the negative/taboo, positive, and neutral categories. Thus, each participant was presented with only four words from each word category in each language in order not to repeat the translation equivalents within a single experiment. Each participant was presented with either two or one taboo word in each language. The order of words within each language block was randomized. Although the words were otherwise randomly assigned to one of the six word lists, in order to control for semantic relatedness, which has been identified as a potentially confounding factor in memory research (Talmi & Moscovitch, Reference Talmi and Moscovitch2004), the word lists were constructed so that semantically related words such as “funeral” and “death” did not appear in the same language block.

Table 1 shows the length of the words, as well as the mean normative ratings of familiarity, emotional valence, arousal, and offensiveness for the Finnish and English words as reported by Eilola and Havelka (Reference Eilola and Havelka2010). The words did not differ significantly in their word length across the three languages, F (2, 66) = 1.43, p = .247, partial η² = 0.04, or the word categories, F (1, 33) = 1.30, p = .287, partial η² = 0.07. There were also no differences in familiarity ratings across Finnish and English words, F (1, 33) = 0.51, p = .480, partial η² = 0.06, nor the word categories, F (2, 33) = 0.80, p = .459, partial η² = 0.05. Furthermore, no differences in familiarity existed within the languages.

Table 1.

Mean word length of the words and the mean normative ratings of familiarity, emotional valence, arousal, and offensiveness for the Finnish and English words as reported by Eilola and Havelka (Reference Eilola and Havelka2010)

As for emotional valence, significant language differences were detected across Finnish and English F (1, 33) = 5.98, p = .020, partial η² = 0.15. Post hoc tests revealed that Finnish words were rated as slightly more emotional than English words. Furthermore, significant differences were found between the word categories, F (1, 33) = 245.35, p < .001, partial η² = 0.94. This reflected that negative/taboo, positive, and neutral words differed significantly from each other in both Finnish and English (p < .001 for all comparisons).

For emotional arousal, significant language differences were detected, F (1, 33) = 8.74, p = .006, partial η² = 0.21: English words were rated slightly less arousing than Finnish words. Moreover, there were significant differences across the word categories, F (2, 33) = 38.05, p < .001, partial η² = 0.70: neutral words were rated as less arousing than negative/taboo and positive words (p < .001 for both comparisons). In both languages, neutral words were rated as significantly less arousing than negative/taboo and positive words (p < .001 for all comparisons).

As for offensiveness, no significant language differences were found in the ratings F (1, 33) = 0.02, p = .888, partial η² = 0.001. However, significant differences were found across the word categories F (2, 33) = 42.19, p < .001 partial η² = 0.72: negative/taboo words were rated as significantly more offensive than positive and neutral words (p < .001 for both comparisons). Furthermore, in both Finnish and English, negative/taboo words were rated as significantly more offensive than any other word types (p < .001 for all comparisons).

Other materials consisted of the word-rating sheet, and the emotionality and the frequency of everyday use of language questionnaires. In the orienting word-rating task, participants either rated the words for how emotionally intense they considered the words to be, or how easily they could pronounce the word aloud on a 7-point Likert scale. These rating tasks were included in order to prevent the participants from explicitly trying to memorize the words, and to see if recall would be differentially affected by deep (emotionality rating) and shallow (pronunciation rating) processing tasks. Within each word list, each language block was assigned one of the two rating tasks, and because there were three language blocks within each word list, participants completed either two emotionality ratings and one ease of pronunciation rating, or one emotionality rating and two ease of pronunciation ratings. This ensured that each word was rated for its emotionality and ease of pronunciation in each language across participants. The emotionality questionnaire had 12 questions adapted from the web questionnaire by Dewaele (Reference Dewaele2004b), and included questions such as “How difficult is it for you to talk about emotional topics in English/Swedish?” rated on a 5-point Likert scale. The frequency of everyday use questionnaire was similar to that of Eilola et al. (Reference Eilola, Havelka and Sharma2007). The native language Finnish was not included in the questionnaires, as it was participants’ dominant language, and participants were living in a native-language environment. Participants would therefore be expected to rate Finnish as very emotional and very frequently used. The summary results of the frequency of everyday use questionnaire are provided in Appendix B.

Procedure

Participants completed the experiment individually. Oral instructions were given in Finnish, but the questionnaires were in English. Participants were given an opportunity to ask advice from the experimenter if they did not understand the questions asked. It was not expected that the findings would be affected by having the questionnaires in English, as the participants were living in their native-language environment, and were highly unbalanced multilinguals with the native-language Finnish as the dominant language. For the orienting rating task, participants were told to either decide how emotional the word was, or how easy or difficult the word's pronunciation was on a 7-point Likert scale. The instructions for the ratings were first provided orally before the beginning of the experiment, and on the rating sheet. For the emotionality rating task, the question was, “How would you rate the intensity of the emotion associated with the meaning of the word?” For the ease of pronunciation task, the question was, “How easily or with how much difficulty could you pronounce the word aloud?” The 7-point scales went from unemotional to emotional and easy to say to hard to say, respectively.

The words were then presented one by one on a 17-in. laptop screen (HP 6730) using Microsoft Powerpoint 2007. Each word was written in black on a white background (66-point Calibri) and was visible on the screen for 11 s. Immediately after all the words had been presented, participants were given an unexpected free recall test. They were simply asked to recall and write down the words they had just rated. They were reminded that the correct spelling was not important, but they should try to write down all of the words that they could remember. Finally, participants completed the perceived emotionality and frequency of everyday use of language questionnaires, the order of which was counterbalanced across participants, after which they were debriefed. The experiment took 15–20 min to complete. Both the perceived emotionality and the use of language questionnaires were analyzed for reliability using the Cronbach alpha and the split-half test. These demonstrated satisfactory reliability for the scales (see Table 2).

Table 2.

The reliability analysis of the perceived emotionality and use of the language questionnaires

RESULTS

Perceived emotionality and the everyday use of English and Swedish

As the scores for the perceived emotionality and frequency of everyday use of Swedish were not normally distributed, the nonparametric Wilcoxon signed-rank test was used to compare the scores for English and Swedish. This analysis revealed that English (Mdn = 39.00) was perceived to be more emotional than Swedish (Mdn = 20.00; z = −5.38, p < .001). The strength of this difference was r = −.59, which represents a large-sized effect. Moreover, participants reported using English (Mdn = 33.00) significantly more often in everyday life than Swedish (Mdn = 15.00; z = −5.58, p < .001, r = −.61), suggesting a greater exposure to English than Swedish.

Analysis using Pearson's correlation coefficient revealed that perceived emotionality of English and the extent to which English was used in everyday life were significantly correlated (r = .69, p < .001, two tailed). The Pearson correlation between the everyday use of Swedish and its perceived emotionality was also significant (r = .66, p < .001, two tailed). These correlations suggest that the frequency of everyday use of a language is significantly related to the perceived emotionality of that language, which is consistent with previous research (e.g., Dewaele, Reference Dewaele2004b). The relationship between these two variables is illustrated in Figure 1 and Figure 2.

Figure 1.

The correlation between frequency of use and perceived emotionality of English.

Figure 2.

The correlation between frequency of use and perceived emotionality of Swedish.

Word ratings

In the orienting word-rating tasks, participants rated each word presented to them on a 7-point Likert scale for its emotionality (1 = unemotional; 7 = emotional) or ease of pronunciation (1 = easy to say; 7 = hard to say). Table 3 shows the mean ratings for these two rating tasks across languages and word categories. To see if there were any significant differences in emotionality ratings between languages and word categories, a 3 × 3 (Language × Word Category) analysis of variance (ANOVA) was conducted. This revealed a significant main effect of language, F (2, 142) = 13.20, p < .001, partial η² = 0.16. Post hoc comparisons (all comparisons made using Bonferroni adjustments) indicated that, on average, Finnish words (M = 3.94, SE = 0.13) were rated as significantly more emotional than Swedish words (M = 3.23, SE = 0.13; p = .001). The emotionality ratings for English words (M = 4.03, SE = 0.12) did not differ from the ratings for Finnish words (p = 1.00), but they were significantly higher than the ratings for Swedish words (p < .001).

Table 3.

Mean (standard deviation) emotionality and ease of pronunciation ratings across languages and word categories

Furthermore, a significant main effect of word category was found, F (2, 142) = 158.75, p < .001, partial η² = 0.69. In post hoc tests, it was revealed that higher emotionality ratings were given to negative/taboo (M = 4.51, SE = 0.13) and positive (M = 4.40, SE = 0.11) words as opposed to neutral words (M = 2.29, SE = 0.11; p < .001 for both comparisons).

Finally, the ANOVA revealed a significant Language × Word Category interaction, F (4, 284) = 16.69, p < .001, partial η² = 0.19. Post hoc analyses revealed that, in Finnish, negative/taboo words (M = 5.24, SE = 0.18) were rated as being significantly more emotional than positive (M = 4.58, SE = 0.18) and neutral (M = 2.01, SE = 0.17) words (p < .001 for both comparisons). Furthermore, positive words were rated as more emotional than neutral words (p = .007). Similarly for English, neutral words (M = 2.29, SE = 0.18) were rated as significantly less emotional than negative/taboo (M = 4.92, SE = 0.22) and positive words (M = 4.88, SE = 0.19; p < .001 for both comparisons), but no significant difference was found between negative/taboo and positive words. As for Swedish, a similar pattern emerged: neutral words (M = 2.56, SE = 0.19) were rated as less emotional than negative/taboo (M = 3.75, SE = 0.19) and positive words (M = 3.39, SE = 0.19; p < .001 for both comparisons). Furthermore, negative/taboo words in Swedish were rated as less emotional than negative/taboo words in Finnish (p < .001) and English (p < .001). Similarly, positive words in Swedish were rated as less emotional than positive words in Finnish (p = .013) and English (p < .001). No differences were found in the emotionality ratings of neutral words across the three different languages.

Another 3 × 3 (Language × Word Category) was conducted to look at the differences in ease of pronunciation ratings between languages and word categories. This ANOVA revealed that there was no main effect of language, F (2, 118) = 0.16, p = .852, partial η² = 0.03. Thus, there were no differences in how difficult to pronounce the participants perceived the words to be in Finnish (M = 2.98, SE = 0.16), English (M = 2.94, SE = 0.13), and Swedish (M = 2.87, SE = 0.13).

However, there was a significant main effect of word category, F (2, 118) = 6.32, p = .002, partial η² = 0.10. Positive words (M = 2.66, SE = 0.12) were rated as being significantly easier to pronounce than negative/taboo (M = 3.15, SE = 0.12) and neutral (M = 2.98, SE = 0.11; p = .001 and .047, respectively). Negative/taboo words were rated as being most difficult to say out loud, but they differed significantly only from the ratings of positive words.

There was also a significant interaction between language and word category, F (4, 236) = 4.72, p = .001, partial η² = 0.07. Pairwise comparisons revealed that in Finnish, negative/taboo words were rated as being most difficult to pronounce (M = 3.32, SE = 0.19), although this difference was only marginally significantly different from positive words (M = 2.73, SE = 0.23; p = .051). In English, positive words (M = 2.50, SE = 0.20) were rated as significantly easier to say out loud than neutral words (M = 3.48, SE = 0.23; p < .001). In Swedish, negative words (M = 3.32, SE = 0.19) were rated as significantly more difficult to pronounce than positive (M = 2.75, SE = 0.19) and neutral (M = 2.55, SE = 0.16; p = .043 and p = .002, respectively). In addition, neutral words were rated as being significantly more difficult to pronounce in Swedish (M = 3.32, SE = 0.19) than in English (M = 2.83, SE = 0.18; p = .006).

Unexpected free recall

In order to test whether there were any emotionality effects within and between the first- and later-learned languages, and across the two rating tasks, the unexpected free recall data were analyzed. The mean percentage of words recalled across languages, word categories, and rating tasks was 25.5% (SD = 1.5%). Preliminary analysis indicated that a similar pattern of recall emerged for all the six word lists. Very few recall errors were made (N = 3), suggesting that recall of words is language-specific, which is consistent with previous memory studies (e.g., Aycicegi-Dinn & Caldwell-Harris, Reference Aycicegi-Dinn and Caldwell-Harris2009). The three incorrectly recalled words were all from the neutral word category, and they were Finnish translation equivalents of English target words. Table 4 displays the mean correct recall percentages and their standard deviations for each language/word category/rating task combination. It is clear from Table 4 that the emotion-word memory advantage was different across the three languages. Although the distribution of some of the data was somewhat skewed, nonparametric inferential tests resulted in similar conclusions to the equivalent parametric tests; therefore, only the parametric tests are reported below.

Table 4.

Mean (standard deviation) recall percentages by language, word category, and rating task

To test for effects of language, word category, and rating task on the recall of words, a mixed-design 3 × 3 × 2 (Language × Word Category × Rating Task) ANOVA was conducted. All post hoc tests were adjusted using Bonferroni corrections.

No significant differences in recall between the two rating tasks were found in the ANOVA. Thus, there was no main effect of rating task, F (1, 31) = 2.57, p = .12, partial η² = 0.08, nor any significant interactions between Rating Task × Language, F (2, 62) = 2.70, p = .075, partial η² = 0.08; Rating Task × Word Category, F (2, 62) = 0.17, p = .84, partial η² = 0.006; or Rating Task × Language × Word Category, F (4, 124) = 1.73, p = .15, partial η² = 0.05. However, the Rating Task × Language Interaction was not far from being significant, and from Table 2, it can be seen that the highest recall rate was for Finnish negative/taboo words when they were rated for their emotionality.

A significant main effect for language was found, F (2, 62) = 9.22, p < .001, partial η² = 0.23. Post hoc tests revealed that the participants recalled significantly more Finnish (M = 31.7%, SE = 2.9%) than Swedish words (M = 17.0%, SE = 1.7%; p = .001), and significantly more English words (M = 27.9%, SE = 2.8%) than Swedish words (p = .01).

The main effect of word category was also significant, F (2, 62) = 26.33, p < .001, partial η² = 0.46. The percentage of negative/taboo words recalled (M = 38.1%, SE = 2.3%) was significantly higher (p < .001) than the percentage of neutral words recalled (M = 17.7%, SE = 2.5%). Furthermore, fewer positive words (M = 20.8%, SE = 2.0%) were recalled than negative words (p < .001).

In addition to these significant main effects, a significant interaction between language and word category was found, F (4, 124) = 6.88, p < .001, partial η² = 0.18, which suggests that emotionality effects were different depending on the language in which the words were presented. This interaction is illustrated in Figure 3. The error bars show the standard errors of the mean, and overlapping bars are likely to differ from each other only by chance.

Figure 3.

The interaction between language of presentation and word category.

Paired comparisons revealed that, for words presented in Finnish, significantly more negative/taboo words were recalled than neutral or positive words (p < .001). The difference between positive and neutral words was also significant (p = .019), with more positive words being recalled.

For words presented in English, a similar pattern emerged: negative/taboo words had the highest recall rate, and this was significantly higher than the number of positive (p = .001) and neutral (p = .010) words recalled.

However, for the third language, Swedish, all the comparisons between the different word categories were nonsignificant, which suggests that there were no emotionality effects in Swedish.

Furthermore, significantly more negative/taboo words were recalled in Finnish than in English (p = .016) or Swedish (p < .001). Moreover, the recall of English negative/taboo words was significantly better than that of Swedish negative/taboo words (p < .001), which provides further evidence for the graded nature of emotionality effects, with the effect being greatest in Finnish, smaller in English, and smallest in Swedish.

To summarize, participants recalled significantly more words overall in Finnish and English than in Swedish. Across languages, significantly more emotionally negative/taboo words were recalled compared to positive and neutral words. This provides support for the view that negative/taboo words are processed differently than positive or emotionally neutral words. These two language and word category factors interacted so that, in Finnish, participants recalled significantly more negative/taboo than positive or neutral words. Moreover, the recall of positive words was significantly better than the recall of neutral words. Participants had superior recall in English of negative/taboo words compared to positive and neutral words. No significant emotionality effects could be detected for words presented in Swedish. Furthermore, recall of negative/taboo words was significantly better in Finnish than in English and Swedish, and significantly more negative/taboo words were recalled in English than in Swedish. These results provide support for the idea that emotionality effects are graded rather than an all-or-none phenomenon. Such effects do not seem to be completely restricted to the first language but can also be seen in languages learned after early childhood. However, they seem to become less strong the later the language is acquired and the less it is used.

Additional analyses: Taboo words as a separate category

Because previous research has indicated that taboo words may be processed differently compared to other negatively valenced words, an additional 3 × 4 × 2 (Language × Word Category × Rating Task) ANOVA was conducted in which negative and taboo words were treated as two separate categories. As in the main analyses, a significant main effect of language was found, F (2, 62) = 10.58, p < .001, partial η² = 0.26, with significantly more Finnish and English words recalled than Swedish words (p < .001 and p = .003, respectively). The main effect of word category was also significant, F (3, 93) = 10.78, p < .001, partial η² = 0.26. Post hoc tests revealed that taboo words elicited the highest recall rate, which was significantly different from both neutral and positive words (p < .001 and .008, respectively). The percentage of negative words recalled was significantly higher (p = .001) than the percentage of neutral words recalled. Furthermore, fewer positive words were recalled than negative words, although this difference was only marginally significant (p = .077).

In addition to the main effects, a significant interaction between language and word category was found, F (3.50, 108.37) = 4.25, p = .005, partial η² = 0.12, which suggests that emotionality effects were different depending on the language in which the words were presented. This interaction is illustrated in Figure 4. The error bars show the standard errors of the mean, and overlapping bars are likely to differ from each other only by chance.

Figure 4.

The interaction between language of presentation and word category (taboo words as a separate category).

Paired comparisons revealed that, for words presented in Finnish, significantly more negative words were recalled than neutral or positive words (p < .001). Significantly fewer neutral words than taboo words were recalled (p = .002), and the difference between positive and neutral words was also significant (p = .019), with more positive words being recalled.

For words presented in English, a slightly different pattern emerged: taboo words had the highest recall rate, and this was significantly higher than the number of negative (p = .035), positive (p = .001) or neutral words (p = .002) recalled. For Swedish, no significant differences were found between the different word categories.

To summarize the results of the additional analyses, participants recalled significantly more words overall in Finnish and English than in Swedish. Across languages, significantly more taboo words were recalled compared to positive and neutral words, and the recall of negative words was also significantly higher than the recall of neutral words. This provides support for the view that taboo and negative words are processed differently than positive or emotionally neutral words. Language and word category interacted so that participants recalled significantly more negative than positive or neutral words in Finnish. Moreover, the recall of taboo and positive words was significantly better than the recall of neutral words. In English, taboo words had superior recall compared to negative, positive, and neutral words. No emotionality effects could be detected for words presented in Swedish. Furthermore, recall of negative words was significantly better in Finnish than in English and Swedish and significantly more negative words were recalled in English than in Swedish.