2.1 Research on learner emotions in WEIRD and non-WEIRD contexts

Henrich et al. (2010) pointed out that social scientists are often not aware of the sample bias in their research. They make broad claims about human psychology and behaviour based on data collected from participants from WEIRD societies, which “are among the least representative populations one could find for generalizing about humans” (p. 2). The field of FL learning emotions has largely avoided the WEIRD bias because its researchers operate in different parts of the world, including non-WEIRD societies, and collected data across continental borders. Dewaele and MacIntyre (2014) was the first study to consider FLE and FLCA using a sample of 1742 FL learners from all over the world, but a large majority (85%) came from WEIRD countries. The surge in interest in learner emotions in China meant that data were collected from several thousands of Chinese EFL learners in the late 2010s, which compensated the initial bias (Li et al., 2022, 2023). Studies in Kazakhstan and Iran also added a welcome non-WEIRD pool of FL learners (Dewaele, Özdemir et al., 2022; Elahi Shirvan & Talebzadeh, 2018; Kruk et al., 2022).

2.2 The measurement of learner emotions

The boost in research in learner emotions is undoubtedly linked to the development of solid instruments to measure various emotions, allowing replication studies, translations and adaptations of the original instruments. MacIntyre (2017) argued that the publication of the psychometrically validated FLCAS by Horwitz et al. (1986) marked the end of the chaotic “Confounded period” in FL anxiety research and heralded the “Specialized approach”. The original FLCA Scale contains 33 items that capture a FL learner’s tendency (trait) to feel anxious (state) when using or learning a FL. It focuses on both psychological and physiological symptoms of anxiety, including sweating and increased heart rate. An 8-item version of the scale was psychometrically validated in Botes, van der Westhuizen et al. (2022). FLCA was found to be unidimensional. FLCA can be described as a negative, medium to high-arousal emotion.

The FLE scale contained 21 items with 5-point Likert scale ratings reflecting positive emotions towards the FL learning experience, towards peers and teacher (Dewaele & MacIntyre, 2014). The authors collected quantitative and qualitative data simultaneously and carried out separate analyses before combining the findings of both types of data. In a follow-up study on the same dataset, Dewaele and MacIntyre (2016) investigated the factor structure of the FLE and FLCA items in more detail. The eight FLCA items loaded on a first dimension, the 21 FLE items loaded on a second dimension that reflected the social dimension of enjoyment and a third one that reflected the private dimension of enjoyment.

Botes et al. (2021) used the database on which Dewaele and MacIntyre (2014) was based to investigate the factor structure in more detail with the aim of constructing a Short Foreign Language Enjoyment Scale. A principal components analysis revealed a 3-factor hierarchical model, with a higher-order FLE factor. The first dimension reflected FL learners’ appreciation of their FL teacher. The second dimension showed their personal enjoyment in the FL class; and the third factor reflected their social enjoyment, namely good relations with peers. FLE is a positively valenced medium to high-arousal emotion.

FL Learning Boredom (FLLB) has only recently emerged in SLA research (Li, 2021). FLLB induces sleepiness among students and hence reduces both their FL input and output, negatively affecting their performance and progress. Li et al. (2023) defined FLLB as a learner’s emotion characterised by a combination of negative valence and extremely low arousal when engaging in under-challenging or over-challenging activities. They collected data from 808 Chinese EFL students in a pilot study and a further 2223 Chinese EFL students in the main study which allowed them to describe the factor structure of FLLB. The authors developed a 32-item scale with seven factors. The first factor (8 items), FLB, has been used independently. Li’s (2021) mixed-methods study found that FLB is more prevalent among learners who report a lack of both control and value. In other words, learners who did not feel very competent, disliked their FL classes and failed to perceive their intrinsic value, were more likely to be bored. These findings were confirmed in studies on Polish EFL learners (Kruk et al., 2021; Pawlak et al., 2022).

Zhou et al. (2021) introduced the emotion Foreign Language Peace of Mind (FLPOM) 外语平和心态 (waiyu pinghe xintai), which they drew from the Chinese cultural context. It refers to a state of peacefulness and harmony (Lee et al., 2013) and can be seen as “an indicator of psychological well-being in collectivist cultures” (p. 8). Zhou et al. (2021) collected data from 158 Chinese EFL students in a pilot study and a further 441 Chinese EFL students in the main study to develop and validate the 8-item FLPOM scale. Factor analysis revealed a single factor structure. FLPOM turned out to be a better predictor of self-perceived FL proficiency of 400 Chinese EFL learners than FLE. FLPOM and FLE were positively linked but they are arguably separate emotion constructs as FLPOM is a low-arousal emotion while FLE is a higher arousal emotion.

Gardner’s (1985) socio-educational model dominated the field of applied linguistics until the arrival of Dörnyei’s (2009) L2 selves model. The socio-educational model was presented as “a dynamic causal interplay of individual difference variables interacting with environmental and acquisition contexts resulting in both linguistic and non-linguistic outcomes” (Gardner, 1985, p. 165). It includes four dimensions: integrativeness, attitudes towards the learning situation, language anxiety and language attitudes and motivation. The importance of the local and wider societal learning context makes the socio-educational model particularly useful for our study. Dörnyei (2019) underlined that Gardner’s model is “still relevant” (p. xxi).

2.3 The effect of skill level on learner emotions

The finding in Dewaele and MacIntyre (2014) that more advanced (and proficient) FL learners reported more FLE and less FLCA was confirmed in later studies (Jiang & Dewaele, 2019; Li et al., 2020). Botes et al. (2020) speculated that with increased skill comes increased self-confidence and an ability to project a more sophisticated and accurate sense of self. The more positive emotions may in turn push students to work harder and improve their skill further, creating a positive feedback loop. A similar pattern emerged in Dewaele and MacIntyre (2022) who reported that beginners spend less time in a state of flow during the FL class than more advanced FL learners. Flow experiences become more habitual, more intense, more social and longer in duration as FL learners’ skill level increases, which could boost their positive emotions and control their negative ones.

2.4 Relationships between learner emotions

The first research question in Dewaele and MacIntyre (2014) was the relationship between FLE and FCLA. The significant negative correlation, representing a small effect size, allowed the authors to claim that FLE and FLCA are distinct emotions – a finding confirmed in the recent meta-analysis (Botes et al., 2022) where the average relationship for 46 effect sizes was a moderate r = –.31 (p. 214).

With the introduction of “new” learner emotions like FLB and FLPOM, researchers are establishing a growing “map” of relationships between FL learner emotions. Li and Han’s (2022) study of the emotions of 348 Chinese EFL learners in online Chinese settings found FLE to be negatively correlated with both FLCA and FLB. FLCA and FLLB were positively correlated. A similar pattern emerged in Li (2022) who found a strong negative correlation between levels of FLE and FLLB among 868 EFL students in China. She speculated that while FLE and FLBB are opposites in terms of valence, they share an object focus on learning activities. Both are also strongly affected by attitudes toward the teacher.

Following the same avenue, Li and Wei (2022) examined to what extent FLE, FLCA, and FLB affected achievement of 954 junior EFL learners in China. The longitudinal study revealed that the three emotions at Time 1 predicted achievement at Time 2 but that FLE was the strongest predictor across time, with FLCA being a significant negative predictor at Times 2 and 3. The negative effect of FLB faded completely by Time 3.

Dewaele et al. (to appear) used the same database of 502 Moroccan EFL learners as the one on which the present paper is based, to establish which learner emotions (FLE, FLCA and FLB) were the best predictors of EFL achievement. Latent dominance analysis showed that FLCA had the strongest (negative) effect on EFL achievement, followed by FLB, and that FLE had a significant -but weaker- positive effect.

Dewaele, Albakistani et al. (2022) collected data from 168 Arab and Kurdish EFL learners and found a significant positive correlation between FLCA and FLB, and significant negative relationships between FLE and both FLCA and FLB. In other words, anxious learners also tend to be bored, and bored students are unlikely to enjoy their FL classes. FLCA in in-person classes was the strongest negative predictor of English academic achievement but FLE and FLB had no effect.

Adopting a Structural Equation Modelling (SEM) approach, Dewaele, Botes et al. (2022) used an international sample of 332 learners to investigate the relationship between FLE, FLCA, and FLB. The authors found a close-fitting SEM in which higher levels of FLE were correlated with significantly lower FLCA and FLB levels. A positive relationship emerged between FLCA and FLB.

Adding motivation in the picture, Dewaele and Proietti Ergün (2020) focused on relationships and differences in levels of FLE, FLCA and motivation in Italian FL and English FL classes of 110 Turkish pupils in an immersion school in Istanbul. Motivation was significantly positively correlated with FLE in Italian and English (large effect sizes), and negatively correlated with FLCA in Italian and English (medium effect sizes). FLE and FLCA were negatively correlated in Italian and English (small-to-medium effect sizes).

Similar patterns emerged in Zhang et al. (2020) who found significant positive correlations between FLE and both instrumental motivation and integrative motivation in a group of 598 Chinese EFL university students. Finally, using Dörnyei’s model of L2 self, Papi and Khajavy (2021) investigated, among others, the relationships between motivation, FLE, FLCA and FL performance among 324 Iranian EFL university learners. Ideal self was a positive predictor of FLE and a negative predictor of FLCA. Ought-other and ideal-other were positive predictors of FLCA.

2.5 Longitudinal designs in FL learner emotions research

Researchers have considered fluctuations in learner emotions and motivation on various time-scales, ranging from seconds, minutes to days, weeks and months (Boudreau et al., 2018; Dewaele, Saito et al., 2023; Elahi Shirvan & Talebzadeh, 2018; Guedat-Bittighofer & Dewaele, to appear; Pawlak et al., 2022). Both Boudreau et al. (2018) who looked at second-by-second fluctuation in FLE and FLCA among ten Anglo-Canadian leaners of French during a 2-minute task, and Pawlak et al. (2022) who analysed minute-by-minute fluctuation in FLB of eleven Polish EFL learners during a 75-minute class, found huge interindividual variation. Pawlak et al. (2022) concluded that variation at group-level obscures individual trajectories of FLB which are affected by a wide range of individual and contextual factors.

Adopting a slightly longer timeframe, Dewaele, Saito et al. (2023) looked at variation in FLE, FLCA and motivation over the course of one semester among 360 FL learners of English, German, French and Spanish in a Kuwaiti university. Little variation was found for FLCA and FLE (including the dimensions FLE Personal and FLE Social) but values for the dimension FLE Teacher and motivation dropped significantly (specifically the dimensions Motivation and Attitudes toward the Learning Situation), which the authors attribute to possible disenchantment with the teacher and the course, possibly because of disappointing test results and a realisation that progress in the FL requires constant effort.

Using a similar timeframe, Elahi Shirvan and Taherian (2021) found that 367 Iranian EFL students’ levels of FLE increased significantly while their level of FLCA decreased significantly towards the end of the semester. Adopting a different method, Elahi Shirvan, Taherian et al. (2021) used a longitudinal confirmatory factor analysis-curve of factors model to investigate fluctuations in FLE of 437 Iranian EFL students between the onset and the end of the course (four data collections with two-week intervals). The authors found a general increase in FLE over time with interesting individual differences: students who scored highest on FLE at the start of the course experienced less change in FLE over time while those with lower initial levels of FLE increased more by the end of the course. Finally, Elahi Shirvan, Yazdanmehr et al. (2021) used longitudinal confirmatory factor analysis-curve of factors model to investigate change in levels of boredom among 412 Iranian EFL learners enrolled in an online course during the COVID pandemic. They found that levels of boredom tended to decrease between the onset and the end of the course. The sharpest drop in boredom happened to students with higher levels of boredom at the onset. Combining multiple timeframes, Elahi Shirvan, Taherian et al. (2020) used the ecological momentary assessment method with two Iranian EFL students, collecting data on their FLE through monthly open-ended interviews, weekly reports in journals, enjoymeters¹ across minutes, and variation across seconds. The authors found that FLE fluctuated at different rates over different periods. The two participants exhibited similar patterns of FLE under influence of the same ecological factors but reacted in their unique ways: one participant’s FLE was more linked to social factors whereas fluctuation in the other learner depended more on her personal FLE.

Finally, Pan and Zhang (2021) investigated the relationships between motivation, personality and fluctuations in FLE and FLCA of 55 Chinese EFL learners over a 14-week period. They found a gradual drop in FLCA and a gradual increase in FLE (with a lot of small ups and downs and a lot of inter-individual variation). FLE was significantly positively correlated with “Ideal L2 Self, Ought-to L2 Self, Family Influences, English Learning Attitudes, Cultural Interest, and Linguistic Self-confidence” (p. 12). FLCA was only negatively correlated with Criterion Measure and Ought-to L2 Self. Students who invested most energy in their FL learning were found to experience greater fluctuations in FLE and FLCA.

2.6 Pseudo-longitudinal designs in FL learner emotions research

The first study to adopt a pseudo-longitudinal design to look at fluctuation of learner emotions was Dewaele and Dewaele (2017). The aim was to establish the effect of age group on levels of FLE and FLCA among 189 British FL pupils in two London schools. The three age groups (12–13 year olds, 14–15 year olds and 16–18 year olds) showed similar levels of FLCA but levels of FLE differed significantly, with relatively high levels of FLE in the youngest age group, dropping for the middle age group, and increasing to its highest level for the oldest age group. The predictors of FLE and FLCA changed over time with FLE being increasingly linked to the teacher and FLCA being increasingly linked to the peers. Interestingly, the relationship between FLE and FLCA differed in the three groups.

Li et al. (2020) also adopted a pseudo-longitudinal design to investigate variation in emotions according to achievement group (though they did not use the word “pseudo-longitudinal”). They collected data from 1307 EFL Chinese students and used average English test scores to create a low achievement group (more than 1 standard deviation below the mean), a middle group (from – 1 SD to +1 SD above the mean-, and a high achievement group (higher than +1 SD above the mean). They found that the high achievement group had the highest levels of FLE and the lowest levels of FLCA. The relationship between FLE and test scores was found to be strong in the high-achievement group and non-significant in the low-achievement group.

Dewaele (2022b) followed a similar avenue in comparing the emotional, sociobiographical and contextual variables that predicted exam results of 275 secondary school FL pupils and 317 university learners of Turkish in Kazakhstan. Levels of FLCA of secondary school FL pupils were non-significantly different from those of university learners but levels of FLE were significantly higher among university learners. FLCA was the only significant (negative) predictor of exam results among the secondary school pupils whereas both FLE and FLCA were significant predictors of the university students’ results. One possible explanation for this finding is that the university learners, just like the oldest age group in Dewaele and Dewaele (2017), had consciously decided to attain an advanced level in the FL and that this explained their increased emotional involvement and motivation.

Summing up, the literature review showed that learner emotions and motivation fluctuate to a greater or lesser extent, depending on the timeframe and the degree of granularity in the observation. No research so far has measured the effect of skill level (Beginner, Intermediate, Advanced) on learner emotions and motivation and investigated whether their relationships evolve over what would be a period of around six years. To do so, we will use a pseudo-longitudinal design (Ellis & Barkhuizen, 2005).

The following research questions will therefore be explored:

1. How do levels of FLE, FLPOM, FLCA, FLB and motivation change from Beginner to Advanced learners?
2. How do the relationships between FLE, FLPOM, FLCA, FLB and motivation change from Beginner to Advanced learners?

3.1 Participants

Participants were 502 Moroccan EFL learners from various secondary schools, universities and language centres based in Morocco. The average age was 22.4 years, with 228 male (45.9%) and 269 female participants (54.1%).² Participants could choose an English or an Arabic version of the questionnaire. The second author translated the questionnaire from English into Arabic and discussed the translation with a fellow English/Arabic teacher.

All participants were multilingual, reflecting the multilingualism in the Moroccan population (Ennaji, 2005). Participants consisted of 83 bilinguals, 258 trilinguals, 134 quadrilinguals, 19 pentalinguals and 8 participants indicated six or more languages in their repertoire. Besides English and Arabic, participants listed French (n = 415), Amazigh (n = 146), Spanish (n = 39), Turkish (n = 14), and German (n = 12).

3.2 English in the Moroccan education system

Foreign language education is compulsory in Moroccan public high schools (English, Spanish or German depending on the region). English is introduced in middle school around the age of 13 (2 hours a week) and it is taught for another three years in high school (3, 4 or 5 hours a week depending on the stream). English is also a popular course at university. The EFL curriculum in public secondary schools is based on a standards-based approach (Ministry of National Education, 2007). The focus is on teaching key competencies and the four skills so learners develop communicative competence. Private schools follow their own EFL curricula.

3.3 Instruments

The first section of the questionnaire concerned participants’ sociobiographical and linguistic profile background and their latest English major test results. It contained one item about their general English language proficiency (ranging from Beginner to Advanced), which is the independent variable in the present study. A group of 85 participants (16.9%) reported being Beginners, 322 participants (64.2%) identified as Intermediate learners, and 95 participants (18.9%) identified as Advanced learners. The second section of the questionnaire focused on four FL classroom emotions: FLE, FLPOM, FLCA, FLB and motivation.

Short-Form Foreign Language Enjoyment Scale (α = .925). The multidimensional nine-item scale was developed by Botes et al. (2021). It provides a score for the higher-order factor of FLE and for three lower-order factors, namely Personal Enjoyment (3 items, e.g. “I enjoy my English class”), Social Enjoyment (3 items, e.g. “There is a good atmosphere in my English classroom”), and Teacher Appreciation (3 items, e.g. “My teacher is supportive”). Items were measured on a five-point Likert scale from “strongly disagree” to “strongly agree”.

Foreign Language Peace of Mind Scale (α = .94). FLPOM was measured with the eight-item scale developed in Zhou et al. (2021). Items included “My mind is free and at ease in the English class” and “I have an inner sense of tranquility and harmony in the English class”. Items were measured on a five-point Likert scale from “strongly disagree” to “strongly agree”.

Short-form Foreign Language Classroom Anxiety Scale (α = .81). The unidimensional eight-item measure is a shortened form of the original 33-item Foreign Language Classroom Anxiety Scale designed by Horwitz et al. (1986), shortened in Dewaele and MacIntyre (2014) and recently validated by Botes, van der Westhuizen et al. (2022). Items included “It embarrasses me to volunteer answers” and “Even if I am well prepared for English, I feel anxious about it”. Items were measured on a five-point Likert scale from “strongly disagree” to “strongly agree”.

Foreign Language Boredom Scale (α = .95). FLB was measured through the 8-item unidimensional scale measuring boredom in the EFL classroom. The measure is a subscale of the broader FLLB Scale (Li et al., 2023), with the subscale specifically examining boredom within the EFL classroom and as such is equivalent in context to the scales used to measure FLA and FLE. Items included were “It is difficult for me to concentrate in the FL class” and “My mind begins to wander in the FL class”. Items were measured on five-point Likert scale from “strongly disagree” to “strongly agree”.

Mini-Attitudes/Motivation Test Battery (AMTB) (α = .68). Attitudes/Motivation (in short: “motivation”) was measured with the slightly adapted (for the EFL context) 12-item scale developed by Tennant and Gardner (2004). Items include: “My attitude toward English-speaking people is…”, “My motivation to learn English for practical purposes (e.g., to get a good job) is…”, “My desire to learn English really well is…”: Items were measured on a 7-point Likert scales (later re-calculated for a 5-point scale), with anchors including ‘weak to strong’, ‘unfavourable’ to ‘favourable’, ‘very low/little’ to ‘very much/high’.

3.4 Data collection

Ethical approval for this study was obtained from the ethics committee at the authors’ research institution. A snowball sampling strategy was adopted. A call for participation was issued on professional networks and social media (LinkedIn, Facebook, Instagram). Personal contacts were used to reach a wide range of Moroccan EFL teachers and learners across Morocco. The questionnaire was posted online using Google Forms and was accessible in June 2022 – when all students had reverted to in-person classes after the COVID pandemic. Participants’ individual consent was obtained at the start of the survey, and no names of students, teachers or schools were collected. The anonymity of participants, their rights, and the confidentiality of their information were explained and guaranteed at the start of the questionnaire. Participants were encouraged to forward the link to their peers.

3.5 Data Analysis

The calculation of the Q-Q (quantile-quantile) showed plots for the four emotions and motivation that follow a normal distribution reasonably well (available from the authors on request). Skewness values range from .49 to –1.40 which falls within the range for data to be considered normally distributed, namely –2.0 to 2.0 (Byrne, 2010). Kurtosis values range from –.01 to 5.88 which also falls within the range for data to be considered normally distributed, namely –7.0 to 7.0 (Byrne, 2010) (see Table 1). A decision was thus made to use parametric statistics, and more specifically ANOVAs, which allow post hoc tests, using SPSS 28.0.

A Pearson correlation analysis showed positive relationships between FLE, FLPOM, and Motivation. Negative relationships emerged between these two positive emotions and motivation and the two negative emotions (see Table 2). Although the correlation between FLE and FLPOM is high, it remains below the threshold of .70 for collinearity. In other words, they do not represent the same construct (Dormann et al., 2013, p. 32). The other correlation values are much lower.

3.6 Composition of the groups of Beginners, Intermediate and Advanced learners

A series of one-way ANOVAs revealed that the three groups differ on a number of background variables. As could be expected, Advanced learners are older, more multilingual, reported having a higher relative standing in their FL class and reported higher results on the latest test (see Table 3). This makes sense as only the best students would pursue the study of English to an Advanced level and weaker students would drop out. An additional Pearson Chi² test showed that the gender distribution was similar across the three skill levels (Chi² = 5.34, p = .25).