Another interesting point in language anxiety research is the question of how motivation and anxiety are connected. Gardner and MacIntyre (1993) propose that these two variables are in a reciprocal relationship. What is more, MacIntyre (2002) suggests that emotion, in the given case language anxiety, can be a motivating factor in itself.
Spearman’s rho was calculated to investigate the mutuality of the relationship between language anxiety and motivation. Five dimensions of motivation were looked at: ideal self, ought-to self/own, ought-to self/other, motivated behavior and language learning attitude (all of these dimensions form part of or are derivates of the motivational construct as measured in Kormos and Csizér (2008, 2009))
Slight but significant negative correlations were found between learners’ ideal L2 self and class anxiety (rs = -.22 p < .01), and ideal L2 self and foreign language classroom anxiety (rs = -.23 p < .01). Interestingly, the expected influence of the ought-to self from the learners’ own aspect and from the aspect of significant others on language anxiety was not reinforced by the obtained results, except for a low correlation between FLCAS scores and ought-to self/own (rs = -.17 p < .05).
Motivated language learning behavior depicted significant and somewhat higher correlations with the anxiety measures: rs use = -.22, rs class = -.28, rs flcas = -.36 p < .01 respectively. Finally, language learning attitude and language anxiety proved to have the strongest link out of the five motivational dimensions. In the case of the three anxiety measures, correlations were found to be rs use = -.30, rs class = -.33, rs flcas = -.46 respectively. All of these findings suggest that the more motivated one is, either in the sense that the ideal L2 self image of the learner or the ought-to self provides the
impetus, the less language anxiety the person is likely to experience. Likewise, the more positive attitude someone has towards language learning and exhibits motivated language learning behavior, the less likely they are to experience language anxiety.
The above correlations are not strong enough for us to declare that language learning motivation and language anxiety are in a linear reciprocal relationship.
Previous empirical studies have suggested that the interaction between the two variables is more complex than correlations could adequately describe. Clément (1986) postulated that language anxiety and perceived language proficiency are the key determinants of self-confidence, which influence learners’ motivation (Clément, 1986). While Gardner, Masgoret, and Tremblay (1999) in testing their causal model of second language learning found that motivation (attitude and desire to learn the second language) exerts its influence on the level of anxiety experienced, which in turn affects learners’ self-perceptions of their level of language proficiency. What is more, they also found a significant but low correlation of r = -.34 p < .05 between class anxiety and learners’ attitudes towards the learning situation. All in all, the relationship of language anxiety and language learning motivation seems to be rather indirect and complex.
Going back to MacIntyre’s (2002) proposition of emotion also being a driving force, the low negative correlations could be the result of a joint effect of the interference of a facilitating type of anxiety and a debilitating one. That is, up to a certain point language anxiety can be facilitating, while after a certain threshold it becomes inhibiting (see Yerkes-Dodson Law, Levitt(1980) and Elliot’s (2006) approach and avoidant motivation). To follow up this line of reasoning, two variables with normal distribution were selected to see whether an inverted U relationship is what characterizes best the link between motivation and language anxiety rather than
a linear correlation. To this means nonlinear regression analysis was employed with the inclusion of the variables of motivated behavior and language use anxiety.
Analogous to the Yerkes-Dodson Law, language use anxiety was designated as the independent variable and motivated language learning as the dependent one.
Contrary to expectations, the best fitting curve was the cubic curve (lying S shape) and not the quadratic (inverted U shape).The cubic regression (see Figure 4) curve depicted a relatively low explanatory value of R = .33, accounting for 9.5% of the variance, F(3, 21) = 8.45, p < .01. The standardized beta weights for the variable of language use anxiety were buseanx = -3.62, buseanx2 = 6.71, and buseanx3 = -3.47, with corresponding tuseanx = -3.28, tuseanx2 = 2.89, and tuseanx3 = -2.69 with associated probabilities p < .01 which suggest the following: First, there is a negative relationship between anxiety and motivation where as anxiety increases, motivation falls; then, a stagnation of a moderate level of anxiety is coupled with a moderate level of motivation; this is proceeded by a decline of motivation as anxiety further increases. Non-linear regression was able to detect a more subtle interaction between anxiety and motivation, but as the graphic representation (Figure 4) and low albeit significant R value illustrate, a curvilinear regression model is visually not very convincing.
Figure 4. Cubic regression curve depicting the relationship between language use anxiety and motivated language learning behavior.
Note: mlbeha - motivated language learning behavior useanx – language use anxiety
Because motivation is a multidimensional construct, the dimensions related to the L2 motivational self system (Dörnyei, 2005) and their relationship to language anxiety were also investigated to uncover more of the two variables’ interaction. The question was to what extent do L2 ideal self, ought-to self/own and ought-to self/other explain motivated language learning behavior and to what extent do they explain language use anxiety. The criterion variables of motivated language learning behavior and language use anxiety both met the assumption of normal distribution according to the Kolgoromov-Smirnov Test of normality. With this, the assumption of normality of the criterion variable for regression analysis to be carried out was sufficiently met (Dancey & Reid, 2004) and the regression models’ fit could be calculated.
In the first multiple regression model motivated language learning behavior was the dependent variable and ideal-self, ought-to self/other and ought-to self/own were the independent variables. A strong associations (R = .70) between the criterion and explanatory variables was found. Ideal self and ought-to selves explain 49% of the variance in the dependent variable, F = 66.28 p < .01. Ideal self seems to relate positively to motivated behavior as the regression coefficient b = .59 t = 8.70, p < .01) depicts; ought-to self/own also explains the criterion variable b = .22, t = 3.51, p < .01 and implies that the ideal L2 self is a stronger predictor of motivated language learning behavior than ought-to self/own. A surprising result is the negative regression coefficient of ought-to self/other which does not contribute to explaining the criterion variable of motivated language learning behavior to a significant extent.
On the other hand, in the second regression model, where the dependent variable was language use anxiety, results were the opposite. In this case, ideal self and ought-to self/own did not show a significant relationship towards language use anxiety, but ought-to self/other seemed to have a positive and significant explanatory relationship with the criterion variable. Consequently, the overall association between language use anxiety and the independent variables was rather weak at R = .22 and explaining a mere 5% of the variance. It is noteworthy to mention that the standardized regression coefficient of ought-to self/other was found to be significant:
b = .23, t = 2.95, p < .01, whereas the other variables of ideal self and ought-to/own were not related to language use anxiety neither positively nor significantly in the given sample.
These are interesting findings in that until now, in applied linguistics research, the ideal and ought-to selves have been considered as constructs relevant to language learning motivation. Results here, however, suggest that the link between motivation
and language anxiety could be through the L2 motivation self system. The ought-to self/other component was found to be linked more to language anxiety than to motivated learning behavior. In motivation research, it has been suggested that goals relevant to the ought-to self, are associated with a prevention focus rather than a promotional focus, and thus are more closely related to avoiding negative feelings than promoting goal attainment (Elliot, 2006). What is more, motivation with a prevention focus is linked to higher levels of anxiety (Higgins, 1987).
The final group of statistical tests concerned the idea of whether or not the difference between ought-to/own and the actual self and the difference between ought-to/other and the actual self are related to language anxiety. First of all, the actual self was based on learners’ identifying themselves as beginner (kezdő), intermediate (középhaladó), or advanced (haladó) learners. These categorical variables were assigned values of 1, 2, and 3 respectively. Based their ought-to/own and ought-to/other aggregate scores, learners were assigned to groups 1, 2, 3 with 1 being the group of learners with the lowest scores. These two values ranging from 1 to 3 were subtracted and squared which yielded three categories of the level of differences between the ought-to and actual selves. The FLCAS scores were also categorized into low, medium and high anxiety groups. Then using chi-square statistic the relationship of the discrepancies and anxiety was sought. Upon computing the chi-squarestatistic, the assumption of the minimum cell counts to be above five was not met. The conclusion therefore was that there is no evidence to suggest an association between ought-to/own and actual discrepancy and language anxiety.
Similar procedures were followed to test the existence of an association between language anxiety and ought-to/other actual self discrepancy. Here, the expected cell count was met. Nevertheless, the χ2=.42 value was not found to be
significant p = .52. The value of Cramer’s V was .05 showing that only 2.5% of the variance in language anxiety can be explained by ought-to/other and actual self discrepancy. Again, there is little evidence to suggest that ought-to/other and actual discrepancy is associated with language anxiety.
These results go against what was expected. Strauman and Higgins (1988) found in their causal model that discrepancy between the ought-to and actual selves is significantly associated with social anxiety (R = .33, p < .01). The analogue of this association in the language learning context did not come to light, most probably due to the lack of instruments that could reliably and in a valid way measure incongruities between learners ought-to and actual L2 selves.