6 Study One: Variability and Stability of L2 WTC during a Semester
6.3 Results
6.3.2 Variability in trait and state WTC
To show how state WTC fluctuated at the aggregate level during the semester, state WTC on each measurement occasion was averaged across all individuals. As depicted in Figure 6.3 (panel 1), the aggregated trajectory fluctuates little, although showing a roughly decreasing trend. It indicates that state WTC did not significantly vary across different measurement occasions. However, when looking at each participant individually (panel 2), not a single individual’s trajectory was the same as the aggregated trajectory (i.e. fluctuating around 5 on a scale from 1 to 7). Panel 2 shows that individuals’ state WTC changed dramatically across different measurement occasions, and they differed from each other on each measurement occasion. If focus was only on aggregate level, then both between- and within-person variability could be overlooked.
Figure 6.3. Aggregated and individual trajectories of state WTC over one semester To quantify variability in trait and state WTC and to compare between different variability quantities, in the following sections I will follow Fleeson’s approach of using standard deviations to estimate the amounts of between- and within-person variability (see Fleeson, 2001).
6.3.2.1 Within-person variability in state WTC
the standard deviation of scores for state WTC across all momentary responses collected over the semester (N = 1118), regardless of whether the responses were from the same individual. By doing this, it is assumed that individuals overlapped completely in their distributions of state WTC over time, and there were few between-person differences. This figure indicates the maximum amount of within-person variability.
Figure 6.4. Variability in non-English major students’ trait and state WTC
The amount of within-person variability in WTC was quantified by calculating the standard deviation of scores for state WTC across all measurement occasions for each individual separately, each deviation representing an individual’s amount of within- person variability across the semester (the participants who had only one report were excluded from this analysis). To show the typical amount of within-person variability in WTC across the semester, a mean score was then calculated. As shown in the second bar of Figure 6.4, the average amount of within-person variability in state WTC accounts for two thirds of total variability. This indicates that individuals’ density distributions of state WTC were wide, and there might be a large degree of overlap between individuals. Figure 6.5 provides three density distributions of state WTC as examples. The distribution represented by the blue line is like the typical individual’s distribution of state WTC during one semester, indicating relatively high state WTC and moderate within-person variability (mean = 5.00; SD = 0.82). Compared to the typical individual, the individual represented by the yellow line has a similar level of state WTC but a much higher level of within-person variability (mean = 5.23; SD = 2.09), whereas the individual represented by the green line has higher but less variable state WTC (mean = 6.00; SD = 0.41).
1.54 0.95 1.07 1.19 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 Total Variation (State Level) Within-person Variation (State Level) Between-person Variation (Trait Level)
Between-person Variation (State Level)
S tan d ard De v iatio n Willingness to Communicate
However, according to Figure 6.5, these three individuals’ distributions overlap considerably: (a) like the one whose state WTC is less flexible (the green line), the individual who varies greatly (the yellow line) also tends to be very willing to communicate on most occasions; and (b) for the one whose state WTC is relatively fixed, state WTC may also vary on some occasions. In this circumstance, the means (i.e. central tendencies) are not adequate descriptions of individual distributions, and standard deviations (i.e. within-person variability) need to be considered.
Figure 6.5. Three density distributions of state WTC
6.3.2.2 Between-person variability in trait and state WTC
The amount of variability between different individuals was quantified by two methods of analysis. Firstly, the amount of between-person variability was calculated at the trait level. The standard deviation of scores for trait WTC was calculated across individuals, to show how much individuals differed between each other in WTC at the trait level. Another way to show how much people differed in WTC was to quantify the amount of between-person variability at the state level. By averaging each participant’s scores for state WTC across all measurement occasions, everyone received a mean score representing average level of state WTC across the semester. Then, the standard deviation of all individuals’ mean scores for state WTC was calculated, representing the amount of between-person variability in WTC at the state level. The right two bars in Figure 6.5 show the amounts of between-person variability in WTC at trait and state levels,
respectively. The amounts of between-person variability in trait and state WTC were similar.
6.3.2.3 Within- vs between-person variability comparison
Results of the descriptive analysis show that the amount of within-person variability in state WTC was about the same as the amount of between-person variability in trait and state WTC. To further compare these amounts, they were estimated by using an unconditional model of HLM (see section 6.2.8.1). Results show that nearly half (46%)
of the total amount of variability in state WTC occurred within individuals (e = 1.12)6.
This is in line with the earlier findings using Fleeson’s approach based on the calculation of standard deviations, suggesting that students’ state WTC varied across different lessons during the semester, and the amount of this within-person variability was nearly as much as the amount of observed between-person variability in state WTC.
6.3.2.4 Variability in self-reported communication behaviour
For comparison purposes, amounts of between- and within-person variability in self- reported communication behaviour were estimated. Because the participants had not been asked to report their communication behaviour at the trait level on the baseline measurement, the amount of between-person variability in communication behaviour was only quantified at the state level, followed by the same procedure for state WTC. It was found that the amounts of within-person variability in communication behaviour were higher than the amounts of within-person variability observed in state WTC (see Figure 6.6). Standard deviations show that the average amount of within-person variability in communication behaviour was nearly 20% higher than the average amount of between- person variability. The result of HLM analysis was similar, in that nearly 70% of the total variability in communication behaviour was accounted by within-person variability (e = 2.31).
6 Level 1 model: State WTC = π0 + e, where e refers to within-person variability in state WTC. Level 2 model: π0 =
β00 + r0, where r0 refers to between-person variability in state WTC. The total amount of variability in state WTC is e
+ r0.
Figure 6.6. Variability in non-English major students’ communication behaviour