Test Instructions Do Not Moderate the Indirect Effect of Perceived Test Importance on Test Performance in Low-Stakes Testing Contexts

Development and Internal Validity of the Student Opinion Scale: A Measure of Test-Taking Motivation

“…It can be administered after a single test (test-specific SOS) or after a series of tests administered in a single testing session (session-specific SOS). Multiple studies indicate SOS scores operate as expected based on theory and previous research (e.g., Finney et al, 2018;Myers & Finney, 2021;Thelk et al, 2009;Zilberberg et al, 2014).…”

“…Additionally, immekus & McGee, (2016) found strong support for Model 2, but did not consider the full set of 10 items in their analysis. Only Finney et al (2016) supported Model 2 without modifications for the full set of 10 items. Interestingly, Finney et al (2016) was able to support Model 2 whereas other studies (Thelk et al, 2009) conducted in similar context and with similar examinees were not.…”

“…Given the presence of negatively worded items (items 3, 4, 7, and 9), items that might be conceptually distinct from others (items 3, 4, and 8), and items with complex wording (items 3, 9, and 10), it is not surprising that more recent studies of the internal structure of the SOS yielded inconsistent findings. Table 2 summarizes results of four CFA studies (Finney et al, 2016; immekus & McGee, (2016); Sessoms & Finney, 2015; Thelk et al, 2009) using some or all of the final set of items and one or more of the models in Figures 2 and 3. One study was conducted using high-school students, with the remaining studies using college students.…”

“…Many of those studies modeling all 10 items simultaneously suggest the one-factor model (Model 1) is not plausible (Finney et al, 2016; Thelk et al, 2009). A two-factor model (Model 2) yielded improved fit over Model 1, but Model 2 was not consistently championed (Finney et al, 2016; Thelk et al, 2009).…”

“…Many of those studies modeling all 10 items simultaneously suggest the one-factor model (Model 1) is not plausible (Finney et al, 2016; Thelk et al, 2009). A two-factor model (Model 2) yielded improved fit over Model 1, but Model 2 was not consistently championed (Finney et al, 2016; Thelk et al, 2009). For example, Thelk et al (2009) championed a modified version of Model 2 that included two residual correlations (Model 2b) based on the similar wording between items 1 and 5 and the negative wording of items 7 and 9.…”