Immediacy and Learning Meta-Analysis 18
best represented student affect, and despite the findings of Witt and Wheeless (2001), who
observed a meaningfully higher r with affective than cognitive learning, we classified the
learning loss measure in the perceived learning category. In the absence of complete norms (e.g.,
skewness, kurtosis) for a cognitive test that correlated well with learning loss in a single study
(Chesebro & McCroskey, 2000), we were unable to determine if the reported r was artifactual.
The general absence of reliability assessment and reporting forced the use of the .60 reliability
convention (Hedges & Olkin, 1985) in most of the meta-analysis involving this variable. Since
the two items in learning loss measured different things—predicted learning from an ideal
teacher and perceived learning from the current teacher—alpha reliabilities reported in two
dissertations (Baker, 2001; Schaller, 1993) of .71 and .63 are apparently invalid. Moreover, the
subtraction of the second item from the first produces a single learning loss score equivalent to a
single-item measure that requires test-retest reliability estimates. Because they measured
students' attitudes or opinions about certain types of learning, five other measures utilized in
these studies were classified as perceived learning measures for the purpose of meta-analysis: the
Learning Indicators instrument (Frymier & Houser, 2000), the "cognitive" section of the
Learning Variables Survey (Allen & Shaw, 1990; Shaw, 1988), and 3 course-specific perceived
learning measures (Frymier & Schulman, 1994; Jordan, 1989; Roach & Byrne, 2001). The
average reliability for these measures was .86.
Our categorization of cognitive learning was generally restricted in an attempt to more
closely to match instances in Bloom's (1956) cognitive learning category. All of the included
studies that measured cognitive learning performance assessed recall, recognition, test grades, or
course grades. While test or course grades may reflect some types of higher-order learning,
levels of learning involving analysis, synthesis, and problem-solving were not identified or
Convention