The Relationship Between Pedagogical Content Knowledge and Reading Achievement

• In Event: Evidence Use in Professional Learning Communities

Abstract

Objectives
This paper analyzes the relationship between pedagogical content knowledge generated collectively (CPCK) in a PLC, and student achievement. The focus is on PLCs where data are discussed to improve achievement.
Perspective(s)/Theoretical framework
PCK is the knowledge of a subject matter and how to teach it. In a PLC, PCK is collective in nature, as knowledge is shared and distributed across the PLC (Wenger, 1998). In discussions aimed at analyzing data to improve student learning, collective knowledge building occurs when participants become more specific about the instructional practices that will address an achievement problem e.g., describe more details about the practice and under what conditions it is best implemented. Empirical studies suggest that increased PCK specificity is linked to improved achievement (e.g., Lai, Wilson, McNaughton & Hsiao, 2014).

Methods and data sources
Two groups of schools (n=8 schools) participated. One meeting per school was observed. Achievement data (STAR assessment) were collected from 573 students in three year-level cohorts at two time points (start of two consecutive academic years). STAR is a nationally normed standardized reading assessment (Elley, 2001). The test-retest reliability r is 0.91. To control for age-specific variations over time, STAR was analyzed as norm-adjusted scale scores by subtracting the national norms (of each year level at each time of administration) from student scale scores.
The unit of analysis for the PLC meetings was a turn, which was each time a person spoke. PCK-related turns were scored from 1-3. A higher score indicated that more specific instructional practices were discussed to solve an achievement problem. Non-CPCK turns were analyzed thematically into Data Trend (discussing patterns in the data) and Other. Inter-rater reliability was 86%. PCK scores were aggregated to form a CPCK score for each PLC.
We applied Hierarchical Linear Modelling (HLM). ANOVAs showed that CPCK and Data Trend, when analyzed as the total percentage of turns per meeting, was significant for modelling. Three three-level models (Student: gender, cohort, initial achievement; School; Group: A, B) were built. Models only varied at the school level (Model 1: CPCK, Model 2: Data trend, Model 3: CPCK+ data trend). Akaike information criterion (AIC), the Bayesian information criterion (BIC), and comparison of the log likelihood were used as the model selection criteria.

Results
A one-way analysis of covariance (ANCOVA) with random effects model with cohort (p-value < 0.001), gender (p-value = 0.07), and initial achievement (p-value < 0.001) at student level, and total CPCK and total Data Trend (p-value = 0.05) at school level was the best-fitted model. On average, a 10% increase in total CPCK and Data Trend increased achievement by 1.9 units (STAR reading scale scores) i.e., one fifth and one sixth of expected gains in one year.

Significance
This study is one of the few that has statistically modelled CPCK measured via observations with achievement. It extends current models of data use (Gummer & Mandinach, 2015), by demonstrating that neither knowledge nor discussing data alone is related to student achievement; rather the combination of both is critical.

Authors

• Mei Kuin Lai, The University of Auckland

• Stuart Mcnaughton, The University of Auckland

• Tong Zhu, The University of Auckland