Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)

Graduate Group


First Advisor

John W. Fantuzzo


Children from low-income households are at risk for entering school behind their more economically advantaged peers across major domains of school readiness. The Head Start program represents the federal government’s response to these achievement gaps by mandating the use of scientifically based assessments and curricula to provide children with the necessary school readiness skills. Routine teacher-report assessment of children’s school readiness using scientifically validated assessments is key to effectively guide early childhood education. Approaches to Learning is one of the five domains of school readiness targeted by Head Start. The Learning-to-Learn Scales (LTLS) is currently the only multidimensional, teacher-report assessment of Approaches to Learning that has been validated for use with Head Start students using traditional statistical methods used to identify the dimensions of the LTLS. These methods, however, do not address the multilevel nature of children nested within teacher assessors and therefore do not account for assessor variance that may compromise the validity of teacher-report child assessments. The present study applies the most advanced, multilevel factor analytic methods to examine how assessor variance impacts the validity of the LTLS dimensions. The results of this study revealed a substantial level of assessor variance was founded associated with every item of the LTLS. Accounting for assessor variance changed both the number of dimensions identified and the nature of the dimensions. Furthermore, the multilevel dimensions had greater capacity to explain variance in important external outcomes compared to dimensions identified by traditional factor analysis. The present study was the first to investigate assessor variance in teacher-report assessment of preschool-aged Head Start children. This research calls into question the validity of widely used preschool, teacher-report assessment based solely on traditional statistical methods. It, therefore, sounds an alarm to alert the early childhood education community to the need to examine assessor variance in its widely used, teacher-report assessments and where necessary use multilevel statistical methods to produce more scientifically valid assessments, especially if these assessments are used to inform decision making for young children from low-income households.