|
Article Excerpt
Abstract. The purpose of this followup study was to determine the long-term predictive validity of theoretically coherent reading measures administered during fall and winter of kindergarten. Seventy-nine children were screened using measures representing letter identification, phonological awareness, phonological memory, and rapid automatized naming. Reading achievement was measured at the end of grades 1 through 4 on passage comprehension, oral reading fluency, sight-word recognition, and phonemic decoding. A multivariate screening model incorporating letter identification, phonological awareness, and rapid automatized naming emerged as the most parsimonious model for predicting long-term reading achievement. This screening model yielded the highest correlations with oral reading fluency as the outcome measure. There was no practical significance between the fall and winter screening over the four-year period using this model. Results are interpreted as confirming the importance of a multivariate screening approach using letter identification, phonological awareness, and rapid automatized naming at the earliest time frame in kindergarten.
**********
A monumental effort is underway in this country to ensure all children are proficient readers before exiting elementary school. Under the No Child Left Behind (NCLB) Act of 2001 every child must meet challenging literacy standards by the end of the 2013-2014 academic year or face unprecedented consequences. Over half of the states currently report that graduation from high school is contingent on performance on exit exams (Skinner & Staresina, 2004). Other states, such as Florida, have imposed retention requirements on third grade students not reading at grade level, regardless of any special needs or circumstances (Florida Department of Education, 2004).
Schools are responding to this sweeping legislative mandate, implementing the use of evidence-based practices as required by the NCLB legislation. Fortunately, the reading research community has amassed a sound research base on the effective elements of early reading instruction (National Reading Panel, 2000; Snow, Burns, & Griffin, 1998) and efforts are underway to provide training to teachers to deliver scientifically based reading instruction (U.S. Department of Education, 2002).
Success reported from multiple studies of early intervention programs underscores the effort that is being made by schools and researchers to ensure that every child becomes a reader. Most children who receive intensive early intervention tailored to their specific deficits become functional readers (Foorman, Francis, Fletcher, Schatschneider, & Mehta, 1998; O'Connor, 2000; Torgesen, Wagner, & Rashotte, 1997). However, challenges remain. Intervention programs are costly, particularly for schools with a high concentration of students who enter school behind their peers. With limited resources, it is important that educators be able to accurately determine which children would be best served to receive early intervention instruction.
For the past eight decades, the research community has attempted to identify appropriate screening measures, investigating multiple constructs that have ranged from mental maturity (Deputy, 1930) to phonological processing skills (e.g., Wagner et al., 1997) to a combination of such variables as phonological awareness, rapid naming, letter identification, language factors, and the mother's education (Catts, Fey, Zhang, & Tomblin, 2001). While reading researchers are rapidly developing screening measures to identify students with potential reading difficulties, consensus has not been reached on the optimal combination of predictive measures (Scarborough, 1998; Schatschneider, Fletcher, Francis, Carlson, & Foorman, 2004), and issues related to matching the most appropriate reading outcome measures with screening measures are still under investigation (Speece, Mills, Ritchey, & Hillman, 2003). Further, additional longitudinal data are needed to predict outcomes beyond the primary grades (Good, Simmons, & Kame'enui, 2001) and to identify the earliest time frame to screen for potential reading problems (Schatschneider et al., 2004).
The optimal time for administering screening measures remains an unanswered question in the reading community. Prominent researchers in the field state the obvious: The most accurate way to identify a successful reader is to examine the ease with which he or she achieves early reading (Adams, 1990; Scarborough, 1998; Torgesen, 1998). As a case in point, O'Connor and Jenkins (1999) determined that assessments administered later in first grade were more predictive than those given in kindergarten. However, it must be taken into consideration that 15 minutes of instruction in phonologicial awareness yielded significant gains for kindergarten children (Foorman, Fletcher, & Francis, 1997). As such, early kindergarten screening for phonological deficits may be critical. In fact, kindergarten has repeatedly been identified as a critical time frame for screening (Adams, 1990; Snow et al., 1998; Torgesen, 1998).
Researchers typically recommend screening at mid-kindergarten based on the assumption that kindergarten students without early home reading instruction may be inaccurately identified prior to the opportunity for school instruction (O'Connor & Jenkins, 1999; Snow et al., 1998; Torgesen, 1998). In contrast, McCardle and colleagues' (2001) summary of prediction research called for preschool and kindergarten screening, positing that early identification prior to kindergarten would provide opportunities to narrow the literacy gap. Through support such as home literacy programs and language intervention, children could be better prepared to meet early literacy demands.
Choosing the most appropriate combination of skills for screening is also important. Scarborough's (1998) review of the predictive literature revealed a wide variety of constructs used as screening measures, ranging from reading and language skills to family, neighborhood, and school-based factors. Her review also suggested that researchers consider a multivariate approach to screening, and Scarborough's more recent work promotes identification at the earliest possible time frame. This research base is key in moving forward with identifying theoretically coherent, efficient screening measures.
One methodologically sound approach would be to identify measures that directly relate to the reading process and provide a cognitive view of reading (Goswami & Bryant, 1990). Specifically, using Adams' (1990) model of the reading process and Wagner and Torgesen's (1987) phonological processing model as a springboard could serve as a theoretical framework to identify key constructs for a pre-reading screening measure. To further address the issue of early identification, it would be important to determine if fall administration was as effective as waiting until mid-kindergarten. This is particularly worthy of consideration as few studies have administered the same predictive measures at two separate time frames in kindergarten (O'Connor & Jenkins, 1999). Identification of the optimal combination of predictive measures that reflect the reading process and an examination of their long-term outcomes using this approach is warranted.
The goal of this research was to identify a combination of predictive measures based on a cognitive view of reading and determine the earliest, most accurate time frame for test administration in kindergarten (fall or mid-kindergarten). The study (Bishop, 2003) was originally conducted using 103 students after correlating kindergarten screening measures to end-of-first-grade reading measures. However, the need for predicting outcomes beyond the primary grades led to additional data collection with the original sample. Longitudinal data are reported for the original full sample (N=103), the devolving group of students during grades 2 and 3, and the core sample (N=79). The core sample includes students who were present at the beginning of the study and remained through the end of grade 4. In sum, the purpose of this study was to (a) examine the long-term predictive validity of theoretically coherent reading measures, and (b) determine the stability of the fall and winter time frames.
MODEL DEVELOPMENT
The first step in conducting this research was to develop the theoretical models based on a review of the literature. The springboard was Scarborough's (1998) analysis of predictive studies, suggesting the use of a multivariate approach to early, accurate screening of children. Further, relating the screening measures to the reading process was deemed warranted (Goswami & Bryant, 1990). Therefore, an understanding of Adams' (1990) model of the reading process was core to the development of the theoretical framework. Measures that represented letter identification (orthography) and Wagner and Torgesen's (1987) phonological processing model were chosen as key constructs, incorporating phonological awareness, rapid automatized naming, and phonological memory into the models.
Letter Identification Model
Letter identification (LI) has been identified as the best single predictive reading measure in kindergarten (Adams, 1990; Scarborough 1998). From a theoretical perspective, letter identification is one the first steps in early literacy as a young child focuses on print in order to link letters, identify letter patterns, and recognize whole words. Establishing a model represented by letter identification alone to verify these findings was a reasonable goal; if the finding was valid, schools would save valuable instructional time by administering only one test.
Letter Identification and Phonological Awareness Model
The relationship between phonological awareness and reading disabilities is compelling, and the ability of phonological awareness to predict future reading achievement cannot be ignored. A number of substantial studies have confirmed the predictive strength of phonological awareness in the early grades and considerable evidence has demonstrated its predictive nature beyond the early years (Blachman, 2000). For example, differences in phonological awareness skill prior to beginning reading instruction predict later reading growth. The subsequent inability to acquire phonological awareness serves as a diagnostic tool in identifying children with reading disabilities beyond the early grades. Further, while letter identification and phonological awareness are closely related, each captures different abilities that can exist independent of each other, representing orthographic processing and phonological processing, respectively.
Letter Identification, Phonological Awareness, and Rapid Automatized Naming Model
There is growing consensus that rapid automatized naming (RAN) may serve as an accurate, early, time-efficient predictor of reading achievement (e.g., Wolf, 1997) as the rapid naming of visual stimuli resembles the reading process. As a phonological subprocessor, RAN relates to the efficiency with which the reader accesses or retrieves information. While there is disagreement on whether it is a subprocessor of phonological awareness or a separate entity that contributes to the reading process, experts agree that incorporating rapid automatized naming and phonological awareness tasks into predictive measures is warranted (Torgesen & Burgess, 1998; Wolf, Bowers, & Biddle, 2000). Hence a model that incorporated RAN was developed to see if it had a value-added effect in addition to letter identification and phonological awareness.
Letter Identification, Phonological Awareness, Rapid Automatized Naming, and Phonological Memory Model
This model adds phonological memory and represents all the phonological processing tasks with the added element of letter identification. Clearly, phonological processing is critical to the reading process; indeed, longitudinal predictive data have demonstrated a causal relationship between phonological processing and future word reading ability (e.g., Torgesen et al., 1997). However, the role of...
|
