By: Connie Barroso
For those parents who are gearing up to get young ones into elementary schools, many questions may be running through their minds. What schools are closest to their neighborhoods? Are they good schools? What are the best schools in the county? Answering whether or not a school is “good” or the “best” can be a pretty difficult endeavor, so many may choose to take a straightforward answer even if it may not be an accurate one. In Florida, school grades are that straightforward answer.
Although letter grades (“A”, “B”, etc.) are easy to remember, they don’t quite mean that the “A” schools are the best, and the “F” schools are the worst. These grades are determined by student performance on state tests and student academic gains throughout the school year. These scores and gains do not just represent something about the students, but rather they represent something about the intersection between students’ abilities, the teachers, the curriculum, and other aspects of the school environment. The question now becomes: is there something different between school’s who receive higher school grades versus lower school grades that changes how students’ genes and environments influence their abilities? Research scientists can answer this question by analyzing behavioral genetic models.
Although the name sounds daunting, behavioral genetics is really just the science that teases apart the “nature”, or genetics, and “nurture”, or environment, in human behavior. These models are interesting because they often need a certain type of participant: twins! Scientists compare identical and fraternal twins to better understand the influence of genes, shared environments—like the home or schools twins share, and non-shared environments—like twins placed in different classrooms or having different friends. So, identical twins appearing more similar than fraternal twins would suggest that there are genetic factors influencing the given outcome. If fraternal twins and identical twins are more similar, this suggests that there are shared environmental influences or something in the environment making the twins appear more similar. Finally, if identical twins differ on a given outcome, this suggests that there are non-shared environmental factors at play.
Besides individual genetic and environmental influences on behavior, recent behavioral genetic research has found that different environments may interact with a person’s genetics, also called a gene-by-environment interaction. For student achievement, this suggests two things: first, genetic factors tend to explain more of the differences we see between students in academic achievement when they are in enriched environments, such as schools that provide high quality, research-supported instruction. Second, environmental factors, like a student’s school or neighborhood, tend to explain more achievement differences when environments are poorer and less stable. The recent study by Haughbrook, Hart, Schatschneider, & Taylor (2016) used a twin study to investigate the role of school quality, measured by school grade, on early literacy skills. They expected their results to mirror the previous assumptions: that genetic influences on early literacy skills would increase as the school quality increased, while decreases in school quality would be associated with increased environmental influences on early literacy skills.
Using reading data available from 1,313 twin pairs in kindergarten through third grade from the Florida Twin Project on Reading, Haughbrook et al. (2016) analyzed five reading measures testing different early literacy skills. The twins’ schools were split up into two groups: “A” schools and “non-A” schools (i.e., all schools with a “B”, “C”, “D” or “F” grade).
Results showed that among all the schools, there were significant genetic and environmental influences on each of the five early literacy skill measures. Interestingly, there were differences in the contributions of genetic and environmental factors on early literacy skills between “A” schools and “non-A” schools, but only for pre-reading skills, such as knowing letter-names.
These findings supported the researchers’ hypotheses, indicating that there are more genetic factors influencing pre-reading skills in schools assigned the grade of “A” than for schools with lower grades. Perhaps the “A” schools provide a more consistent environmental element so that natural genetic abilities can flourish. On the other hand, shared and non-shared environments are more influential for pre-reading skills in schools not graded as “A”. This can mean that for schools with lower grades, things like lesson plans and proper implementation of these lesson plans matter more for developing proficient pre-reading skills, which are important for future reading achievement.
Interestingly, schools with better grades are often in neighborhoods with less diversity on both racial and economic fronts and are more likely to be located in wealthier communities. They also receive more government funding as a reward for their good grades as opposed to those schools with lower grades. Financial incentives allow schools to keep better teachers and purchase more educational resources for their students. On the contrary, low-performing schools do not get these financial rewards, are more likely to be in poorer communities, and tend to have a lot more diversity in the socio-economic and racial status of their neighborhoods. This work shows us that it would potentially be more beneficial for lower performing schools to get more resources, as aspects related to the instructional environment seem to be more important for these students. More financial resources could allow these non-A schools to invest in their school instruction and create a more stable school environment in order to help their students.
The evidence from Haughbrook et al. (2016) is telling more than just the story of school quality; it’s suggesting a greater need for change in the use of school grade. This indicator of school quality may be more useful instead as a tool for improving educational environments in lower performing schools, rather than punishing them. If the school quality in lower-performing schools is improved, their learning environments may likely turn more stable. Then, maybe parents won’t have to worry as much about whether the school they are sending their children to is “good” or the “best”; schools can instead focus on improving their student’s natural abilities.
Full citation: Haughbrook, R., Hart, S. A., Schatschneider, C., Taylor, J. (in press). Genetic and Environmental Influences on Early Literacy Skills Across School Grade Contexts. Developmental Science.
Incoming graduate student Mia presented her honors thesis at the Association for Psychological Science conference in Chicago! It was her first conference presentation, congrats Mia!
Getting It Right!
The classification of a reading disability has been the battle at the front and center of current reading research. Although many researchers and educators have found ways to identify children with reading problems, the current issue at war here is the lack of consensus among the many different ways to identify reading disabilities. Changes in government policy have been made to make things easier for multiple classification schemes. The previously federally mandated aptitude-by-achievement standard, which used the difference between student IQ and achievement scores as a way to classify for reading disability, was been replaced by the Individuals with Disabilities Education Act (IDEA). This act lets individual states dictate their own standards for identifying reading disabilities.
Allowing states to make their own decisions for classifying reading disabilities may sound like the next best thing compared to ineffectively identifying students with only one standard; however, this isn’t exactly the case. Research has found that beyond any cognitive or developmental issues, around 10% of the world’s population is likely to be affected with reading problems. So imagine this scenario: at Reading S. Fun Elementary School, there are 100 students. According to the previous statistic, around 10 students will likely have issues learning to read out of this group of 100 students. How do we figure out which students are the ones that need help?
Here is where the problem lies. Under classification scheme A, 20 students would meet the diagnosis of a reading disability. However, classification scheme B would diagnose only 12 children. From the students found to have a reading disability under both classification schemes A and B, only 2 children overlap. What’s worse is that a year later, both A and B are identifying different kids with a reading disability than from the previous year.
So why does this happen? Is it due to the differences in how reading is measured by the reading tests? What if the students are actually regressing back to average reading levels regardless of their true reading skill, as a result of repeatedly taking similar reading tests? Or is it that students are truly changing categories, from reading disabled to average reader and vice-versa? Either way, low agreement and unstable classifications over time raises red flags to developmental scientists in the reading disability field.
Prompted by these concerns, a study done by Schatschneider, Wagner, Hart, & Tighe (2016) used data simulations to examine these issues surrounding several of the current classifications for reading disability. Simulating data is a useful technique when manipulating things that can’t be manipulated- things like natural learning patterns in children, for example. One can intervene and provide varying lesson plans, but, naturally, children’s learning will be changing. “Forcing” students to not grow as fast as other students, or to grow at the same speed, is both inorganic and unethical. This is where simulated data comes in handy: by using an existing data set from 31,339 first and second grade students in Florida schools as a base, Schatschneider and his colleagues were able to create new data sets that resembled realistic possible reading growth patterns in children and manipulate those to test different classification schemes with specific types of growth.
In their study, they created three simulated data sets and manipulated them to represent three student growth patterns: a fan-spread pattern, where students with the highest starting points increase the most over time, while the students with the lowest starting point increase the least over time; a mastery-learning pattern, where students who start below average grow at a faster pace than above-average students; and a stable-growth pattern, where students all have the same growth no matter where they are on initial reading levels. They used 6 commonly used classification schemes to classify students as reading disabled in each pattern of growth. These classification schemes were (1) low achievement on end-of-year oral reading fluency (ORF) tests, (2) low achievement on end-of-year non-word reading fluency (NWF) tests, (3) unexpected low achievement on ORF as predicted by scores on a verbal aptitude measure, (4) same as previous but with unexpected low NWF achievement, (5) dual discrepancy classification, which means that students initial scores and scores over time on the ORF are similarly low, and (6) a constellation model, which is a combination of two or more of the previous 5 classifications as a diagnosis of reading disability. Additionally for each classification scheme, three cut-score values were tested at the 25th, 15th, and 5th percentiles for reading scores.
The researchers used these six classification schemes and three patterns of growth to investigate several questions: first, they wanted to find the pattern of growth that produced the most stable classifications of reading disability over time; second, they asked which cut-score values in each classification system produced the most stable results; and third, they wanted to know which classification scheme produced the most stable classifications of reading disability.
Results found that the stability of reading disability identification over time was highest in the fan-spread growth pattern. This means that for the classification schemes, the best growth pattern for students to be classified in was when students with higher reading levels increased at faster rates than students with low reading levels (consistency range from 44%-74%). This was the case for all 5 classification schemes except for the dual-discrepancy classification, which was more stable over time with the stable-growth pattern (consistency = 84%). Regarding the cut points, they found that the higher cut-score values were more stable and consistent across classification schemes over time than the lower cut points. As for the most stable classification scheme, they found that the rudimentary form of the constellation model (i.e., comprising of two or more of the other 5 classification schemes) produced the most stable classifications (consistency range from 68%-70%).
What might this mean for the concerns of not successfully identifying and treating students with a reading disability? Although this study does not necessarily give one definitive answer for the best way to identify a reading disability, it does provide evidence on the current state of affairs when examining classification schemes. The current stability rates can be considered low, even for the most stable classifications. Even still, a hopeful silver lining does appear in the findings from the constellation model. Similar to medical practice where multiple measures such as family history, x-rays, and blood tests are taken into account for a diagnosis, these findings suggest that even in the educational system, a reading diagnosis that takes into account many different sources of component skill levels in reading may be the best option for helping struggling readers.
Dr. Hart won the American Psychological Society Rising Star Award. Check it out! http://www.psychologicalscience.org/rising-stars/stars.cfm#H
This summer has been a BIG one for the IDCd Lab! We had two weddings and a baby (making two lab babies)! While in Hawaii, Callie got herself hitched :)
Callie presented a talk on the biometric dual change score model of reading development, and Sara presented on family history as a moderator of response to intervention!
IDCd lab updates
Here you'll find out about lab new, both research and life success