Policy Implications of Select Student Characteristics and Their Influence on the Florida Biology End-Of-Course Assessment

MISSING IMAGE

Material Information

Title:
Policy Implications of Select Student Characteristics and Their Influence on the Florida Biology End-Of-Course Assessment
Physical Description:
1 online resource (193 p.)
Language:
english
Creator:
Bertolotti, Janine C
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ed.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Educational Leadership, Human Development and Organizational Studies in Education
Committee Chair:
OLIVER,BERNARD
Committee Co-Chair:
ELDRIDGE,LINDA BURNEY
Committee Members:
WOOD,R C
ADAMS,ALYSON JOYCE

Subjects

Subjects / Keywords:
education -- leadership -- policy -- science
Human Development and Organizational Studies in Education -- Dissertations, Academic -- UF
Genre:
Educational Leadership thesis, Ed.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
In an attempt to improve student achievement in science in Florida, the Florida Department of Education implemented end-of-course (EOC) assessments in biology during the 2011-2012 academic school year. Although this first administration would only account for 30% of the student's overall final course grade in biology, subsequent administrations would be accompanied by increasing stakes for students, teachers, and schools. Therefore, this study sought to address gaps in empirical evidence as well as discuss how educational policy will potentially impact on teacher evaluation and professional development, student retention and graduation rates, and school accountability indicators. This study explored four variables- reading proficiency, ethnicity, socioeconomic status, and gender- to determine their influence and relationship on biology achievement on the Biology I EOC assessment at a Title 1 school. To do so, the results of the Biology I EOC assessment administered during the Spring 2012 school year was obtained from a small, rural Title 1 high school in North Florida. Additional data regarding each student's qualification for free and reduced-price lunch, FCAT Reading developmental scale scores, FCAT Reading level, grade level, gender, and ethnicity were also collected for the causal-comparative exploratory study. Of the 178 students represented, 48% qualified for free and reduced-price lunch, 54% were female, and 55% scored at FCAT Reading level 3 or higher. Additionally, 59% were White and 37% Black. A combination of descriptive statistics and other statistical procedures such as independent samples one-tailed t-test, one-way ANOVAs, ANCOVAs, multiple-regression, and a Pearson r correlation was utilized in the analysis, with a significance level set at 0.05. Results indicate that of all four variables, FCAT Reading proficiency was the sole variable, after adjusting for other variables; that had a significant impact on biology achievement. Students with higher FCAT Reading developmental scores scored significantly higher on the Biology I EOC assessment than their peers with lower FCAT Reading scores. Additionally, FCAT Reading developmental scale scores were significantly correlated with Biology I EOC scores. The significant predictors for biology scores included FCAT Reading developmental scale scores, grade level, and eligibility for free lunch, which collectively explained 60% of the variability.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Janine C Bertolotti.
Thesis:
Thesis (Ed.D.)--University of Florida, 2013.
Local:
Adviser: OLIVER,BERNARD.
Local:
Co-adviser: ELDRIDGE,LINDA BURNEY.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0046255:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 POLICY IMPLICATIONS OF SELECT STUDENT CHARACTERISTICS AND THEIR INFLUENCE ON THE FLOR IDA BIOLOGY END OF COURSE ASSESSMENT By JANINE CECELIA BERTOLOTTI A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF F LORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR TH E DEGREE OF DOCTOR OF EDUCATION UNIVERSITY OF FLORIDA 2013

PAGE 2

2 2013 Janine Cecelia Bertolotti

PAGE 3

3 To my parents, Michael and Orlene Ireland, my husband, Jo shua Bertolotti, and my daughter, Brooke Cecelia Bertolotti

PAGE 4

4 ACKNOWLEDGMENTS "I can do all things through Christ who strengthens me" Philippians 4:13 This endeavor would be unsuccessful had it not been for the encouragement of family, friends, and c olleagues. I would like to thank my parents, Michael and Orlene Ireland, for encouraging me in my academic pursuits and the endowment of financial support at both the undergraduate and graduate level. My brother, Garth Ireland, who has always been my cheer leader and friend. Most of all, my success has been dependent on the unwavering love and support of my beloved husband, Joshua Bertolotti. Your support has been invaluable, despite the subsequent imposed move and deployment that accompanied it. I am so tha nkful we have made it through this journey together and am very grateful that you will always be in my life. Last, I thank my daughter, Brooke Cecelia Bertolotti, whose recent addition within the past year has broadened my perspective on life and has serve d as a motivator to complete this venture. I am forever indebted to each of you, especially my husband. Dr. Bernard Oliver, my doctoral committee chair and advisor, has provided helpful guidance and patience throughout this process, ensuring that this stud y would be completed in time for the execution of military orders. Drs. Wood and Eldridge have provided direction to my cohort throughout the past three years, challenging us and guiding us throughout this entire process. Dr. Adams has also been valuable i n sharing of both time, resources, and expertise on my committee. Dr. Mousa for always supporting me and providing me with feedback during my entire academic life at Florida. Collectively, you have all played a role in shaping my personal and academic grow th as a researcher and educator at The University of Florida. Thank you all.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 11 ABSTRACT ................................ ................................ ................................ ................... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 14 Histori cal Context of School Accountability ................................ ............................. 14 Education Policy ................................ ................................ ................................ ..... 16 Factors Affecting Student Achievement Scores on Standardized Tests ................. 20 End of Course Assessments ................................ ................................ ............ 27 Reasons cited for implementation of the EOC ................................ ........... 31 The implementation of the end of course assessments ............................. 31 Student P erformance on S tandardized S cience T ests ................................ ..... 32 Sta tement of the Problem ................................ ................................ ....................... 36 Research Questions ................................ ................................ ............................... 37 Statement of Hypothesis ................................ ................................ ......................... 38 Significance of the Study ................................ ................................ ........................ 39 Limitations ................................ ................................ ................................ ............... 42 2 REVIEW OF THE LITERATURE ................................ ................................ ............ 44 Standardized Testing ................................ ................................ .............................. 44 Test Pollution ................................ ................................ ................................ .... 45 Florida Biology I End Of Course (EOC) Assessment ................................ ....... 47 Resul ts of first administration of the Biology I EOC ................................ .... 50 Other science district administration results ................................ ............... 52 Computerized Testing ................................ ................................ ...................... 5 4 Comparability of computer based testing ................................ ................... 54 Other empiric al evidence on computerized testing ................................ .... 56 Computer platform for the Biology EOC ................................ ..................... 57 The Importance of Science Education ................................ ................................ .... 58 Reasons Cited for Shortage in Science Personnel ................................ ........... 59 Catal yst for the Improvement of Science Education ................................ ......... 60 Reading and Science Achievement ................................ ................................ ........ 64 Importance of Reading ................................ ................................ ..................... 64 Literacy Rates of American Students ................................ ............................... 65

PAGE 6

6 Trends in Student Literacy ................................ ................................ ................ 67 International comparison of student literacy ................................ ............... 68 Trends in student literacy in Florida ................................ ........................... 68 Results of Research on Reading Proficiency and Science Achievement ......... 70 Background knowledge ................................ ................................ .............. 76 Vocab ulary ................................ ................................ ................................ 81 Ethnicity and Science Achievement ................................ ................................ ........ 82 Proposed Explanations for Disparities in Achievement ................................ .... 82 The Current Status of Minorities in Science ................................ ..................... 89 Socioeconomic Status and Science Achievement ................................ .................. 92 Free and Reduced Price Lunch ................................ ................................ ........ 93 Research on the Impact of socioeconomic on Achievement ............................ 93 Gender and Science Achievement ................................ ................................ ....... 100 Proposed Explanations for Disparity in Achievement ................................ ..... 103 Representation of Women in Science ................................ ............................ 108 Sum mary ................................ ................................ ................................ .............. 110 3 METHODOLOGY ................................ ................................ ................................ 112 Research Design ................................ ................................ ................................ .. 112 Population ................................ ................................ ................................ ............. 112 Sample and Sampling Procedures ................................ ................................ ....... 114 Procedure for Data Collection ................................ ................................ ............... 115 Measures ................................ ................................ ................................ .............. 115 Reading Proficiency ................................ ................................ ....................... 115 Socioeconomic Status ................................ ................................ .................... 116 B iology Proficiency ................................ ................................ ......................... 116 Participants ................................ ................................ ................................ ........... 116 Method of Data Analysis ................................ ................................ ....................... 116 4 RESULTS AND ANALYSIS ................................ ................................ .................. 119 Research Questions ................................ ................................ ............................. 119 Summary ................................ ................................ ................................ .............. 132 5 DISCUSSION ................................ ................................ ................................ ....... 148 Implications ................................ ................................ ................................ ........... 155 Limitations of the Study ................................ ................................ ......................... 162 Rec ommendations for Future Research ................................ ............................... 164 Summary ................................ ................................ ................................ .............. 165 APPENDIX A DISTRICT CONSENT FORM AND IRB CORRESPONDENCE ........................... 167 B DISTRICT APPROVAL LETTER AND DATA DICTIONARY ................................ 169

PAGE 7

7 LIST OF REFERENCES ................................ ................................ ............................. 172 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 193

PAGE 8

8 LIST OF TABLES Table page 3 1 Student performance by grade level from the Title 1 school population ........... 118 3 2 Sample of student performance by grade level in the study ............................. 118 4 1 FCAT Reading level distribution by grade ................................ ........................ 136 4 2 Descriptive statistics on the influence of FCAT Reading level on Biology I EOC T scale scores ................................ ................................ ......................... 136 4 3 One way ANOVA results on FCAT reading levels and Biology I EOC T scale scores ................................ ................................ ................................ ............... 136 4 4 Tukey post hoc analysis of the influence of FCAT Reading developmental scale scores on Biology I EOC T scale scores ................................ ................. 136 4 5 ANCOVA results for FCAT Reading level and Biology I EOC T scale scores .. 137 4 6 Tukey post hoc analysis of the influence of ethnicity on FCAT Reading developmental s cale scores ................................ ................................ ............. 138 4 7 ANCOVA results for ethnicity and FCAT Reading developmental scale scores ................................ ................................ ................................ ............... 138 4 8 Tukey post hoc analysis o f the influence of socioeconomic status on FCAT Reading developmental scale scores ................................ ............................... 139 4 9 ANCOVA results for socioeconomic status and FCAT Reading developmental scale scores ................................ ................................ ............. 139 4 10 Independent samples t test to determine the influence of gender on FCAT Reading developmental scale scores ................................ ............................... 139 4 11 Descriptive stati stics on the influence of ethnicity on Biology I EOC T scale scores ................................ ................................ ................................ ............... 140 4 12 One way ANOVA results on ethnicity and Biology I EOC T scale scores ........ 140 4 13 Tukey post hoc analysis of the influence of ethnicity on Biology I EOC T scale scores ................................ ................................ ................................ ...... 141 4 14 Descriptive statistics on the influence of socioeconomic status on B iology I EOC T scale scores ................................ ................................ ......................... 141 4 15 One way ANOVA results on socioeconomic status and Biology I EOC T scale scores ................................ ................................ ................................ ...... 142

PAGE 9

9 4 16 T ukey post hoc analysis of the influence of socioeconomic status on Biology I EOC T scale scores ................................ ................................ ......................... 142 4 17 Descriptive statistics on the influence of gender on Biology I EOC T scale scores ................................ ................................ ................................ ............... 142 4 18 Independent samples t test to determine the influence of gender on Biology I EOC T scale scores ................................ ................................ ......................... 142 4 19 Correlation matr ix indicating the relationship between possible predictive variables and Biology I EOC T scale scores ................................ ..................... 143 4 20 Results of stepwise regression model indicating significant predictive variables on Biology I EOC T scale scores ................................ ...................... 145 4 21 ANOVA results of stepwise regression model indicating significant predictive variables on Biology I EOC T scale scores ................................ ...................... 145 4 22 Correlation matrix results of regression model indicating significant predictive variables on Biology I EOC T scale scores ................................ ...................... 146 4 23 Results of regr ession model indicating significant predictive variables on Biology I EOC T scale scores ................................ ................................ ........... 147 4 24 ANOVA results of regression model indicating significant predictive variables on Biology I EO C T scale scores ................................ ................................ ...... 147 4 25 Results of regression model indicating significant predictive variables on Biology I EOC T scale scores ................................ ................................ ........... 147

PAGE 10

10 LIST OF FIGURES Figure page 4 1 The relationship between FCAT Reading developmental scale scores and Biology I EOC T scale scores ................................ ................................ ........... 134 4 2 A histogram of the frequency of regression standardized residuals obtained from the initial stepwise regression analysis ................................ ..................... 134 4 3 A Normal P P plot of regression standardized residual that w as used to determine how well a specific distribution fits the observed data ...................... 135 4 4 A scatterplot used to visually determine homoscedasticity of the data used in the initial stepwise regress ion analysis ................................ ............................. 135

PAGE 11

11 LIST OF ABBREVIATIONS A CHIEVEMENT LEVEL This defines the level of success a student has with the objectives of the curriculum dictated by the Next Generation Sunshine State Standards (NGSSS). They ran ge from level 1 (inadequate level of success) to level 5 (demonstration of mastery). Level 3 is considered a satisfactory level of success. Achievement levels exist for FCAT 2.0 Reading but not for Biology I end of c ourse assessments which will be develop ed in Spring 2013. E ND OF COURSE A comprehensive test taken during the academic year that covers the objectives of the curriculum derived from the Next Generation Sunshine State Standards (NGSSS). F CAT 2.0 The newly revised version of Florida Comprehensi ve Assessment Test (FCAT) that measures student achievement based on content standards derived from the Next Generation Sunshine State Standards (NGSSS). T ITLE 1 SCHOOL This refers to a school receiving Federal Title 1 funds which provides financial assis tance to schools with high numbers or high percentages of children from low income families to help ensure that all children meet challenging state academic standards.

PAGE 12

12 Abstract of Dissertation Presented to the Graduate School of the University of Flor ida in Partial Fulfillment of the Requirements for the D egree of Doctor of Education POLICY IMPLICATIONS OF SELECT STUDENT CHARACTERISTICS AND THEIR INFLUENCE ON THE FLOR IDA BIOLOGY END OF COURSE ASSESSMENT By Janine Cecelia Bertolotti December 2013 Chair: Bernard Oliver Major: Educational Leadership In an attempt to improve student achievement in science in Florida, the Florida Depar tment of Education implemented e nd of c ourse (EOC) assessments in b iology during the 2011 2012 academic school year. Although this first administration would only account for 30% of the student's overall final course grade in b iology, subsequent administrations would be accompanied by increasing stakes for students, teachers, and school s. Therefore this st udy sought to address gaps in empirical evidence as well as discuss how educational policy will potentially impact on teacher evaluation and professional development, student retention and graduation rates, and school accountability indicators. This study explored four variables reading proficiency ethnicity, socioeconomic status, and gender to determine their influence and relationship on b iology achievement on the Biology I EOC assessment at a Title 1 school. To do so, the results of the Biology I EOC assessment administered during the Spring 2012 school year was obtained from a small, rural Title 1 high school in North Florida. Additional data regarding each student's qualification for free and reduced price lunch, FCAT Reading developmental scale sco res, FCAT Reading level, grade level, gender, and ethnicity

PAGE 13

13 were also collected for the causal comparative exploratory study. Of the 178 students represented, 48% qualified for free and reduced price lunch, 54% were female, and 55% scored a t FCAT Reading l evel 3 or higher. Additionally, 59% were White and 37% Black. A combination of descriptive statistics and other statistical procedures such as independent samples one tailed t test, one way ANOVAs, ANCOVAs, multiple regression and a Pearson r correlation was utilized in the analysis, with a signifi cance level set at 0.05 Results indicate that of all four variables, FCAT Reading proficiency was the sole variable, after adjusting for other variables; that had a significant imp act on b iology achievement. St udents with higher FCAT Reading developmental scores scored significantly higher on the Biology I EOC assessment than their peers with lower FCAT Reading scores. Additionally, FCAT Reading developmental scale scores were significantly correlated with Biolo gy I EOC scores. T he significant predictors for b iology scores include d FCAT Reading developmental scale scores, grade level, and eligibility for free lunch which collectively explained 60 % o f the variability

PAGE 14

14 CHAPTER 1 INTRODUCTION Historical Context of School Accountability Since the 1983 publication of A Nation at Risk urged greater accountability from schools and higher academic standards (National Commission on Excellence in Education, 1983), the standards based reform movement has gained momentu m and exerted considerable influence on both state and federal educational policy. D ismal student achievement data on national and international assessments served as a catalyst for the exigency of school reform. One such reform, the No Child Left Behind A ct (NCLB) was signed into law on January 8, 2002, by President George W. Bush NCLB was based on the premise that by holding America's public schools accountable for student performance, student achievement would increase. This reform movement assumed that the foundation for academic achievement rests primarily within the public schools whose failure to educate American's children in basic reading and math literacy resulted in a future citizenry that was unprepared for employment and advanced educational op portunities. The proposed solution involved ensuring school accountability through : educational standards in each grade and subject, testing, and accountability through standardized assessments in core subject areas Despite statewide variations in policy it was common for states to create content standards in the core academic subject areas and test regularly to measure mastery of standards. These tests were developed by each state and approved at the federal level for public schools to obtain federal f unding. Some states awarded special diploma s for students who scored exceptionally

PAGE 15

15 high while students receiving low scores might be retained or denied a diploma (American Education al Research Association, 2000 ). The standardized tests were also used t o supply objective performance indicators that determined whether schools made "adequate yearly progress" in educating their students. The scrutiny of published test results served to ensure that administrators and teachers would improve educational deliv ery in a manner that would prevent "failing" and achieve the rewards associated with high achievement (Carnoy & Loeb, 2002). High scores could bring public praise or financial rewards; low scores could bring public embarrassment and heavy sanctions. Failur e to make adequate progress over a period of time resulted in the implementation of sanctions such as allowing students the option to matriculate at other better performing private or public schools (No Child Left Behind Act of 2001, 2002). Prior to the pr ominence of the accountability movement, the association between assessment and accountability was laissez faire (Carnoy & Loeb, 2002 ; Kress, Zechmann, & Schmitten, 2011) Accountability was traditionally community oriented and relied on the participation of parents and local school boards. Schools were "accountable to district administrators, who, in turn, answer to elected boards" (Carnoy & Loeb, 2002, p.306). Assessments were primarily used by administrators and teachers for diagnostic purposes, to deter mine student proficiency with loosely defined state curricula, and to segregate students into academic tracks (Carnoy & Loeb, 2002 ; Kress, Zechmann, & Schmitten, 2011 ). However, after NCLB, the competency of teachers was no longer the primary judge of stu dent accomplishment ; instead, it became increasingly commonplace to rely on testing (Baker, 2001).

PAGE 16

16 This routine use of tests has gained traction due to escalating public views of test credibility as the primary source of information regarding individual s tudent performance, the quality of educational institutions, and the quality of innovations in education (Baker, 2001). Marchant (2004 identified a shift from testing being used for diagnostic/ prescriptive purposes that allow teachers to evaluate student proficiency to becoming a yardstick to evaluate the success of students, teachers, schools, districts, and even state s. Thus testing has become a high stakes process in which achievement tests are associated with serious consequences for educators and scho ols Education Policy Pub lic educational policy is based on the premise "that all children... obtain a high quality education and reach a minimum proficiency on challenging state academic achievement standards and state academic assessments" (U.S Congress, 2002, p. 15). I n Florida, this public policy has been adopted through the implementation of the Next Generation Sunshine State Standards (NGSSS) and an increase in the minimum passing requirements for graduation as well as the adoption of Common Core standards. Common Core standards are English and mathematics educational standards for kindergarten through 12 th grade that had been adopted by 45 states, including Florida by the end of 2010 ( Florid a Department of Education, 2010 ) ( Co mmon Core Standards Initiative, 2012 a). However, these standards will not be fully implemented in Florida until the 2013 2014 school year ( Florid a Department of Education, 2010 ) The Common Core standards are considered to promote equity by ensuring that all students, regardless of geographic location, are provided with the same clear set of expectations for knowledge and skills at each grade level ( Co mmon Core Standards

PAGE 17

17 Initiative, 2012 a). The standards are designed to be robust and relevant to the real world, reflecting the knowledge and ski lls that our young people need for success in college and careers" so that students can effectively compete locally and globally ( Co mmon Core Standards Initiative, 2012 c ). According to the Common Core Standards Initiative (2012 b ) Common Core standards use : increasing levels of complexity in student reading and reading comprehension ; writing standards heavily focused on developing and conveying logical arguments based on substantive claims and sound reasoning ; techniques to develop student vocabulary ; and p rovide for a solid foundation in whole numbers, addition, subtraction, multiplication, division, fractions, and decimals Thus, collectively the Common Core and the Next Generation Sunshine State Standards serve as additional new accountability indicator s that strive to improve the academ ic achievement of all students. policy centers on improving the rigor, skills, and core content knowledge through the implementation of increasing number of standards and asse ssments to measure student learning. Although the Common Core is meant to improve student p erformance through systematic i ncorporation of core standards in reading and math, the extent of the effect of this policy on science education is subject to debate. Education p olicies such as Common Core and NCLB which do not specifically target science education, still exert an influence. Johnson (2013) rev e ale d that the significant emphasis placed on reading and math had resulted in significant reductions in the emphasis on science education. For example, middle schools were burdened with trying to bridge the deficits in student knowledge and experiences due to the fai lure of elementary schools to teach science

PAGE 18

18 Th erefore, t he practic es within science education are influenced by education policy especially ones spec ifically targeted for science education ( Fensham 2009) This influence is in part, due to the endorsement of some practices over others as well as the differential emphasis of the what and how of science teaching and learning Th us education policy "are operational statements of va lues" (Fensham, 2009, p. 1080). O ne such example is Thailand's policy decision to increase student achievement in the sciences. The policy require d high school students focusing on sciences, to study physics, chemistry, and biology for three years each. Even students who focused on the humanities, were required to complete 2 years of physical and biological sciences. This policy of "no choice resulted in the reduction in gender inequity in student performance in the sciences (Fensham, 2009) Despite th is positive outcome, such a policy decision within the United States may not be a national possibility. T he strong cultural emphasis in American education on the provision of "choice" may inadvertently result in less choices at the postsecondary level and future careers (Fen sham, 2009). Thus the effect of policy on practice is a significant one This effect is particularly pronounced in federally mandated school accountability situations that has influenced school funding and district strategic planning (Johnson, 2013). As state developed multiple ch oice assessments have increased in popularity, emphasis on student gains on such assessm ents has become top priority thereby affecting funding, personnel, scheduling, curriculum and instruction, the learning environment, and accountability (Johnson, 2013). The collective effect of both the macro and micro level education policy pro duces educational turbulence and stifles systemic educ ationa l reform

PAGE 19

19 (Johnson, 2013) as long as test based accountability systems maintain their current theoretical underpinnings ( Anderson, 2011). According to Anderson (2011), accountability policies alter the practice of science educ atio n because they limit the effectiveness of other reform movements" (p. 116). The use of innovative practices such as project based learning, student centered teaching, inquiry based instruction, and constructivist le arning; that research has identified as improving student engageme nt and understanding of science ; are typically sidelined to fo cus on improving standardized test scores. Research has also rev eale d that due to this accountability test system, the highest performing students may have received a reduction in attention and opportunities due to teacher focus on strategically improving the scores of a few st udents to pass state tests (Anderson, 2011). Furthermore the availability and funding of professional development for teachers in science was also negatively impacte d, so as to address the accountability requirements in math and reading (Anderson, 2011). Therefore, although expectations of science education nationa l l y and state wide are high, there exists a disc onnect in what takes place in America's classrooms. Thus education policy as it relates to the curriculum and m easurement of student learning, has served to alte r science educa tion ; in part, du e to its significant emphasis on student achievement as indicated by standardized assessments. Nevertheless, tensions exist within the public educational policy community about current approaches to the measur ement of learning (Baker, 2001). Public policy is shaped by three competing desires : a) the creation of clear goals and appropriate measures so that students, teachers, and schools can meet such expectations, b) local control in deciding what the

PAGE 20

20 goals and measures should be, and c) using the least expensive way of comparing students, schools, districts, and states (Baker, 2001). According to Baker (2001), these three conflicting goals lead to the reliance on inappropriate tests to measure progress and thus public policy promulgates faulty and premature conclusions from the tests Factors Affecting Student A chievement Scores on Standardized T ests N o test is capable of perfectly measuring the knowledge of a student, hence why educators have consistently expressed concerns about the conclusions derived from standardized testing. In fact, test scores fail to rev eal other factors that influence student achievement. Research has revealed a variety of other characteristics that have a negative impact on student performance on standardized tests include, but are not limited to: low levels of reading proficiency (Haug ht & Wall, 2004; Maerten Rivera, Myers, Lee, & Penfield, 2010; O'Reilly & McNamara, 2007; Romance & Vitale, 2008), limited background knowledge (Cromley, Snyder Hogan, & Luciw Dubas, 2010; Ozuru, Dempsey, & McNamara, 2009), limited vocabulary (Fang, 2006), negative attitudes toward science (Singh, Granville, & Dika, 2002), and student participation in less academically rigorous courses (Moore & Slate, 2008; National Science Foundation, 2012a; Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997; Ru mberger & Palardy, 2005; Strand, 2012). Research has also revealed that student backgrounds have a significant impact on student achievement scores However, that impact is not a simple direct product of social background, but instead reveals itself indire ctly through complex stratific ation mechanisms (Mostafa, 2010 ). For example, research has revealed that students from families with lower levels of parental educational attainment as well as single parent families, have lower levels of educational achievem ent (Hines& Holcomb McCoy, 2013;

PAGE 21

21 Kahlenberg, 2001; Okpala, Smith, Jones, & Ellis, 2000; Qui & Wu, 2011). Research also shows that these famil ies are more likely to have lower earning potential ( National Center for Children in Poverty, 2007 ; U.S. Departmen t of Labor (2010 ). Since the socioeconomic composition of the local community largely determines the socioeconomic composition of schools schools are usually composed of other individuals of similar socioeconomic standing (Mostafa, 2010). Thus students o f a similar socioeconomic background enroll in schools with others of similar backgrounds Thus, the research indicates that the background of the student plays a significant role in influencing academic achievement. However, achievement scores are not so lely a function of the students' background, but instead represent a confluence of other variables, of which the educational system is paramount. Therefore, student scores also fail to reveal the influence of institutional effects on student achievement. D isparities in educational funding from community to community results in reduced resources for less privileged students (Mostafa, 2010). Furthermore, those disparities are compounded by difficulties recruiting and retaining effective teachers (Puma, Karwei t, Price, Ricciuti, Thompson, Vaden Kiernan, 1997; Tillman, 2005 ), thereby increasing the proportion of minority and low income students exposed to less effective teachers (Murphy, DeArmond &Guin, 2003; National Science Foundation, 2012a). Research has c onsistently indicated the importance of effective teachers in improving student achievement for all stude nts ( Allen, Gregory, Mikami, Lun, Hamre & Pianta, 2013; Konstantopoulos & Chung, 2011 ). Studies have indicated that lower teacher expectations often le ad to weakened academic performance in students

PAGE 22

22 (Gregory & Weinstein, 2008; Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997; Rumberger & Palardy, 2005; Strand, 2012 For example, research by Konstantopoulos and Chung (2011) demonstrated that teachers have a significant influence on student achievement particularly in reading and science. T his effect is cumulative H aving effective teachers year after year produced larger gains in achievement for all students, an impact which was particularl y pronounced in schools with high minority populations According to the authors, a single effective teacher increased student achievement by approximately one tenth of a standard deviation (Konstantopoulos & Chung, 2011). Therefore, considering th at effective teachers are more likely to be concentrated in schools with students of high socioeconomic standing (Bankston III & Caldas, 1998; Caldas & Bankston III, 1997; Condron & Roscigno, 2003; Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1 997; Rumberger & Palard y, 2005; Ryabov &Van Hook, 2007 ), the experience of schooling differs remarkably based on one's socioeconomic background. Everson and Millsap (2004) revealed that those differences between schools matter as those differences are res ponsible for producing and promulgating differences in student achievement as a factor of school size, the ethnic and racial composition of schools, and the proportion of students in pov erty Okpala, Smith, Jones, and Ellis (2000) and Perry and McConney (2 010) also revealed that schools with a high composition of students of low socioeconomic status ( SES ) have lower levels of achievement in math and reading. In fact, r esearch has indicated that when individual family backgrounds are controlled, the socioeco nomic composition of schools remains a strong predictor of student achievement (Kahlenberg, 2001 ; Perry & McConney, 2010 ).

PAGE 23

23 T eacher expectations of student performance and the type of curriculum provided are also influenced by the socioeconomic composition of schools (Gregory & Weinstein, 2008; Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997; Rumberger & Palardy, 2005; Strand, 2012). However, a highly cohesive school characterized by a supportive and inviting environment was found to mitigate the educational ills that are pervasive in large, urban, primarily minority schools (Stewart, 2007). Thus, research on factors affecting student achievement as reflected by standardized test scores, is clear ; student achievement is a multifaceted outcome of an amalgamation of both student social background and systemic/institutional effects. Th e results of these differences are readily apparent in scores obtained from standardized measures that elucidate how educational attainment is a product of variables most of which are beyond the control of the student. For example, an analysis of students results on the National Assessment of Educational Progress ( NAEP) revealed that m inority students attained significantly lower reported scores than their W hite or Asian/Pacific Islander counterparts while female students scored lower than their male counterparts and students from lower income households scored significantly lower than their peers from more affluent households (National Center for Education Statistic s, 2012). Science proficiency particularly demonstrates t his disparity in academic achievement (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008; Florida Department of Education, 2012b). C haracteristics such as ethnicity, socioeconomic status, and gender have a significant influence on achievement scores in science

PAGE 24

24 While these student characteristics have been deemed influential on student achievement in science, current research has failed to investigate their impact on b iology at the seconda ry level. S ome studies have utilized b iological content on assessments of scien tific knowledge ( along with questions from other branches of science ) but there is a dearth of research about the extent that socioeconomic and other factors influence on b iolo gy achievement. T he limited research available is dated, focus es on affective or metacognitive factors, only evaluate s online instruction, or concentrates specifically on particular instructional strategies However researchers have largely ignored inves tigating these student characteristics possibly due to the assumption that since these student characteristics are influential in other branches of science and academic domains, they probably are influential in b iology achievement. Therefore, a formal inve stigation would be futile and redundant. However, recent developments may have a significant impact on b iology student achievement and are worth investigating First, there has been an increase in the implementation of state mandated high school b iology a ssessment s nationwide (Blazer, 2012), thus prompting the allocation of more fiscal and personnel resources to b iology instruction. Other secondary science subjects have not received the same degree of increase in resources, so it could have a significant i mpact specifically on biology. Second, the re has been an increas e in the number of secondary minority a nd female science teachers (Wood, 2002) which may potentially increase female and minority student identification with and interest in the sciences

PAGE 25

25 Other significant developments include: the increased literacy rate of minority and low socioeconomic students ( National Center for Education Statistics, 2011 ); increased allocation of fiscal resources and personnel to improve literacy outcomes ( Golden, 20 03; Winstead, 2011); the majority of high school students nationwide who take b iology by their high school graduation ( Blank, Langesen, & Petermann, 2007 ); an increase in the number of students taking advanced coursework in b iology (Blank, Langesen, & Pet ermann, 2007 ; National Science Board, 2012 ); an increase in the number of minorities and low socioeconomic students enrolling in advanced science courses (National Science Board, 2012); the increas ed number of female secondary students opting for more adva nced coursework i n science (Moore & Slate, 2008); the increas ed number of certified b iology teachers (Blan k, Langesen, & Petermann, 2007); the increas ed number of science teachers with advanced degrees and full certification (National Science Board, 2012); the impact of the majority (93%) of b iology teachers who teach in field (National Science Board, 2012); the fact that more than half of secondary b iology teachers who are female with 54% possessing a master's degree and more than 10 years teaching experie nce (Wood, 2002); a higher proportion of females undertaking b iological collegiate coursework (National Science Foundati on 2011; Sonnert & Fox, 2012); a higher representation of minorities obtaining bachelor's and master's degrees in science since 1989 ( National Science Foundation, Division of Sci ence Resource Statistics, 2011); more women than men attaining undergraduate scientific degrees (Natio nal Science Foundation, 2012b ); and greater numbers of women earning doctoral degrees in the sciences since 2004 (National Science Foundation, 2011). T hese developments undoubtedly have some impact on student

PAGE 26

26 achievement in biology, but the question remains of how influential socioeconomic and other factors are on student achievement. This is especially notewort hy due to the increased emphasis on b iology achievement nationwide as measured by e nd of c ourse tests instead of the reliance on comprehensive science assessments. T his study was developed to address this gap in the empirical evidence base It also seeks t o address the paucity of research on the importance of reading proficiency for b iology achievement particularly at the secondary level While numerous studies have been conducted on the importance of reading for comprehensive science assessments, only one to date has specifically addressed the role of reading on b iology achievement. However, this study was conducted at the tertiary level thereby underscoring the importance of secondary level specific studies. Consequently researchers are relying heavily on results obtained from comprehensive science assessments to make inferences on the influence of reading on b iology achievement. While such inferences are not erroneous, each scientific discipline varies in its literacy demands and the magnitude of read ing on b iology achievement at the secondary level has yet to be determined. The disparities in literacy demands and the degree of reading influence are particularly pertinent considering the need to develop appropriate professional development opportunities for teachers as well as the significant ramifications of assessment scores on school accountability. Thus, the intent of this study is to address the gaps in extant literature so that instructional leaders, administrators, teachers, stakehold ers, and policymakers can make more informed decisions

PAGE 27

27 End of Course Assessments In the past most states utilized comprehensive assessments to satisfy school accountability requirements Recently, more states have shifted from utilizing comprehensive exi t exams that assess multiple subjects on a singular test, to e nd of c ourse (EOC) assessments that measure student mastery of course content ( Center on Educational Policy, 2008; McIntosh, 2012). EOC assessments are comparable to the final exams that are com pleted at the end of a course, but they are statewide standardized tests (Blazer, 2012; McIntosh, 2012). Florida, for example, has moved from requiring 11 th grade students to complete the Florida Comprehensive Assessment Test (FCAT) Science which cover s v arious science course content from grades 9 through 11, to mandating that all students enrolled in b iology I or an equivalent course complete the Biology EOC assessment in order to receive the course credit required for graduation. Twenty two states admini stered EOC exams to students in associated courses as of s pring 2012 (Blazer, 2012). Over the next 10 years, that number is expected to increase to 26 states (Blazer, 2012). The number of EOC assessments administered per state varies from 1 to 16, with Flo rida electing to implement five (Blazer, 2012; McIntosh, 2012). The nationwide popularity of EOC assessments is not without its detractors. Although the EOC movement is designed to connect course curriculum standards more closely with high school assessme nts so that students are increasingly prepared for postsecondary education or entry into the workforce, critics are concerned about the costs associated with statewide implementation and about the possibility of excessive testing in schools (Blazer, 2012). Other opponents include teachers' unions and advocates of students with special needs (McIntosh, 2012) who cite the

PAGE 28

28 disproportionately negative effects of testing on subsets of the student population and the increasing consequential inequality between gro ups. For example, research has revealed that due to the increased accountability measures, Hispanics and African Americans have demonstrated smaller academic gains relative to W hite students, thus widening the achievement gap (Hanushek & Raymond, 2005). R esearch has also indicated that minorities pass the standardized tests required for graduation at significantly lower rates According to Borg, Plumlee, and Stranahan (2007), "a n average White student has a 65% probability of passing the FCAT graduation re quirement on the first try as compared to a 34% probability for an identical African American student... An average Hispanic student has a 54% probability of passing on the first try 11 percentage points lower than an identical White student" (p. 712). Fur thermore, research has also exposed that standardized testing in Florida has a disproportionate effect on African American an d Hispanic students from low socioeconomic less educated, and high mobility households who are less likely to satisfy graduation r e quirements than their higher socioeconomic White, suburban peers living in educated households (Borg, Plumlee, & Stranahan, 2007). Failure to pass standardized tests can have lasting consequences. M inority and low income students who barely failed New J ersey's exit exam were more likely to drop out of high school ( especially those students who failed on their first attempt ) than students who barely passed (Ou, 2009). Papay, Murnane, and Willett (2010) found similar results. L ow income students from urba n environments were more susceptible, on average, than their wealthier, suburban peers, to the effects of failing an exit exam. They are more likely to drop out of high school if they barely fail the assessment on their

PAGE 29

29 first attempt than their peers who b arely passed. McIntosh (2012) also revealed that high stakes testing affects 7 of every 10 public school students, particularly students of color (71% of African American and 85% of Hispanic students), English language learners (83%), and those o f low soci oeconomic status (71%) that reside in the 26 states that administer or will imple ment exit exams Thus these assessments are more commonplace in states that contain higher proportions of economically disadvantaged and minority students (Center on Education al Policy, 2008) so the consequences for these demographic groups are magnified. In Texas, EOC assessments serve as a graduation requirement for students enrolled in grade 9 during the 2011 2012 school year (Blazer, 2012). According to Blazer (2012), stu dents would be required to pass four EOC assessments : English, mathematics, science, and social studies. In response to the adoption of this graduation requirement, superintendents from various districts testified at a Texas House Public Education Committe e m eeting that dropout rates might increase because students would believe that failing two or three EOC exams would make it impossible to catch up. In fact 75% of students who failed their EOC assessments in s pring 2012 were considered at risk of droppin g out of school, according to the Texas s uperintendents (Blazer, 2012). Therefore, the increased focus on stronger accountability measured through standardized tests has unintentionally increased gaps in achievement and opportunity. Rather than narrowin g the chasms between student s of varying ethnicities and socioeconomic in student achievement and graduation rates (Hanushek & Raymond, 2005), they are being exacerbated. Since graduation from high school is associated

PAGE 30

30 with varied positive life outcomes, e ducational policymakers must understand the impact of high stakes testing on high school completion (Papay, Murnane, & Willett, 2010). Despite the legitimacy of these concerns, advocates of the EOC assessments have remained vocal. Higher education official s have supported the implementation of these assessments, citing the considerable time and costs associated with student remediation of incoming freshman who are unprepared for collegiate level work (McIntosh, 2012). The business community supports assessm ents contend ing that th ey help overcome the lack of basic skills that some recent graduates display in entry level positions ( Center on Educational Policy, 2008; McIntosh, 2012). Other advocates of EOC assessments believe that these assessments are bette r predictors of success at the collegiate level and readiness for work than grades obtained in the classroom or comprehensive exit exams (Blazer, 2012). Research has substantiated some of these claims. According to Hanushek and Raymond (2005), research ha s indicated that since the introduction of accountability systems during the 1990s, student achievement, as measured by state achievement growth data from the National Assessment of Educational Progress (NAEP), has increased. This is particularly true for states that implemented consequential accountability earlier, as they exhibited more rapid gains in NAEP performance, particularly among Hispanic students (Hanushek & Raymond, 2005). Carnoy and Loeb (2002) also revealed that in states with strong state acc ountability students averaged significantly greater gains across all groups on the 8th grade math NAEP and also

PAGE 31

31 achieved at higher levels of math than students in states with little or no accountability measures. Reasons cited for implementation of the EO C T he purpose of implementing the EOC assessment varies according to state (Blazer, 2012). McIntosh (2012) suggests that some states adopt EOC exams in subjects other than English and math in order to evaluate teachers. As EOC assessments allow for easier associations between test scores and specific teachers, they serve as better gauges of teacher performance than other assessments (McIntosh, 2012). McIntosh (2012) also hypothesized that EOC assessments afford greater flexibility at the secondary level sin ce they can test a wide variety of course content that is a function of varied course selection by students. Hence, the connection between curriculum and instruction is improved compared to other kinds of assessments (McIntosh, 2012). In any case, states have reported that they use the assessment to determine student mastery of state standards or as a tool to identify students who are at risk of dropping out of school (Blazer, 2012; McIntosh, 2012). Others use the EOC results to calculate final course grad es and or to require a passing score as a graduation requirement (Blazer, 2012; McIntosh, 2012; Zinth, 2012). In Florida, EOC assessments are criterion referenced tests that measure student proficiency with the Next Generation Sunshine State Standards for specific courses and passing this assessment is a graduation requirement (Blazer, 2012). The implementation of the end of c our se a ssessments D uring the 2011 2012 school year 25 states administered comprehensive or EOC exams on some variation of English l anguage content and 23 assessed math

PAGE 32

32 (McIntosh, 2012 ), but currently only 21 states are administering EOC assessments in b iology (Blazer, 2012). The implementation of EOC assessments in science, specifically b iology is relatively new as past assessments f ocused primarily on reading and mathematics. This addition has been welcomed within the scientific community. Advocates of science education, such as Bybee (2010) campaigned for the equal treatment of science in education reform by pointing out how the No Child Left Behind act inadvertently reduced or eliminated science from some school programs ( particularly at the elementary level ) by failing to include science test scores as a significant part of computing a dequate y early p rogress Student P erformance o n S tandardized S cience T ests Such shifts in priorities demonstrate how Florida divert ed resources and curriculum emphasis as a result of No Child Left Behind mandates regarding standardized test performance on mathematics and reading. Research has indicat ed that educators have diverted resources or increased funding to focus on reading and mathematics while neglecting non tested subjects and reducing funding for gifted students ( Darling Hammond, 2006; Doppen, 2007, Golden, 2003 Marchant, 2004, Winstead, 2 011 ). Research on the effects of school decisions to focus primarily on assessed subjects has also revealed : the inclusion of reading and mathematics in the subject area classroom (Velde Pederson, 2007; Winstead, 2011) ; a reduction in instructional time fo r other subjects (Velde Pederson, 2007; (Winstead, 2011) a decline in resource availability for other subject areas ; the allocation of additional periods of reading instruction to students who fail to attain proficiency on standardized tests ; and decreased professional development in subjects other than reading and math (Winstead, 2011).

PAGE 33

33 Thus, the extent to which those curriculum changes and diverted resources has affected b iology achievement has not been documented. Such changes could potentially result i n increased b iology achievement due to improved student reading comprehension or a reduction in b iology achievement due to loss of instructional time, resources, and professional development opportunities for teachers. Further research is needed to explore this outcome The need for higher quality science instruction is readily apparent when reviewing U.S. student performance on various international and national assessments. One such assessment, the Trends in International Mathematics and Science Study (T IMSS) measure s science knowledge and skills of fourth and eighth graders internationally TIMSS results in 2007 indicated that U.S. fourth graders ranked eighth out of thirty five countries in science achievement and our eighth graders eleventh out of for ty seven countries in 2007 (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008). U.S. students also performed in higher percentages in each of the benchmarks regarding proficiency levels at both grade levels; thus there were proportionally more U.S. students who performed at the advanced level than the international mean and concurrently more U.S. students underperforming compared to the international mean (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008). When U.S. fourth and eigh th grade student performance was compared over a twelve year period ( using data from 1995, 2003, and 2007 ), results indicated no significant change in student performance at both grade levels (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008).

PAGE 34

34 Another international assessment, the Program for International Student Assessment (PISA) measures scientific literacy of 15 year olds in select countries. U.S. students ranked number 23 among both 65 Organisation for Economic Co operation and Development (OECD) and Non OECD countries in 2009 ( Fleischman, Hopstock, Pelczar, & Shelley, 2010). Only 29% of U.S. students scored at the literacy level associated with high order tasks such as "select[ing] and integrat[ing] explanations from different disciplines o f science and technology" and "link[ing] those explanations directly to...life situations" ( Fleischman, Hopstock, Pelczar & Shelley, 2010, p. 26). Eighteen percent of U.S. students scored at a level considered below baseline level of proficiency in which students begin to demonstrate the science competencies that will enable them to participate effectively and productively in life situations related to science and technology" ( Fleischman, Hopstock, Pelczar & Shelley, 2010, p. 26). Nevertheless, the average U.S. science literacy score increased between 2006 and 2009, although 71% of U.S. students in 2009 still fall within the category of low to moderate levels of proficiency in science ( Fleischman, Hopstock, Pelczar, & Shelley, 2010). When the National Asses sment of Educational Progress (NAEP ) assessed the scientific knowledge of U.S. elementary and secondary students , 65% of eighth graders performed at or above basic level, 32% at or above proficient, and 2% at or above advanced in 2011 ( National Center fo r Education Statistics, 2012) When compared with results obtained in 2009, the percentage of students at or above basic and proficient levels increased by 2 points, although there was no significant change in the percentage of students performing at the a dvanced level ( National Center for Education Statistics, 2012).

PAGE 35

35 Additionally, an examination of student results on the ACT indicate s that t he average ACT score in science has remained relatively stable since 2006 ( ACT, 2010). Furthermore, only 29% of ACT tested high school graduates in 2010 met the College Readiness Benchmarks score in science. Since 2006, the percentage of students who met the College Readiness Benchmarks in Science remained relatively stable with a 2% increase in 2010 from 2006 ( ACT, 201 0 ). W hen the results of all these assessments are collectively examined, one can readily identify the consistent underperformance of U.S. students in science A lthough there are indicators of improvement significant inroads need to be made in improving st udent achievement nationally. An examination of Florida's results on the NAEP reveals that although average 8th grade science scores improved by 3 points from 2009 to 2011, the average score was below the national average by 3 points (National Center for Education Statistics, 2012). Furthermore, 72% of Florida's students were categorized as below basic or basic in science proficiency, with only 1% scoring at the advanced level (National Center for Education Statistics, 2012). Thus, the statistic s paint a gloomy picture of student preparation for collegiate and science careers as well as overall These statistics also allude to the increased need to identify the student character istics that play a significant role in affecting student achievement in science Although the aforementioned assessments were not tied to school accountability indicators, they nevertheless suggest that a significant proportion of Florida's students may n ot achieve the passing score required for graduation assuming the student

PAGE 36

36 performance on Florida's B iology end of c ourse is relatively similar Thus it is essential to identify those students at risk of failing to meet Florida Department of Education's accepted standard of achievement in b iology Therefore, although student achievement by increasing the rigor, skills, and core content knowledge by the implementation of increasing number of standards and as sessments t his assertion is, based on the research, highly improbable given the confluence of variables that impact on student achievement. However, since more changes can be made to teaching rather than altering students' personal characteristics or fami ly circumstances; educational policy discussions will continue to remain focused on what is within a school's control. Therefore an investigation of the student characteristics that influence b iology achievement is a timely one. It is essential that educat ion personnel are aware of the role of select student characteristics, such as reading proficiency, ethnicity, socioeconomic status, and gender; that research has consistently shown to influence student achievement, and their impact o n student achievement on the B iology I end of course assessment. Statement of the Problem Due to the increasing high stakes associated with student performance on the B iology I EOC assessments, schools need to be able to predict and determine the influence of student character istics on student performance prior to the administration of the test. Although a significant number of studies in science, technology, engineering, and math ( STEM ) have produced viable answers, there is a paucity of research specifically on the predictors and differences in student achievement in b iology specifically T hus stakeholders are limited when making decisions about biology

PAGE 37

37 education By exploring four variables reading proficiency ethnicity, socioeconomic status, and gender the researcher can address the gap in extant literature and stakeholder s will be able to potentially make instructional changes that can improve student performance on the assessment and reduce achievement gaps. Such changes could have a profound impact on the success of in dividual schools. S tudent outcomes affect a school's grade which determines whether it is considered high performing and receives additional funding by the federal government If it is labeled underperforming it could eventually be closed or taken over. t As the B iology e nd of c ourse assessments are tied to graduation sanctions, the graduation rate w ould also be affected by changes in biology achievement. Graduation rate is itself another accountability indicator which has a significant impact on schools. Moreover, the performance of students also serve s as the primary basis for teacher evaluations (The Florida Legislature, 2010) and thus ha s significant ramifications on the recruitment, retention, and morale of highly qualified and skilled teachers, espec ially those teaching in high poverty schools. Research Questions 1. Is there a significant difference in student performance on the e nd of c ourse assessment in b iology associated with reading proficiency ? After adjusting for student gender, race/ethnicity, an d socioeconomic status, is there a significant difference in student performance on the e nd of c ourse assessment in b iology associated with reading proficiency ? 2. Is there a significant difference in student performance on the e nd of c ourse assessment in b i ology associated with race/ethnicity ? After adjusting for reading proficiency student gender, and socioeconomic status, is there a significant difference in student performance on the e nd of c ourse assessment in b iology associated with race/ethnicity ? 3. Is there a significant difference in student performance on the e nd of c ourse assessment in b iology associated with socioeconomic status? After adjusting for reading proficiency race/ethnicity, and student gender is there a significant

PAGE 38

38 difference in student performance on the e nd of c ourse assessment in b iology associated with socioeconomic status ? 4. Is there a significant difference in student performance on the e nd of c ourse assessment in b iology associated with gender ? After adjusting for reading proficienc y race/ethnicity, and socioeconomic status is there a significant difference in student performance on the e nd of c ourse assessment in b iology associated with gender ? 5. What are the best predictors of student performance on the b iology e nd of c ourse asses sment? Statement of Hypothesis It was hypothesized that there will be a significant difference in student performance on the e nd of c ourse assessment in b iology associated with reading proficiency It is hypothesized that students who are more proficient r eaders will score significantly higher than their less proficient peers on the Biology I EOC assessment Furthermore, more proficient students will score significantly higher that their less proficient peers on the Biology I EOC assessment after adjusting for student gender, race/ethnicity, and socioeconomic status. It was hypothesized that there will be a significant difference between the race/ethnicity of a student and their performance on the e nd of c ourse assessment in b iology. It is hypothesized that students from Caucasian backgrounds will score significantly higher than those from other ethnic backgrounds. It is also hypothesized that students from Caucasian backgrounds will score significantly higher than those from other ethnic backgrounds after ad justing for reading proficiency gender, and socioeconomic background. It was hypothesized that there will be a significant difference in student performance on the e nd of c ours e assessment in b iology associated with socioeconomic status. It is hypothesize d that students from higher socioeconomic

PAGE 39

39 backgrounds will score significantly higher than those from lower socioeconomic backgrounds. It is also hypothesized that students from higher socioeconomic backgrounds will score significantly higher than those fr om lower socioeconomic backgrounds after adjusting for student gender, race/ethnicity, and reading proficiency It was hypothesized that there will be a significant difference in student performance on the e nd of c ourse assessment in b iology associated wit h gender. It is hypothesized that males will score significantly higher than females o n the Biology I e nd of c ourse assessment. It is hypothesized that males will score significantly higher on the e nd of c ourse assessment in b io logy after adjusting for rea ding proficiency race/ethnicity, and socioeconomic background. It was hypothesized that reading proficiency ethnicity, and socioeconomic status will serve as significant predictors of student performance on the B iology e nd of c ourse assessments. Signific ance of the Study The purpose of this study is two fold. First, it seeks to determine how student performance is influenced by four variables -reading proficiency ethnicity, socioeconomic status, and gender -and examines their impact on the results of the Florida Biology I EOC assessment at a Title 1 school. The study is intended as an initial source of feedback to instructional leaders, teachers and administrators regarding the status of presumed achievement gaps based on student characteristics and it m ay prompt further discussion on institutional practices to alleviate those gaps. Second, it aims to predict student performance in b iology. Thus, it aims to address the gap s in empirical studies regarding the student characteristics that impact b iology ach ievement.

PAGE 40

40 T h findings will contribute to research on student achievement in b iology and the effect of e nd o f c ourse asse ssments on graduation rates, school accountability indicators professional development, and teacher evaluations Previous re search findings on many of these variables have been limited or nonexistent, and therefore this study can contribute to the existing literature. These findings may also be useful within the state of Florida and other states that have implemented or plan on implementing e nd of c ourse science assessments instead of minimum competency based graduation tests. T he majority of existing literature on student achievement in b iology has focused on a single predictor. T his study is unique in examin ing four variables simultaneously each of which has been individually identified as very influential on student achievement in science B y including these four variables simultaneously, the model is capable of determining the ir cumulative effect on student achievement rathe r than focusing on and overemphasiz ing one variable. T his study will analyze individual student scores while simultaneously controlling for other variables through adjusting. Undertaking this task is important as influential variables due to diverse stude nt backgrounds (rea ding proficiency, ethnicity, socioeconomic status and gender) intersect with each other. Thus the results of the study can be utilized as a source of insight in designing a comprehensive plan to address the achievement gap in science at the school site level. Unlike other assessments of student performance in science such as the National Assessment of Educational Progress (NAEP), Program for International Student Assessment (PISA), the Third International Mathematics and Science Study

PAGE 41

41 (T IMMS), and FCAT Science the Florida Biology I end of c ourse assessment serves as the basis of not only school accountability, but also that of student s and teacher s The previously adopted FCAT Science assessments were solely utilized to evaluate schools T he se tests are not utilized by teachers to assign course grades and scores are not recorded on high school transcripts, so there is no motivation or incentive on the student's part to do well on such tests ( unless they are concerned about the statewide p restige of their school and potential funding opportunities from governmental agencies ) (Stansfield, 2011). However, this does not invalidate the usefulness of the other aforementioned assessments in providing data that clearly indicates discrepancies in student performance as a factor of ethnicity, socioeconomic standing, and gender at the state, national, and international level s Thus, using data derived from the Florida Biology EOC assessment will serve as a useful tool for comparison with previous nat ional and international findings to determine whether similar results will be replicated. Other studies have not made use of the state's reading assessment (FCAT Reading and FCAT 2.0 Reading) as the measure of reading proficiency. Thus for administrators a nd teachers within the state of Florida, this data is more familiar and readily accessible. Depending on the outcome, this source of data can become more significant as a potential means of making initial decisions regarding student achievement in b iology. Furthermore, the future of educational policy decisions by site administrators and district and state personnel may be potentially affected by outcomes of the study.

PAGE 42

42 Finally, these findings have greater implications for the future of this nation a nd its ability to educate all students in a challenging manner that enables each student to obtain the knowledge and sk ills needed to become scientifically literate. The results can serve as one of many indicators to gauge trends in scientific proficiency, collegiate preparation, and the status of achievement gaps based on ethnicity, socioeconomic status, and gender. Thus, the results of this study can be used to determine instructional deficiencies, instructional practices, or systemic issues that contribu t e to an achievement gap so that stakeholders can potentially rectify such issues. Educators cannot afford to be cavalier about what student achievement data reveals about the degree of preparedness of our students for a more advanced technological society It has dire consequences for our ability to compete globally in the future and influences the degree to which an informed citizenry can contribute to the democratic process There are additional implications regarding national immigration and security polic ies Limitations One of the most important limitations is part of the nature of a causal comparative study. C ausal comparative stud ies aim to determine the cause for existing differences in the behavior or status of groups or individuals (Gay Mills, & Airasian, 2009). Because the cause is preexisting, the researcher has no control over the conditions of this ex post facto study and cannot draw cause effect conclusions with any degree of confidence due to the lack of randomiz ation and manipula tion Thus, the interpretation of findings in this study requires considerable caution. Additionally, the use of convenience sampling means that the sample size is limited and not representative of either the demographic composition or the academic proficiency of students in the state of Florida. Furthermore, due to the use of

PAGE 43

43 data obtained from a Title 1 school, the effect of school context on student achievement is not addressed by the study.

PAGE 44

44 CHAPTER 2 REVIEW OF THE LITERAT URE A study on variable s that impact student achi evem ent on the Florida Biology end o f c ourse assessment would not be complete without a review of related literature. Thus the following discussion will present an overview of research related to this study. The following subsecti ons that comprise this chapter include: (a) Standardized Testing, (b ) The Imp ortance of Science Education, (c ) Rea ding and Science Achievement, ( d ) Ethni city and Science Achievement, (e ) Socioeconomic Status and Science Achievement, and (f ) Gender and Scie nce Achievement. For purposes of this study, the literature review subsection on ethnicity will focus primarily on African American, Hispanic, and White populations as the researcher has elected to discuss the major demographic groups that are representati ve of the school under investigation. Standardized Testing According to Haladyna, Nolen and Haas (1991), standardized tests are universally and uncritically accepted by the public and many educators as a valid increased perceived need to evaluate education at virtually all units of analysis (i.e., individuals, classes, schools, school because tangible consequences depend o n them and they are used to: rank schools and their districts to bemoan the failure of education in newspapers, determine merit pay and make other personnel decisions by school district personnel, used to rate l estate agents, and rank states and their effectiveness on state and national levels; to name a few (Haladyna, Nolen & Haas, 1991 ; Madaus & Russell, 2010 ; Wiliam, 2010 ). Thus, these high stakes tests are

PAGE 45

45 probably going to remain a thing of the future as l ong as test scores remain the single most important index of educational effectiveness (Haladyna, Nolen & Haas, 1991). Although these high stakes assessments provide an insight into the academic achievement of students nationally and at the state level, no single standardized achievement test represents a complete mapping of the content of school achievement domain (Haladyna, Nolen & Haas, 1991 ; Madaus & Russell, 2010 ; Wiliam, 2010 ). Often test results are used without considering the complexity of achievem ent and its causes ( Madaus & Russell, 2010) For example, attributing the level of achievement based on test scores to the influence of a single teacher, school, or school district erroneously oversimplifies the nature of the scores, thus contributing to t he loathing of teachers and administrators toward standardized testing (Haladyna, Nolen & Haas, 1991). Furthermore, student achievement is a function of a variety of factors, of which only some are under the influence of schools (Haladyna, Nolen & Haas, 19 91 ; Madaus & Russell, 2010; Qiu & Wu, 2011 ). Schools can influence the quality and quantity of instruction, motivation, and the learning environment but they have little to no effect on family and home environment, maturity, and mental ability (Haladyna, N olen & Haas, 1991). Test P ollution Typically in response to the high stakes nature of standardized testing, test score pollution occurs. Test score pollution refers to factors affecting the truthfulness of a test score interpretation and is very pervasive in America (Haladyna, Nolen & Haas, 1991). The main sources of test pollution are: the way schools and its personnel prepare students for tests, test administration activities or conditions, and exogenous factors that are beyond the control of schools and school personnel (Haladyna, Nolen & Haas,

PAGE 46

46 1991). Polluting practices include: teaching test taking skills, preparing teaching objectives and developing a curriculum to match the test, using commercial materials specifically designed to improve test perfor mance, presenting test items before the test, and interfering with response (e.g. giving hints or answers to students or altering motivation which all impede student p erformance. In regards to exogenous factors, the reporting of test scores without acknowledging the influence of the family, family mobility, economic environment, proficiency with the English language, and other such factors results in the depiction of i nvalid inferences from the test scores (Haladyna, Nolen & Haas, 1991 ; Madaus & Russell, 2010 ). In fact, p ast research has documented the significant impact of student family background on test scores, in which single parent backgrounds had a negative effec t ( Bankston III & Caldas, 1998; Pong, Dronkers, & Hampden Thompson, 2003; Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiern an, 1997; Qiu & Wu, 2011; Rathbun & West, 2004 ), and increasing income and education levels were associated with positive effects ( Bankston III & Caldas, 1998; Education Information and Accountability Services, 2009; Liu & Lu, 2008; Qiu & Wu, 2011; Rathbun & West, 2004; Woods, Kurtz Costes, & Rowley, 2005 ). Therefore, the socioeconomic status of the student affects educational oppor tunities and accessibility to a variety of resources while simultaneously the home environment affects the degree of student aspiration in learning and interest in school ( Qiu & Wu, 2011). Research has also indicated that the validity of standardized test s is always a source of concern as although they are carefully designed to measure predetermined

PAGE 47

47 constructs, the readabi lity of test items is often considered as a threat to the validity of the test ( American Educational Research Association, American Psyc hological Association, National Council on Measurement in Education, 1999). Therefore, according to the Standards for Educational and Psychological Testing (American Educational Research Association, et al., 1999), the possibility that the test is assessin g reading proficiency and reading comprehension in addition to its intended concepts, is increased if test items are not examined for readability and comprehension to the test takers. It is therefore recommended that "in testing applications where the leve l of linguistic reading ability is not part of the construct of interest, the linguistic or reading demands of the test should be kept to the minimum necessary for valid assessment of the intended construct" (American Educational Research Association et al ., 1999, p. 82). Thus, the length of sentences, selected vocabulary, and the direct nature of questioning should be relatively easy in communicating the construct so that student knowledge of the construct, not reading proficiency, is assessed (Visone, 200 9). Florida Biology I End Of Course (EOC) Assessment According to Goodwin, Englert, and Cicchinelli (2003), the original purpose for outcome oriented accountability systems is based on the assumption of improving schools by increasing academic performance for all students. This premise serves as the basis for the i mplementation of the Biology I end of c ourse (EOC) assessment as part of Florida's Next Generation Strategic Plan (Florida Department of Education, 2005a). This Next Generation Strategic Plan oper ates under the vision that "Florida will have an efficient world class education system that engages and prepares all students to be globally competitive for college and careers" (Florida Department of Education, 2005b) and which would be measured by EOC a ssessments. According to the Florida

PAGE 48

48 Legislature (2012), the Florida Statutes state that, "the primary purpose of the student assessment program are to provide information needed to improve the public schools by enhancing the learning gains of all students and to inform parents of the educational progress of their public school children." The program was designed to satisfy six conditions: 1) to assess the learning aims of each student annually in achieving the Sunshine State Standards appropriate to the g rade level of the student, 2) to provide data for school accountability and recognition so that schools can make decisions, 3) to determine the educational strengths and needs of students as well as their readiness to be promoted to the next grade level or to graduate from high school with either a standard or special high school diploma, 4) to determine how well educational goals and curricular standards are achieved at the school, district, and state level, 5) to provide information that can be used in th e evaluation and development of educational programs and policies, and 6) to determine student performance in Florida against that of other students nationally (The Florida Legislature, 2012). One such assessment, the Biology I end of c ourse assessment; w as administered for the first time in Ma y 2012 to students enrolled in b iology I or an equivalent course to determine student proficiency with the Next Generation Sunshine State Standards (NGSSS) and expectations for student learning outlined in the course descriptions (Florida Department of Education, 2005a). Student performance during this academic year on the Biology I EOC would constitute 30% of the student's final course grade (The Florida Legislature, 2012). In addition to this and other student perfo rmance indicators, student performance contributed to the school's overall grade, and served as

PAGE 49

49 the chief criteria for b iology teacher evaluations (The Florida Legislature, 2010; 2012). For the first administration of the assessment, student achievement wa s categorized in achievement levels represented by thirds based on comparisons with other students in Florida that range from 1 (lowest third) to 3 (highest third) (Florida Department of Education, 2012i). During the 2012 2013 school year, all students en tering grade 9 would be required to earn a passing score on the EOC assessment in b iology to earn course credit (The Florida Legislature, 2012). This passing score indicated by obtaining a minimum scale score in Achievement Level 3, will be used to fulfill one of the requirements for high school graduation ( Florida Department of Education 2012f). Future test administrations starting during the 2012 2013 academic year will also have a similar effect on the overall school grade and teacher evaluations (The F lorida Legislature, 2012). Thus this new current policy of science accountability will not only present a new set of challenges: one of which is due to the 9th grade reading level in which the test is written and its incorporation of short reading passages (Florida Department of Education, 2011c), and the historical differing levels of achievement as an outcome of the diversity of the student body; but also opportunities in science education. One such opportunity is the reduction in the achievement gap bet ween subgroups of the student population (Penfield & Lee, 2010). Additionally, test based accountability in science forces schools, districts, and states to ensure that the educational needs of all students, especially those in historically low performing subgroups; are addressed (Penfield & Lee, 2010). Thus, this commonly overlooked

PAGE 50

50 subset of the school population will be able to gain access to educational opportunities and the associated potential for careers in science, thus reducing inequalities already present within the educational sector. Last, science will be integrated within the school curriculum after traditionally being ignored in order to address the urgency associated with basic literacy and numeracy ( Doppen, 2007; Penfield & Lee, 2010 ; Winstea d, 2011 ). Thus, additional resources and instructional time for science may now be potentially allocated by districts and schools (Penfield & Lee, 2010). Results of first a dministration of the Biology I EOC As previously discussed, student achievement leve ls for the May 2012 assessment administration was categorized in thirds in which student T scores were compared with other students in Florida (Florida Department of Education, 2012i). Scores ranged from 1 or lowest third to 3 or highest third. Of the 190, 344 students who took the Biology EOC assessment in May 2012, 35% scored in the top third while 31% scored in the middle third and 34% the lowest third (Florida Department of Education, 2012a). Using scores represented on the T scale ranging from 20 80, th e statewide mean was approximately 49 (Florida Department of Education, 2012a). Results of the assessment revealed that only 59% of students would have obtained a passing score on the assessment, of which 11% would score at mastery level 5, 11% at the abo ve satisfactory level 4, 37% at the satisfactory level 3, 27% at the below satisfactory level 2, and 14% at the inadequate level of 1 (Florida Department of Education, 2012m). Furthermore, as student grade level increased (beginning with the 7th grade and ending with 12th grade), the mean scale score for each grade level decreased. Thus 7th grade students had the highest mean scale score of 60 whereas their 11th and 12th grade counterparts obtained a mean scale score of 43 (Florida

PAGE 51

51 Department of Education, 2012a). Additionally, as the student grade level increased, the percentage of students scoring in the top third decreased significantly (Florida Department of Education, 2012a). For example, at least 50% of students enrolled in grades 7,8, and 9 scored in the top third whereas less than 25% of 10th, 11th, and 12th grade students earned a comparable score (Florida Department of Education, 2012a). Furthermore, students enrolled in higher grade levels were disproportionately represented in the lowest third tha n their peers enrolled in lower grades (Florida Department of Education, 2012a). When the test results were further analyzed, certain patterns became apparent regarding ethnicity, socioeconomic status, and gender. According to Florida Department of Educat ion (2012b), a greater percentage of Asian and White students scored in the top third than Black and Hispanic/Latino students. In fact, 58% of Asian and 45% of White students scored in the top third compared to 17% of Black students and 29% of Hispanic/Lat ino. Black and Hispanic students were also disproportionately represented in the lowest third. 53% of Black and 39% of Hispanic students comprised the lowest third compared with 23% of White students (Florida Department of Education, 2012b). When data was disaggregated regarding socioeconomic status, other trends surfaced. For each grade level, higher percentage of students who were ineligible for free or reduced lunch scored in the top third than eligible students (Florida Department of Education, 2012b). In fact, 23% of those eligible for free and reduced price school lunch scored in the top third compared to 48% of those ineligible. Additionally, 45% of those eligible for free and reduced price lunch scored in the lowest third compared to 22% of those in eligible (Florida Department of Education, 2012b).

PAGE 52

52 In terms of gender, simple patterns as well as the intersection of gender and ethnicity surfaced. For each grade level, higher percentage of males scored in the top third than females (Florida Department of Education, 2012b). When ethnicity was analyzed with gender, data revealed that only 16% of Black males compared to 47% of W hite and 31% of Hispanic/Latino males scored in the top third Of those scoring at the lowest third, the majority were Black, spec ifically 55% of Black males compared to 39% of Hispanic/Latino, and 23% of W hite males (Florida Department of Educatio n, 2012b). For females, 43% of W hite, 28% of Hispanic/Latino, and 18% of Black females scored in the top third compared to 23 % W hite, 39% Hispanic, and 51% Black females scoring in the lowest third (Florida Department of Education, 2012b). Other s cience d istrict administration r esults When the past academic performance for the North Florida school under investigation was examined during the 2010 2011school year, FCAT Science scores from the 137 11th graders, showed that the average mean scale score was 278, which is below the state average of 307 (Florida Department of Education, 2011a). While only 26% of students scored at proficiency or hi gher compared to the state average of 40% at the North Florida school (Florida Department of Education, 2011a), student data indicates an increase in student performance by 9% from the previous 2010 spring administration (Florida Department of Education, 2 010). This performance is not particularly surprising when the academic performance of the lower grades that feed into the North Florida high school is examined. According to Florida Department of Education (2012d), only 22% of students scored at proficien cy or above compared to 46% of eighth grade students statewide during the 2012 academic year administration of the 8th grade science FCAT 2.0.Results from the fifth

PAGE 53

53 grade assessment also revealed that 26% of students scored at proficiency or above, compare d to 51% statewide (Florida Department of Education, 2012c). Thus the future students in the North Florida high school have underperformed in science in general when compared with the state and have also underperformed in the domains of the foundations of science and biological content (Florida Department of Education, 2012c, 2012d). Thus at the 5th and 8th grade level, students already exhibit a deficit in content area knowledge that serve as the background knowledge needed for more advanced understanding of biological content. What is concerning is that it is these students that will be t ested upon their enrollment in b iology I or an equivalent course at the secondary level, and their deficits in lower grades will serve as a detriment to their performance Statewide, student performance in science as measured by the FCAT Science in grades 5, 8, and 11 has improved in the state of Florida between 2003 and 2011 (Florida Department of Education, 2011). The percent of students scoring at proficiency and above levels at the fifth grade increased from 28% in 2003 to 51% in 2011, at the eighth grade level from 28% in 2003 to 46% in 2011, and from 33% in 2005 to 40% in 2011 for eleventh graders (Florida Department of Education, 2011b). Despite improvements on this indicator, another indicator of student achievement in science fails to illuminate such improvements. According to the National Center for Education Statistics (2012), the 2011 data on 8th grade science performance indicates that 72% of Florida students sc ored at or below basic. Furthermore, the average score of Floridian 8th graders did not change significantly between 2009 and 2011 and remained below the national average, resulting in Floridian students scoring below 37 other states

PAGE 54

54 (National Center for E ducation Statistics, 2012). Therefore these indicators of student performance signify that nationally, and particularly in the state of Florida, that students are grossly unprepared for future science coursework or more advanced technological careers that will become increasingly commonplace in the future. Computerized Testing One form of test pollution that has been documented by empirical research which ma y skew the accuracy of Biology end of c ourse assessments is the use of computerized testing. To addr ess stakeholder concerns regarding computer based assessments, Florida Department of Education (2006) published a paper on the empirical comparability of paper and computer administered assessments. According to Florida Department of Education (2006), tes t scores derived from computer administered and paper administered tests are comparable. Although the state agency cited research that indicates that some students perform better on a computer administered test and others on its paper based counterpart, th e inconsistency was explained by highlighting the dissimilar computerized test administration systems (Florida Department of Education, 2006). Comparability of computer based t esting The state agency then further cited previous research studies, the major ity (seven) of which indicate the comparability of test scores in b iology attained via computer and paper based administration, one citing computers as the more difficult administration mode (Cerillo & Davis, 2004), and three citing paper based as more dif ficult (Russell, 1999; Russell & Haney, 1997, 2000). Pomplum, Ritchie, and Custer (2006) revealed that students eligible for free lunch performed better on paper and pencil tests. However, the discrepancy in scores between paper and pencil tests and

PAGE 55

55 comput erized tests decreased as student grade level increased. Pomplum, Ritchie, and Cruster (2006) explained that this reduction in the score discrepancy may be due to increased student familiarity with computers at each grade level. Despite this result, this s tudy has limited applicability to the Biology 1 end of c ourse assessment as the participants were enrolled in the lower grades at the elementary level where reading proficiency serves as a confounding variable due to the age of the participants. Russell ( 1999) and Russell and Haney (2000) however, focused on computer based administration of open ended test questions; a format that is not utilized in the B iology end of c ourse assessment; thus both studies are not applicable. On the other hand, Russell and H aney's 1997 study focused on the comparability of computer and paper and pencil test administration on multiple choice test items and revealed that the mode of administration is inconsequential. Mead and Drasgow's (1993) meta analysis of 29 studies also fo und no significant difference between paper based and computer based test administration of multiple choice tests, particularly for timed pow er tests such as the Biology I end of c ourse assessments. More recent studies however have indicated the equivalen ce in student performance on online and paper based tests (Escudier, Newton, Cox, Reynolds, & Odell, 2011; Frein, 2011; Horne, 2007; Kingston, 2009). Research has also indicated that some students have performed better in the online medium, completed more questions, and the majority of students expressed a preference for the online medium (Escudier, Newton, Cox, Reynolds, & Odell, 2011; Frein, 2011; Horne, 2007). According to Kingston (2009), the discrepancies in studies on the comparability of computer bas ed and pencil and paper tests may be partially due to the changes in the quality of

PAGE 56

56 computer based test administration systems as well as the computer experience of students. For example, older computer systems required that students scroll up and down an d from side to side to read an entire passage or large item which made the computer platform less user friendly and more difficult. Newer systems however, have minimized scrolling of stimuli to a single dimension at most and the majority of newer systems a void the scrolling of items (Kingston, 2009). According to Kingston (2009), reading from a page is different than reading while scrolling as while reading a page, spatial memory clues are used in which students remember seeing information regarding a parti cular question in a specific location on a page. Therefore, the student can quickly return to that spot to find the information. However, in a traditional computer administered system, these parallel clues are unavailable as the spatial frame of reference is changed while scrolling. Thus some newer systems enable students to highlight text as they read or provide other cues that reduce score differences due to task differences (Kingston, 2009). Other empirical evidence on computerized testing Other factors that were examined empirically in the Florida Department of Education (2006) publication include: student preference for computer based testing, inconclusive evidence regarding student computer experience and its effect on academic performance, no gender based comparability differences, and text presentation and slow response times due to internet connection speed or school network constraints has a negative effect on comparability (Florida Department of Education, 2006). However, caution must be employed when examining student results from the empirical analysis that was used as the basis of the decision making process

PAGE 57

57 employed by the Florida Department of Education (2006) because a paucity of empirical research exists specifically regarding b iology multip le choice test items, some of which contain long passages, that are timed and non adaptive. Most research regarding computer based assessments fail to focus on the specifics regarding this type of assessment although the t est is considered high stakes. Co mputer p latform for the Biology EOC In fact, the Biology I end of c ourse assessment provides the highlighting function for students in addition to other tools such as: a reset button that removes the selected answer, a review button that enables the st udent to flag a question for review at a later time, a go to button that allows the student to access the question marked for review on the item review screen, an eliminate choice tool that enables the student to cross out answer choices deemed incorre ct, an eraser tool to remove highlighting or remove an X from an eliminated answer choice made with the eliminate choice tool, a calculator, and a help button (Pearson Education, 2013). To further ensure that students are not negatively impacted from the computerized platform, students are also provided four page, hard copy work folders as scratch paper (Florida Department of Education, 2012i). Students also can utilize the practice test available online to gain familiarity with the software (Florida Depar tment of Education, 2012i). For students at a disadvantage, paper based testing is available for students with disabilities based on their Individual Educational Plans (IEPs) or Section 504 plans (Florida Department of Education, 2012i). Thus, concessions have been made by the Department of Education to ensure that scores are representative of student proficiency by: allowing students who need the alternative option are pr ovided with paper based testing; and that those that are not only familiar with the so ftware, an

PAGE 58

58 option to gain familiarity with it, while ensuring that the software utilized is user friendly and contains tools that minimize the change in testing format. The Importance of Science Education Second only to a weapon of mass destruction d etona ting in an American city, we can think of nothing more dangerous tha n a failure to manage properly science, technology, and education for the common good over the next quarter century (The United States Commission on National Security/21 st Century, 2001a, p. 30). The aforementioned quote from the United States Commission on National Security/21st Century underscores the Commission's view of education as critical to ensuring the nation's security and its future. That perspective has become increasingly promi nent due to the September 11, 2001 attacks in which the strategic role of science and technology in the post Cold War era has changed (National Science Board, 2004). Of concern has been the role of foreign students, scientists, and engineers within the Uni ted States science and engineering system; determining the appropriate degree of balance between security and openness within scientific communication; the degree of contributions that research and development are capable of making in domestic security; an d the course of federal research and development initiatives (National Science Board, 2004). Thus, in order to sufficiently meet the scale and nature of scientific and technological advances in society, the need for human capital is critical (The United St ates Commission on National Security/21 st Century, 2001b; Trefil, 2008). Furthermore, the ability of Americans to compete globally is highly dependent upon their knowledge of science and math and their ability to apply this knowledge to emerging technologi es (U.S. Department of Education, 1989). Thus, to be competent, citizens and workers require a sound understanding of science and mathematics (National Science

PAGE 59

59 Board, 2004; Trefil, 2007; 2008). Science and mathematics are considered essential to the decisi ons made regarding jobs, use of resources, health, and everyday activities of consumers (U.S. Department of Education, 1989). Furthermore, the products, services, standard of living, and economic and military security needed to sustain Americans at home an d globally will be derived from mathematics and the sciences (National Commission on Mathematics and Science Teaching for the 21st Century, 2000). Consequently, in an integrated and global economy, math and science will serve as the core forms of knowledge that are required by innovators, producers, and workers to solve unforeseen problems and determine America's future (National Commission on Mathematics and Science Teaching for the 21st Century, 2000; Trefil, 2007). Reasons C ited for S hortage in S cience P ersonnel However, despite this need for the highest quality human capital in science, mathematics, and engineering by the nation, it is currently not being met by the educational system (The United States Commission on National Security/21 st Century, 2001b ) This system operates under the notion that the elementary and secondary schools are ensuring that students acquire knowledge in science and mathematics (National Science Board, 2004). One reason cited as contributing to this problem of an unprepared cit izenry is the view that the American K 12 educational system fails to prepare students as it should for college or the commercial sector system (The United States Commission on National Security/21 st Century, 2001b) due to the minimization of science and m athematics instruction, particularly in the early grades, and the lack of qualified personnel in these content areas in teaching positions (U.S. Department of Education, 1989).

PAGE 60

60 Additionally, the lack of importance placed on education, according to Christe nsen, Horn, and Johnson (2011), can be partially understood when one understands the effect of prosperity on the nation. The authors explained that the and prosper ity (Christensen, Horn, & Johnson, 2011). For students residing in a developing country that is becoming an increasingly industrial based economy, studying subjects such as science, math, and engineering offer rewards in which the individual can escape fro m poverty. The freedom that results from living in a prosperous nation results in students having greater freedom to study subjects they find fun and intrinsically motivating. This proposed explanation appears fitting based on results of surveys that indic ate that positive student attitudes toward science (Gibson and Chase, 2002; U.S. Department of Education, 1989) and mathematics decline as students advance in coursework (U.S. Department of Education, 1989). Additionally, most parents and students subscrib e to the notion regarding the unimportance of science and mathematics for the majority of students and that high achievement in these content areas are due to factors other than effort and hard work (U.S. Department of Education, 1989). Catalyst for the I mprovement of Science E ducation This may, in part, explain why the percent of students scoring at proficiency levels in science decreases from the 4rth to the 12th grade ( National Science Foundation, 2011) and why foreign students account for 57% of all do ctorates in engineering, 54% in computer science, and 51% in physics (National Science Foundation, 2012a). Moreover, of the 5 million degrees earned in science and engineering worldwide, 23% of those were obtained by students in China, 19% from

PAGE 61

61 those in th e European Union, and only 10% within the United States (National Science Foundation, 2012a). China, in fact, has awarded the most number of doctoral degrees in natural sciences and engineering since 2007, thus overtaking the United States as the world lea der (National Science Foundation, 2012a). Consequently the dissatisfaction of parents, policymakers, legislatures, and educators regarding mathematics and science underachievement, has resulted in numerous efforts to both reform and improve schools (Nation al Science Board, 2004). National effort had been galvanized since the appearance of Sputnik to improve mathematics and science education (National Commission on Mathematics and Science Teaching for the 21st Century, 2000). The publication of A Nation At R isk urged greater accountability of schools, as well as higher academic standards, and better teacher preparation as a means of improving schools (National Commission on Excellence in Education, 1983). Other reports and commissions thereafter established l ofty goals, such as stating that by the year 2000, U.S. students would rank "first in the world in science and mathematics achievement" by strengthening math and science education in early grades, increasing the number of math and science teachers with sub stantive backgrounds in those fields by 50%, and significantly increasing the number of U.S. undergraduate and graduate students with degrees in math, science, and engineering; especially women and minorities (U.S. Department of Education, 1989). In respon se to this announcement, the National Commission on Mathematics and Science Teaching for the 21st Century announced in 2000 that U.S. students were "devastatingly far from this goal by the time they finish high school" (p. 10).

PAGE 62

62 According to American Assoc iation for the Advancement of Science (AAAS) Project 2061 (1990), two disparate and growing public concerns have motivated the majority of education reports in the 1980s. One such concern is the seeming decline of America's economy. When the domestic afflu ence and international power, as evident in scientific and technological domination of America is compared with other countries, particularly Japan; they appear to be weakening. The other concern is based on trends in U.S. public education that include low test scores, the avoidance of mathematics and science by students, a weakened teaching staff characterized by demoralization, a comparatively low learning expectation with other technologically advanced nations, and the ranking of U S students near the b ottom in mathematics and science on international studies. Both the reports and coverage by the mass media of such reports have highlighted the deficiencies in education, and thus the nation became aware of the crisis in American education. However, the mo st powerful argument to improve science education is its role in "liberating the human intellect" although much of the public discussion has been focused on more tangible, utilitarian, and immediate explanations (American Association for the Advancement of Science (AAAS) Project 2061, 1990). School reform thereafter became increasingly strengthened under the Federal No Child Left Behind (NCLB) Act of 2011 in which school accountability measures were implemented by requiring that schools demonstrate progress in student achievement in core subject areas using high stakes testing to measure learning. The act specified to each state, the immediate development of standards in math and science by the academic year 2005. Furthermore, the NCLB Act mandates students in grades 3 through 8 to be assessed every year beginning in the 2005 academic year for math and

PAGE 63

63 2007 for science. Sanctions will be imposed on schools that fail to show improvement in student achievement for all students after they initially received assi stance (No Child Left Behind Act of 2001, 2002). Despite such attempts, U.S. students continued to perform poorly in both science and mathematics. In his attempt to address the underachievement of American students in math and science, President Obama inst ituted the first of two White House Science Fairs as a means of fulfilling his commitment to improve American student performance from the middle to the top of the pack over the next decade (The White House, Office of the Press Secretary, 2010) The Presid ent also indicated the importance of Science, Technology, Engineering, and Math (STEM) education in providing the new foundation that will determine America's future prosperity (The White House, Office of the Press Secretary, 2010) In order to achieve thi s goal, his administration allocated $ 4 billion to Race to the Top (RTT) initiatives in which states were empowered to develop a comprehensive strategy to improve student achievement in STEM subjects, especially for women and underrepresented minorities ( The White House, Office of the Press Secretary, 2010). This decision by the Obama administration to allocate fiscal resources to address the issues of student underachievement in science represents one of many attempts by the federal government to improve the science performance of American students (U.S. Department of Education, 2004). This underperformance of American students was significant enough to warrant a statement within the 2011 State of the Union address (The White House, Office of the Press Se cretary, 2011) and is viewed as limiting the nation's ability required for global economic leadership as well as homeland security in

PAGE 64

64 the 21st century (U.S. Department of Education, 2004). Thus, America's children are viewed as failing to adequately respon d to the challenges for the 21st century but its potential as well (National Commission on Mathematics and Science Teaching for the 21st Century, 2000). Reading and Science Achievement Of the many challenges within the education sector, the proficiency lev el of U.S. students in reading is a disconcerting one. Considering that reading is a requirement on assessments, o ne can reasonably assume that r eading proficiency can impact test scores. What remains unclear is whether test scores are affected by student inability to understand the question being asked or in the case of an incorrect response, whether the reason is due to lack of knowledge or an inability to comprehend the test item (Homan, Hewitt, & Linder, 1994). Importance of Reading Of the many indicato rs of reading proficiency at various grade levels, U.S. students have consistently engaged in paltry performance (ACT, 2010; College Board, 2011; Cromley, 2009; Fleischman, Hopstock, Pelczar, & Shelley, 2010; Florida Department of Education, 2012e; Nationa l Center for Education Statistics, 2011; Torgesen, 2006 ). Considering such dismal student performance on fundamental literacies, it is feasible to state that the majority of high school students and graduates are lacking the reading skills needed to succes sfully traverse the Information Age as well as navigate content area learning. Furthermore, it is argued that basic skills are no longer sufficient to successfully navigate the post secondary world (Meltzer, Cook Smith, &Clark, 2002) and consequently multi ple indicators of student performance such as the criticisms of employers, standardized test scores, standards based assessment

PAGE 65

65 results, and the 2011 NAEP (National Center for Education Statistics, 2011) paint a melancholy picture on the competencies of ou r students. Strong literacy skills are necessary for adolescents to navigate the world, communicate effectively, understand academic content, and partake in cultural communities (Meltzer, 2001). This concept was held by one of the greatest lyricist of Ame rican Democracy, Thomas Jefferson, who noted the direct relationship between literacy, citizenship, and successful self government. Good citizenship, in his opinion, was not capable without the knowledge and discernment that accompany literacy. In addition to participating in society, literacy enables the individual to access opportunities for economic vitality, and individual or personal fulfillment. Although various cultural and language minority groups have rich written and verbal literacy indicated in r esearch, the lack of literacy in the larger society fails to create equal opportunities of access to resources in that society and consequently to the privileges, rights, and responsibilities of American citizenship in a democratic society (Meltzer, Cook S mith, & Clark, 2002). Literacy Rates of American S tudents An examination of various assessments reveals that in general, literacy levels of all subsets of the school population is low, and particularly among certain subsets. For example, one test, the ACT defines "college readiness" as the probability of students earning a gr ade of "C" or higher at 75% or a 50% chance of earning a B or higher in first year college courses such as College Algebra; English Composition; History, Psychology, Sociology, Politi cal Science or Economics; and Biology (ACT, 2010). Of the 2010 ACT tested high school graduates, only 24% met all four ACT College Readiness Benchmarks thus only 24% of students were academically prepared for first year college courses such as College Alg ebra, English Composition, social sciences, and

PAGE 66

66 Biology (ACT, 2010). Of the students who completed the 2010 examination, only 62% of White, 61% of Asian, 34% of Hispanic, and 21% of African American high school graduates met College Readiness Benchmarks in Reading. Thus, the access to economic vitality obtained through obtaining a undergraduate degree is differentially available to students due to academic performance in reading and its effect on subsequent collegiate course grades. Other academic indicator s paint a similar picture of general paltry performance in reading and differential performance based on ethnicity, gender, and socioeconomic status. The SAT, which is a predictor of student performance in college; indicates that the average reading score has dropped since the 1986 1987 administration (U.S. Department of Education, National Center for Education Statistics, 2012). Since the 1986 administration, White students consistently perform higher than the average score while their Black and Hispanic c ounterparts score lower than the average (U.S. Department of Education, National Center for Education Statistics, 2012). Specifically, the average score of Black (428) and Hispanic (451) students was lower than the national average (497) on SAT critical re ading although their White (528) and Asian (517) peers scored higher (College Board, 2011). Furthermore, as family incomes increased, the average score on the critical reading section also increased (College Board, 2011). The poor academic performance of high school students however, is not unforeseen when one considers the performance of students at the elementary level. According to the National Center for Education Statistics (2011), since 2007, 67% of students were scoring at the basic level. When eigh th graders were assessed, the

PAGE 67

67 percent of students scoring at basic levels increased to 76% (National Center for Education Statistics, 2011). Therefore since 2007, student performance at the fourth grade level has remained relatively unchanged, thus the maj ority of the students continue to lack proficiency in reading. At the eighth grade level, even greater numbers of students are failing to demonstrate proficiency. Trends in Student L iteracy Despite this alarming trend, students from high and lower income families increased academic performance (National Center for Education Statistics, 2011), thus there is some improvements on academic gauges worth celebrating. When disaggregated, certain trends become readily apparent. Students who read more often score h igher than their peers who read less often (National Center for Education Statistics, 2011). Additionally, students who scored above the 75th percentile were predominantly White (71%) and read for fun almost every day (60%). For students who scored below t he 25th percentile, 38% read for fun almost every day and 74% were eligible for free/reduced lunch (National Center for Education Statistics, 2011). Another readily discernible trend is the difference in achievement obtained by students of different ethni cities. For example, the mean score of White fourth grade students (231) was higher than Black (205) and Hispanic (206) students in 2011 (National Center for Education Statistics, 2011). When the performance of fourth grade Black students was further exami ned, 84% of Black students scored below or at the basic level. When compared with their White counterparts, only 57% scored below or at the basic level (National Center for Education Statistics, 2011). Thus Black fourth grade students accounted for proport ionately higher levels of students scoring at or below basic levels. They also accounted for proportionately lower levels of students scoring at

PAGE 68

68 advanced levels when compared with their White counterpar ts. At the eighth grade level, similar results in perf ormance were also revealed (National Center for Education Statistics, 2011). Other discernible trends regarding socioeconomic status were also revealed. When socioeconomic status data was also analyzed at both grade levels, students eligible for free lunch scored significantly lower than those eligible for reduced price lunch and lower than those not eligible (National Center for Education Statistics, 2011). International comparison of student literacy When viewed internationally, U.S. 15 year old students obtained an average score of 500 on the combined reading literacy scale, higher than the 493 OECD average (Fleischman, Hopstock, Pelczar, & Shelley, 2010). Although this may indicate that students are performing at relatively proficient levels, when compar ed with other countries, 15 OECD and non OECD countries earned higher average scores. Furthermore, only 30% of U.S. students scored at or above proficiency in locating and organizing "several pieces of embedded information", "interpreting the meaning of nu ances in language", "understanding and applying categories in an unfamiliar context", and demonstrating "an accurate understanding of long or complex tests whose content or form may be unfamiliar" (Fleischman, Hopstock, Pelczar, & Shelley, 2010, p. 10). T rends in student literacy in Florida Specifically in the state of Florida, the State Board of Education established new Achievement Level standards for FCAT 2.0 Reading on December 19, 2011 (Florida Department of Education, 2012e). Florida Comprehensive As sessment (FCAT) 2.0 is considered to be more demanding in its content standards and more rigorous in its achievement standards than the former FCAT version of the Reading test (Florida

PAGE 69

69 Department of Education, 2012e). Consequently, the scores reported are different from previous scales for the 10th grade assessment. In response to the change, new Achievement Levels were developed as follows for all FCAT 2.0 and EOC assessments: Level 5: Students at this level demonstrate m astery of the most challenging cont ent of the Next Generation Sunshine State Standards. Level 4: Students at this level demonstrate an above satisfactory level of success with the challenging content of the Next Generation Sunshine State Standards. Level 3: Students at this level demonstr ate a sati sfactory level of success with the challenging content of the Next Generation Sunshine State Standards. Level 2: Students at this level demonstrate a below satisfactory level of success with the challenging content of the Next Generation Sunshin e State Standards. Level 1: Students at this level demonstrate an in adequate level of success with the challenging content of the Next Generation Sunshine State Standards. (Florida Department of Education Office of Assessment, 2012j). In its first year of implementation, only 51% of 10th grade students scored at proficiency or above on the FCAT Reading 2.0 assessment (which is considered level 3 or higher) (Florida Department of Education, 2012e). Specifically, 20% scored at level 1 (inadequate level of success), 30% at level 2 (below satisfactory), 22% at level 3 (satisfactory level), 19% at level 4 (above satisfactory), and 10% at level 5 (mastery of most challenging content) (Florida Department of Education, 2012e). Therefore one can derive a general consensus that on scales of a statewide, national, and international level, American students are failing in disproportionate numbers to achieve basic literacy skills regardless of years spent in formal education. This failure to successfully navigate bas ic literacy skills has dire consequences on content area learning. Although content areas vary in their literacy demands (Grossman & Stodolsky, 1995), the areas of knowledge and skills necessary to advance literacy

PAGE 70

70 between grade 4 and 12 include: a) automa ticity of word recognition, b) the dramatic expansion of vocabulary, c) the growth of conceptual knowledge and understanding, d) the increase of thinking and reasoning skills, e) the development of self regulated reading comprehension strategies, and f) th e acquisition or maintenance of a motivation and interest in broad and deep reading (Torgesen, 2006). Results of Research on Reading Proficiency and Science Achievement Currently, a limited number of studies exist regarding reading proficiency and its subs equent effect on science proficiency, yet assessments and activities in science generally require that students read text ( Norris & Phillips, 2003; Yore, Bisanz, & Hand, 2003) Furthermore, students may fail assessments due to their inability to read or un derstand the test, and not because of lack of knowledge with the content (Roe, Stoodt, & Burns, 1991). Studies on reading and science achievement generally fall into one of three categories: a few correlational studies, research on language based science i nstruction, and research on the comprehension of science text (Cromley, 2009) However, despite the available research on the impact of reading on science achievement, only one study to date has specifically investigated the impact of reading proficiency o n b iology achievement. Thus researchers are relying primarily on results obtained from comprehensive science assessments to make plausible inferences regarding the influence of reading proficiency on b iology achievement. The only study that focused primar ily on b iology content by Haught and Wall (2004) of 730 medical school students revealed an interesting finding. The methodology involved participants completing the Nelson Denny Reading Test that assess es reading vocabulary, reading comprehension, and rea ding rate. Prior scores on the Medical College Admissions Test (MCAT) that included a verbal reasoning and biological

PAGE 71

71 science section were collected. Results from the study reported that the Nelson Denny Reading Test and its 3 sections were significant pre dictors of MCAT verbal reasoning. The vocabulary portion of the Nelson Denny Reading Test was positively correlated with the MCAT biological science score ( r = 0.27, p < .01), the MCAT verbal reasoning score, ( r = 0.53, p < .01), and the United States Medi cal Licensing Examination (USMLE) score ( r = 0.14, p < .01). An increase in the vocabulary score on the Nelson Denny Reading Test was commensurate with an increase in the MCAT and USMLE exam scores. The comprehension portion of the Nelson Denny Reading Te st was positively correlated with the MCAT biological science score ( r = 0.12, p < .01) and the MCAT verbal reasoning score, ( r = 0.41, p < .01). The total reading portion of the Nelson Denny Reading Test was also positively correlated with the MCAT biolog ical science score ( r = 0.23, p < .01) and the MCAT verbal reasoning score, ( r = 0.56, p < .01). Additionally, reading vocabulary was a significant predictor variable of MCAT biological science scores. Therefore, vocabulary, among all other variables in th e study, served as the most significant predictor of scores on other assessments such as MCAT and USMLE as well as proficiency in the biological sciences. This finding, while consistent with other reports using comprehensive science assessments, were not r eplicated in most studies of reading proficiency and general science achievement. A cursory examination of anecdotal evidence of reading and b iology assessments reveals a pattern of reading scor es being relatively similar to b iology scores. For example, o n the SAT Biology subject test, out of a possible 800, the average score was 604 for the Ecology focus and 635 for the Molecular focus ( College

PAGE 72

72 Board, 2011) Of students who took those subject tests, the average score on the critical reading section was 596 and 609, respectively out of a possible 800 ( College Board, 2011). Despite the relative similarity in scores from an observational account, a limited number of studies have used statistical methods in a more objective manner. Visone (2009) investigated th e relationship between reading and science achievement on a standardized test. Results indicated a moderate to strong positive relationship between the variables, with a range from 0.41 to 0.74. Another study by Visone (2010) sought to determine the readin g issues associated with a standardized test in science by analyzing the self reports of 11th grade students. Qualitative results revealed that some of the science questions contained too much information that served to distract the students from their pur pose, the use of sophisticated vocabulary, failure to understand scientific terminology or concepts, lack of background knowledge (Visone, 2010). Maerten Rivera, Myers, Lee, and Penfield (2010) revealed that reading and mathematics accounted for 58% of th e variation in science achievement among student level factors. According to the researchers, for every one unit increase in reading achievement, a resulting average increase of 0.51 points occurred in science performance. Additionally, reading performance accounted for 25% of the variation in science achievement (Maerten Rivera, Myers, Lee, & Penfield, 2010). Romance and Vitale (2008) reported an increase in student achievement on both the Iowa Test of Basic Skills comprehension subtest and the MAT science achievement test by increasing science content knowledge through reading, writing, inquiry, and discussion without any explicit instruction in vocabulary or comprehension strategies.

PAGE 73

73 Other increases in science achievement have also been identified due to the implementation of broad reading interventions in which students are instructed in reading strategies, vocabulary, background knowledge, and writing (Fang & Wei, 2010; Guthrie, Wigfield, Barbosa, Perencevich, Taboada, Davis, Scafiddi, &Tonks, 2004; Rea ves, 2000; Shymansky, Yore, & Anderson, 2004; Vitale & Romance, 2012; Yore, Bisanz, & Hand, 2003 ). When the results of these two studies are analyzed, it is therefore not startling that Dempster and Reddy's (2007) study revealed that science performance is relatively dependent on reading skills. One study, conducted by O'Reilly and McNamara (2007) examined how well science knowledge, reading skill, and reading strategy knowledge predicted science achievement. Using data from 1,433 students, they revealed t hat the correlation of science knowledge and reading skill was 0.577. Additionally, science knowledge, reading skill, and reading strategy knowledge accounted for 36% of the variance in answers to multiple choice questions, with reading skill and science k nowledge contributing significantly to the model. The model of science knowledge, reading skill, and reading strategy knowledge also contributed to 46% of the variance in which reading skill and science knowledge served as significant predictors in the mod el. Reading strategy knowledge however was not significant. Thus, reading skill and science knowledge served as significant predictors of science achievement (O'Reilly & McNamara, 2007). Reading skill also significantly improved science proficiency for st udents with low levels of science knowledge so that they scored as high or higher than their less skilled, higher knowledge readers; thus compensating for low knowledge (O'Reilly & McNamara, 2007). For higher knowledge

PAGE 74

74 students, a high level of reading ski ll significantly improved science achievement (O'Reilly & McNamara, 2007). Another study conducted by Medina and Mishra (1994) examined the relationship between the reading achievement and academic performance in science of 518 Mexican American students re siding in Arizona. Although the tests were conducted in the native language of the students, results nevertheless indicated a significant positive relationship between native reading and native science performance (0.61, p <0.001 ) (Medina & Mishra, 1994). Three other studies of English speaking high school students also similarly reported high correlations. Nolen (2003) reported a strong correlate between scores on the district science achievement test and reading comprehension of r = 0.60. Demps and Onwueg buzie (2001) revealed a correlation of 0.80 between the scores on the Iowa Test of Basic Skills (ITBS) reading assessment and the high school science graduation test. Additionally, a correlation of 0.78 between the high school science graduation test and t he high school language arts graduation test was also found (Demps & Onwuegbuzie, 2001). Last, another study sought to analyze the relationship between reading comprehension and science proficiency in more than 40 countries using 15 year old students who completed the PISA reading and science literacy assessments in 2000, 2003, and 2006 (Cromley, 2009). Using both scores obtained from each year of the reading and scientific literacy assessments, Cromley (2009) determined a mean correlation of 0.840 betwee n reading and science and a range of 0.675 to 0.916 at a significance level of 0.001 across 43 countries using PISA 2000 data. This significant variation among the 43 countries was as a result of correlations that were the highest in

PAGE 75

75 countries that were in the top one half on mean reading achievement, such as The Netherlands, the United States, and the United Kingdom; and low among the six countries that were the lowest achieving in reading such as Indonesia and Peru, except Latvia. Therefore, a stronger re lationship between reading scores and science scores occurred in countries with higher mean reading performance than countries with lower mean reading performance. When Cromley (2009) replicated the study using 2003 data from 276,192 students from 41 coun tries, a correlation of 0.805 was determined as well as a range of 0.599 to 0.892 at a 0.001 significance level. Data obtained from the 2006 administration of 389,750 15 year olds also revealed a correlation of 0.819 at a 0.001 significance level (Cromley, 2009) despite the test designers deliberately requiring less reading (OECD, 2007) Cromley (2009) also found a range of 0 .603 to 0.902 at a 0.001 significance level among 56 countries. Similar to previous findings in the study, high correlations were pres ent from countries in the top one half on mean reading achievement and low correlations among the lowest achieving. Thus, reading and science scores were stronger in countries with higher mean reading performance (Cromley, 2009). Other research using PISA data indicated a 0.889 correlation between latent reading factors and latent science factors for retrieving information, 0.890 for interpreting texts, and for reflection and evaluation, a 0.840 correlation (OECD, 2002). Thus of the existing literature on th e relationship between reading proficiency and science proficiency, all studies at the secondary level indicate a strong positive relationship.

PAGE 76

76 Cromley (2009) concluded that the significant correlation between reading comprehension and science proficiency is due to good reading comprehension causing scientific proficiency as good reading comprehension should result in a good understanding of scientific texts and tests. Other research findings concur with this conclusion as acquiring scientific information t hrough knowledge and skills is undertaken through reading (Ediger, 2005). Cromley (2009) also proposed that increases in science proficiency could result from the products of extensive reading experience such as background knowledge, reading comprehension strategies, inference, and general vocabulary. Other research has also indicated that because proficient readers engage in more reading (Bray, Pascarella, & Pierson, 2004; Torgesen, 2006) and read more widely (Smith, 2000; Torgesen, 2006; Wingfield & Guthr ie, 1997), their exposure to scientific text would be more than their less proficient peers. Hence a myriad of factors contiguous with reading contribute to general scientific proficiency. For purposes of this review, I will focus solely on the two readi ng requirements that limited studies in b iology have demonstrated are responsible for directly improving b iology achievement: background knowledge and vocabulary. Background k nowledge B ackground knowledge serves as an important impetus of increasing readi ng literacy. Marzano (2004) indicated that the ability to learn new information is dependent upon what students already know. Thus, background knowledge is necessary to activate existing knowledge needed integrate new material into preexisting information (Feathers, 2004). Utilizing background experiences and knowledge occurred more often in the better readers who used the text in conjunction with their experiences and

PAGE 77

77 knowledge to reformulate old ideas. Poor readers were less likely to use the text to reth ink their experiences (Langer, 2001). Based on these findings, it is therefore not surprising that students from minority and low socioeconomic backgrounds typically were ch aracterized as poor readers as they lacked ample background knowledge, and thus ear ned paltry scores on reading assessments ( Fisher, Ross, & Grant, 2010; Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997; Reeves, 2000). Consequently, l ack of use of this salient element in reading is attributable to deficiencies in reading com prehension (Best, Rowe, Ozuru, & McNamara, 2005; Cromley, Snyder Hogan, & Luciw Dubas, 2010; Guthrie, Wigfield, Barbosa, Perencevich, Taboada, Davis, Scafiddi, &Tonks, 2004; Ozuru, Dempsey, & McNamara, 2009). In fact, research has shown that the better pre dictor of science text comprehension is prior knowledge, not reading skill ( Cromley, Snyder Hogan, & Luciw Dubas, 2010; Ozuru, Dempsey, & McNamara, 2009). Prior knowledge about b iology was positively correlated with overall comprehension of science text an d accounted for a large portion of variance in performance on comprehension questions above and beyond reading skills (Ozuru, Dempsey, & McNamara, 2009). The effect of prior knowledge is also larger on questions that require more extensive integration of i nformation while the effect of reading skill was larger on questions that were text based (Ozuru, Dempsey, & McNamara, 2009). Thus scientific learning and performance is hampered by limited background knowledge. Learning has also been cited as a challenge due to the nature of scientific texts. Research has indicated that students have difficulties understanding expository texts and in particular those scientific in nature (Best, Rowe, Ozuru, & McNamara, 2005;

PAGE 78

78 Ozuru, Dempsey, & McNamara, 2009) Some of the p roposed explanations for this difficulty is due to student difficulties in generating inferences required to understand science as they possess a low level of prior knowledge and inadequate reading strategies (Best, Rowe, Ozuru, & McNamara, 2005; Cromley, Snyder Hogan, & Luciw Dubas, 2010; Ozuru, Dempsey, & McNamara, 2009). Consequently, readers are often deficient in a sufficiently developed mental model that is representative of the overall conceptual relationship between relevant concepts such as hormone s and tropism (Kendeou & Van Den Broek, 2007; Ozuru, Dempsey, & McNamara, 2009). Another reason cited for the difficulty in understanding scientific text is due to the manner in which experts on the topic convey information to less knowledgeable readers ( Best, Rowe, Ozuru, & McNamara, 2005; Norris & Phillips, 2003). The process of creating a textbook involves one or a few expert teachers who write the text and subject matter experts and veteran teachers who edit and review it. Hence the textbook by design ( Christensen, the type of brain whose wiring i s most consistent with the methods used to solve problems in the field, as the domain ( Christensen, Horn, & Johnson, 2011, p. 128 129). Furthermore, the writing style in science texts is characterized by "low cohesion" in which s tudents must generate many inferences and fill in conceptual gaps ( Cromley, Snyder Hogan, & Luciw Dubas, 2010; Ozuru, Dempsey, & McNamara, 2009). According to Best, Rowe, Ozuru, & McNamara (2005), many science textbooks often omit information that the auth ors assume are prior knowledge for their readers; thus

PAGE 79

79 exacerbating previously existing reading deficiencies due to poor background knowledge. Since students lack specific knowledge on concepts such as osmosis, and gravity; their ability to generate infere nces that links concepts across and within sentences is limited (Best, Rowe, Ozuru, & McNamara, 2005; Kendeou & Van Den Broek, 2007; Ozuru, Dempsey, & McNamara, 2009). Thus students experience a fragmented and isolated understanding of the text which preve nts students from obtaining a coherent mental representation of the overall information (Best, Rowe, Ozuru, & McNamara, 2005; Kendeou & Van Den Broek, 2007; Ozuru, Dempsey, & McNamara, 2009). However, this omission of background knowledge by authors has be en shown to be beneficial with reading comprehension with highly skilled readers and detrimental to their less skilled peers (Ozuru, Dempsey, & McNamara, 2009). Research has in fact indicated that the most common problem for older readers is comprehension of text because some students may not possess sufficient fluency to aid in comprehension, others are lacking the strategies to comprehend the text, while others are not versed in employing the use of these strategies in a variety of situations (Biancarosa & Snow, 2004). Thus, sophisticated reading comprehension skills are necessary at the secondary school level to successfully navigate the demands of more challenging academic expectations (Meltzer, Cook Smith, & Clark, 2002). Linguistic features in academi c subject texts, particularly science, contribute to the abstract and dense nature of these texts that are unfamiliar to the texts used by students at earlier ages. These features contribute to comprehension challenges, especially for struggling readers an d English language learners; and may be overlooked by proficient adults (Fang, 2006; Ozuru, Dempsey, & McNamara, 2009).

PAGE 80

80 These challenges are evident in the research conducted by Torgesen (2006) that investigated passage length according to grade level on the FCAT. Results indicated that the ability of students to understand longer and more complex ideas and sentences as grade levels increased was poor ( Torgesen, 2006) Using 2005 FCAT data, Torgesen (2006) reported that as Floridian students increased in g rade level, the proficiency levels in reading decreased. This is partially due to the increased skills and knowledge required to meet each grade level standard, such as more extensive use of background knowledge, being able to handle lengthier sentences wi th complex ideas, increased vocabulary demands, and the automaticity required in recognizing new vocabulary ( Torgesen, 2006). Thus the comprehension process consists of two salient features in which background knowledge is activated and the conveyance of ideas occurs through printed words. Therefore limited background knowledge and good word level processes (such as quick and effortless word identification and the meanings of keywords) if absent, serve as impediments in the comprehension process (Feathers, 2004; Neufeld, 2005; Marzano, 2004; Marzano & Pickering, 2005; Ozuru, Dempsey, & McNamara, 2009) and has been found to be an indicator of how well students learn content information (Marzano, 2004). Therefore, the following are erroneous independent assum ptions that poor readers are incapable of deriving meaning from text, or that they are unaware of utilizing a variety of sources to obtain knowledge, or that poor vocabulary is the basis of the differences in comprehension between low and high performing students. Instead, better readers have the knowledge of how to build on the information that they have grasped or understood (Langer, 2001).

PAGE 81

81 Vocabulary Vocabulary is also important in improving proficiency in reading as it involves the study of multiple re lationships or the connection of ideas. Thus, previous vocabulary serves as a framework for future vocabulary (Smith, 1990). The acquisition of some vocabulary may occur more readily than others due to the frequency of its use in everyday scenarios and rar ely require explicit instruction. These words constitute Tier 1 words which are comprised of words that commonly appear in spoken language. Tier 2 words consists of more sophisticated vocabulary of written text (academic words) that are used with regularit y with mature language users but are encountered less frequently with students. Vocabulary that is limited to specific domains and appear only in isolated situations comprise Tier 3 words. As a result of the limited exposure of these words, students have d ifficulties internalizing the meanings of these words that are present in content area texts (Harmon, Hedrick, & Wood, 2005). Examples of these words include medical, legal, and biology terms that are central to building knowledge and conceptual understan ding within various academic domains although they are rarely used in the general vocabulary (Beck, McCaslin, & McKeown, 1980). To enhance vocabulary acquisition for Tier 3 words, instruction should encompass effective practices for general vocabulary as w ell as the unique features of the language of various content areas (Harmon, Hedrick, & Wood, 2005). It is therefore recommended in science that students should learn the specialized vocabulary of science when they are used in a unit of study (Fang, 2006). In fact, students need to learn to be attentive to the salient features inherent in the language of school science by creating an awareness of the similarities and differences between the

PAGE 82

82 language of school science and the language of everyday life as wel l as understand how the language of science contributes to scientific thinking and thought (Fang, 2006). Ethnicity and Science Achievement According to measures of science achievement, students of minority backgrounds, on average; score lower than their Wh ite and Asian/ Pacific Islander counterparts and are disproportionately represented in scores of lower achievement (ACT, 2010; Bankston & Caldas, 1998; Florida Department of Education, 2012b; Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008; M aerten Rivera, Myers, Lee, & Penfield, 2010; National Center for Education Statistics, 2012; Rathbun & West, 2004; Riegle Crumb, Moore, & Ramos Wada, 2010; Strand, 2012) For example, for ACT tested high school graduates in 2010, only 44% of Asian, 36% of White, 14% of Hispanic, and 6% of Black students met ACT College Readiness Benchmarks in Science. Additionally, for U.S. fourth and eighth graders, average science scores of Black students were below the U.S. national average in 2007 and were the lowest of all ethnic groups while the average scores of their White and Asian counterparts were higher than the U.S. national average (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008). Hispanic students also scored below the U.S. national average at b oth the fourth and eighth grade level (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008). Proposed Explanations for Disparities in Achievement Based on these findings and other similar reports of underachievement in minority populations on ev ery academic measure, some have wondered whether such differences are cultural or genetic in nat ure. The notion that underachievement is inherent within minority populations, especially within the Black population, and that

PAGE 83

83 these populations are incapable of achieving at high levels due to their ethnic background is grossly incorrect. Instead, both cultural factors and systemic issues are being blamed for the achievement gap. According to Butchart (2010), despite the implementation of Jim Crow policies that mandated racial segregation, the legacy of slavery, the denial of resources particularly social, intellectual, and financial; incongruent pedagogy and curriculum, cultural values antithetical to the Victorian values of school, and the ubiquitous segregat ed schools; African Americans performed exceptionally well in formal education during the 1860s and 1870s. Instead, the demise of African American educationa l progress was due to the open W hite terrorism that aimed to destroy "the black dream of intellect ual emancipation through education" (Butchart, 2010, p. 37) through resistance to black educational freedom via education related violence such as harassment, ostracism, denial of physical resources to black schools, burning of schools or churches that hou sed schools, targeting the living quarters of teachers with violence terrorism of black students, W hite indifference to the atrocities, and the use of violence, sometimes deadly; on teachers. Easton Brooks and Davis (2007) further corroborates this thesi s by indicating that the economic and political history of African Americans is the root of the low educational performance of this subset of the population which schools cannot repair. Furthermore, socioeconomic accounted for significant variance in educa tional outcomes of African American students but wealth accounted for greater variance in educational outcomes (Easton Brooks & Davis, 2007). Thus, since more African Americans occupy low socioeconomic and posses less wealth than their W hite counterparts ( Caldas &

PAGE 84

84 Bankston III, 1997; Snyder, Dillow, & Hoffman, 2008), the underachievement by African Americans is grounded in poverty rather than an outcome of race. In fact, research has indicated the interconnected nature of socioeconomic status and race (Aud, Fox, & KewalRamani, 2010; Caldas & Bankston III, 1997; House & Williams, 2000). Furthermore, within the Black community, achievement varies. Pinder (2012) discovered that Afro Caribbean students statistically scored significantly higher, on average, than their African American peers in science. The study also revealed that on average, students from Afro Caribbean households received more assistance with homework, li ved with their father, and had more books at home (Pinder, 2012). The study further reveale d that the strongest predictors of science achievement were similar in both groups as reflected in the statements "parents discuss school progress" and "time spent in extra science lessons." However, the Afr o Caribbean group also cited, performance motiva tion to do science" whilst the African Americans additionally cited, "time studying science", and "science attitudes" (Pinder, 2012). Thus the study emphasized the importance of parental involvement in determining academic achievement in science. This fin ding regarding parental involvement is consistent with Lee, Bryk and Smith's 1993 study that found that parental expectations for their children's achievement and the importance that parents place on education are positively and strongly related to academi c performance. Even in high school, parents' beliefs, goals, and values concerning education and achievement strongly influence adolescents, who tend to incorporate these standards into their own. Of course, such influence may also be

PAGE 85

85 negative, when parent s do not value education or do not regularly enforce standards (Lee, Bryk & Smith, 1993). Another interpretation of the res ult of Pinder's 2012 and Ogbu and 1998 studies states that the difference in immigration patterns (voluntary or involuntary) is responsible for part of the difference in educational outcomes. According to Ogbu and Simons (1998), voluntary (immigrant) minorities willingly immigrated to the U.S. unlike their African American counterparts who view their permanent p resence as force d upon them by W hite Americans or the U.S. government during slavery. Furthermore, voluntary immigrants have a more positive view of American schooling as a means of gaining access to the greater opportunities available within the United States than their frame of reference which is "back home." Involuntary immigrants howeve r hold a more negative view of W hite controlled American schooling and a mistrust of teachers and the schools who are viewed as discriminatory. They view American society as not fully re warding or accepting their education and hard work and thus are ambivalent about the notion that education is a tool that results in success or aids in overcoming barriers to upward mobility and economic success (Ogbu & Simons, 1998). Empirical research o n the other factors that contribute to the academic underperformance of minorities have focused on, family/personal and community as well as institutional/structural variables. Family/personal and community factors include: the disproportionate representa tion of Black and Hispanic students from low income households (Education Information and Accountability Services, 2009; Woods, Kurtz Costes, & Rowley, 2005) ; family structure that is dominated by single parent households (Bankston III & Caldas, 1998; Pong Dronkers, & Hampden Thompson, 2003; Puma,

PAGE 86

86 Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997; Rathbun & West, 2004) ; high rates of teen pregnancy (Pong, Dron kers, & Hampden Thompson, 2003); low family social statu s (Caldas & Bankston III, 1997); les s stimulating home environment, particularly during the preschool and elementary years (Baharudin & Luster, 1998); highest maternal educational level at less than a high school diploma or its equivalent (Bankston III & Caldas, 1998; Liu & Lu, 2008; Rathbun & West, 2004); poor parental involvement (Lee, Bry k & Smith, 1993; Tillman, 2005); lowered parental academic expectations (Lee, Bryk & Smith, 1993; Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997); incongruence between home culture and mains tream school culture (Lovelace & Wheeler, 2006; Ogb u & Simons, 1998; Strand, 2012); lower proficiency with the English language (Haladyna, Nolen & Haas, 1991); less positive attitudes toward science ( Singh, Granville, & Dika, 2002); less positive views of science ability ( Sikora & Pokropek, 2012); lower levels of student motivation (Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997; Singh, Granvil le, & Dika, 2002; Strand, 2012); limited background knowledge ( Fisher, Ross, & Grant, 2010); poor li teracy rates (Puma, Karweit, Price, Ricciuti, Thompson, Vad en Kiernan, 1997; Reeves, 2000); lower levels of student attendance (Puma, Karweit, Price, Ricciuti, Thompson, Vad en Kiernan, 1997; Strand, 2012); student participation in less academically rigorou s courses (Moore & Slate, 2008; National Science Foundation, 2012a; Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997; Rumberger & Palardy, 2005; Strand, 2012); lower degree of fitness and higher proportion of fatness (Chomitz, Slining, McGowan Mitchell, Dawson, & Hacker, 2009; Davis & Cooper, 2011; Gurley Calvez & Higginbotham, 2010; Hollar, Messiah, Lopez Mitnik, Hollar, Almon, & Agatston, 2010 );

PAGE 87

87 and poor community emphasis on the importance of education (Ogbu & Simons, 1998; Price, 2005). In stitutional/structural factors include: staffing problems at schools with high minority populations (Puma, Karweit, Price, Ricciuti, Thompson, Vade n Kiernan, 1997; Tillman, 2005); shortage of qualified teachers in the classroom (Murphy, DeArmond &Guin, 200 3; National Scien ce Foundation, 2012a); lowered teacher expectations (Gregory & Weinstein, 2008; Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997; Rumberger & Palardy, 2005; Strand, 2012); poorer level of instruction from t eachers (Desimone & Long, 2010); receipt of instruction from struggling or teachers new to the teaching field ( Desimone & Long, 2010 ) ; the disproportionate representation of Black students in discipline referrals and its subsequent effect on student removal from the classroom (Aud, Fox, & KewalRamani,. 2010; Gregory & Weinstein, 2008; Skiba, Michael, Nardo, & Peterson, 2002; Strand, 2012; Vincent, Tobin, Hawken, & Frank, 2012); negative racial climate in heterogeneous schools denoted by lack of racial fairness and experiences of racism (Mattison & Aber, 2007); lower degree of safety expressed by students at sch ool (Rumberger & Palardy, 2005); disproportionate representation of students from similar low socioeconomic standing attending schools with those from similar backgrounds (Bankston III & Caldas, 1998; Caldas & Bankston III, 1997; Condron & Roscigno, 2003; Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997; Rumberger & Palardy, 2005; Ryabov &Van Hook, 2007); similar ethnic composition of sch ools (Aud, Fox, & Kewa lRamani, 2010; Bankston III & Caldas, 1998; Condron & Roscigno, 2003; Lee, Bryk & Smith, 1993; Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997); disparity in educational funding in which high

PAGE 88

88 poverty districts receive less resources (Butchart 2010; Condron & Roscigno, 2003; Chambers, Levin, & Parrish, 2006; Ryan, 1999; Tho mpson, Wood, & Crampton, 2008); and the negative impact of stereotypes about a group's intellectual ability or competence based on ethnicity ( Steele & Aronson, 1995; Strand, 2012; Walton & Spencer, 2009; Woods, Kurtz Costes, & Rowley, 2005). T he disparity in scores has in part, been attributed to the concept known as stereotype threat. "Stereotype threat is conceived as a state of psychological discomfort that, if sufficien tly acute, can impair performance. It is thought to arise when students are confronted with an evaluative situation, in which a stereotype regarding a particular ability is relevant" (Appel, Kronberger, & Aronson, 2011, p. 904). According to Hoy and Hoy (2 they bear an extra emotional and cognitive burden. The burden is the possibility of This bur den can result in anxiety and thus undermine performance (Hoy & Hoy, 2009) by negatively affecting the intellectual functioning of these students, particularly on standardized tests ( Steele & Aronson, 1995) and increasing negative domain specific thinking ( Cadinu, Maass, Rosabianca, & Kiesner, 2005) If this threat continues, it may result in students disassociating with achievement in school and related intellectual domains as a protective mechanism; thus altering their self concept in a manner in which sc hool achievement does not serve as the basis of self evaluation or personal identity ( Steele & Aronson, 1995). The negative impact of stereotypes is not solely limi ted to race, however, but instead to any categorical grouping of individuals in

PAGE 89

89 which a ster eotype exists including those from low socioeconomic backgrounds and of female performance in m ath and science. The Current Status of Minorities in Science Despite the paltry scores earned by Black and Hispanic students, trends in academic performance all ude to the notion that there is hope as the average score of Black and Hispanic students have increased since 1995 at both the fourth and eighth grade level (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008) The caveat however, is that the av erage scores of both Black and Hispanic students continue to lag behind their White and Asian peers. In fact, the average scores between White and Black students in 2007 resulted in a 79 point and 96 point difference at the fourth and eighth grade level re spectively, and a difference of 65 and 71 points between fourth and eighth grade White and Hispanic students, respectively (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008). However, recent assessments have indicated that the achievement gap between the different ethnicities are narrowing (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008; National Center for Education Statistics, 2012). For example, the NAEP Science 2011 data indicates that when compared with 2009 data, the gap in scores was reduced from 36 points to 35 points between White and Black students, and from 30 points to 26 points between White and Hispanic students (National Center for Education Statistics, 2012). Therefore although an achievement gap exists between Whi te and Asian students and their Black and Hispanic peers, student achievement for Black and Hispanic students are improving. Based on these staggering statistics, it is not surprising that a disparity in ethnic representation in the sciences exist. Althou gh Blacks and Hispanics comprised 12%

PAGE 90

90 and 15% of the total U.S. population, they comprise smaller percentages of engineering and science degree recipients and employed scientists and engineers compared to the population; and thus are considered underrepres ented in science and engineering (National Science Foundation, Division of Science Resources Statistics, 2011; National Science Foundation, 2012a). In fact, the science and engineering work force is primarily composed of W hite males (National Science Founda tion, Division of Science Resources Statistics, 2011; National Science Foundation, 2012a) and that demographic is also primarily represented in graduate students and post doctorates in science and engineering in 2010 (National Science Foundation, 2012b). A frican American and Hispanics represent substantially lower levels of participation in science and engineering and other professional and related occupations than the U.S. workforce as a whole (National Science Foundation, Division of Science Resources Sta tistics, 2011) although at the elementary level, White and Black males express similar levels of positive attitudes towards science (Riegle Crumb, Moore, & Ramos Wada, 2010). In fact, Hispanic students are more likely to attain associate degrees than bache lor's degrees (National Science Board, 2004). However, underrepresented minority students obtaining bachelor's and master's degrees in science and engineering have increased over two decades since 1989, although those earning doctorates in these fields fla ttened since 2000 (National Science Foundation, Division of Science Resources Statistics, 2011). Of the science and engineering degrees earned at the bachelor's level by underrepresented minorities, the greatest rise has been in social, computer, and medic al sciences fields of study (National Science Foundation, Division of Science Resources Statistics, 2011). In fact,

PAGE 91

91 the percent of minorities in the doctoral science and engineering labor force increased from 4% in 1990 to 6% in 2007 (National Science Foun dation, 2011). While this increase in minority representation in the sciences represents a step forward in the right direction, significant inroads still need to be made to address student achievement disparities in light of the current demographic shift w ithin the U.S. population. According to United States Census Bureau, Population Division (2012), minorities such as Hispanics, Blacks, and Asians represent the largest net increase in the populat ion between 2000 and 2009. The United States Census Bureau (2 012) also reported that 50.4% of the nation's population under the age of 1 was comprised of minorities, as of July 1, 2011. As of 2011, there were 114 million minorities, primarily composed of Hispanics who numbered 52 million in 2011 (United States Censu s Bureau, 2012). Hispanics also represented the fastest growing population with a growth rate of 3.1% since 2010; thus comprising 16.7% of the natio n's total population in 2011 (United States Census Bureau, 2012). African Americans however, represented the second largest minority group at 43.9 million in 2011, a n increase of 1.6% from 2010 (United States Census Bureau, 2012). While the shift in demographics is not particularly troubling in itself, when viewed within the context of an increase in the retire ment of workers in the science and engineering workforce, a reduction in White male participation in science and engineering, more foreigners attaining U.S. science and engineering degrees, particularly advance d degrees in the past 2 decades; and an increa sing minority population whose participation rate is half or less than their White peers and have been traditionally underrepresented in science and engineering (National Science Board,

PAGE 92

92 2004); the need for minorities pursuing degrees in science and enginee ring becomes of increasing importance. Socioeconomic Status and Science Achievement Empirical research has conclusively indicated that student achievement is negatively affected by the low socioeconomic status of students (Caldas & Bankston III, 1997; Des imone & Long, 2010; Fleischman, Hopstock, Pelczar, & Shelley, 2010; Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008; Liu & Lu, 2008; Maerten Rivera, Myers, Lee, & Penfield, 2010; Maerten Rivera, Myers, Lee, & Penfield, 2010; National Center f or Education Statistics, 2012; National Science Foundation, 2012a; Rathbun & West, 2004). Thus the condition of poverty has greater implications that go beyond mere access to financial resources. According to the United Nations Children's Fund [UNICEF] (20 12), failing to protect children from poverty is a costly mistake borne by not only children, but the nation of those children due to the reduction in skills and productivity, lower levels of health and academic achievement, increased probability of unempl oyment and dependence on welfare, increased judicial and social protection systems, as well as the loss of social cohesion. According to the United Nations Children's Fund [UNICEF] (2012), 23.1% of American children are living in relative poverty (defined as "living in a household in which disposable income, when adjusted for family size and composition, is less than 50% of the national median income" (p. 3). When compared internationally, the U.S. ranked as the 2nd highest on relative child poverty rates o ut of the 35 economically advanced countries (UNICEF, 2012).

PAGE 93

93 Free and Reduced Price Lunch One service provided to children living in poverty is the National School Lunch Program (NSLP) which is federally assisted and operates within approximately 100,000 public and nonprofit private schools and residential child care institutions (United States Department of Agriculture, 2012). NSLP provides "nutritionally balanced, low cost or free lunches to more than 31 million children each school day in 2011" (United States Department of Agriculture, 2012, para. 1). According to the Education Information and Accountability Services (2009), the NSLP was established in 1946 to provide free and reduced price lunches to students from economically disadvantaged families. Ch ildren eligible for free meals are from families with incomes at or below 130 percent of the poverty level and those eligible for reduced price meals are from families with incomes between 130 percent and 185 percent of the poverty level (United States Dep artment of Agriculture, 2012). In the 2006 2007 school year, 1.2 million students in Florida qualified for free and reduced price lunch (Education Information and Accountability Services, 2009). This figure represents a 24.69% increase or an increase of 2 57,214 students since the 1999 2000 and 2008 2009 academic school years. Of students eligible, Black and Hispanic students represented the largest proportions of all ethnicities in which 69.59% of Black and 62.38% of Hispanic students were eligible; a sign ificant quantity compared to the 43.98% average of all ethnicities combined (Education Information and Accountability Services, 2009). Research on the Impact of socioeconomic on Achievement According to Caldas and Bankston III (1997), research has consis tently demonstrated since the mid 1960s that individuals with a low socioeconomic standing

PAGE 94

94 are more likely to engage in lower levels of academic achievement than individuals of a higher socioeconomic standing, even while controlling for a variety of other factors. One such study by the National Center for Education Statistics (2012), categorized students as coming from lower income families based on student eligibility for the NSLP for either free or reduced price school lunch. The study indicated that stud ents not eligible for free or reduced price school lunch obtained an average science score that was 27 points higher than their lower income peers. The study also indicated that when compared with 2009, the average scores of both eligible and non eligible students increased by 4 points for those eligible and 3 points for non eligible students. However, no significant changes between 2009 and 2011 in the score gap between both groups occurred. Empirical studies also have indicated that students from low soc ioeconomic status score lower than those from higher socioeconomic standing in combined reading literacy (Fleischman, Hopstock, Pelczar, & Shelley, 2010) and science measures (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008). Additionally, st udents attending the highest poverty public schools (in which 75% of students were eligible for free or reduced price school lunch) had the lowest average science score in both fourth and eighth grade than students attending schools in which the percent of students eligible for free or reduced price lunch was 50 74.9%, 25 49.9%, 10 24.9%, and less than 10% respectively (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008). This finding is consistent with Puma, Karwe it, Price, Ricciuti, Thompson, a nd Vaden Kiernan's 1997 study that indicated that the academic performance of children from high poverty schools was lower than those in low poverty schools. Children from low income households started school, on average,

PAGE 95

95 academically behind s tudents in l ow poverty schools and were unable to close the gap in achievement as they progressed through school (Puma, Karwe it, Price, Ricciuti, Thompson, Vaden Kiernan, 1997). Thus although several studies have highlighted the negative impact of low socioeconomic st atus on student achievement, other studies have alluded to a similar, yet nuanced variable. According to Easton Brooks and Davis (2007), it is wealth, not socioeconomic status that accounts for the more significant variance in educational outcomes, particu larly for African American students. It is wealth that confers advantages beyond income, occupation, and par ental education (Easton Brooks & Davis, 2007). It provides psychological benefits as well as the financial aspects that affect the degree of persona l security experienced, one's ability to finance a collegiate education, allows for a variety of options in places to live, as well as affords the protection from the stress of short term unemployment and other emergencies (Easton Brooks & Davis, 2007). Ca ldas and Bankston III (1997) also revealed that the family social status has a more significant effect on student achievement than socioeconomic status as measured by qualification fo r free and reduced price lunch. They noted that students who attended sch ools with high minority concentration tended to contain peers of relatively low family social status backgrounds and were marked by higher levels of poverty. They also discovered that low socioeconomic has a small, negative impact on individual student ach ievement. In fact, attending school with students who are from higher socioeconomic backgrounds has a positive impact on increasing student achie vement, independent of one's socioeconomic background, race, and other factors.

PAGE 96

96 This is because the more afflu ent peers tend to bring additional resources associated with higher family educational and occupational status. Additionally, students of all ethnicities are negatively affected by poverty although African American achievement is more negatively affected t hat their White peers. This supports what proponents for heterogeneous schools along socioeconomic lines such as Coleman, Campbell, Hobson, McPartland, Mood, Weinfield, and York (1966) have long advocated. According to them, the economic diversity of the s chool body serves as a means of enabling those from low socioeconomic lines to benefit from contact with those more socially advantaged as a means of acquiring "social capital." Empirical studies have consistently noted the intersection between socioecono mic standing and ethnicity, as previously discussed in the "Ethnicity and Student Achievement" section of Chapter 2 (Aud, Fox, & KewalRamani, 2010; Caldas & Bankston III, 1997; Easton Br ooks & Davis, 2007; Snyder, Dillow, & Hoffman, 2008) as proportionatel y m ore African Ameri cans occupy low socioeconomic than their W hite counterparts (Caldas & Bankston III, 1997; Snyder, Dillow, & Hoffman, 2008). Thus, race and ethnicity are difficult to conspicuously distinguish from socioeconomic status as impacting on st udent achievement due to the disproportionate representation of racial or ethnic minorities in low socioeconomic settings (Maerten Rivera, Myers, Lee, & Penfield, 2010). According to Snyder, Dillow, and Hoffman (2008), 33% of all Black families with childr en under 18 reside below the poverty level compared to 9.5% of Whites and 26.6% for Hispanics in 2006. The statistic is significantly higher in single parent households lacking a husband as the poverty level for Whites with no husband increased to 32.9%, 4 9.7% for Blacks, and 47.2% for Hispanics (Snyder, Dillow, &

PAGE 97

97 Hoffman, 2008). Thus it is not surprising that African American children are more likely to live in poverty than their peers of other ethnicities. In fact, 34% of black children live in poverty, c omp ared to 27% Hispanics, and 10% W hites (Aud, Fox, & KewalRamani, 2010). Additionally, of the 48% of fourth graders in public school that were eligible for free reduced price lunch in 2009, 77% were Hispanic, 74% Black, and 29% White (Aud, Fox, & KewalRam ani, 2010). Therefore, since ethnicity and socioeconomic status are intertwined, some of the some of the cited reasons for student underperformance in Chapter 2 under the subsection, "Ethnicity and Science Achievement" applies. Research has also indicated the intersection between socioeconomic standing and the number of parents in the household (Puma, Karweit, Price, Ricciuti, Thompson, Vaden Kiernan, 1997; Rathbun & West, 2004). Research has indicated that 55% of children living in poverty reside in single parent households (Rathbun & West, 2004). One particular study, Rathbum and West's 2004 study used data from the Early Childhood Longitudinal Study, Kindergarten Class of 1998 99 (ECLS K) that examined specific risk factors on student achievement. These r isk factors included living in a single parent household, living below the federal poverty level, living in a household whose primary home language is other than English and the highest level of maternal education less than or below a high school diploma o r its equivalent. The results indicated that as the number of risk factors of a child's household increase, student gains between the start of kindergarten and the end of third grade decrease in reading and math when compared with children with fewer fami ly risk factors, after controlling for other child, family, and school characteristics. Furthermore, students in such households are less likely to be proficient in reading, mathematics, and

PAGE 98

98 science than those with less risk factors. Children with no or lo w associated risk factors scored significantly higher than those with more risk factors. Puma, Karwe it, Price, Ricciuti, Thompson, and Vaden Kiernan's 1997 study also demonstrated that students performing poorly academically were "participants [that]were m ore economically disadvantaged. They were more likely to be living with a single parent, have a total family income under $10,000, receive public assistance, and have poorly educated parents whose natural language is not English. T hey were more likely to b e non W hite and live in urban areas with the concomitant increased likelihood of being exposed to high rates of crime, physical violence, drug abuse, and substandard living conditions" (p.22). Thus the impact of the environment is a significant one in the development of the cognitive capabilities of the child. One such environment that has received attention is the verbal environment during the early stages of child development. According to Christensen, Horn, and interactions. Merely speaking with a child was however not indicated to improve intellectual capacity but instead, research indicates that child ren in pre K need discussions with the infant in which the parent speaks in a fully adult, sophisticated, chatty language. It is characterized by deliberate, uncompromised personal adult what needs to be done are often characterized by simple, direct, here and now conversations.

PAGE 99

99 Thus their ability to improve cognitive development is limi ted. According to Zegiob and Forehand (1975), the socioeconomic status of the mother was a significant determinant of maternal behavior. The study found that less directive and more social interchange, especially verbally, were common among middle class mo thers than their lower class counterparts. Lower class mothers also gave more commands than middle class mothers and they were more directive and controlling. Also, less verbal exchanges occurred between lower class mothers and their children. Tulkin and K agan's 1972 study also examined how the different social class backgrounds of individuals affected the experiences of infants. The study also concluded that middle class mothers vocalized every verbal behavior more frequently than lower class mothers. Thus by improving the linguistic environment of the home, lower class mothers can help to improve the cognitive development of their children as a means of improving student academic performance in the future. Research has also indicated that poverty and being a minority does not have to negatively impact on student performance (Reeves, 2000). Reeves's 2000 study revealed a low ( 0.2) correlation between student performance and being poor. In this study 90/90/90 schools were examined that is, schools in which approximately 90% of the school's composition is ethnic minorities, a minimum of 90% are eligible for free and reduced price lunch, and approximately 90% of its students are achieving at proficiency levels based on independently conducted tests of student academic achievement. Reeves (2000) realized that these schools focused on a limited number (in particular 5) areas in which to improve and made curriculum choices whereby students spent more time in core subjects such as reading, writing, and mathematics and less time on other

PAGE 100

100 subjects. Consequently, although significant instruction was lacking in science, these students nevertheless outperformed their peers of similar backgrounds thus highlighting that reading and writing are critical in determining the a cademic achievement of students. Furthermore, they changed the school schedule so that at the elementary level students were engaged in 3 hours of literacy and at the secondary level they provided double periods of English and mathematics (Reeves, 2000). Gender and Science Achievement Research has consistently demonstrated that female performance in science continues to lag behind their male peers as measured through various scientific assessments ( Florida Department of Education, 2012b; Gonzales, Williams Jocelyn, Roey, Kastberg, & Brenwald, 2008; Maerten Rivera, Myers, Lee, & Penfield, 2010; O'Reilly & McNamara, 2007; National Center for Education Statistics, 2012; Riegle Crumb, Moore, & Ramos Wada, 2010; Shymansky, Yore, & Anderson, 2004). According to National Center for Education Statistics (2012), male students scored higher than females in science. The average male science score was higher than those of females in grade four by 5 points (although not statistically significant) and increased in grade 8 to 12 points (statistically significant at p <0.05). Although there was no significant difference in science performance of male and female fourth graders, males outperformed females by 5 points only in Earth Sciences, thus excluding all other science co ntent domains such as Life Sciences and Physical Sciences (National Center for Education Statistics, 2012). In the eighth grade however, males outperformed females in Biology, Physics, and Earth Sciences ( p <0.05). When the 2003 and 2007 average scores wer e analyzed, data indicated that the difference in scores between males and females decreased from

PAGE 101

101 12 scale score points to 5 scale score points for fourth graders. At the eighth grade level however, no significant difference in changes in the gap between a verage science scores from 1995 and 2007, thus maintaining the male advan tage in average science scores (Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008). Considering that individuals in science related fields earn considerably more on averag e than other fields (National Science Foundation, 2011), women are at a disadvantage in future earnings due to their decreased proficiency levels in the sciences. However, a study of Iranian secondary level girls and boys revealed otherwise (Nasr & Solta ni, 2011). According to Nasr and Soltani (2011), Iranian secondary level girls tend to have better achievements than their male counterparts. While some may account for this difference due to discrepancies in student attitudes based on gender, the study in dicated no significant difference on attitudes towards biology between secondary level girls and boys (Nasr & Soltani, 2011). Additionally, according to Nasr and Soltani (2011), there is no significant difference between attitude towards biology and studen ts' achievement in biology courses ( p <0.05, r = 0.12). However, of the many attitudes towards biology dimensions studied, the only positive and significant dimension was "biology is fun for me." George (2006) found contradictory results, revealing an init ial gender difference in the attitude toward science and its utility. According to George (2006), in the 7th grade, boys have a more favorable attitude toward science and its utility than girls. Bhanot and Jovanovic (2009) also revealed that boys were more confident in their science abilities although no gender differences existed in actual science ability or science performance. Greenfield (1997) on the other hand, failed to find any gender

PAGE 102

102 differences in attitudes. According to Greenfield (1997), elementa ry girl students liked science more than elementary boys, and although attitudes for both genders declined during middle school years, high school boys' attitudes rose again but not the girls. Thus, when all grades are combined, no significant gender diffe rence in attitude exists (Greenfield, 1997) George (2006) also noted that the attitude of boys changes faster than girls as they advance in coursework. Furthermore, a higher proportion of students expressed a more favorable attitude toward the utility of science during their 10th grade while taking b iology and lower favorability while taking earth or physical science (George, 2006). Prokop, Prokop, and Tunnicliffe (2007) also revealed no significant gender differences in preferences toward biology lessons although Slovakian girls reported higher interest in biology than Slovakian boys ( p <0.05). Boys also considered biology lessons to be more difficult than girls ( p <0.05) while the mean score for girls reporting that biology lessons were more important was significantly higher than in boys ( p <0.001). Biology was also the preferred science course for 4th and 6th graders (ages 9 10 and 11 12) though, younger (between ages 6 8) and older (between ages 13 15) students reported lower rates of course preference (P rokop, Prokop, & Tunnicliffe, 2007). Thus, despite the trend of the changing reports of the utility of science based on the science subject matter being learned, attitude towards science declines during the middle and high school years (George, 2006; Green field, 1997) although attitudes regarding its utility is positive (George, 2006). This waning positive attitude toward science as students advance in science coursework is partly responsible for the decline in science learning (as evident in scores) and in terest in science related careers

PAGE 103

103 (George, 2006). Riegle Crumb, Moore, and Ramos Wada (2010) also revealed that enjoyment of science serves as an important factor in pursuing science careers specifically for White and Hispanic females although achievement in science served as a more significant influence for Black students at younger ages. Proposed Explanations for Disparity in Achievement Research has documented that attitudes and achievement among adolescents in science is influenced by a myriad of factor s (Taltont & Simpson, 1985). Factors include parental background and family environment, individual characteristics such as locus of control, achievement motivation, and self concept; as well as the influence of the school such as teacher and administrativ e styles as well as class climate (Taltont & Simpson, 1985). For example, Hidi and Harackiewicz's 2000 study that revealed that learning and academic performance is strongly influenced by interest, goals, and motivation. Velayutham, Aldridge, and Fraser (2 012) revealed that motivational beliefs regarding learning goal orientation, task value, and self efficacy influenced student learning. Research also indicated that student self concept is also a strong predictor of attitude towards science and its utility (George, 2006) and males posses a higher self concept in science than females (Sikora & Pokropek, 2012). Additionally, teachers who encouraged students to pursue opportunities in science and work hard in it were more likely to have students with positive attitudes towards the field (George, 2006). Friends also play a role in influencing course enrollment. Riegle Crumb, Farkas, and Muller (2006) revealed that females with female friends who excelled academically in both traditionally male and female subject s positively affected advanced course taking in those subjects.

PAGE 104

104 Research has also exposed the effect of stereotype threat on female performance in math and the sciences. Considering that the gender gap in performance does not emerge until high school (Nat ional Center for Education Statistics, 2012), gender differences in ability and preference may not represent sex differences due to genetic, chemical, or biological factors ( Good, Woodzicka, & Wingfield, 2010), but instead learned through socialization (Bh anot & Jovanovic, 2009; Bleeker & Jacobs, 2004; Cadinu, Maass, Rosabianca, & Kiesner, 2005; Walton & Spencer, 2009). Research has indicated that regardless of age or gender, implicit stereotyping of women in science predicted the sex differences in perform ance on the TIMMS assessment of eighth graders in 1999 and 2003 (Nosek, Smyth, Sriram, Lindner, Devos, Ayala, Bar Anan, Bergh, Cai, Gonsalkorale, Kesebir, Maliszewski, Neto, Olli, Park, Schnabel, Shiomura, Tudor Tulbure, Wiers, Somogyi, Akrami, Ekehammar, Vianello, Banaji, & Greenwald, 2009). Haworth, Dale, and Plomin (2009) reported that genetic and environmental influences are important for high science performance in both genders but the effect of the environment becomes increasingly more influential wi th age. Additionally, Bhanot and Jovanovic (2009) revealed that the science ability of boys were overestimated by their parents than parents of girls, thus partly explaining the tendency of girls to undervalue their science ability and boys in overestimati ng it. Furthermore, parents of boys perceived that their sons liked science more than parents of girls (Bhanot & Jovanovic, 2009). However, when the mother or father of a girl believed in the importance of science, girls valued science more although parent s played no role in influencing the beliefs of boys (Bhanot & Jovanovic, 2009).

PAGE 105

105 Informal discussions by mothers regarding the importance of science and possible careers in science also served in the creation of positive attitudes by girls (Bhanot & Jovano vic, 2009; Dewitt, Archer, Osborne, Dillon, Willis, & Wong, 2011 ). Notable however, is the finding that mothers are more likely to encourage science interest in their sons who are not doing well and only become involved with girls when they fare well acade mically (Bhanot & Jovanovic, 2009). Bleeker and Jacobs (2004) and Dewitt, Archer, Osborne, Dillon, Willis, & Wong (2011 ) also revealed that the self efficacy of adolescents in math and science as well as their career choices were related to their mother's beliefs about their adolescents' abilities. Thus early maternal expectations have an enduring effect on children and their career decisions later in life as children are "taught" which academic pursuits are considered appropriate based on their gender and learn role expectations which may lead to the gender gap (Bleeker & Jacobs, 2004). Research has also indicated that the school context may also serve as a contributor to the effects of stereotype threat, even in Non Western locations (Pico & Stephens, 201 2). Women attending single sex schools exhibit higher levels of motivation, domain identification, self efficacy, and performance in male dominated domains such as mathematics than those attending coed institutions (Pico & Stephens, 2012). Even when stereo types were applied, attending a single sex school served as a buffer, thus failing to negatively affect performance, although those attending coed institutions reported significantly worse performance (Pico & Stephens, 2012). The proposed difference in the impact on student academic performance as a factor of gender composition of the school is due to the salience of stereotypes in the

PAGE 106

106 environment. According to Pico and Stephens (2012), environments in which stereotypes are salient serve as grounds for ster eotypes to be triggered, thus exerting its impact. Thus the chronic exposure of females to stereotype laden environments results in increased susceptibility to its effects by (1) increasing accessibility and potential believability of gender stereotypes, a nd (2) serving as an incubator in which girls adapt to dispositional factors such as stigma consciousness in such contexts over a period of time; thus catalyzing stereotype threat and subsequently degrading performance (Pico & Stephens, 2012). Unfortunatel y, science textbooks have also served to perpetuate gender stereotypes via the images and languages utilized ( Good, Woodzicka, & Wingfield, 2010). According to Good, Woodzicka, and Wingfield (2010), "girls may begin to believe in the stereotype, attributin g their inferior performance to a supposed natural sex difference, and ultimately leading them to avoid majors and professions involving math and science. Thus, textbooks, considered positive instruments of learning, may ironically teach girls that they ha ve no place in the academic areas of math and science and thereby reduce their achievement and enjoyment within these disciplines" (p. 145). Although the authors do not advocate that the elimination of biased textbook images will assuage stereotype threat, they instead advocate that the gender gap in science performance will be lessened ( Good, Woodzicka, & Wingfield, 2010). In fact, a study conducted by Appel, Kronberger, and Aronson (2011) revealed that stereotype threat even impairs the ability to learn a nd acquire knowledge in preparation for a test which serves as a precondition for the potential occurrence of stereotype threat during testing.

PAGE 107

107 According to their study, women who were aware of the stereotype that women were less proficient in science, te chnology, engineering, and mathematics (STEM) test preparation than men were impaired in their note taking activities as well as evaluating the notes of others, especially if they identified with the female gender (Appel, Kronberger, & Aronson, 2011). Acco rding to Appel, Kronberger, and Aronson (2011), "if stereotype threat also impairs learning activities (at least among those who are domain identified), then, over time, targets not only will demonstrate impaired test performance but will actually learn co ntent in less efficient ways as well. Gradually, the knowledge gaps between targets and nontargets will widen" (p. 911). During the learning process as well, stereotype threat surfaces in other ways. There are reports of female students not receiving as m uch teacher attention in science classes although they initiated as many teacher interactions as their male peers ( Greenfield, 1997). During the test taking process, research has shown that females perform better academically in science when exposed to cou nter stereotypic images than stereotypic images ( Good, Woodzicka, & Wingfield, 2010). Thus, the behaviors that develop student capabilities are negatively affected by stereotypes, thus underscoring the importance of gender neutral textbooks, language, and activities within the classroom. Thankfully, t he developers of the Biology I end of c ourse assessment have intentionally ensured that test questions are equal in its specification of gender and ethnicity (Florida Department of Education, 2011c) as one mean s of reducing the impact of stereotype threat on student performance. According to American Association for the Advancement of Science (AAAS) Project 2061 (1990), when the science education of any student (particularly girls and

PAGE 108

108 minorities) are neglected, students are deprived of a basic education which handicaps them for life as well as deprives the nation of a talented workforce and informed citizens; a loss that the nation cannot afford. Representation of Women in Science Historically, women have been un derrepresented in the scientific fields and scientific careers ( National Science Foundation, 2012a ) According to Long and Fox (1995), the institution of science is marred with inequality in career attainments whereby women and most minorities, as groups, participate at lower levels, have lower positions, lower productivity levels, and less recognition than W hite men. Thankfully, this male stereotypical view of science and scientists is less a part of the female consciousness than males (Greenfield, 1997). Despite the stereotypical male view of science and scientists, there was a significant difference in GPA between men and women in any field of the sciences, in which women had higher GPAs than their male counterparts (Sonnert & Fox, 2012) although they wer e more likely to underestimate their science ability (Sikora & Pokropek, 2012). Additionally, according to Sonnert, Fox, and Adkins, (2007), the percentage of women majors enrolled in majors in the scientific and engineering fields has increased steadily i n a linear manner over a 16 year period (between 1984 and 2000). In fact, since 2004, females have attained more doctoral degrees in science and engineering than males ( National Science Foundation, 2011). Moreover, in 2010, more women than men were confe rred with scientific degrees (National Science Foundation, 2012b ) and the percent of women holding doctoral degrees in science and engineering has increased from 23% of the science and engineering doctoral labor force in 1990 to 34% in 2007 ( National Scien ce Foundation, 2011). Even more so, more women than

PAGE 109

109 men enroll in college and graduate with a bachelor's degree irrespective of all racial/ethnic groups (National Science Foundation, Division of Science Resources Statistics, 2011). Despite this trend of i ncreasing women graduating with undergraduate degrees, a higher proportion of degrees in science and engineering fields of study were earned by proportionately more males (National Science Foundation, 2011; National Science Foundation, Division of Science Resources Statistics, 2011; National Science Foundation, 2012a) which is not alarming considering that at the elementary level, White, Hispanic, and Black females reported significantly lower rates of interest in pursuing science careers than their White m ale counterparts and also reported lower levels of science enjoyment and science self concept overall (Riegle Crumb, Moore, & Ramos Wada, 2010). At the secondary level, although more women are enrolled in Advanced Placement courses than their male counterp arts, fewer women are enrolled in science courses and experienced less success, as a lower percentage scored at or above the criterion level in those subjects (Moore & Slate, 2008). Consequently, there is an uneven distribution of women in the scientific f ields. According to National Science Foundation (2011) and Sonnert and Fox (2012) women have a higher percentage of representation in life sciences. The participation of women among other scientific fields also varies although within fields it has remained relatively consistent over every level of degree. For example, women participate at higher levels in psychology and medical sciences. Women's participation in biosciences has also increased at all degree levels although at the master's and bachelor's lev el, the growing trend has stabilized over the

PAGE 110

110 last 5 years (National Science Foundation, Division of Science Resources Statistics, 2011). At the secondary level, Bleeker and Jacobs (2004) and Sikora and Pokropek (2012) also reported that the life sciences were selected by more females interested in pursuing careers in the sciences than other scientific fields such as physical sciences. When compared with their male counterparts in mathematics and physical sciences, women earn degrees at medium to low levels (National Science Foundation, Division of Science Resources Statistics, 2011) as more than 80% of bachelor's degrees in computer science, physics, and engineering were earned by males in 2007 ( National Science Foundation, 2011; National Science Foundation 2012a). Therefore, research has indicated the increasing participation and achievement of women in the Biological sciences, which stimulates discussion on whether a gender gap even exists within b iology National Center for Education Statistics (2012), indicated that males outperformed females in b iology at the eighth grade level, but the gap in scores decreased dramatically among fourth graders to only 5 points. Those fourth graders in 2007 would represent the cohort of ninth graders in this study, if t he traditional educational trajectory is followed. Thus, the possibility exists that these students may represent a cohort in which gender is negligible in student achievement in b iology This does not imply that gender differences may no longer exist amon g other sciences, but instead may not be applicable to b iology Therefore, the context of this study will test such an assertion. Summary The purpose of this study is to determine the influence and relationship of four variables FCAT Reading level, ethni city, socioeconomic status, and gender on b iology achievement as reflected on the results from the Florida Biology I EOC assessment at a

PAGE 111

111 Florida Title 1 high school Results from previous studies have indicated that achievement in science is differentiall y affected by each of the four variables. Research has consistently indicated that students with low levels of reading proficiency score significantly lower on science assessments than their peers who are proficient in reading. Research also indicates that students from minority backgrounds and students who are eligible for free and reduced price lunch score lower than students who are White and who are not eligible for free and reduced price lunch. Last, research has also indicated that women score lower o n science assessments than their male counterparts. The proposed explanations that account for these differences reveal that this phenomenon of underperformance is a complex one and interdependent on several other factors Consequently there are significan t implications not only on graduation rates but also on the ability of the education sector to prepare students for a more technologically advanced and global society.

PAGE 112

112 CHAPTER 3 METHODOLOGY This chapter will present an overview of the quantitative met hods utilized in this study. The subsections that comprise this chapter include research design, population, sample and sampling procedures, procedure for data collection, measures, participants, and method of data analysis. Research Design This study ado pted a causal comparative exploratory research or ex post facto design. According to Gay, Mills, and Airasian (2009), causal comparative research describes preexisting conditions and therefore is considered descriptive research. However, this methodology a lso seeks to deter mine the causes or reasons for the existing differences in the status of groups Thus, the researcher aims to identify the major factor that has led to the difference between groups as "both the effect and the alleged cause have already o ccurred and must be studied in retrospect (Gay, Mills, & Airasian, 2009 p. 218 ). This methodology was deemed appropriate as the study collected preexisting data from the sample being investigated and therefore given the ex post facto design; the research er did not consciously or deliberately manipulate any of the variables of interest in the study. The researcher also did not deliberately control variables, including confounding variables, and failed to utilize random sampling. However, an analysis of cov ariance will be utilized in the study in an attempt to control for extraneous variables. Population Originally, the researcher wanted to utilize data obtained from two schools for the study However, due to the extensive data collection protocols that requ ired

PAGE 113

113 considerable time significant delays in the study would have occurred. Consequently the re searcher elected to focus on Title 1 school as the focus of this study. Thus, t he target population for this study included the high school students matriculat ed at a rural North Florida Title 1 school. The school was purposely sampled as it is a co educational institution whose demographics contain relatively equal proportions of males and females (52% males and 48% females) enrolled in Biology I (Florida Depar tment of Education, 2012, o, p), and is reflective of the major ethnic groups (White, Black, and Hispanic) represented within the state of Florida. The ethnic composition of the student body is comprised of 41.7% White, 54.3% Black, 2.3% Hispanic, 0.3% Asi an, 0.9% American Indian or Alaskan native, and 0.4% two or more races (Florida Department of Education, 2012p). Thus, the minority rate is 59%. Furthermore, the high school is comprised of 683 students ( Florida Department of Education, 2012p), of which 64 % of the students qualified for free and reduced price lunch, thus the North Florida school is categorized as a high poverty or a Title 1 school. Academically, 39 % of students scored at proficiency or higher (3 or higher) in reading on the state mandated a ssessmen t and only 26% of students obtained proficiency on the Science FCAT in 2011 (Florida Department of Education, 2011a), the highest reported percentage since the inception of the Science FCAT. Last, the overall school grade has varied throughout the years and consequently the accountability penalties or financial rewards. The school grade since 2000 has varied from attaining an A to earning Cs, Ds, and Fs (Florida Department of Education, 2012n). The administrative team is composed of three males that each h ave attained a Master's degree and have between 7 to 14 years of experience as an administrator

PAGE 114

114 (Florida Department of Education, 2011d) Of the 41 instructional staff present, 19.5% had 1 5 years of experience teaching, 39% had 6 14 years, and 41.5 % had 15 o r more years of experience (Florida Department of Education, 2011d) Of the instructional staff, 28.6 % held advanced degrees (Florida Department of Education, 2012p) 99.3 % were considered highly qualified under NCLB (Florida Department of Educat ion, 2012p) 12.2% were endorsed in reading, 15% were newly hired (Florida Department of Education, 2012p), and 4.9% were National Board Certified teachers (Florida Department of Education, 2011d) Within the district, 54% of students were male 37.5% wer e White, 56% A frican American, 4.4% Hispanic, 0.2% Asian and 78.2% were economically disadvantaged (Florida Department of Education, 2012p). Of the 2,715 students within the district, 2,049 students were eligible for free and reduced price lunch in 2008 2 009 or 75.47% of the district student population (Education Information and Accountability Services, 2009). When compared with the 1999 2000 data, a 15.11% increase in eligibility occurred during the 2008 2009 school year (Education Information and Accoun tability Services, 2009). Additionally, n ewly hired instructional staff comprised 26.9% of the district 29% of the instructional staff within the district contained advanced degrees, and 80.9% of classes a re taught by teachers in field (Florida Department of Education, 2012p). Sample and Sampling Procedures This research adopted convenience sampling due to the relatively easy accessibility for data collection but was purposely selected because the school under investigation possessed unique characteristics in terms of the geographic location, demographic composition of the student population, and the academic performance of

PAGE 115

115 students historically. Of the students who completed the Biology I EOC assessment, all students were sampled. Procedure for Data Collec tion After having received IRB approval (see Appendix C ) and permission from the appropriate school district personnel (see Appendix D ) student data was collected at the school site level. Specific information from the 2011 2012 administration of the Flor ida Biology I EOC exams was limited to Biology I EOC T scale scores and Biology I student EOC performance comparable to the state. FCAT Reading Retakes Developmental scale scores, FCAT Reading Retakes Achievement levels, FCAT Reading 2.0 Developmental scal e scores, FCAT Reading 2.0 Achievement levels, student ethnicity, eligibility for free and reduced price lunch, gender, and student grade level were also collected. No individual student names or personal information was collected. Researcher records of st udent data were then compiled on a computer and a paper copy was also kept in a secure location. Once all student data was compiled, data was then coded and entered into the software program, SPSS. Twenty one (21) students with missing information from the North Florida student population were then removed from the sample. Measures Reading Proficiency Reading proficiency was assessed using developmental scale scores from the FCAT Reading Retakes for 11th grade students and grade level specific FCAT Reading 2.0 for 9th and 10th grade students.

PAGE 116

116 Socioeconomic Status Socioeconomic status was measured by eligibility for free and reduced price lunch. Biology Proficiency Biology proficiency was assessed using T scores obtained from the Spring 2012 a dministration of the Biology I end of c ourse assessment. Students that were categorized in the top third by the state were considered proficient or obtaining a passing score in b iology in this study. Participants The study was conducted in a North Florida school distric t that contains one Title 1 high school. Participants of this study were comprised of 43 ninth grade, 130 tenth grade, and 5 eleventh grade students; for a total of 178 students. Additionally, of the 178 students, 48% qualified for free and reduced price l unch (38.8% free and 9.6% reduced), 59% were White, 37% were Black, 2.2% were Hispanic, and 1.7% were American Indian or Alaskan native. 54 % were female, and 55% scored 3 or higher on the FCAT Reading (30.3% level 3, 20.2% level 4, and 4.55% level 5). Tab le 3 1 represents the results of student performance by grade level from the Title 1 school population. Table 3 2 represents the sample of academic performance of students by grade level in the study. Method of Data Analysis A combination of descriptive st atistics and other inferential statistical procedures such as a correlation (Pearson r ), on e way ANOVAs, ANCOVAs, multiple regression analysis and independent samples one tailed t tests, were utilized in the analysis of student data. Each appropriate tes t was set at a significance level of 0.05

PAGE 117

117 The researcher used a multiple regression process to explore the relationship of predictive variables ( reading proficiency ethnicity/race, socioeconomic status, gender, and grade level) as they relate to the dep endent variable ( b iology T scale scores) in the study. Due to the categorical nature of some of the independent variables, coding was used to create vectors to represent the levels of categorical independent variable. The number of vectors needed for each independent variable is one less than the number of levels in the categorical independent variable. Therefore, for example, if there were 4 levels of an independent variable, 3 vectors were created Then dummy coding was used in which the values entered i n the vectors are 1s and 0s. Each level was represented by one vector, either 0 or 1. The 4 categorical variables that were dummy coded included gender, eligibility for free and reduced price lunch, ethnicity, and grade level. Appendix D represents the cod ing and dummy coding of variables in the study. In order to initially determine the predictive variables that had a statistically significant relationship to b iology T scale scores, a stepwise regression procedure was utilized. This procedure was used to fine tune the model by providing the data needed to determine the removal of variables that did not significantly contribute to b iology achievement. Based on the initial data, variables were then added or removed to determine the best model that had the l argest R squared and adjusted R squared values which wa s statistically significant and removed highly correlated predictors The statistical significance and relative importance of each predictive variable was checked by examining the unstandardized coef ficient beta weights and the standardized beta weights of each pred ictive variable.

PAGE 118

118 Scatterplots and histograms based on the data were create d as well as correlation matrices, multicollinearity statistics, and a simultaneous regression analysis using all variables. A histogram of the residuals was examined to determine whether the data is relatively normally distributed. A normal P P plot was also used to determine how well a specific distribution fits the observed data An analysis of the scatterplots wer e then conducted to determine the strength of the linear line or whether the scatterplots were unrelated to the dependent variable. Furthermore, a scatterplot of the regression standardized predicted value and regression standardized residual was used to d etermine whether homoscedasticity is present in the data. Table 3 1. Student performance by grade level from the Title 1 school population Grade level Number of Students Mean Scale Score Percent scoring in the bottom third Percent scoring in the middle t hird Percent scoring in the top third 9 43 57 12 19 70 10 140 46 47 30 23 11 16 40 88 6 6 All grades 199 48 42 26 31 Table 3 2. Sample of student performance by grade level in the study Grade level Number of Students Mean Scale Score Percent scoring in the bottom third Percent scoring in the middle third Percent scoring in the top third 9 43 56.8 1 1.6 18.6 69.8 10 13 0 47.1 4 4.6 30 .8 24.6 11 5 38 .0 8 0 .0 20 .0 0 .0 All grades 178 4 9.2 37.6 2 7.5 34.8

PAGE 119

119 CHAPTER 4 RESULTS AND ANALYSIS As previously di scussed in Chapter 2, a myriad of factors contribute to the performance of students in science. The purpose of this study was to determine the influence and relationship of four variables reading proficiency ethnicity, socioeconomic status, and gender; o n biology achievement as reflected on the results from the Florida Biology I EOC assessment at a Title 1 school. The following paragraphs will include a restatement of the research questions and hypotheses, and the corresponding statistical analysis with f urther interpretation. At the conclusion of the chapter, a summary of the answers to each question will be provided. Figures 4 1 to 4 4 and t ables 4 1 to 4 25 are also provided at the end of the chapter with a more detailed or visual depiction of the resul ts. Research Questions Research Question 1 Is there a significant difference in student performance on the end of c ourse assessment in b iology associated with reading proficiency ? After adjusting for student gender, race/ethnicity, and socioeconomic statu s, is there a significant difference in student performance on th e end of c ourse assessment in b iology associated with reading proficiency ? It was hypothesized that there will be a significant difference in student performance on th e end of c ourse assess ment in b iology associated with reading proficiency It is hypothesized that students that are more proficient readers will score significantly higher than their less proficient peers on the Biology I EOC assessment Furthermore, more proficient students w ill score significantly higher that their less proficient peers on the Biology I EOC assessment after adjusting for student gender, race/ethnicity, and socioeconomic status.

PAGE 120

120 Descriptive statistics of student performance on the FCAT Reading produced the fol lowing results: the mean FCAT Reading developmental scale score was 241.01 ( SD = 20.11), with a minimum score of 188 and a maximum score of 302 55% of students scored a 3 or higher on the FCAT Reading assessment. Table 4 1 summarizes the FCAT reading leve ls of students by grade. With alp ha equal to 0.05 there was a significant relationship found between FCAT Reading developmental scale scores and Biology I EOC T scale scores, r (178) = 0.73, p < 0.05 Figure 4 1 illustrates the strong positive linear rela tionship between FCAT Reading developmental scale scores and Biology I EOC T scale scores. Mean Biology I EOC T scale scores was 56.49 ( SD = 8.78) for students who scored 3 or higher on FCAT Reading and 40.25 ( SD = 8.99) for students who scored as a level 1 or 2 on FCAT Reading. With alpha equal to 0.05, a one tailed independent samples t test indicated a significantly higher Biology I EOC T scale score for students scoring 3 or higher on FCAT Reading compared to students scoring less than 3 on FCAT Reading t ( 176 ) = 12.15 p < 0.05 Table 4 2 provides descriptive statistics on the influence of FCAT Reading level on Biology I EOC T scale scores. Mean Biology I EOC T scale scores for students scoring level 1, level 2, level 3, level 4, and level 5 were 37.39 ( SD = 7.88), 42.83 ( SD = 9.23), 52.96 ( SD = 8.24), 59.50 ( SD = 7.37), and 66.75 ( SD = 4.37), respectively. With alpha equal to 0.05, a one factor between subjects ANOVA indicated a significant effect of reading achievement and Biology EOC T scale scores, F (4, 173) = 53.16, p < 0.05 Table 4 3 provides additional data on the results of the one way ANOVA on FCAT reading levels and Biology I EOC T scale scores.

PAGE 121

121 Tukey post hoc comparisons of the five FCAT Reading levels (1, 2, 3, 4, and 5) indicate that the s tudents who attained a level 5 on FCAT Reading ( M = 66.75, 95% Cl [63.10, 70.40]) scored significantly higher on the Biology I EOC than students who attained a level 3 ( M = 52.96, 95% Cl [50.71, 55.21]), p < 0.05 level 2 ( M = 42.83, 95% Cl [39.96, 45.71]) p < 0.05 and level 1 on FCAT Reading ( M = 37.39, 95% Cl [34.81, 39.98]), p < 0.05 Comparisons between students who attained a level 5 and level 4 on the FCAT Reading failed to indicate a statistically significant difference, p > 0.05. Students who atta ined a level 4 on FCAT Reading also scored significantly higher on the Biology I EOC than students who attained a level 3, level 2, and level 1, p < 0.05 Students who attained a level 3 on FCAT Reading also scored significantly higher on the Biology I EOC than students who attained a level 2 and level 1, p < 0.05 Additionally, students who attained a level 2 on FCAT Reading also scored significantly higher on the Biology I EOC than students who attained a level 1, p < 0.05. Table 4 4 provides additional d ata on the Tukey post hoc analysis of the influence of FCAT Reading developmental scale scores on Biology I EOC T scale scores. The analysis of covariance (ANCOVA) was used to investigate the hypothesis that the observed difference in mean scores of the B iology I EOC T scale scores is influenced by the differences in FCAT reading levels, with the covariates, ethnicity, socioeconomic status, and gender. The ANCOVA found that the means for students of varying FCAT reading levels was significant, ( F = 29.87, p < 0.05 ) that is, after the effect of the ethnicity, socioeconomic status, and gender has been accounted for. Table 4 5 provides the results of an ANCOVA for FCAT Reading level and Biology I EOC T scale scores. Based upon the findings; hypothesis one wa s supported.

PAGE 122

122 A further analysis of FCAT Reading developmental score data revealed interesting insights. Mean FCAT Reading developmental scale scores for 105 White, 66 Black, 4 Hispanic, and 3 American Indian students were 249.17 ( SD = 16.93), 228.05 ( SD = 18.65), 237.25 ( SD = 15.13), and 245.33 ( SD = 11.59), respectively. With alpha equal to 0.05, a one factor between subjects ANOVA indicated a significant effect of ethnicity and FCAT reading developmental scale scores, F (3, 174) = 19.78, p < 0.05 Tukey post hoc comparisons of the four groups (White, Black, Hispanic, and American Indian or Alaskan native) indicate that the White students ( M = 249.17, 95% Cl [245.90, 252.45]) scored significantly higher on the FCAT Reading assessment than Black students ( M = 228.05, 95% Cl [223.46, 232.63]), p < 0.05 Comparisons between the Hispanic ( M = 237.25, 95% Cl [213.17, 261.33]) and the other three groups were not statistically significant, p > 0.05. Comparisons between the American Indian or Alaskan native ( M = 24 5.33, 95% Cl [216.54, 274.13]) and the other three groups were not statistically significant, p > 0.05. Table 4 6 presents the results of the Tukey post hoc analysis of the influence of ethnicity on FCAT Reading developmental scale scores. The analysis of covariance (ANCOVA) was used to investigate the hypothesis that the observed difference in mean scores on the FCAT Reading is influenced by the differences in ethnicity, with the covariates, socioeconomic status and gender. The ANCOVA, found that the means for students of varying ethnicities was significant, ( F = 9.0, p < 0.05 ) that is, after the effect of the socioeconomic status and gender has been accounted for. Table 4 7 provides additional ANCOVA results for ethnicity and FCAT Reading developmental s cale scores.

PAGE 123

123 With alpha equal to 0.05, a one factor between subjects ANOVA indicated a significant effect of free/reduced price lunch eligibility and FCAT reading developmental scale scores, F (2, 175) = 21.16, MSE = 329.33, p < 0.05 Tukey post hoc compar isons of the three groups (free, reduced, and not eligible) indicate that the students not eligible for free or reduced price lunch ( M = 249.16, 95% Cl [245.51, 252.82]) scored significantly higher on FCAT Reading than students who were eligible for free l unch ( M = 230.38, 95% Cl [225.77, 234.99]), p < 0.05 Comparisons between the reduced price lunch ( M = 240, 95% Cl [231.66, 248.34]) and the other two groups were not statistically significant, p > 0.05. Table 4 8 provides the data for the Tukey post hoc a nalysis of the influence of socioeconomic status on FCAT Reading developmental scale scores. The analysis of covariance (ANCOVA) was used to investigate the hypothesis that the observed difference in mean scores of the FCAT reading developmental scale sco res is influenced by the differences in socioeconomic status, with the covariates, ethnicity and gender. The ANCOVA found that the means for students eligible for free or reduced price lunch and those ineligible are significant, ( F = 11.67, p < 0.05 ) tha t is, after the effect of ethnicity and gender has been accounted for. Table 4 9 represents the results of the ANCOVA for socioeconomic status and FCAT Reading developmental scale scores. Mean FCAT reading developmental scale scores was 242.46 ( SD = 18.56) for 96 females and 239.30 ( SD = 21.77) for 82 males. With alpha equal to 0.05, a one tailed independent samples t test indicated no significant effect of gender on FCAT reading developmental scale scores, t (176) = 1.043 p > 0.05 Table 4 10 represents the results of the independent samples t test to determine the influence of gender on FCAT

PAGE 124

124 Reading developmental scale scores. Based upon these findings, ethnicity and socioeconomic status of a student have a significant influence on FCAT reading developme ntal scale scores. Research Question 2 Is there a significant difference in student performance on the e nd o f c ourse assessment in b iology associated with race/ethnicity ? After adjusting for reading proficiency, student gender, and socioeconomic status, i s there a significant difference in student performance on th e end of c ourse assessment in b iology associated with race/ethnicity ? It was hypothesized that there will be a significant difference between the race/ethnicity of a student and their performan ce on th e end of c ourse assessment in b iology. It is hypothesized that students from Caucasian backgrounds will score significantly higher than those from other ethnic backgrounds. It is also hypothesized that students from Caucasian backgrounds will scor e significantly higher than those from other ethnic backgrounds after adjusting for reading proficiency gender, and socioeconomic background. Mean Biology I EOC T scale scores for 105 White, 66 Black, 4 Hispanic, and 3 American Indian students were 53.77 ( SD = 10.88), 42.23 ( SD = 10.53), 44.75 ( SD = 10.15), and 48 ( SD = 8.66), respectively. Table 4 11 provides descriptive statistics on the influence of ethnicity on Biology I EOC T scale scores. With alpha equal to 0.05, a one factor between subjects ANOVA indicated a significant effect of ethnicity and Biology EOC T scale scores, F (3, 174) = 15.92, p < 0.05 Table 4 12 provides the results of the one way ANOVA on ethnicity and Biology I EOC T scale scores. Tukey post hoc comparisons of the four groups (Whi te, Black, Hispanic, and American Indian or Alaskan native) indicate that the White students ( M = 53.77, 95% Cl [51.67, 55.88]) scored significantly higher on the Biology I EOC than Black students ( M = 42.23, 95% Cl [39.64, 44.82]), p < 0.05 Comparisons b etween the Hispanic ( M = 44.75, 95% Cl

PAGE 125

125 [28.61, 60.89]) and the other three groups were not statistically significant, p > 0.05. Comparisons between the American Indian or Alaskan native ( M = 48, 95% Cl [26.49, 69.51]) and the other three groups were not st atistically significant, p > 0.05. Table 4 13 provides data on the Tukey post hoc analysis of the influence of ethnicity on Biology I EOC T scale scores. The analysis of covariance (ANCOVA) was used to investigate the hypothesis that the observed differenc e in mean scores of the Biology I EOC T scale scores is influenced by the differences in ethnicity, with the covariates, reading proficiency, socioeconomic status, and gender. The ANCOVA, however, found that the means for students of varying ethnicities wa s not significant, ( F = 0.63, p > 0.05) that is, after the effect of reading proficiency, socioeconomic status, and gender has been accounted for. Based upon the findings; hypothesis two was rejected. Research Question 3 Is there a significant differenc e in student performance on t he end of c ourse assessment in b iology associated with socioeconomic status? After adjusting for reading proficiency, race/ethnicity, and student gender is there a significant difference in student performance on th e end of c ourse assessment in b iology associated with socioeconomic status? It was hypothesized that there will be a significant difference in student performance on the end of c ours e assessment in b iology associated with socioeconomic status. It is hypothesized th at students from higher socioeconomic backgrounds will score significantly higher than those from lower socioeconomic backgrounds. It is also hypothesized that students from higher socioeconomic backgrounds will score significantly higher than those from l ower socioeconomic backgrounds after adjusting for student gender, race/ethnicity, and reading proficiency Mean Biology I EOC T scale scores for students eligible for free lunch, reduced price lunch, and not eligible for free/reduced price lunch were 42.5 7 ( SD = 10.99) for 69

PAGE 126

126 students, 48.59 ( SD = 9.31) for 17 students, and 54.27 ( SD = 10.72) for 92 students, respectively. Table 4 14 provides additional descriptive statistics on the influence of socioeconomic status on Biology I EOC T scale scores. With al pha equal to 0.05, a one factor between subjects ANOVA indicated a significant effect of free and reduced price lunch eligibility and Biology EOC T scale scores, F (2, 175) = 23.60, MSE = 114.65, p < 0.05 Table 4 15 provides further data on the one way AN OVA results on socioeconomic status and Biology I EOC T scale scores. Tukey post hoc comparisons of the three groups (free, reduced, and not eligible) indicate that the students not eligible for free or reduced price lunch ( M = 54.27, 95% Cl [52.05, 56.49] ) scored significantly higher on the Biology I EOC than students who were eligible for free lunch ( M = 42.57, 95% Cl [39.92, 45.21]), p < 0.05 Comparisons between the reduced price lunch ( M = 48.59, 95% Cl [43.80, 53.37]) and the other two groups were not statistically significant when alpha was set at 0.05 ( p > 0.05), although students eligible for reduced price lunch scored significantly higher than those eligible for free lunch at the p < 0.10 level. Table 4 16 provides additional data on a Tukey post hoc analysis of the influence of socioeconomic status on Biology I EOC T scale scores. The analysis of covariance (ANCOVA) was used to investigate the hypothesis that the observed difference in mean scores of the Biology I EOC T scale scores is influenced by the differences in socioeconomic status, with the covariates, reading proficiency, ethnicity, and gender. The ANCOVA, however, found that the marginal means for students eligible for free and reduced price lunch and those ineligible are not significant, ( F = 1.84, p > 0.05) that is, after the effect of reading proficiency, ethnicity,

PAGE 127

127 and gender has been accounted for. Based upon the findings; hypothesis three was rejected. Research Question 4. Is there a significant difference in student performance o n t he end of c ourse assessment in b iology associated with gender ? After adjusting for reading proficiency, race/ethnicity, and socioeconomic status is there a significant difference in student performance on th e end of c ourse assessment in b iology associ ated with gender ? It was hypothesized that there will be a significant difference in student performance on th e end of c ourse assessment in b iology associated with gender. It is hypothesized that males will score significantly higher than females o n the B iology I end of c ourse assessment. It is hypothesized that males will sco re significantly higher on the e nd of c ourse assessment in b io logy after adjusting for reading proficiency race/ethnicity, and socioeconomic background. Mean Biology I EOC T scale scores was 50.68 ( SD = 11.53) for 96 females and 47.45 ( SD = 12.36) for 82 males. Table 4 17 provides additional data regarding descriptive statistics on the influence of gender on Biology I EOC T scale scores. With alpha equal to 0.05, a one tailed indepe ndent samples t test indicated no significant effect of gender on Biology I EOC T scale scores, t (176) = 1.80 p > 0.05 Levene's Test for Equality of Variances indicates variances for males and females do not differ significantly from each other ( p = 0. 63). Table 4 18 provides additional data on the independent samples t test to determine the influence of gender on Biology I EOC T scale scores. The analysis of covariance (ANCOVA) was used to investigate the hypothesis that the observed difference in mean scores of the Biology I EOC T scale scores is influenced by gender differences and the covariates, reading proficiency, ethnicity, and socioeconomic status. The ANCOVA indicated that the marginal means for both male

PAGE 128

128 and female students are not significant ( F = 1.93, p > 0.05) that is, after the effect of reading proficiency, ethnicity, and socioeconomic status had been accounted for. Based upon the findings; hypothesis four was rejected. Research Question 5 What are the best predictors of stud ent perfo rmance on the Biology end of c ourse assessment? It was hypothesized that reading proficiency, ethnicity, and socioeconomic status will serve as significant predictors of stud ent performance on the Biology end of c ourse assessments. During the stepwise regr ession procedure, a correlation matrix was produced that indicates the relationship between each variable and the dependent variable (see Table 4 19). Statistically significant correlations between Biology T scale scores and the following variables were fo und: FCAT Reading developmental scale scores (0.73) student gender (0.13), White students (0.46) Black students ( 0.45) students not eligible for free and reduced price lunch (0.44) students eligible for free lunch ( 0.44) minority students (0.46) an d students enrolled in the 9th and 10th grade ( 0.29), and the 11th grade ( 0.16) p < 0.05 Other notable significant correlations existed between FCAT reading developmental scale scores and grade level ( 0.24), White students (0.49), Black students ( 0.5 0), students not eligible for free or reduced price lunch (0.42), students eligi ble for free lunch ( 0.42), minority students (0.49) and students enrolled in the 9th and 10th grade ( 0.15), and the 11th grade ( 0.15), p < 0.05 Table 4 20 repo rts the su mmary results of the four models as an outcome of the stepwise multiple regression procedure. As the number of predictors increased, the R value also increased from 0.73 to 0.77. Therefore, model 4 (which includes the

PAGE 129

129 predictive variables FCAT reading deve lopmental scale scores, grade level s 9 11 and students eligible for free lunch) had a statistically significant correlation of 0.77. The R square for this model was 0.60 therefore indicating that 60 % of the variance in the Biology T scale scor es can be e xplained by the four predictive variables (FCAT reading developmental scale scores, grade level s 9 11 and students eligible for free lunch). The adjusted R square value (which takes into consideration the sample size and number of predictors ) remained at 0.59, thus illustrati ng that 59% of the variance in b iology scores is still attributable to the four aforementioned variables. Other analysis using an ANOVA indicated a significant effect of FCAT Reading developmental scale scores, grade level, and student s eligible for free lunch and Biology EOC T scale scores, F (4, 173) = 63.58 p < 0.05 (see Table 4 21) After adding and removing other predictive variables that were statistically significant on Biology T scale scores in the previous correlation matrix ( from Table 4 19) such as White students (RegressionWhite), Black students (RegressionBlack), students not eligible for free and reduced price lunch (RegressionNOTeligible), and minority students (Regressionminority); the researcher selected the model that had the largest R squared and adjusted R squared values, which was statistically significant, and removed highly correlated predictors Notable significant correlations that affected the selection of variables for th e model included: the high correlation b etween students not eligible for free and reduced price lunch and White students (0.50) Black students and eligibility for free lunch (0.42), White students and FCAT Reading developmental scale scores (0.49), and Black students and FCAT Reading developmen tal scale scores ( 0. 50); thereby indicating that there are intercorrelations or multicollinearity amongst the

PAGE 130

130 predictor variables. The researcher, in order to eliminate redundancy of predictor variables or m ulticollinearity as well as achieve parsimony of the regression model removed the following predictive variables from the model White students, Black students, and students not eligible for free and reduced price lunch. Consequently, four predictive variables remained in the model, notably FCAT Readin g developmental scale scores, grade level s 9 and 10, and grade level 11 and students eligible for free lunch. Table 4 22 summarizes the relationship of each of these variables to each other and the dependent va riable, Biology T scale scores, all of wh ich were significant at the 0.05 level. According to the correlation matrix, Biology T scale scores were significantly related to FCAT Reading developmental s cale scores (0.73), eligibility for free lunch ( 0.44) students in the 9th and 10th grade ( 0.29), an d students in the 11th grade ( 0.16) Therefore indicating that students with higher FCAT Reading developmental scale scores had higher Biology T scale scores. Furthermore, students eligible for free lunch and those in higher grades earned lower Biology T scale scores. Based upon the findings; hypothesis five was rejected A summary of the model (Table 4 23) indicates that all four predictive variables collectively contribute to 60 % of the variance in Biology T scale scores. The adjusted R squared value is also 59%, after factoring in the sample size and number of predictive variables There is also a significa nt correlation between the four predictor variables and the dependent variable ( R = 0.77). Last, t he Durbin Watson test is a test that is used to dete rmine whether the residuals from a multiple regression are independent. The statistic of 1.52 falls within the critical values of the Durbin Watson statistic based on

PAGE 131

131 sample size and number of regressors and therefore the residuals are independent and the re are no meaningful serial correlations. The results of an ANOVA (Table 4 24) show that with an alpha equal to 0.05 a significant effect of FCAT Reading developmental scale scores, grade level, and eligibility for free lunch and Biology EOC T scale scor es, F (4, 177) = 63.58 MSE = 3790.25 p < 0.05 Table 4 25 provides additional information regarding the contribution of each predictor variable to the model. According to the results using standardized beta coefficients, FCAT Reading developmental sc ale scores contribute 0.63 to the predictive model, whereas student s eligible for free lunch 0.13 students enrolled in the 9th and 10th grade 0.20, and students enrolled in the 11th grade 0.11, p < 0.05. Therefore, students with higher FCAT Reading dev elopmental scale scores had higher Biology T scale scores, after controlling for the other variables in the model. The eligibility for free lunch and grade level had a significant negative weight, indicating that students who were eligible for free lunch a nd were in higher grades had lower Biology T scale scores (a suppressor effect). Unstandardized B coefficients also indicate that FCAT Reading developmental scale scores contributes 0.38 to the predictive model studen ts eligible for free lunch 3.09, stud ents enrolled in 9th and 10th grade 5.36, and students enrolled in the 11th grade 8.01 Therefore, for each added point or every unit increase on the FCAT Reading developmental scale score, Biology T scale scores will increase by 0.38 (holding all other variables constant) Conversely, eligibility for free lunch enrollment in the 10th and 11th grade reduces b iology scores by 3.09 5.36, and 8.01 points respectively (holding all other variables constant)

PAGE 132

132 To ensure that all the assumptions of a multiple regression were met that is, that variables are normally distributed, an assumption of a linear relationship between the independent and dependent variable, the m easurement of variables without error and the assumption of homoscedasticity; a combination of various visual aids were created. To test the normal distribution of the variables, a histogram of the frequency of regression standardized residuals was created (Figure 4 2). The visual inspection indicates that the distribution is indeed relatively no rmal. To test whether a linear relationship between the independent and dependent variables existed, a normal P P plot was used, which indicates a linear relationship (Figure 4 3) To test whether the variables were measured without error a reduction in t he number of independent variables in the model was used as well as the inclusion of the adjusted R 2 To test the assumption of homoscedasticity, a visual inspection of a plot of the standardized residuals by the regression standardized predicted value was undertaken (Figure 4 4) which indicates a relatively even distribution, thus satisfying this assumption. Collectively, all assumptions were met, thus reducing the probability of Type I and II errors. The regression equation therefore is: Biology score = ( 0.379 *FCAT reading developmental scale score) + ( 5.356* student enrolled in the 9th or 10th grade) + ( 8.009 student enrolled in the 11th grade) + ( 3.085 student eligible for free lunch) 36.697 Summary The purpose of this section is to describe the quantitative results of the data analysis of this stu dy. The findings indicated that FCAT Reading developmental scale scores are strongly correlated with Biology I EOC T scale scores. Students who

PAGE 133

133 attained high FCAT Reading developmental scale scores score d significantly higher on the Biology I EOC assessment than students who attained low FCAT Reading developmental scale scores. However, there were no significant differences found on the influence of ethnicity on Biology I EOC T scale scores when scores we re adjusted for reading proficiency socioeconomic status, and gender. No significant differences were also found on the influence of socioeconomic status on Biology I EOC T scale scores when scores were adjusted for reading proficiency ethnicity, and gen der. Last, n o significant differences were found on the influence of gender on Biology I EOC T scale scores when score s were adjusted for reading proficiency ethnicity, and socioeconomic status. The significant predictors for b iology scores on the end of c ourse assessment include d : FCAT Reading developmental scale scores, grade level, and eligibility for free lunch. FCAT Reading developmental scale scores, grade level, and eligib ility for free lunch explained 60 % of the variance in b iology scores. FCAT Rea ding developmental scale scores, grade level, and eligibility for free lunch collectively is also significantly correlated to b iology scores. FCAT Read ing developmental scale scores wa s the most significant contributor to the predictive model followed by enrollment in either the 9th or 10th grade, eligibility for free lunch, and enrollment in the 11th grade. Students with higher FCAT Reading developmental scale score s are predicted to earn higher b iology scores whilst students enrolled in higher grades an d eligible for free lunch are predict ed to earn lower b iology scores, after controlling for other variables in the study. Additionally, f or every added point or every unit increase on the FCAT Reading developmental scale score, Biology T scale scores will increase by 0.38

PAGE 134

134 (holding all other variables constant). Conversely, eligibility for free lunch and enrollment in the 10th and 11th grade reduces b iology scores by 3.09, 5.36, and 8.01 points, respectively (holding all other variables constant). The regress ion equation is: Biology score = (0.379*FCAT reading developmental scale score) + ( 5.356* student enrolled in the 9th or 10th grade) + ( 8.009 student enrolled in the 11th grade) + ( 3.085* student eligible for free lunch) 36.697 Figure 4 1. The r elationship between FCAT Reading developme ntal scale scores and Biology I EOC T scale scores Figure 4 2 A histogram of the fre quency of regression standardized residuals obtained from the initial stepwise regression analysis

PAGE 135

135 Figure 4 3 A Normal P P plot of regression standardized residual that was used to determine how well a specific distribution fits the observed data Figure 4 4 A scatterplot used to visually determine homoscedasticity of the data used in the initial stepwise regression an alysis

PAGE 136

136 Table 4 1. FCAT Reading level distribution by grade Grade Level Total 9 10 11 FCAT Reading Level 1 2 34 2 38 2 7 32 3 42 3 15 39 0 54 4 16 20 0 36 5 3 5 0 8 Total 43 130 5 178 Table 4 2. Descriptive statistics on the influence of FCAT Reading level on Biology I EOC T scale scores Table 4 3 One way ANOVA results on FCAT reading levels and Biology I EOC T scale scores ANOVA Biology T scale scores Sum of Squares df Mean Square F Sig. Between Groups 14046.167 4 3511.542 53.162 .000 Within Groups 11427.338 173 66.054 Total 25473.506 1 77 Table 4 4. Tukey post hoc analysis of the influence of FCAT Reading developmental scale scores on Biology I EOC T scale scores Multiple Comparisons Dependent Variable: Biology T scale score Tukey HSD Descriptives Dependent variable: Biology T scale sc ore FCAT Reading level N Mean Std. Deviation Std. Error 95% Confidence Interval for Mean Minimu m Maximu m Lower Bound Upper Bound 1 38 37.39 7.876 1.278 34.81 39.98 20 50 2 42 42.83 9.234 1.425 39.96 45.71 20 59 3 54 52.96 8.242 1.122 50.71 55. 21 20 67 4 36 59.50 7.374 1.229 57.01 61.99 43 78 5 8 66.75 4.367 1.544 63.10 70.40 58 72 Total 178 49.19 11.997 .899 47.42 50.97 20 78

PAGE 137

137 Table 4 4. Continued. (I) FCAT Readin g Level (J) FCAT Reading Level Mean Difference (I J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound 1 2 5.439 1.820 .026 10.45 .42 3 15.568 1.721 .000 20.31 10.82 4 22.105 1.890 .000 27.32 16.89 5 29.355 3.161 .0 00 38.07 20.64 2 1 5.439 1.820 .026 .42 10.45 3 10.130 1.672 .000 14.74 5.52 4 16.667 1.846 .000 21.76 11.58 5 23.917 3.135 .000 32.56 15.27 3 1 15.568 1.721 .000 10.82 20.31 2 10.130 1.672 .000 5.52 14.74 4 6.537 1.749 .002 11.36 1.72 5 13.787 3.079 .000 22.27 5.30 4 1 22.105 1.890 .000 16.89 27.32 2 16.667 1.846 .000 11.58 21.76 3 6.537 1.749 .002 1.72 11.36 5 7.250 3.177 .156 16.01 1.51 5 1 29.355 3.161 .000 20.64 38.07 2 23.917 3.135 .000 15.27 32. 56 3 13.787 3.079 .000 5.30 22.27 4 7.250 3.177 .156 1.51 16.01 *. The mean difference is significant at the 0.05 level. Table 4 5. ANCOVA results for FCAT Reading level and Biology I EOC T scale scores Tests of Between Subjects Effects Dependen t Variable: Biology T scale score Source Type III Sum of Squares df Mean Square F Sig. Partial Eta Squared Corrected Model 14538.355 a 7 2076.908 32.288 .000 .571 Intercept 14911.548 1 14911.548 231.818 .000 .577 Gender 112.080 1 112.080 1.742 .189 .0 10 Ethnicity 89.755 1 89.755 1.395 .239 .008 Free and Reduced Lunch Eligibility 196.811 1 196.811 3.060 .082 .018 FCAT Reading Level 7686.254 4 1921.564 29.873 .000 .413 Error 10935.150 170 64.324 Total 456190.000 178 Corrected Total 25473.506 177 a. R Squared = .571 (Adjusted R Squared = .553)

PAGE 138

138 Table 4 6. Tukey post hoc analysis of the influence of ethnicity on FCAT Reading developmental scale scores Table 4 7. ANCOVA results for ethnicity and FCAT Reading developmental scale scores Tests of Between Subjects Effects Dependent Variable: FCAT Reading Developmental Scale Score Source Type III Sum of Squares df Mean Square F Sig. Partial Eta Squared Corrected Model 21870.421 a 5 4374.084 15.139 .000 .306 Intercept 404168.234 1 404168.23 4 1398.828 .000 .891 Gender 34.044 1 34.044 .118 .732 .001 Free and Reduced Lunch Eligibility 3603.971 1 3603.971 12.473 .001 .068 Ethnicity 7800.229 3 2600.076 8.999 .000 .136 Error 49696.574 172 288.934 Total 10410467.00 0 178 Corrected Total 71566.994 177 a. R Squared = .306 (Adjusted R Squared = .285) Multiple Comparisons Dependent Variable: FCAT Reading Developmental Scale Score Tu key HSD (I) Ethnicity (J) Ethnicity Mean Difference (I J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound White Black 21.126 2.751 .000 13.99 28.26 Hispanic 11.921 8.922 .541 11.22 35.06 American Indian or Alaskan native 3.8 38 10.255 .982 22.76 30.44 Black White 21.126 2.751 .000 28.26 13.99 Hispanic 9.205 9.018 .738 32.60 14.19 American Indian or Alaskan native 17.288 10.338 .342 44.11 9.53 Hispanic White 11.921 8.922 .541 35.06 11.22 Black 9.205 9.018 .73 8 14.19 32.60 American Indian or Alaskan native 8.083 13.376 .931 42.78 26.61 American Indian or Alaskan native White 3.838 10.255 .982 30.44 22.76 Black 17.288 10.338 .342 9.53 44.11 Hispanic 8.083 13.376 .931 26.61 42.78 *. The mean differ ence is significant at the 0.05 level.

PAGE 139

139 Table 4 8. Tukey post hoc analysis of the influence of socioeconomic status on FCAT Reading developmental scale scores Table 4 9. ANCOVA results for socioeconomi c status and FCAT Reading developmental scale scores Tests of Between Subjects Effects Dependent Variable: FCAT Reading Developmental Scale Score Table 4 10. Independent samples t test to determine the influence of gender on FCAT Reading developmental scale scores Indepe ndent Samples Test Multiple Comparisons Dependent Variable: FCAT Reading Dev elopmental Scale Score Tukey HSD (I) What is the eligibility of the student regarding free/reduced lunch? (J) What is the eligibility of the student regarding free/reduced lunch? Mean Difference (I J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound Free Reduced 9.623 4.914 .126 21.24 1.99 Not eligible 18.786 2.890 .000 25.62 11.95 Reduced Free 9.623 4.914 .126 1.99 21.24 Not eligible 9.163 4.791 .138 20.49 2.16 Not eligible Free 18.786 2.890 .000 11.95 25.62 Redu ced 9.163 4.791 .138 2.16 20.49 *. The mean difference is significant at the 0.05 level. Source Type III Sum of Squares df Mean Square F Sig. Partial Eta Squar ed Corrected Model 17768.985 a 4 44 42.246 14.285 .000 .248 Intercept 593592.022 1 593592.022 1908.8 33 .000 .917 Ethnicity 3697.918 1 3697.918 11.892 .001 .064 Gender 135.660 1 135.660 .436 .510 .003 Free and Reduced Lunch Eligibility 7260.215 2 3630.108 11.673 .000 .119 Error 53798.010 173 310.971 Total 10410467.000 178 Corrected Total 71566.994 177 a. R Squared = .248 (Adjusted R Squared = .231)

PAGE 140

140 Table 4 10 Continued. Levene's Test for Equality of Variances t test for Equality of Means F Sig. t df Sig. (2 tailed) Std. Error Differe nce 95% Confidence Interval of the Difference Lower Upper FCAT Reading Develop men tal Score Equal variances assumed 3.197 .075 1.043 176 .298 3.023 9.119 2.812 Equal variances not assumed 1.030 160.158 .304 3.061 9.199 2.892 Table 4 11. Descriptive statistics on the influence of ethnicity on Biology I EOC T scale scores Table 4 12. One way ANOVA results on ethnicity and Biology I EOC T scale scores Descr iptives Biology T scale scores Ethnicity N Mean Std. Deviation Std. Error 95% Confidence Interval for Mean Minimu m Maxi mum Lower Bound Upper Bound White 105 53.77 10.883 1.062 51.67 55.88 20 78 Black 66 42.23 10.532 1.296 39.64 44.82 20 64 Hi spanic 4 44.75 10.145 5.072 28.61 60.89 35 59 American Indian or Alaskan native 3 48.00 8.660 5.000 26.49 69.51 43 58 Total 178 49.19 11.997 .899 47.42 50.97 20 78 Biology T scale scores Sum of Squares df Mean Square F Sig. Between Groups 5486.650 3 1828.883 15.922 .000 Within Groups 19986.855 174 114.867 Total 25473.506 177

PAGE 141

141 Table 4 13. Tukey post hoc analysis of the influence of ethnicity on Biology I EOC T s cale scores Table 4 14. Descriptive statistics on the influence of socioeconomic status on Biology I EOC T scale scores Biology T scale score Categories N Mean Std. Deviation Std. Error 95% Confidence Interval for Mean Minimu m Maxim um Lower Bound Upper Bound Free 69 42.57 10.994 1.324 39.92 45.21 20 63 Reduced 17 48.59 9.308 2.257 43.80 53.37 32 64 Not eligible 92 54.27 10.720 1.118 52.05 56.49 20 78 Total 178 49.19 11.997 .899 47.42 50.97 20 78 Multiple Comparisons Dependent Variable: Biology T scale score Tukey HSD (I) Ethnicity (J) Ethnicity Mean Difference (I J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound White Black 11.544 1.684 .000 7.18 15.91 Hispanic 9.021 5.460 .352 5.14 23.18 American Indian or Alaskan native 5.771 6.276 .794 10.51 22.05 Black White 11.544 1.684 .000 15.91 7.18 Hispanic 2.523 5.519 .968 16.84 11.79 American Indian or Alaskan native 5.773 6.327 .798 22.18 10 .64 Hispanic White 9.021 5.460 .352 23.18 5.14 Black 2.523 5.519 .968 11.79 16.84 American Indian or Alaskan native 3.250 8.186 .979 24.48 17.98 American Indian or Alaskan native White 5.771 6.276 .794 22.05 10.51 Black 5.773 6.327 .798 10. 64 22.18 Hispanic 3.250 8.186 .979 17.98 24.48

PAGE 142

142 Table 4 15. One way ANOVA results on socioeconomic status and Biology I EOC T scale scores ANOVA Biology T scale score Sum of Squares df Mean Square F Sig. Between Groups 5410.225 2 2705.112 23.595 .000 Within Groups 20063.281 175 114.647 Total 25473.506 177 Table 4 16. Tukey post hoc analysis of the influ ence of socioeconomic status on Biology I EOC T scale scores Tabl e 4 17. Descriptive statistics on the influ ence of gender on Biology I T scale scores Table 4 18. Indepen dent samples t test to determine the influence of gender on Biology I EOC T scale scores Multiple Comparisons Dependent Variable: Biology T scale scores Tukey HSD (I) What is the eligibility of the student regarding fre e/reduced lunch? (J) What is the eligibility of the student regarding free/reduced lunch? Mean Differenc e (I J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound Free Reduced 6.023 2.899 .097 12.88 .83 Not eligible 11.707 1.705 000 15.74 7.68 Reduced Free 6.023 2.899 .097 .83 12.88 Not eligible 5.684 2.827 .113 12.37 1.00 Not eligible Free 11.707 1.705 .000 7.68 15.74 Reduced 5.684 2.827 .113 1.00 12.37 *. The mean difference is significant at the 0.05 level. Group Statistics Gender N Mean Std. Deviation Std. Error Mean Biology T scale score males 82 47.45 12.361 1.365 females 96 50.68 11.533 1.177 t test for Equality of Means t df Sig. (2 tailed) Mean Difference Std. Error Difference 95% Confidence Interval of the Difference Lower Upper

PAGE 143

143 Table 4 18. Continued. Table 4 19. Correlation matrix indicating the relationship between possible predictive variables and Biology I EOC T scale scores BioTs core RDevS core Regres White Regres sBlack Regres sHispa nic Regre ssAmI nAlNa Reg re sGEN DER Regres sNOTeli gible Regres sreduc edeligi ble Pe ars on Co rre lati on BioTsc ore 1.000 .733 .459 .447 .056 .013 .134 .439 .016 RDevS core .733 1.000 .488 .496 .028 .028 .078 .421 .016 Regres sWhite .459 .488 1.000 .921 .182 .157 .077 .497 .1 57 Regres sBlack .447 .496 .921 1.000 .116 .101 .107 .491 .146 Regres sHispa nic .056 .028 .182 .116 1.000 .020 .064 .005 .049 Regres sAmIn AlNa .013 .028 .157 .101 .020 1.000 .033 .048 .106 Regres sGEN DER .134 .078 .077 .107 .064 .03 3 1.000 .076 .006 t df Sig. (2 tailed) Mean Difference Std. Error Difference 95% Confid ence Interva l of the Differe nce Lower Upper Biolog y T scale score Equal variances assumed 1.799 176 .074 3.226 1.793 6.764 .312 Equal variances not assumed 1 .790 167.354 .075 3.226 1.802 6.784 .333

PAGE 144

144 Table 4 19. Continued BioTs core RDevS core Regres White Regres sBlack Regres sHispa nic Regre ssAmI nAlNa Regre sGEN DER Regres sNOTeli gible Regres sreduc edeligi ble Regres sNOTel igible .439 .421 .497 .491 .005 .048 .076 1.000 .336 Regr es sreduc edeligi ble .016 .016 .157 .146 .049 .106 .006 .336 1.000 Regres sfreelu nch .441 .422 .415 .416 .035 .015 .074 .823 .259 Regres sminori ty .459 .488 1.000 .921 .182 .157 .077 .497 .157 Regres s10 .287 .148 .121 .178 .079 .117 .257 .208 .025 Regres s11 .159 .149 .204 .221 .026 .022 .048 .176 .176 RDevS core .000 .000 .000 .353 .354 .149 .000 .415 Regre ssWhit e .000 .000 .000 .008 .018 .153 .000 .018 Regre ssBlac k .000 .000 .000 .061 .091 .077 .000 .026 Reg re ssHisp anic .228 .353 .008 .061 .396 .198 .473 .257 Regre ssAmI nAlNa .431 .354 .018 .091 .396 .329 .262 .080 Regre ssGEN DER .037 .149 .153 .077 .198 .329 .156 .466

PAGE 145

145 Table 4 19. Continued. Table 4 20 Results of s tepwis e regression model indicating sig nificant predictive variables on Biology I EOC T scale scores Model Summary e Model R R Square Adjusted R Square Std. Error of the Estimate Durbin Watson 1 .733 a .538 .535 8.179 3 .764 c .584 .577 7.800 4 .771 d .595 .586 7.721 1.513 a. Predictors: (Co nstant), RDevScore c. Predictors: ( Constant), RDevScore, Regress 10 Regress freelunch d. Predictors: (Constant), RDe vScore, Regress10, Regressfreelunch, Regress 11 e. Dependent Variable: BioTscore Table 4 21 ANOVA r esults of stepwise regression model indicating significant predictive variables on Biology I EOC T scale scores ANOVA a BioTs core RDevS core Regres White Regres sBlack Regres sHispa nic Regre ssAmI nAlNa Regre sGEN DER Regres sNOTeli gible Regres sreduc edeligi ble Regre ssNOT eligible .000 .000 .000 .000 .473 .262 .156 .000 Regre ssredu cedelig ible .414 .415 .018 .026 .257 .080 .466 .000 Regre ssfreel unch .000 .000 .000 .000 .322 .423 .1 62 .000 .000 Regre ssmino rity .000 .000 .000 .000 .008 .018 .153 .000 .018 Regre ss10 .000 .024 .054 .009 .148 .060 .000 .003 .369 Regre ss11 .017 .023 .003 .001 .366 .384 .264 .009 .009

PAGE 146

146 Table 4 21. Continued. Model Sum of Squares df Mean Square F Sig. 1 Regression 13700.338 1 13700.338 204.810 .000 b Residual 11773.168 176 66.893 Total 25473.506 17 7 2 Regression 14524.794 2 7262.397 116.079 .000 c Residual 10948.711 175 62.564 Table 4 21. Continued. Total 25473.506 177 3 Regression 14886.382 3 4962.127 81.553 .000 d Residual 10587.124 174 60.846 Total 25473.506 177 4 Regressio n 15161.011 4 3790.253 63.584 .000 e Residual 10312.494 173 59.610 Total 25473.506 177 Table 4 22 Correlation matrix results of regression model indicating significant predictive variables on Biology I EOC T scale scores Correlations BioTscor e RDevSc ore Regress freelun ch Regress 1 0 Regress 1 1 Pearson Correlati on BioTscore 1.000 .733 .441 .287 .159 RDevScore .733 1.000 .422 .148 .149 Regress freelun ch .441 .422 1.000 .198 .074 Regress 10 .287 .148 .198 1.000 .280 Regress 1 1 .159 .149 .074 .280 1.000 Sig. (1 BioTscore .000 .000 .000 .017 tailed) RDevScore .000 .000 .024 .023 Regress freelun ch .000 .000 .004 .163 Regress 10 .000 .024 .004 .000 Regress 11 .017 .023 .163 .000 a. Dependent Variable: BioTscore b. Predictors: (Constant), RDevScore c. Predictors: ( Constant), RDevScore, Regress 10 d. Predictors: ( Constant), RDevScore, Re gress10, Regress freelunch e. Predictors: (Constant), RDe vScore, Regress10, Regress free lunch, Regress 11

PAGE 147

147 Table 4 23 Results of regress ion model indicating significant predictive variables on Biology I EOC T scale scores Model Summary b Model R R Square Adjusted R Square Std. Error of the Estimate Durbin Watson 1 .771 a .595 .586 7.721 1.513 a. Predictors: (Con stant), Regress11, Regressf reelunch, Regress 10, RDevScore b. Dependent Variable: BioTscore Table 4 24 ANOVA r esults of regression model indicating significant predictive variables on Biology I EOC T scale scores ANOVA a Model Sum of Squares df Mean Square F Sig. 1 Regression 1 5161.011 4 3790.253 63.584 .000 b Residual 10312.494 173 59.610 Total 25473.506 177 a. Dependent Variable: BioTscore Table 4 24. Continued. b. Predictors: (Con stant), Regress11, Regressfreelunch, Regress 10, RDevScore Table 4 25 R esults of reg ression model indicating significant predictive variables on Biology I EOC T scale scores Model Unstandardized Coefficients Standa rdized Coeffic ients t Sig. Collinearity Statistics B Std. Error Beta Tolerance VIF 1 (Constant) 36.697 8.219 4.465 .00 0 RDevScore .379 .032 .634 11.706 .000 .797 1.255 Regress freelun ch 3.085 1.328 .126 2.323 .021 .800 1.250 Regress 10 5.356 1.405 .199 3.812 .000 .861 1.161 Regress 11 8.009 3.731 .111 2.146 .033 .881 1.135

PAGE 148

148 CHAPTER 5 DISCUSS I ON Florid a's Next Generation Strategic Plan operates under the premise that measuring student achievement through EOC assessments would serve as a tool of ensuring students are globally competitive for college and future careers (Florida Department of Education, 20 05b). Starting in the 2012 2013 academic year, the state requires students to pass the one such assessment, the Biology EOC, with a satisfactory score or higher to obtain the course credit required to graduate (The Florida Legislature, 2012). Like similar assessments the Biology EOC influences student outcomes such as retention and graduation rates -which are data indicators for school accountability that determines overall school grade s (The Florida Legislature, 2012) and serve as the basis of teacher ev aluations (The Florida Legislature, 2010). T his increased accountability makes the ability to make more informed decisions increasingly important. Thus, the purpose of this study is t wo fold: (1) t o examine four variables -reading proficiency ethnicity, socioeconomic status, and gender -to determine their influence on b iology achievement as reflected on the results from the Florida Biology I EOC assessment at a Title 1 school, and (2) to predict which student characteristics are most influential on b iolo gy scores. Of all four variables identified, reading proficiency (which is determined by FCAT Reading developmental scale scores) was the sole variable that had a significant influence on Biology I EOC T scale scores. Results indicated that students scori ng 3 or higher on FCAT Reading scored significantly higher on the Biology EOC than their peers who obtained lower FCAT reading scores. This finding regarding the importance of reading proficiency on science achievement is consistent with previous research that

PAGE 149

149 more proficient readers obtained significantly higher scores on science assessments than their less proficient peers (Cromley, 2009; Dempster & Red dy, 2007; Haught & Wall, 2004; Medina & Mishra, 1994 ; O'Reilly & McNamara, 2007). However, identifying the mechanism involved in reading that contributes to the disparity in scores is beyond the scope of this research. N or can any conclusions be unequivocally made that improving FCAT Reading scores will increase Biology EOC scores. Because the causal compa rative approach of this study relies on determining the reason for differences in Biology I EOC scores after the fact no cause and effect conclusions can be definitively drawn due to the lack of randomization, manipulation, and control. Although reading proficiency, as indicated by FCAT Reading levels, was found to be the sole variable that influences student achieveme nt in b iology, results indicate that reading proficiency itself is influenced by ethnicity and socioeconomic status. In the sample group, 8 0% of African Americans qualified for free and reduced price lunch compared to 28% of the White students and 50% of Hispanic students Thus the group demonstrated a correlation of ethnicity and socioeconomic status. Furthermore, the study revealed the hig h correlation between students not eligible for free and reduced price lunch and White students, Black students and eligibility for free lunch, White students and FCAT Reading developmental scale scores, and Black students and FCAT Reading developmental sc ale scores Thus the research revealed strong correlations between ethnicity, socioeconomic status, and reading performance. Nevertheless, r esults indicated that White students scored significantly higher than Black students on reading as sessments and tha t students in eligible for free and

PAGE 150

150 reduced price lunch scored significantly higher than eligible students This finding is consistent with previous studies (College Board, 2011; National Center for Education Statistics, 2011 ; U.S. Department of Education, National Center for Education Statistics, 2012 ). Therefore, although ethnicity and socioeconomic status did not have a direct impact on b iology achievement, those variables indirectly influence it by directly affecting reading proficiency. Again, no indis putable conclusions can be made regarding how or why those effects exist since such questions lie outside the scope of this research. This research also revealed that collectively as a model FCAT Reading developmental scale scores, grade level, and eligi bility for free lunch were factors that could predict b iology achievement and explain 60 % of the variance in b iology scores. However, when each variable except reading was examined independently, each failed to re ach a significance of 0.05 Nevertheless t hese results emphasize the need for effective teacher evaluation design that takes into account student demographic composition so as not to penalize teachers who serve a student population that has historically underperformed academically Additionally, a lthough extant literature fails to quantify the institutional effect on b iology achievement, the significant variance in b iology scores highlights the potential challenges as well as opportunities involved in improving b iology achievement. One challenge i s to make substantial improvements in reading proficiency to minimize the negative effects of socioeconomic status However, this is also an opportunity because improvement in reading proficiency levels will reduce the negative effects of some student char acteristics and will result in improved b iology achievement.

PAGE 151

151 According to the model, reading proficiency accounts for 6 3% of the variation in b iology scores t herefore, it is recommended that schools provide adequate and quality professional development in reading strategies to b iology teachers The recent implementation of Common Core standards that emphasize the development of student vocabulary and reading comprehension may serve as a catalyst in the process especially as those standards will be tested in the future Thus, schools will need to divert additional resources to ensure teachers are not only adequately prepared for biology instruction, but also in incorporating reading strategies within the classroom. Biology achievement was significantly corr elated with ethnicity, but ethnicity was also significantly correlated w ith eligibility for free lunch The inclusion of ethnicity in the model resulted in statistically insignificant values particularly compared to the greater statistical strength of soci oeconomic status and of free lunch as the best predictor of socioeconomic status The choice to include grade level in the model may seem unexpected; however study results indicate d that as grade level increased, reading proficiency declined, there were l ess White and students ineligible for free and reduced price lunch and the sample of Black students increased Therefore grade level represents a conglomeration of the variables associated with ethnicity, reading proficiency, and socioeconomic status. Con sequently, the results provide evidence supporting the significant impact of socioeconomic status on overall b iology and reading achievement. Another important finding in this study is the strong, positive, linear relationship between FCAT Reading scores and Biology I EOC scores. A s FCAT Reading scores increased, Biology I EOC scores also increased. In fact, FCAT Reading developmental

PAGE 152

152 scale scores served as the most significant predictor of b iology achievement in the regression analysis. A lthough correlati ons refrain from making causal statements, they do indicate the presence of a relationship between the two variables. These relationships may be mediated by no, a sole, or a combination of covariates which serve to influence either variable. It is uncertai n whether high FCAT Reading developmental scale scores results in higher Biology I EOC T scale scores or if students with higher Biology I EOC T scale scores perform better in reading assessments than their peers with lower Biology I EOC T scale scores. T hese results beg the question: D oes increased reading proficiency help to accurately predict b iology proficiency? Considering that the Biology I EOC test items are written at a 9th grade reading level (Florida Department of Education, 2011c), a certain lev el of reading proficiency is required. Furthermore, i t is unclear whether the Biology EOC assessment is comprised of compound sentences and complex vocabulary that collectively increase sentence complexity. Conseque ntly, in an attempt to measure b iology pr oficiency, the test may inadvertently be assess ing reading proficiency. In that case, the EOC would fail to provide a valid and fair assessment of the b iology knowledge and skills of less proficient readers Until the test developers address this issue, t h is researcher proposes that increased reading proficiency may increase the probability that scores obtained are representative of stud ent b iology proficiency. However, it is possible that increased reading proficiency may improve scores beyond actual biolo gy proficiency through the use of covariates such as background knowledge, ability to use context clues, and enhanced vocabulary. Thus test creators must ensure that the Biology EOC assesses

PAGE 153

153 student knowledge of b iology concepts and skills rather than undu ly assess ing reading proficiency This research would suggest that test creators should use simple sentence structure s with basic domain general vocabulary. The nature of the Biology EOC sentence complexity is however, beyond the scope of this study. On t he other hand, educators in Florida can make use of the familiarity of the FCAT Reading assessment and resulting score interpretations indicated by the varying achievement levels identified in the study. Since FCAT Reading data is more easily accessible to teachers and administrators that Biology EOC assessment data educators can use reading scores make better instructional decisions in biology to improve student comprehension and achievement. These results also provide an initial basis for identifying stu dents at risk of failing the Biology I EOC assessment who, according to the data, are struggling readers with low FCAT Reading developmental scale scores In addition, instructional activities that improve FCAT Reading developmental scale scores could be i mplemented in the classroom to potentially improve b iology achievement. Based on these results and previous studies, three possible explanations exist regarding the relationship between reading and b iology proficiency : 1) reading proficiency results in b i ology proficiency, 2) b iology proficiency results in reading proficiency, or 3) covariate(s) result in both reading proficiency and b iology proficiency. An unequivocal conclusion to this question is beyond the scope of this study, but extant research may s uggest the influence of covariates (Cromley, 2009). According to Cromley (2009), the high correlations between reading comprehension and science proficiency may be the result of plausible third factors that were not investigated in their

PAGE 154

154 study such as art ifacts of measurement, instructional practices, the familiarity of students with standardized tests, or the alignment of home and school culture s or practices. Cromley (2009) proposed additional covariates that may include the influence of background know ledge, reading comprehension strategies, general vocabulary, inference, and other products of extensive reading experience which increase science proficiency. This hypothesis is supported by recent research which indicates that specific instruction in read ing strategies, vocabulary, and background knowledge have been shown to increase student achievement in science (Fang & Wei, 2010; Guthrie, Wigfield, Barbosa, Perencevich, Taboada, Davis, Scafiddi, &Tonks, 2004; Reaves, 2000; Shymansky, Yore, & Anderson, 2 004; Vitale & Romance, 2012; Yore, Bisanz, & Hand, 2003). Cromley, Snyder Hogan, and Luciw Dubas (2010) also found that background knowledge, inference, and vocabulary had significant effects on b iology achievement for high school students. Additionally, students with more scientific background knowledge exhibit better reading comprehension of scien tific text s due to the direct effect of background knowledge and the indirect effects of strategies and inference (Cromley, Snyder Hogan, & Luciw Dubas, 2010). Consequently, according to Cromley (2009), by increasing reading comprehension of scientific texts, the degree of student proficiency in science increase s (Cromley, 2009). However, the assessments of reading and science used in the Cromley (2009) study we re heavily weighted toward primarily evaluating students utilizing inference s Cromley, Snyder Hogan, and Luciw Dubas's 2010 study also utilized biology text s that relied heavily on inferences. I t is uncertain whether a similar

PAGE 155

155 heavy reliance on inferences is utilized in the Biology I EOC so these results may not be as applicable. Thus the role of inference as a covariate influencing science or reading proficiency in this study is uncertain. Implications Findings of this study indicate that by improving re ading proficiency, all students, regardless of ethnicity, socioeconomic status, or gender will improve their achievement in b iology. Therefore the results can assist educators and policymakers in high stakes testing educational climates to integrat e readi ng in to the content areas at the secondary level. This would require a shift in the allocation of resources and would impact instructional practices. Such a shift could pose a challenge due to limited instructional time, potential inadequate teacher knowle dge of reading pedagogy and its incorporation into the science classroom, and perceived threats to teacher autonomy. However, the recent required incorporation of Common Core standards in Florida classrooms may serve as a catalyst to build teacher profici ency in reading instruction and acquisition of vocabulary Considering that all teachers will be required to incorporate grade level specific reading strategies to satisfy Common Core requirements, only time will reveal whether students will be m ore prepared to handle the levels of reading proficiency required for the biology EOC. The implementation of Common Core standards also raises questions whether the biology EOC will become increasingly more challenging as a means of measuring student profi ciency with the standards. Regardless, s ince student performance serves as a significant influence on teacher evaluations, b iology teachers need to become familiar with and incorporate reading strategies within the classroom so that student achievement is improved. In

PAGE 156

156 fact the study indicates that for every unit increase on the FCAT Reading developmental scale score, b iology scores would increase by 0.38. Therefore, professional development in incorporating reading strategies is highly indicated especial ly for teachers working in Title 1 schools. Of course, solely emphasizing reading to the neglect of b iology content is not advisable In fact, 39 students (22%) in the study who were proficient in reading failed to score in the top third of the biology as sessment so high quality b iology content instruction is clearly essential However, it is recommended that teachers infuse reading instruction into b iology content to improve conceptual understanding and student achievement. Test creators also need to be cognizant of the reading level s of the questions so that they accurately measuring b iology achievement instead of both b iology and reading proficiency. Failure to account for increased reading demands t, may serve to reduce the numbers of African American and low socioeconomic students who graduate from high school and thus limiting their access to beneficial post graduation opportunities such as higher payi ng jobs and advanced education. E ducational policy runs the risk that the goal of c reating high school graduates who are more globally competitive in careers and at the collegiate level has increasingly subjugated a subset of the student population ( Florida Department of Education, 2005b ) W hile legislative directives are aimed at increa sing academic achievement of all students, further analysis must be conducted to determine whether this goal is bei ng met and whether public policy is providing the resources needed to ensure that all students have access to resources as well as highly qua lified and effective b iology teachers.

PAGE 157

157 For school administrators and educational policymakers, the implementation of increased stakes in successive administrations of the Biology end of c ourse assessment may be accompanied by potential drops in school acc ountability indicators due to the performance of struggling readers unless institutional adjustments are made in terms of the allocation of time, resources, and personnel. Thus, it is essential that instructional leaders and administrators utilize data dri ven decision making to analyze both the reading proficiency levels of both the incoming freshman class and all those students currently enrolled in Biology This analysis is particularly important for Title 1 schools that serve populations that have histor ically underperformed in reading. D ue to the sample size and methodology it is unclear whether similar achievement gaps based on gender, race/ethnicity, and socioeconomic status would be found at other schools, particularly Title 1 schools It is therefore also recommended that all schools analyze student achievement data to determine whether all students are being equally served by the instructional personnel and the b iology curriculum. Institutional changes may include providing professional dev elopment in content area reading, culturally appropriate pedagogical practices, gender neutral classrooms, and making instructional personnel changes, etc. The results of the study also highlight potential issues for instructional evaluations and the rec ruitment and retention of b iology teachers in Florida particularly in high poverty schools According to Council of Chief State School Officers (2011), one potential reason for the implementation of e nd of c ourse assessments is to determine teacher effecti veness. However, such a policy is fraught with numerous challenges For example, t his study and countless others have consistently demonstrated the

PAGE 158

158 underperformance of low income and minority students in science ( ACT, 2010; Bankston & Caldas, 1998; Florida Department of Education, 2012b; Gonzales, Williams, Jocelyn, Roey, Kastberg, & Brenwald, 2008; Maerten Rivera, Myers, Lee, & Penfield, 2010; National Center for Education Statistics, 2012; Rathbun & West, 2004; Riegle Crumb, Moore, & Ramos Wada, 2010; Str and, 2012 ) R ecent changes in Florida's instructional evaluations make student achievement scores account for 50% of a teacher's evaluation, so that some teachers could be unfairly penalized for the demographic composition of the student body and its assoc iated challenges. This is particularly critical as all new teachers hired after July 1, 2011 are offered annual contracts in which continued employment and compensation are tied to student outcomes (Harrison & Cohen Vogel, 2012). For this reason any use of test data to determine teacher effectiveness should control for prior student performance ( Council of Chief State School Officers, 2011). However, t he current teacher evaluation tool does not factor in student gains for b iology teachers T her efore a singl e test, the e nd of c ourse assessment, will have a significant impact on a teacher's evaluation continued employment, and opportunities for financial incentives. Research has indicated that an accumulation of classroom and school effects impact on student achievement, therefore relying on such a measure could produce a highly biased estimate of teacher effect s ( Sykes, Schneider, Plank, 2009). Additionally, decisions that impact on student a chievement are made at various levels, such as federal, state, district, school, and classroom. F or example, curriculum decisions may be made at the state and district level as well as other decisions such as

PAGE 159

159 class size teaching personnel, and budgets ; yet such interactions are not considered in a teacher's evaluation ( Sykes, Schneider, Plank, 2009) According to Council of Chief State School Officers (2011), any teacher evaluation measurement should reflect the progress and growth of a student during the period of time the teacher provided instruction. This requires that two elements must be in place : availability of one or more prior scores and a suitable analytic method. However, even when a trustworthy measure of student progress is developed, other challenges have to be addressed in order to associate such results with teacher effectiveness such as extraneous factors that threaten the interpretation that the teacher's behavior led to the observed gains such as truancy, or family crisis during the instructional term (Council of Chief State School Officers, 2011). S tudent characteristics account for the majority of the influence on b iology achievement at the Title 1 school in this study b ut teachers in this environment are ultimately held accountable for student achievement. Considering that r esearch has documented the shortage of qualified instructional personnel in science in rural and Title 1 schools ( Baker & Cooper, 2005 ; Murphy, DeArm ond & Guin, 2003; National Science Foundation, 2012a ), a s well as their high attrition levels ( Barrera, Braley & Slate, 2010 ) compared to their more affluent counterparts policymakers need to devise a more suitable evaluation tool for science teachers work ing at Title 1 schools. Furthermore, policymakers must also ensure that schools serving populations of struggling students also receive adequate resources and effective professional development to address student deficiencies in reading. P oli c ymakers must also ensure that they do not pose

PAGE 160

160 additional burdens on such schools to attract and retain highly qua lified instructional personnel. One proposed solution to the discrepancy in measuring teacher effectiveness is the use of the Value Added Modeling (VAM). T his process utilizes several statistical techniques that seeks to link or establish causality between student test scores on a standardized test and a teacher's performance ( Hill, Kapitula, & Umland, 2011) According to Florid a Department of Education (201 2q), the VAM model takes into account other factors that may affect the learning process and do es not evaluate teachers based on student data derived from a single year or on a comparison of growth from one year to the next. Instead, teachers are evaluated based on the difference between the predicted performance and the actual performance of students. Thus, the model purports to "level the playing field" by statistically accounting for the differences in both the level of student proficiency and student ch aracteristics that are assigned to teachers ( Florid a Department of Education, 2012q). Therefore, no teacher is unfairly advantaged or disadvantaged simply based on student composition. However, Green III, Baker, & Oluwole (2012) cited multiple problems wit h the VAM approach. They noted the "instability of teacher ratings, classification and model prediction error, unreliable results from different 'standardized' tests, difficulties in isolating a single teacher's contribution to students' learning, the non random assignment of students across teachers, schools, and districts, and the struggle for teachers to even receive VAM ratings" (Green III, Baker, & Oluwole, 2012, p. 6). Hill, Kapitula, & Umland 's 2011 study also revealed that although VAM scores conve rged with the ratings from experts regarding their instruction, data also revealed that the

PAGE 161

161 scores correlated with the student composition of a teacher's classroom. Thus the authors advocate the use of VAM scores alongside high quality, discri minating obse rvational systems or as a trigg er for the use of observations, as VAM scores alone are not sufficient to identify teachers in need of remediation, reward, or removal. Florida's current system does utilize a combination of VAM scores and discriminatory obs ervational systems ; however whether this system is equitable in implementation is the subject of further analysis. U nlike reading and mathematics assessments, data is limited at the eighth grade level for FCAT Science 2.0. Thus, the majority of students w ho take the Biology EOC at the tenth grade level have not taken a state mandated science assessment since the eighth grade. Thus, the data being used by policymakers in calculating VAM for b iology teachers is subject to larger errors than their reading and mathematics counterparts. Finally this study reveals that reading proficiency explains a significant portion of the variability in b iology achievement A lthough the role of the tea cher was not directly addressed in this study one can reasonably conclude that b iology achievement is the product of more external factors than just teacher. T he VAM model, based on this finding would unduly penalize teachers who have students with low reading proficiency levels. On a national scale, the results hav e some positive and negative indicators for the future. On the positive note, the reduction in the achievement gap between genders may represent a potential increase in women entering STEM fields in the future. Although this study utilized a small sample f rom one high school, data reported from elsewhere in the state paints a similar picture. However, the improved achievement of one group is not representative of other subsets of the population,

PAGE 162

162 particularly low socioeconomic and African American students. T hese groups were disproportionately represented in the bottom third of the assessment scores. If this trend continues, these groups are at a greater risk of failing this assessment required for graduation, thereby creating gaps in future opportunities. Th e diversity of the future science workforce will therefore continue to be compromised, reducing our nation's viability and global competitiveness while simultaneously increasing both intellectual and fiscal poverty and reduc ing the earning potential of a s ubset of the nation's workforce. Limitations of the Study Caution should also be exercised in interpreting the findings particularly because of the discrepanc ies in the measurement of reading proficiency. While the FCAT Reading assessment was used to eval uate all students, the scores are not reflective of proficiency based on a single and identical test. In fact, although the ninth grade students were tested, their test content varied remarkably from those enrolled at the tenth grade level, as well as thos e in the eleventh grade who may have taken the FCAT Reading test on more than one occasion. Additionally, students enrolled in grades 9 and 10 completed the newly implemented FCAT 2.0 Reading test in Spring 2012 whereas eleventh grade students completed th e FCAT Reading, an older version of the assessment based on different content standards (Florida Department of Education, 2012k) that are less demanding and rigorous ( Florida Department of Education, 2012e) S ince these scores we re obtained from different tests measuring different content standards and may have been taken during different points of the school year, the scores are not directly comparable although all are criterion referenced tests in reading created by the same publisher.

PAGE 163

163 Another discrepancy within the study is the conservative classification of student s achievement on the Biology I e nd of c ourse assessment. For this study, students who performed in the top third in the state were considered as proficient or having passed the Biology I e nd of c ourse assessment ; however the state mandated achievement levels were not determined until s pring 2013 (Florida Department of Education, 2012j). Additionally, the scores provided by the state for the 201 3 assessment administration will be markedly differ ent from the 201 2 assessment administration. For example, the 2012 administration scores were represented using raw scores and scores derived from a T scale range from 20 to 80, whereas the scores from the 2013 administration will be reported on a scale fr om 325 and 475 (Florida Department of Education, 2012l). Additionally, due to the geographic location, demographic composition, and level of academic achievement at the specified Florida Title 1 high school, the results of the study have limited generaliz ability beyond the sample group and ultimately, the school population Of particular concern is the small sample size of Hispanic and American Indian students at the North Florida school and the limited number of eleventh grade students. Additionally, the unbalanced sample based on ethnicity and grade level also reduces the generalizability of the study to a larger population. Consequently, the results are not considered strong enough to inform policy, although the implications for education and re search are worthwhile. Nevertheless caution should be exercised when applying the results of this study to other settings. Finally the lack of c ontrol over the course placement of students served as another limitation At the Title 1 high school, student s enrolled at the h onors level were more similar to each other in socioeconomic standing (high socioeconomic ) than

PAGE 164

164 students enrolled at the regular level. Furthermore, students at the h onors level were disproportionately of White heritage whereas those enr olled at the r egular level were disproportionately Black. Additionally, all students enrolled in the h onors biology course at the North Florida high school were taught by the same teacher, whereas students at the r egular level were taught by one of two tea chers. Thus the quality and type of instruction in b iology serves as a confounding variable. Recommendations for Future Research This study points to the need for further research on the variables that impact student achievement in b iology at the secondar y level. T he influence of covariates on student performance necessitates more research to isolate variables in a quantifiable manner. Future research should include: Specific reading interventions that significantly improve student achievement in b iology a t the secondary level Although most research on science achievement regarding reading interventions are broad, there is a paucity of research about biology at the secondary level. Based on the results of this study, level of reading proficienc y is important as it has an influence on b iology proficiency. T he strength and positive nature of the relationship and findings from previous studies suggest that reading interventions may serve to improve b iology achievement. Current studies indicate that interventions focused on incorporating reading strategies specifically improving background knowledge and vocabulary should result in significant improvements in b iology achievement. However, this assumption is untested at the secondary level An exami nation of the comparability of the scores derived from a computer and paper based administration of the Biology I EOC. Although the Florida Department of Education (2006) cited previous research regarding the comparability of b iology scores attained via co mputer and paper based test administrations, the current computer platform being utilized has yet to be tested with empirical research. Therefore, it is unknown whether certain student populations are disproportionately affected by the use of the computer based testing with the existing platform and whether other factors, such as connection speed or school network constraints have a negative impact on student performance. Empirical research is also needed to investigate b iology multiple choice test items, some of which contain long passages that are timed and non

PAGE 165

165 adaptive. Although study results are inconsistent there is a possibility that reading from a computer is different from paper based reading ; it is slower, result ing in reduced accuracy and reduced reading comprehension (Noyes & Garland, 2008) T hus it is uncertain whether the differences between computer and paper based administrations exacerbates differences in reading proficiency. Nevertheless, the specifics regarding this type of assessment are salient, especially due to the high stakes nature of the assessment. An investigation of the relationship between math proficiency and b iology student achievement. A previous study has indicated that reading and mathematics accounted for 58% of the variati on in science achievement (Maerten Rivera, Myers, Lee, & Penfield, 2010). However, this study was not specific to b iology; therefore it is worth investigating whether mathematics proficiency contributes specifically to b iology achievement. A study on the degree to which institutional factors impact b iology achie vement. In this study, 60 % of the variance in b iology achievement was due to student characteristi cs, but whether the remaining 40 % is due to institutional factors ( such as teacher quality, instruct ional strategies, and classroom environment ) is the subject of further investigation. Research on whether females score significantly higher when instructed by female b iology teachers. Research has documented the negative impact of stereotypes on academic achievement (Appel, Kronberger, & Aronson, 2011; Hoy & Hoy, 2009; Pico & Stephens, 2012; Steele & Aronson, 1995) as well as the positive impact of counter stereotypic images on academic performance (Good, Woodzicka, & Wingfield, 2010). Therefore, female te achers may serve as counter stereotypic role models for female students, and may potentially improve student achievement in b iology. However, research specific to the academic impact of female biology teachers compared to their male counterparts is scant. Summary Th is study investigated the influence and relationship of four variables -reading proficiency ethnicity, socioeconomic status, and gender -on the academic performance of students in b iology. Overall, it determined that reading proficiency ha s a si gnificant effect on b iology achievement, although reading proficiency itself is significantly influenced by ethnicity and socioeconomic status. Therefore, of the four variables investigated, one of them (reading proficiency) could be used in future researc h to predict academic performance in b iology. It is therefore important that this variable be

PAGE 166

166 taken into consideration by education officials, policymakers, and stakeholders in formulating policies teacher evaluation tools, and assessments. F urther resear ch should focus on this area so that the specific factors that account for differences in reading proficiency and its subsequent effects on b iology achievement can be investigated. Consequently, the right measures can then be applied to help all students i mprove their academic achievement in state mandated b iology assessments.

PAGE 167

167 APPENDIX A DISTRICT CONSENT FORM Dear School District Personnel, My name is Janine Bertolotti and I am a graduate student in Educational Administration and Policy at the University o f Florida under the supervision of Dr. Bernard Oliver. I am conducting research on the relationship between four variables FCAT Reading, ethnicity, socioeconomic status, and gender; and their influence on the results of the Florida Biology EOC exam. I wou ld appreciate permission to use information from the district files on your ninth and tenth grade students who were enrolled in Biology 1 during the 2011 2012 school year. Specifically, I would like to collect data on their gender, grade level, Biology 1 E OC scores, FCAT Reading scores, and eligibility for free and reduced price lunch. Only my supervisor and I will have access to this personal information. All data will be coded and kept confidential; no names or any identifying information will be reported If you have any questions, please feel free to contact me or my faculty supervisor, Dr. Bernard Oliver, at (352) 273 4358. I will be contacting your office within the next two weeks to confirm your approval. Sincerely, Janine C. Bertolotti Graduate Student in Educational Leadership

PAGE 168

168 APPENDIX A IRB CORRESPONDENCE

PAGE 169

169 APPENDIX B DISTRICT APPROVAL LETTER May 13, 2013 Janine Bertolotti University of Florida Gainesville, Fl 32605 Dear Ms. Bertolotti, This letter is to inform you that our High School has reviewed and supports your research study titled, An investigation of the student characteristics that influence Florida Biology end of c ourse scores at a Title 1 school ." It is our understanding that the data collection will begin on May 14, 2013. We are very interested in your efforts that may help to improve our understanding of Biology achievement on the end of c ourse exam. If you have any questions or need further assistance, please do not hesitate to contact me. Sincerely, Principa l

PAGE 170

170 APPENDIX B DATA DICTIONARY Variable Description Coding ID Student id number # Gender Gender of student Male = 1 Female = 2 Free/reduced lunch (FRLunch) What is the eligibility of the student regarding free/reduced lunch? Free = 1 Reduced = 2 Inel igible = 3 Free/reduced lunch (Freereducedlunch) Does the student qualify for free/reduced lunch? Yes= 1 No = 2 Ethnicity What is the ethnic background of the student? White = 1 Black = 2 Hispanic = 3 American Indian or Alaskan native = 4 Pass or fail F CAT Reading (PFReading) Did the student receive a grade of 3 or higher on the reading FCAT? Yes = 1 No = 2 FCAT Reading level (RLevel) What is the student's FCAT reading level? 1 = 1 2 = 2 3 = 3 4 = 4 5 = 5 FCAT Reading Developmental Scale Score (RDevSco re) What is the student's FCAT Reading developmental scale score? # Grade level (GradeLevel) What is the student's grade level? 9 = 9 10 = 10 11 = 11 Biology EOC Level (BioLevel) Which third did the student score on the Biology EOC? 1 = 1 2 = 2 3 = 3 Bi ology T score (BioTscore) What T scale score did the student earn on the Biology EOC exam? (20 80) #

PAGE 171

171 Pass or fail BIO EOC (PFBioEoc) Did the student score in the top third on the Bio EOC? Yes = 1 No = 2 Regression gender (Reg r ess gender) How is gender coded in the regression? Male = 0 Female = 1 Regres s ion ine li gible ( Regress NOT eligible) How are students ineligible for free and reduced price lunch coded in the regression ? Eligible = 0 Ineligible = 1 Regression reduced lunch (Reg ress reduced eligible ) Ho w are studen ts who are eligible for reduced price lunch coded in the regression ? Free and not eligible = 0 Reduced =1 Regression free lunch (Regress free lunch ) How are students who are eligible for free lunch coded in the regression ? Reduced and Not eligib le = 0 Free = 1 Regression White students (Regress White) H ow are White students coded in the regression? Non white = 0 White = 1 Regression Black students (Regress Black) H ow are Black students coded in the regression? Non Black = 0 Black = 1 Regression Hispanic students (Regress Hispanic) How are Hispanic students coded in the regression? Non Hispanic = 0 Hispanic = 1 Regression American Indian or Alaskan native students (Regress AmInAlNa) How are American Indian or A laskan native students coded in the re gression? Non American Indian/Alaskan native = 0 American Indian/Alaskan native = 1 Regression minority students (Regress minority) How are minorities coded in the regression? Minority = 0 Non minority = 1 Regression 10th grade students (Regress 10) How are 10th grade students coded in the regression? 9th grade = 0 10th grade = 1 Regression 11th grade students (Regress 11) How are 11th grade students coded in the regression? 9th and 10th= 0 11th grade = 1

PAGE 172

172 LIST OF REFERENCES ACT (2010). The condition of college & career readiness 2010. Retrieved from http://www.act.org/research/policymakers/cccr10/pdf/ConditionofCollegeandCare erReadiness2010.p df Allen, J., Gregory, A., Mikami, A., Lun, J. Hamre, B., & Pianta, R. (2013). Observations of effective teacher student interactions in secondary school classrooms: Predicting student achievement with the classroom assessment scoring system secondary. Sc hool Psychology Review, 42 (1), 76 97. American Association for the Advancement of Science (AAAS) Project 2061 (1990). Science for all Americans. New York, NY: Oxford University Press. American Association for the Advancement of Science (1989). Science for all Americans New York, NY: Oxford University Press. American Educational Research Association (2006, April 9). 43.010. AERA Distinguished lecture: Presidential Session. Retrieved from http://www.softconference.com/MEDIA/WMP/260407/260407 43.010.asp [video] American Educational Research Association (2000). Position statement on high stakes testing. Retrieved from http://www.aera.net/AboutAERA/AERARulesPolicies/AERAPolicyStatements/Pos itionStatementonHighStakesTesting/tabid/11083/Default.aspx American Educational Research Associatio n, American Psychological Association, & National Council on M easurement in Education (1999). Standards for Educational and Psychological T esting. Washington, DC: American Educational Research Association. Anderson, K.J.B. (2011). Science education and test based accountability: Reviewing their relationship and exploring implications for future policy. Science Education, 96 (1), 104 129. Appel, M., Kronberger, N., & Aronson, J. (2011). S tereotype threat impairs ability building: Effects on test preparation among women in science and technology. European Journal of Social Psychology, 41 (7), 904 913. Aud, S., Fox, M., & KewalRamani, A. (2010). Status and Trends in the Education of Racial an d Ethnic Groups (NCES 2010 015). U.S. Department of Education, National Center for Education Statistics. Washington, DC: U.S. Government Printing Office. Baharudin, R. & Luster, T. (1998). Factors related to the quality of the home environment and children 's achievement. Journal of Family Issues, 19 (4), 375 403.

PAGE 173

173 Baker, E.L. (2001). Testing and assessment: A progress report. Educational Assessment, 7 (1), 1 12. Ban kston III, C.L. & Caldas, S.J. (1998). Family structure, schoolmates, and racial inequalities in school achievement. Journal of Marriage & the Family, 60, 715 724. Barrera, A., Braley, R. T., & Slate, J. R. (2010). Beginning teacher success: An investigation into the feedback from mentors of formal mentoring programs. Mentoring & Tutoring: Partnershi p in Learning, 18 (1), 61 74. Beck, I.L., McCaslin, E.S., & McKeown, M.G. (1980). The rational and design of a program to teach vocabulary to fourth grade students (ERDC Publication 1980/25). Pittsburgh: University of Pittsburgh, Learning Research Developm ent Center. Best, R.M., Rowe, M., Ozuru, Y., & McNamara, D.S. (2005). Deep level comprehension of science texts: The role of the reader and the text. Topics in Language Disorders, 25 (1), 65 83. Bhanot, R.T. & Jovanovic, J. (2009). The links between parent behaviors and boys' and girls' science achievement beliefs. Applied Developmental Science, 13 (1), 42 59. Biancarosa, G. & Snow, C.E. (2004). Reading Next A Vision for Action and Research in Middle and High School Literacy: A Report from Carnegie Corporati on of New York. Washington, DC: Alliance for Excellent Education. Blank, R., Langesen, D., & Petermann, A. (2007). State indicators of science and mathematics education. Retrieved from http://programs.ccsso.org/content/pdfs/SM%2007%20report%20part%201.pdf Blazer, C. (2012). National trends in end of c ourse assessment programs. Information capsule. Volume 1202. Miami Dade County. Retrieved from http://www.eric.ed.gov/ERICWebPortal/contentdelivery/servlet/ERICServlet?accn o=ED536522 Bleeker, M. M. & Jacobs, J. E. (2004). Achievement in math and science: Do mothers' belief s matter 12 years later? Journal of Educational Psychology, 96 (1), 97 109. Borg, M. O., Plumlee, J.P., & Stranahan, H.A. (2007). Plenty of children left behind: High stakes testing and graduation rates in Duval county, Florida. Educational Policy, 21 (5), 6 95 716. Bray, G. B., Pascarella, E. T., & Pierson, C. T. (2004). Postsecondary education and some dimensions of literacy development: An exploration of longitudinal evidence. Reading Research Quarterly, 39 (3), 306 330.

PAGE 174

174 Butchart, R. E. (2010). Black hope, w hite power: Emancipation, reconstruction and the legacy of unequal schooling in the US South, 1861 1880. Paedagogica Historica, 46 (1/2), 33 50. Bybee, R.W. (2010). What is STEM Education? Science 329( 5994 ), 996 999. Cadinu, M., Maass, A., Rosabianca, A., & Kiesner, J. (2005). Why do women underperform under stereotype threat? Evidence for the role of negative thinking. Psychological Science, 16 (7), 572 578. Caldas, S.J. & Bankston III, C. (1997). Effect of school population socioeconomic status on indivi dual academic achievement. Journal of Educational Research, 90( 5 ), 269 277. Carnoy, M. & Loeb, S. (2002). Does external accountability affect student outcomes? A crossstate analysis. Educational Evaluation and Policy Analysis 24 (4), 305 332. Center on Edu cational Policy (2008, August). State high school exit exams: A move toward end of course exams. Retrieved from www.cep dc.org%2Fcfcontent_file.cfm%3FAttachment%3DCEP_HSEE08Report_081308. pdf&ei=2WRsUryQNaTC2wWL94GACw&usg=AFQjCNHadAcjYxNKwPmIrle98B 3GkVnxBA&sig2=GX9XJQc0l CmGLaUGLICczQ Chambers, J.G., Levin, J.D., & Parrish, T.B. (2006). Examining the relationship between educational outcomes and gaps in funding: An extension of the New York adequacy study. Peabody Journal of Education, 81 (2), 1 32. Chomitz, V.R., Slining, M.M., McGowan, R.J., Mitchell, S.E., Dawson, G.F., & Hacker, K.A. (2009). Is there a relationship between physical fitness and academic achievement? Positive results from public school children in the Northeastern United States. Journal of School Health, 79 (1), 30 37. Christensen, C. M., Horn, M. B., & Johnson, C. W. (2011) Disrupting Class: How Disruptive Innovation Will Change the Way the World Learns New York: McGraw Hill. Coleman, J.S., Campbell, E., Hobson, C., McPartland, J., Mood, A., Weinfield, F ., & York, R. (1966). Equality of educational opportunity. Washington, DC: U.S. Government Printing Office. College Board (2011). 2011 College Bound Seniors: Total group profile report. Retrieved from http://professionals.collegeboard.com/profdownload/cbs2011_total_group_report. pdf Common Core Standards Initiative (2012 a ) Frequently asked questions. Retrieved from http://www.corestandards.org/resources/frequently asked questions

PAGE 175

175 Common Core Standards Initiative (2012 b ). Key points in English Language Arts. Retrieved from http://www.corestandards.org/resources/key points in english language arts Common Core Standards Initiative (2012 c ). Mission Statement. Retrieved from http://www.corestandards.o rg Condron, D.J. & Roscigno, V.J. (2003). Disparities within: Unequal spending and achievement in an urban school district. Sociology of Education, 76 (1), 18 36. Council of Chief State School Officers (2011). State end of course testing programs: A poli cy brief. Retrieved from http://www.ccsso.org/Documents/2011/State_End of Course_Testing_Programs_2011.pdf Cromley, J.G. (2009). Reading achievement and scie nce proficiency: International comparisons from the programme on international student assessment. Reading Psychology, 30 89 118. Cromley, J.G., Snyder Hogan, L.E., & Luciw Dubas, U.A. (2010). Reading comprehension of scientific text: A domain specific test of the direct and inferential mediation model of reading comprehension. Journal of Educational Psychology, 102 (3), 687 700. Darling Hammond, L. (2006). No Child Left Behind and High School Reform. Harvard Educational Review, 76 (4), 642 663. Davis, C.L & Cooper, S. (2011). Fitness, fatness, cognition, behavior, and academic achievement among overweight children: Do cross sectional associations correspond to exercise trial outcomes? Preventative Medicine, 52 (1), S65 S69 Demps, D. L. & Onwuegbuzie, A. J. (2001). The relationship between eighthgrade reading scores and achievement on the Georgia High School Graduation Test. Research in the Schools, 8 (2), 1 9. Dempster, E. R. & Reddy, V. (2007). Item readability and science achievement in TIMSS 2003 in Sout h Africa. Science Education, 91 (6), 906 925. Desimone, L.M. & Long, D. (2010). Teacher effects and the achievement gap: Do teacher and teaching quality influence the achievement gap between Black and White and high and low SES students in the early grades ? Teachers College Record, 112 (12), 3024 3073. Dewitt, J., Archer, L., Osborne, J., Dillon, J., Willis, B., & Wong, B. (2011). High aspirations but low progression: The science aspirations careers paradox amongst minority ethnic students. International Jou rnal of Mathematics and Science Education, 9 243 271.

PAGE 176

176 Doppen, F. S. (2007). Pre Standards, High Stakes. International Journal of Social Education, 21 (2), 18 45. Easton Brooks, D. & Davis, A.(2007, Fall) Wealth, traditional socioeconomic indicators, and the achievement debt. Journal of Negro Education, 76 (4), 530 541. Ediger, M. (2005, March). Assessing reading in the science curriculum. College Student Journal, 39 (1), 26 30. Education Information and Ac countability Services (2009, September). Data Report: Series 2010 07D Free/reduced price lunch eligibility. Retrieved from http://www.fldoe.org/eias/eiaspubs/pdf/frplunch.pdf Escudier, M .P., Newton, T.J., Cox, M.J., Reynolds, P.A., & Odell, E.W. (2011). University students' attainment and perceptions of computer delivered assessment; a comparison between computer based and traditional tests in a 'high stakes' examination. Journal of Compu ter Assisted Learning, 27 440 447. Everson, H.T. & Millsap, R.E. (2004). Beyond individual differences: Exploring school effects on SAT scores. Educational Psychologist, 39 (3), 157 172. Fang, Z. (2006). The Language Demands of Science Reading in Middle Sc hool. International Journal of Science Education, 28( 5 ), 491 520. Fang, Z. & Wei, Y. (2010). Improving middle school students' science literacy through reading infusion. Journal of Educational Research, 103 262 273. Feathers, K.M. (2004). Infotext: Readi ng and Learning. Toronto: Pippin Publishing. Fen sham, P.J. ( 2009). The link between policy and practice in science education: The role of research. Science Education, 93 (6), 1076 1095. Fisher, D., Ross, D., & Grant, M. (2010, Jan ). Building background knowledge. Science Teacher, 77 (1), 23 26. Fleischman, H.L., Hopstock, P.J., Pelczar, M.P., & Shelley, B.E. (2010). Highlights From PISA 2009: P erformance of U.S. 15 Year Old Students in Reading, Mathematics, and Science Literacy in an International Context (NCES 2011 004). U.S. Department of Education, National Center for Education Statistics. Washington, DC: U.S. Government Printing Office. Flor id a Department of Education (2010 ) Next Generation Sunshine State Standards. Retrieved from http://www.fldoe.org/bii/curriculum/sss/ Florida Department of Education (2012a). Biology I End Of Course Assessment Next Generation Sunshine State Standards: State Report of Districts. Retrieved from http://fcat.fldoe.org/mediapacket/2012/xls/Biology1EOCStateRepo rtDistrictsv3.xls

PAGE 177

177 Florida Department of Education (2012b). Biology I End Of Course Assessment Statewide Demographic Results: Spring 2012 Combined forms. Retrieved from http://fcat.fldoe.org/mediapacket/2012/xls/Spr2012_Bio1EOC_State_Demographi c_Report_Combined_Forms.xls Florida Department of Education (2012c). 2012 Florida Comprehensive Assessment Test 2.0 (FCAT 2.0) Next Generation Sun shine State Standards State Report of District Results Grade 05 Science. Retrieved from http://fcat.fldoe.org/mediapacket/2012/xls/2012_GR05_Sdisv2.xls Florida Department o f Education (2012d). 2012 Florida Comprehensive Assessment Test 2.0 (FCAT 2.0) Next Generation Sunshine State Standards State Report of District Results Grade 08 Science. Retrieved from http://fcat.fldoe.org/mediapacket/2012/xls/2012_GR08_Sdisv2.xls Florida Department of Education (2012e). 2012 Florida Comprehensive Assessment Test 2.0 (FCAT 2.0) Next Generation Sunshine State Standards State Report of District Results G rade 10 Reading. Retrieved from http://fcat.fldoe.org/mediapacket/2012/xls/GR10_Rdis_2012v4.xls Florida Department of Education (2012f, September). Graduation requirements fo r Florida's statewide assessments. Retrieved from http://fcat.fldoe.org/pdf/fcatpass.pdf Florida Department of Education (2012g, May 24). Spring 2012 Biology 1 End of Course Assessment: State Summary. Retrieved from http://fcat.fldoe.org/mediapacket/2012/pdf/Biology1StateSummary.pdf Florida Department of Education (2012h). All districts: Non high schools. Retrieved fro m http://schoolgrades.fldoe.org/xls/1112/All districts 11 12.xls Florida Department of Education (2012i). 2012 Biology 1 End of Course (EOC) Assessment Fact Sheet. Retrieved f rom http://fcat.fldoe.org/eoc/pdf/12beocfs.pdf Florida Department of Education Office of Assessment (2012j). FCAT 2.0 and Florida EOC Assessments Achievement Levels. Retrieved from http://fcat.fldoe.org/fcat2/pdf/achlevel.pdf Florida Department of Education (2012k). Fall 2012: Reading & Mathematics Retake Test administration manual. Retrieved from http://www.pearsonaccess.com/media/fl/docs/FL00005262_F12RMRt_TAM_WT _FINAL_APPROVED.pdf Florida Department of Education (2012l). Statement from Commissioner of Education Pam Stewart on Achi evement Level Recommendations for Geometry and Biology End of Course Assessments and FCAT 2.0 Science. Retrieved from http://www.fldoe.org/news/2012/2012_11_19 2.asp

PAGE 178

178 Florida Department of Edu cation (2012m). Standard setting recommendations for FCAT 2.0 Science (grades 5 and 8), Biology 1 and Geometry EOC assessments. Retrieved from http://fcat.fldoe.org/pdf/StandardsImpact.pdf Flor ida Department of Education (2012n).Detailed information on high schools. Retrieved from http://schoolgrades.fldoe.org/xls/1112/HighSchoolGradesAllDistricts.xls Florid a Department of Education (2012o).School, district, and state public accountability report. Retrieved from http://doeweb prd.doe.state.fl.us/eds/nclbspar /year1112/nclb1112.cfm?dist_schl=13_7361 Florida Department of Education (2012p). School, district, and state public accountability report. Retrieved from http://doeweb prd.doe.state.fl.us/eds/nclbspar/year1112/nclb1112.cfm?dist_schl=40_11 Florid a Department of Education (2012q). Florida's value added model: Overview of the model used to measure student learning growth on FCAT Reading and Mathematics. Ret rieved from http://www.feraonline.org/presentations/FloridaValueAddedModelJuanCopaFLD OE.pdf Florida Department of Education (2011a). Florida Comprehensive Asse ssment Test (FCAT) 2011 Sunshine State Standards State Report of School Results Grade 11 Science. Retrieved from http://fcat.fldoe.org/xls/2011/F11_GR11_Ssch.xls Florida Department of Ed ucation (2011b). Science scores: Statewide Comparison for 2001 to 2011 FCAT Science Sunshine State Standards Test. Retrieved from http://fcat.fldoe.org/mediapacket/2011/p df/2011ScienceComparison.pdf Florida Department of Education (2011c). Biology I End of course assessment test item specifications. Retrieved from http://fcat.fldoe.org/eoc/pdf/BiologyFL11Sp.p df Florida Department of Education (2011d). Florida differentiated accountability program: 2011 2012 School Improvement Plan. Retrieved from http://www.flbsi.org/1112_SIP/Public/pr int.aspx?uid=400011 Florida Department of Education (2010). Florida Comprehensive Assessment Test (FCAT) 2010 Sunshine State Standards State Report of School Results Grade 11 Science. Retrieved from http://fcat.fldoe.org/xls/2010/F10_GR11_Ssch.xls Florida Department of Education (2006). What do we know about choosing to take a high stakes test on a computer? Retrieved from http://www.fldoe.org/asp/k12memo/pdf/WhatDoWeKnowAboutChoosingToTakeA HighStakesTestOnAComputer.pdf Florida Department of Education (2005a). Florida End of Course (EOC) Assessments. Retrieved from http://fcat.fldoe.org/eoc/

PAGE 179

179 Florida Department of Education (2005b). Information about Florida's State of Board of Education Strategic Plan 2012 2018. Retrieved from ht tp://www.fldoe.org/Strategic_Plan/ Florida Department of State (2013). 6A 6.054 K 1 2 Student Reading Intervention Requirements. Retrieved from https://www.flrules.org/gateway/not ice_Files.asp?ID=12969693 Florida Department of State (2011). 6A 5.090 Content Area Reading Professional Development and Next Generation Content Area Professional Development. Retrieved from https://www.flrules.org/gateway/notice_Files.asp?ID=9823789 Florida Legislature (2012). The 2012 Florida Statutes: 1008.22 Student assessment program for public schools. Retrieved from http://www.leg.state.fl.us/Statutes/index.cfm?App_mode=Display_Statute&Searc h_String=&URL=1000 1099/1008/Sections/1008.22.html Florida Legislature (2010). 2010 Florida St atutes (including Special Session A). 1012.34 Assessment procedures and criteria. Retrieved from http://www.flsenate.gov/laws/statutes/2010/1012.34 Frein, S.T. (2011). Comparing in class a nd out of class computer based tests to traditional paper and pencil tests in introductory psychology courses. Teaching of Psychology, 38 (4), 282 287. Gay, L.R., Mills, G.E., & Airasian, P. (2009 ). Educational Research: Competencies for Analysis and Applic ations Upper Saddle River, NJ: Pearson. George, R. (2006). A cross domain analysis of change in students' attitudes toward science and attitudes about the utility of science. International Journal of Science Education, 28 (6) 571 589. Gibson, H. L. & Chas e, C. (2002). Longitudinal impact of an inquiry based science Science Education 86 (5) 693 705. Golden, D. (2003, December 29). Initiative to leave No Child Behind leaves out gifted: Educators divert resources from classes for smartest to focus on basic literacy. The Wall Street Journal, Retrieved from http://online.wsj.com/public/resources/documents/Polk_Gifted.htm Good, J.J., Woodzicka, J.A., & Wingfield, L.C. (2010). The effects of gender stereotypic and counter stereotypic textbook images on science performance. Journal of Social Psychology, 150 (2), 132 147. Goodwin, B., Englert, K., & Cicchinelli, L. (2003). Co m prehensive accountability systems: A framework for evaluation. Aurora, CO: Mid continent Research for Education and Learning.

PAGE 180

180 Gonzales, P., Williams, T., Jocelyn, L., Roey, S., Kastberg, D., & Brenwald, S. (2008). Highlights from TIMSS 2007: Mathematics and Science Achievement of U.S. Fourth and Eighth Grade Students in an International Context (NCES 2009 001 Revised). National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education. Washington, DC. Green III, P.C. Baker, B.D., & Oluwole, J. (2012). The legal and policy implications of value added teacher assessment policies. Brigham Young University Education & Law Journal, 1, 1 29. Greenfield, T. A. (1997). Gender and grade level differences in science interest and participation. Science Education, 81 (3) 259 276. Gregory, A. & Weinstein, R.S. (2008). The discipline gap and African Americans: Defiance or cooperation in the high school classroom. Journal of School Psychology, 46 (4), 455 475. Grissmer, D. W., Kirby S. N., Berends, M., & Williamson, S. (1994). Student achievement and the changing American family: An executive summary. Retrieved from the Rand on line database on the World Wide Web: http://www.rand.org/publications/MR/MR535/MR535.html Grossman, P.L. & Stodolsky, S.S. (1995). Content as context: The role of school subjects in secondary school teaching. Educational Researcher, 24( 8 ), 5 11. Gurley Calvez T. & Higginbotham, A. (2010). Childhood obesity, academic achievement, and school expenditures. Public Finance Review, 38 (5), 619 646. Guthrie, J. T., Wigfield, A., Barbosa, P., Perencevich, K. C., Taboada, A., Davis, M. H., Scafiddi, N.T, & Tonks, S. (2004). Increasing reading compreh ension and engagement through Concept Oriented Reading Instruction. Journal of Educational Psychology, 96 (3), 403 423. Haladyna, T., Nolen, S., & Haas, N. (1991). Raising standardized achievement test scores and the origins of test score pollution. Educati onal Researcher, 20 (5), 2 7. Hanushek, E.A. & Raymond, M.E. (2005). Does school accountability lead to improved student performance? Journal of Policy Analysis and Management, 24 (2), 297 327. Harmon, J.M., Hedrick, W.B., & Wood, K.D. (2005). Research on vo cabulary instruction in the content areas: Implications for struggling readers. Reading & Writing Quarterly, 21, 261 280. Harrison, C. & Cohen Vogel, L. (2012). The politics of teacher reform in Florida: Analyzing causal narrative surrounding state adoptio n of performance based evaluations, performance pay, and tenure elimination. Peabody Journal of Education, 87, 517 534.

PAGE 181

181 Haught, P. A. & Walls, R. T. (2004). Relationships of reading, MCAT, and USMLE Step 1 Test results for medical students. Reading Psychol ogy: An International Quarterly, 25 (2), 83 92. Haworth, C.M.A., Dale, P.S., & Plomin, R. (2009). Sex differences and science: The etiology of science excellence. Journal of Child Psychology and Psychiatry, 509 (9), 1113 1120. Hidi, S. & Harackiewicz, J. M. (2000). Motivating the academically unmotivated: A critical issue for the 21st century. Review of Educational Research 70 (2), 151 179. Hill, H.C., Kapitula, L., & Umland, K. (2011). A validity argument approach to evaluating teacher value added scores. A merican Educational Research Journal, 48 (3), 794 831. Hines, E. & Holcomb McCoy, C. (2013). Parental characteristics, ecological factors, and the academic achievement of African American males. Journal of Counseling & Development, 91( 1 ), 68 77. Hollar, D., Messiah, S.E., Lopez Mitnik, G., Hollar, L., Almon, M. & Agatston, A.S. (2010, April). Effect of a two year obesity prevention intervention on percentile changes in body mass index and academic performance in low income elementary school children. America n Journal of Public Health, 100 (4), 646 653. Homan, S., Hewitt, M., & Linder, J. (1994). The development and validation of a formula for measuring single sentence test item readability. Journal of Educational Measurement, 31 (4), 349 358. Horne, J. (2007 ). Gender differences in computerised and conventional educational tests. Journal of Computer Assisted Learning, 23, 47 55. House, J.S. & Williams, D.R. (2000). "Understanding and Reducing Socioeconomic and Racial/Ethnic Disparities in Health." Pp. 81 124 in Promoting Health: Intervention Strategies From Social and Behavioral Research Eds. B. D. Smedley and S. L. Syme Washington, D.C.: National Academy of Sciences Press. Hoy, A. W. & Hoy, W. K. (2009). Instructional leadership: A research based guide to learning in schools (3rd ed.) Boston: Allyn and Bacon. Johnson, C.C. ( 2013). Educational turbulence: The influence of macro and micro policy on science education reform. Journal of Science Teacher Education, 24 (4), 693 715. Kahlenberg, R.D. ( 2001 ). All together now: Creating middle class schools through public school choice. Washington, DC: Brookings Institution.

PAGE 182

182 Kendeou, P. & Van Den Broek, P. (2007). The effects of pr ior knowledge and text structure on comprehension processes during reading of scientific texts. Memory & Cognition, 35 (7), 1567 1577. Kingston, N.M. (2009). Comparability of computer and paper administered multiple choice tests for K 12 populations: A synt hesis. Applied Measurement in Education, 22 22 37. Konstantopoulos, S. & Chung, V. (2011). Teacher effects on minority and disadvantaged students' grade 4 achievement. The Journal of Educational Research, 104 73 86. Kress, S., Zechmann, S., & Schmitten, J. (2011). When performance matters: The past, present, and future of consequential accountability in public education. Harvard Journal On Legislation 48 (1), 185 234. Langer, J.A. (2001). Approaches toward meaning in low and high rated readers. Retriev ed June 11, 2007, from http://cela.albany.edu/reports/langer/langerapproaches.pdf Lee, V. E., Bryk, A. S., & Smith, J. B. (1993). The organization of effective high schools. In L. Darling Hammond (Ed.), Review of Research in Education, 19, pp. 171 267.Washington, DC: American Educational Research Association. Lesley, M. (2004). Looking for critical literacy with postbaccalaureate content area literacy students. International Reading Association, 48( 4 ), 320 334. Liu, X. & Lu, K. (2008, September). Student performance and family socioeconomic status: Results from a survey of compulso ry education in western China. Chinese education and society, 41( 5), 70 83. Long, J.S. & Fox, M.F. (199 5). Scientific careers: Universalism and particularism. Annual Review of Sociology, 21, 45 71. Lovelace, S. & Wheeler, T. (2006). Cultural discontinuity between home and school language socialization patterns: Implications for teachers. Education, 127 (3) 303 309. Madaus, G. & Russell, M. ( 2010 ). Paradoxes of high stakes testing. Journal of Education, 190( 1/2 ), 21 30. Maerten Rivera, J., Myers, N., Lee, O., & Penfield, R. (2010). Student and school predictors of high stakes assessment in science. Science E ducation, 94, 937 962. Marchant, G. (2004). What is at stake with high stakes testing? A discussion of issues and research. Ohio Journal of Science, 104 (2), 2 7.

PAGE 183

183 Marzano, R. (2004). Building Background Knowledge for Academic Achievement: Research on What Works in Schools Alexandria, VA: Association for Supervision & Curriculum Development. Marzano, R. & Pickering, D (2005). Building academic vocabulary: Teacher's manual Alexandria, VA: ASCD Mattison, E. & Aber, M.S. (2007). Closing the achievement gap: The association of racial climate with achievement and behavioral outcomes. American Journal of Community Psychology, 40, 1 12. McIntosh, S. (2012). State high school exit exams: A policy in transition. Center on Education Policy. Retrieved from http://www.cep dc.org/cfcontent_file.cfm?Attachment=McIntosh%5FReport%5FHSEE2012%5F9 %2E19%2E12%2Epdf Mead, A. D. & Drasgow, F. (1993). Equivalence of computerized and paper and pencil cognitive ability tests: A meta analysis. Psychological Bulletin, 114 (3), 449 58. Medina, M. & Mishra, S. (1994). Relationships among Spanish reading achievement and selected content areas for fluent and limited Spanis h proficient Mexican Americans. Bilingual Review/Revista Bilingue, 19 (2), 134 141. Meltzer, J.(2001). Supporting Adolescent literacy across the content areas: Perspectives on policy and practice. Retrieved from http://www.alliance.brown.edu/pubs/perspectives/adlitcontent.pdf Meltzer, J., Cook Smith, N., & Clark, H. (2002). Adolescent literacy resources: Linking research and practice. Retrieved from http://www.alliance.brown.edu/pubs/adlit/alr_lrp.pdf Moore, G.W. & Slate, J.R. (2008). Who's taking the advanced placement courses and how are they doing: A statewide two year study. The High School Journal, 92 (1), 56 67. Mostafa, T. (2010). Decomposing inequalities in performance scores: The role of student background, peer effects and school characteristics. International Review of Education, 56 (5/6), 567 589. Murphy, P., DeArmond, M., & Guin, K. (2003). A nationa l crisis or localized problems? Getting perspective on the scope and scale of the teacher shortage. Education Policy Analysis Archives 11 (23) Retrieved from http://epaa.asu.edu/ojs/article/ download/251/377 Nasr, A.R. & Soltani K., A. (2011). Attitude towards Biology and its effects on student's achievement. International Journal of Biology, 3 (4), 100 104.

PAGE 184

184 National Center for Children in Poverty (2007). Parents' low education leads to low in come, despite full time employment. Retrieved from http://www.nccp.org/publications/pdf/text_786.pdf National Center for Education Statistics (2012). The Nation's Report Card: Science 2011 (NCES 2012 465). Institute of Education Sciences, U.S. Department of Education, Washington, D.C. National Center for Education Statistics (2011). The Nation's Report Card: Reading 2011 (NCES 2012 457). Institute of Education Sciences, U.S. Department of Ed ucation, Washington, D.C. National Commission on Excellence in Education (1983). A Nation At Risk: The Imperative for Educational Reform Washington, DC: Government Printing Office. National Commission on Mathematics and Science Teaching for the 21st Centu ry (2000). Before It's Too Late: A Report to the Nation from The National Commission on Mathematics and Science Teaching for the 21st Century. Washington, DC: U.S. Department of Education. National Governors Association Center for Best Practices, Council o f Chief State School Officer s (2010). Common Core Standards. Washington, DC: National Governors Association Center for Best Practices, Council of Chief State School Officers. National Research Council (2005). in high sc hool science Washington, DC: National Academies Press. National Science Board (2012). Science and Engineering Indicators 2012 Retrieved from http://www.nsf.gov/statistics/seind12/c1/c1h.htm National Science Board (2004, May). Science and Engineering Indicators 2004 Retrieved from http://www.nsf.gov/statistics/seind04/toc.htm National Science Foundation (2012a). Science and engi neering indicators 2012. Retrieved from http://www.nsf.gov/statistics/seind12/ National Science Foundation (2012b). WebCASPAR Integrated Science and Engineering Resources Data System Table builder: IPEDS Completions Survey by Race. Retrieved from https://caspar.nsf.gov./EmailedTable?table=1016121345202757143 National Science Foundation (2012c). WebCASPAR Integrated Scien ce and Engineering Resources Data System Table builder: NSF NIH Survey of Graduate Students & Postdoctorates in Science and Engineering Retrieved from https://caspar.nsf.gov./ EmailedTable?table=1016121334032321826 National Science Foundation, Division of Science Resources Statistics (2011). Women, minorities, and persons with disabilities in science and engineering: 2011

PAGE 185

185 Special Report NSF 11 309. Arlington, VA. Retrieved fro m http://www.nsf.gov/statistics/wmpd/pdf/wmpd2011.pdf Neufeld, P. (2005). Comprehension instruction in content area classes. The Reading Teacher, 59 (4), 302 312. No Child Left Behind Act of 2001, 20 USC § 6301 (2002). Retrieved from http://www2.ed.gov/policy/elsec/leg/esea02/107 110.pdf Nolen, S. B. (2003). Learning environment, motivation, and achievement in high scho ol science. Journal of Research in Science Teaching, 40 (4), 347 368. Norris, S. P. & Phillips, L. M. (2003). How literacy in its fundamental sense is central to scientific literacy. Science Education, 87 (2), 224 240. Nosek, B.A., Smyth, F.L., Sriram, N., L indner, N.M., Devos, T., Ayala, A., Bar Anan, Y., Bergh, R., Cai, H., Gonsalkorale, K., Kesebir, S., Maliszewski, N., Neto, F., Olli, E., Park, J., Schnabel, K., Shiomura, K., Tudor Tulbure, B., Wiers, R.W., Somogyi, M., Akrami, N., Ekehammar, B., Vianello M., Banaji, M.R., & Greenwald, A.G. (2009). National differences in gender science stereotypes predict national sex differences in science and math achievement. PNAS, 106 (26), 10593 10597. Noyes, J.M. & Garland, K.J. (2008). Computer v s. paper based task s: Are they equivalent? Ergonomics, 51 (9), 1352 1375. Ogbu, J. U. & Simons, H. D. (1998). Voluntary and Involuntary minorities: A cultural ecological theory of school performance with some implications for education. Anthropology and Education Quarterly, 29 (2), 155 188. Okpala, C.O., Smith, F., Jones, E., & Ellis, R. (2000). A clear link between school and teacher characteristics, student demographics, and student achievement. Education, 120 (3), 487 494. & McNamara, D. S. (2007). The impact of science knowledge, reading skill, American Educational Research Journal, 44 (1), 161 196. Organisation for Economic Co operation and Dev elopment. (2007). PISA 2006: Science Organisation for Economic Co operation and Development (OECD). (2002a). Measuring student knowledge and skills: The PISA 2000 assessment of reading, mathematical and sci entific literacy. Paris: Author. Organisation for Economic Co operation and Development (OECD). (2002b). PISA 2000 technical report. Paris: Author.

PAGE 186

186 Ou, D. (2009). To leave or not to leave? A regression discontinuity analysis of the impact of failing the hi gh school exit exam. Economics of Education Review 29 171 186. Ozuru, Y., Dempsey, K., & McNamara, D.S. (2009). Prior knowledge, reading skill, and text cohesion in the comprehension of science texts. Learning and Instruction 19 228 242. Papay, J.P., M urnane, R.J., & Willet, J.B. (2010). The consequences of high school exit examinations for low performing urban students: Evidence from Massachusetts. Educational Evaluation and Policy Analysis 32 (1), 5 23. Pearson Education (2013). Script for administeri ng the Biology I EOC Practice Test (ePAT). Retrieved from http://www.pearsonaccess.com/cs/Satellite?blobcol=urlblob&blobheader=app licat ion%2Fforcedownload&blobheadername1=ContentDisposition&blobheadername 2=ContentTransferEncoding&blobheadervalue1=attachment%3B+filename%3D %222012+Biology+1+EOC+TestNav+ePAT+Script_WT_FINAL.pdf%22&blobhea dervalue2=binary&blobkey=id&blobtable=BinaryData& blobwhere=125504996085 5&ssbinary=true Penfield, R.D. & Lee, O. (2010). Test based accountability: Potential benefits and pitfalls of science assessment with student diversity. Journal of Research in Science Teaching, 47 (1), 6 24. Perie, M., Grigg, W., & D (NCES 2006 451). U.S. Department of Education, National Center for Education Statistics. Washington, D.C.: U.S. Government Printing Office. Retrieved from http://nces.ed.gov/nationsreportcard/pdf/main2005/2006451.pdf Perry, L. & McConney, A. (2010). School socio economic composition and student outcomes in Australia: Implications for educational policy. Australian Journal Of Education (ACER Press) 54 (1), 72 85. Pico, K. & Stephens, J.M. (2012). Culture, context and stereotype threat: A comparative analysis of young Ugandan women in coed and single sex schools. The Journal of Educational Research, 105 52 63. Pinder, P.J. (2012). Afro Caribbean and African American students, family factors, and the influence on science performance in the United States: The untold story. Education, 132 (4), 725 738. Pomplum, M., Ritchie, T., & Custer, M. (2006). Factors in paper and pencil a nd computer reading score differences at the primary grades. Educational Assessment, 11 (2), 127 143.

PAGE 187

187 Pong, S.L., Dronkers, J., & Hampden Thompson, G. (2003). Family policies and children's school achievement in single versus two parent families. Journal o f Marriage and Family, 65, 681 699. Prokop, P., Prokop, M., & Tunnicliffe, S. D. (2007, Winter). Is biology boring? Student attitudes toward biology. Journal of Biological Education, 42 (1), 36 39. Puma, M. J., Karweit, N., Price, C., Ricciuti, A., Thompson W., & Vaden Kiernan, M. (1997, April). Prospects: Final report on student outcomes (Washington, DC: U.S. Department of Education). Retrieved from http://www.eric.ed.gov/PDFS/ED413411.pdf Qiu, X. & Wu, S. (2011). Global and local regression analysis of factors of American College Test (ACT) score for public high schools in the state of Missouri. Annals of the Association of American Geographers, 101 (1), 63 83. Rathbun, A. & West, J. (2004) From Ki Beginning School Experiences (NCES 2004 007). U.S. Department of Education, National Center for Education Statistics. Washington, DC: U.S. Government Printing Office. Reeves, D.B. (2000) The 90/90/90 Schools: A c ase study In Accountability in action: A blueprint for learning organizations, 2 nd ed (pp. 185 208). Denver, CO: Advanced Learning Press. Riegle Crumb, C., Farkas, G., & Muller, C. (2006). The role of gender and friendship in advanced course taking. Soci ology of Education, 79 (3), 206 228. Riegle Crumb, C., Moore, C., & Ramos Wada, A. (2010). Who wants to have a career in science or math? Exploring adolescents' future aspirations by gender and race/ethnicity. Science Education, 95 (3), 458 476. Roe, B. D., Stoodt, B. D., & Burns, P. C. (1991). Secondary school reading instruction: The content areas. Boston: Houghton Mifflin. Romance, N. R. & Vitale, M. R. (2008, March). Science IDEAS: A knowledge based model for accelerating reading/literacy through in dept h science learning. Paper presented at the annual meeting of the American Educational Research Association, New York. Rumberger, R.W. & Palardy, G.J. (2005). Does segregation still matter? The impact of student composition on academic achievement in high s chool. Teachers College Record, 107 (9),1999 2045. Russell, M. (1999). Testing on computers: a follow up study comparing performance on computer and on paper. Education Policy Analysis Archives, 7 (20). Retrieved from http://epaa.asu.edu/epaa/v7n20

PAGE 188

188 Russell, M. & Haney, W. (2000). Bridging the gap between testing and technology in schools. Education Policy Analysis Archives, 8 (19). Retrieved from http://epaa.asu .edu/epaa/v8n19.html Russell, M. & Haney, W. (1997). Testing writing on computers: an experiment comparing student performance on tests conducted via computer and via paper Educational Policy Analysis Archives, 5 (3). Retrieved from http://epaa.asu.edu/epaa/v5n3.html Ryabov, I & Van Hook, J. (2007). School segregation and academic achievement among Hispanic children. Social Science Research 36 (2), 767 788. Ryan, J.E. (1999). The influence of race in sc hool finance reform. Michigan Law Review, 98, 432 481. Sanders, W. L. & Rivers, J. C. (1996). Cumulative and residual effects of teachers on future students academic achievement Knoxville, TN: University of Tennessee Value Added Research and Assessment Center. Retrieved from http://www.westendweb.com/robert/CS302/Sanders.pdf science reform effort on t he achievement and attitudes of third and fourth grade students. Journal of Research in Science Teaching, 41 (8), 771 790. Sikora, J. & Pokropek, A. (2012). Gender segregation of adolescent science career plans in 50 countries. Science Education, 96 (2), 23 4 264. Singh, K., Granville, M., & Dika, S. (2002). Mathematics and science achievement: Effects of motivation, interest, and academic engagement. Journal of Educational Research, 95 (6), 323 332. Skiba, R. J., Michael, R. S., Nardo, A. C., & Peterson, R. L (2002). The color of discipline: Sources of racial and gender disproportionality in school punishment. Urban Review, 34, 317 342. Smith, M. C. (2000). The real world reading practices of adults. Journal of Literacy Research, 32 (1), 25 52. Sonnert, G. & Fo x, M.F. (2012). Women, men, and academic performance in science and engineering: The gender differences in undergraduate grade point averages. Journal of Higher Education, 83 (1), 73 101. Sonnert, G., Fox, M.F., & Adkins, K. (2007). Undergraduate women in science and engineering: Effects of faculty, fields, and institutions over time. Social Science Quarterly 88 1333 1356.

PAGE 189

189 Snyder, T.D., Dillow, S.A., & Hoffman, C.M. (2008). Digest of Education Statistics 2007 (NCES 2008 022). National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education. Washington, DC. Stansfield, W.D. (2011). Educational curriculum standards & standardized educational tests: Comparing apples & oranges. The American Biology Teacher, 73 (7), 389 393 Steele, C.M. & Aronson, J. (1995). Stereotype threat and the intellectual test performance of African Americans. Journal of Personality and Social Psychology, 69 (5), 797 811. Stewart, E.B. (2007). Individual and school structural effects on African Ameri can High School Students' Academic Achievement. High School Journal, 91 ( 2 ), 16 34. Strand, S. (2012). The White British Black Caribbean achievement gap: Tests, tiers and teacher expectations. British Educational Research Journal, 38 (1), 75 101. Sykes, G., Schneider, B., & Plank, D.N. ( 2009). Handbook of education policy rese arch. New York, NY: Routledge. Taltont, E. L. & Simpson, R. D. (1985). Relationships between peer and individual attitudes toward science among adolescent students. Science Education, 69 19 24 Thompson, D., Wood, R. C., & Crampton, F.E. (2008). Money & Schools, 4 th ed. Larchmont, NY: Eye on Educ ation. Torgesen, J.K. (2006). Adolescent Literacy: An Overview of Skill and Knowledge Requirements. Retrieved from http://www.centeroninstruction.org/files/Adolescent%20Literacy%20An%20overvi ew%20of%20skill%20and%20knowledge%20requirements.pdf Trefil, J. (2008). Science education for everyone: Why and what? Liberal Education, 6 11. Trefil, J. (2007). Why science? New York: Teachers College Press. Tulkin, S.R. & Kagan, J. (1972). Mother child interaction in the first year of life. Child Development, 43 (1), 31 41. United Nations Children's Fund (UNICEF) (2012, May). Measuring child poverty: New league tables of child poverty in the world's rich countries. Retrieved from http://www.unicef irc.org/publications/pdf/rc10_eng.pdf United States Census Bureau (2012, May 17). Most children younger than a ge 1 are minorities, Census Bureau reports. Retrieved from http://www.census.gov/newsroom/releases/archives/population/cb12 90.html

PAGE 190

190 United States Census Bureau, Popul ation Division (2012). Table 5. Cumulative estimates of the components of resident population change by race and Hispanic origin for the United States: April 1, 2000 to July 1, 2009 (NC EST2009 05) Retrieved from http://www.census.gov/compendia/statab/2012/tables/12s0005.pdf United States Commission on National Security/21 st Century (2001b, May 15). Journeys through the teacher pipeline: Recapitalizing American education Parti al cost estimates for road map for national security Retrieved from http://www.au.af.mil/au/awc/awcgate/nssg/addendum/Education_Addendum.pdf United States Commission on National Security/21 st Century (2001a, February 15). Road map for national security: Imperative for change The phase III report of the U.S. Commission on National Security/21 st century. Retrieved from http://www.au.af.mil/au/awc/awcgate/nssg/phaseIIIfr.pdf United States Congress (2002). Public Law 107 110. No Child Left Behind Act of 2001. Washington, D.C. Retrieved from http://www2.ed.gov/policy/elsec/leg/esea02/index.html United States Department of Agriculture (2012, August). National School Lunch Program. Retrieved from http://www .fns.usda.gov/cnd/lunch/AboutLunch/NSLPFactSheet.pdf United States Department of Education (2004, February 10). Proven methods: The facts about...science achievement. Retrieved from http ://www2.ed.gov/nclb/methods/science/science.html United States Department of Education (1989). Report on the National Education Goals Panel Washington, DC: Government Printing Office. United States Department of Education, National Center for Education S tatistics (2012). Digest of Education Statistics, 2011 (NCES 2012 001). Retrieved from http://nces.ed.gov/fastfacts/display.asp?id=171 United States Department of Labor (2010). Highlights of Women's Earnings in 2009. Retrieved from http://www.bls.gov/cps/cpswom2009.pdf Velayutham, S., Aldridge, J.M., & Fraser, B. (2012). Gender differences in student motivation and self regulation in scien ce learning: A multi group structural equation modeling analysis. International Journal of Science and Mathematics Education, 10 1347 1368. Velde Pederson, P. (2007). What is measured is treasured: The impact of the No Child Left Behind Act and nonassesse d subjects. The Clearing House, 80( 6 ), 287 291. Vincent, C.G., Tobin, T.J., Hawken, L.S., & Frank, J.L. (2012). Discipline referrals and access to secondary level support in elementary and middle schools: Patterns

PAGE 191

191 across African American, Hispanic America n, and White students. Education and Treatment of Children, 35 (3), 431 458. Visone, J. (2010). Science and reading: What is being measured by standardized tests? American Secondary Education, 39 (1), 95 112. Visone, J. (2009). The validity of standardized testing in science. American Secondary Education, 38 (1), 46 61. Vitale, M.R. & Romance, N.R. (2012). Using in depth science instruction to accelerate student achievement in science and reading comprehension in grades 1 2. International Journal of Science a nd Mathematics Education, 10, 457 472. Walton, G.M. & Spencer, S.J. (2009). Latent ability: Grades and test scores systematically underestimate the intellectual ability of negatively stereotyped students. Psychological Science, 20 (9), 1132 1139. White Hou se, Office of the Press Secretary. (2011, January 25). Remarks by the President in State of Union Address [Press Release]. Retrieved from http://www .whitehouse.gov/the press office/2011/01/25/remarks president state union address White House, Office of the Press Secretary. (2010, October 18). President Obama to host White House Science Fair [Press Release]. Retrieved from http://www.whitehouse.gov/the press office/2010/10/18/president obama host white house science fair Wiliam, D. (2010). Standardized Testing and School Accountabili ty. Educational Psychologist 45 (2), 107 122 Wingfield, A. & Guthrie, J.T. (1997). Relations of children's motivation for reading to the amount and breadth of their reading. Journal of Educational Psychology, 89 (3), 420 432. Winstead, L. (2011). The impact of NCLB and accountability on social studies: Teacher experiences and perceptions about teaching social studies. The Social Studies, 102, 221 227. Wood, K. D. (2002). Status of high school Biology teaching. Retrieved from http://2000survey.horizon research.com/reports/high_biology/high_biology.pdf Woods, T.A., Kurtz Costes, B., & Rowley, S.J. (2005). The development of stereotypes about the rich and poor: Age, r ace, and family income differences in beliefs. Journal of Youth and Adolescence, 34 (5), 437 445. Yore, L. D., Bisanz, G. L., & Hand, B. M. (2003). Examining the literacy component of science literacy: 25 years of language arts and science research. Interna tional Journal of Science Education, 25 (6), 689 725.

PAGE 192

192 Zegiob, L.E. & Forehand, R. (1975). Maternal interactive behavior as a function of race, socioeconomic status, and sex of the child. Child Development, 46 (2), 564 568. Zinth, J.D. (2012). End of Course E xams. Education Commission of the States. Retrieved from http://www.ecs.org/cl

PAGE 193

193 BIOGRAPHICAL SKETCH Janine Cecelia Ber tolotti was born in 1983 in Kingston, Jamaica. She attended Covenant Christian Academy before enrol ling at Sts. Peter and Paul Preparatory School. Subsequently, she enrolled at Immaculate Conception High School in Jamaica before matriculating at Wesleyan College in Macon, Georgia at the tender age of 17. In May 2005, she graduated summa cum laude with a major in p sychology and double minors in b iology and n euroscience. After working for approximately one year at a technical college in Georgia, she enrolled full time in the Master's degree program in curriculum and i nstruction at the University of Central Florida in Orlando, Florida. Before graduating in December 2007, she served as a graduate research assistant before beg inning her first year teaching b iology at a public, Miami high school. In the summer of 2010, she began a part time doctoral progra m whi le working full time as a b iology high school teacher. She graduated in August 2012 with an Education Specialist de gree in educational l eadership from the University of Florida before attaining her Doctor of Education degree in educational leadership in De cember 2013. As a U.S. Air Force wife, she moves frequently with her husband and daughter.