Development of a Psychophysiological Artificial Neural Network to Measure Science Literacy

Development of a Psychophysiological Artificial Neural Network to Measure Science Literacy by Amanda Kavner April 23, 2020 A dissertation submitted to the faculty of the Graduate School of the University at Buffalo, The State University of New York in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Learning and Instruction

ProQuest Number: 27995938 All rights reserved INFORMATION TO ALL USERS The quality of this reproduction is dependent on the quality of the copy submitted. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. ProQuest 27995938 Published by ProQuest LLC (2020). Copyright of the Dissertation is held by the Author. All Rights Reserved. This work is protected against unauthorized copying under Title 17, United States Code Microform Edition © ProQuest LLC. ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, MI 48106 - 1346

Copyright by Amanda Kavner 2020 All Rights Reserved (This page is Required) ii

This dissertation is in loving memory of Dolores Barell and Margo Goldman. I dedicate my dissertation work to my husband, Lee Goldman. Without your support and patience, I could never have accomplished this milestone. And to my parents, Stuart and Fran Kavner, without your loving guidance and encouragement throughout my life, I would not be who I am today. . iii

Acknowledgements I would like to thank my committee for their feedback. In particular my advisor Richard Lamb who has guided me through my Doctoral pursuit and pushed me to accomplish when I did not know I could. I would also like to thank the researchers of Project LENS who made available their dataset for my research; especially Pasha Antonenko, principal investigator for the project. I would like to express my appreciation and love for my family; Herbert Barell, Jonathan, Stefani, Dylan and Hayden Kavner, Paul and Charlie Hornstein, Zoe and Alex Jerva, Adina Dumitrescu, Tracy Goldman, Matthew and Fiona Riddick, Dr. Richard and Carole Kavner, Robert and Rebecca Barell, Roselyn and Robert Gross, Dr. Nicole Miller, and the Steinberg and Jessop families. I would also like to thank the rest of my family, friends, and colleagues who have helped support me throughout this process; to Raymond Mantione, Diane Jurgens, Jessica McGowan, Marissa Hylton, Richard Birmingham, Dr. Jeffery Woodberry, Steven Wilson, Cortney Brandwein, Christian Fogarazzo, Gene Fisch and, Dr. Gabriel and Dorothy Taub. Thank you all for your patience, your understanding, support, friendship, and guidance. I could never have reached this accomplishment without each of you. iv

Abstract The rapid development of other nations' science and technology makes it more difficult to stay competitive internationally without concentrating on how science is taught in US classes. Representative competence, the capacity to generate, transform, interpret and clarify representations, is the primary obstacle to visual literacy in science, technology, engineering and mathematics (STEM) fields and although the relationship between the fundamental visual literacy and domain-specific science literacy is known, how visual science literacy is achieved through science learning is still not fully understood. In order to improve student representational competence skills, the hierarchy of component visualization skills required to interpret these science representations needs to be identified in order to evaluate an individual’s level of visual science literacy and to provide the resources to enable the individual to reach the next literacy level. This involves the construction of instruments capable of assessing visual science literacy as well as a Rasch analysis to rank complexities of science visuals. This research investigates modern methods of assessing visual science literacy with a focus on using artificial neural networks (ANN) to analyze neurocognitive measurements captured during science content related tasks and requiring varying predictable levels of visual science literacy. The method of developing this machine learning tool will be detailed by investigating the ANN, successfully made using the Gradient Boosted Trees algorithm to analyze functional Near-Infrared Spectroscopy (fNIR) data. With an autonomic, neurocognitive, and quantitative scientific literacy assessment, educators and curriculum designers will have the ability to create more targeted classroom v

resources to enhance the visual and spatial cognitive processes behind visual science literacy. Keywords: visual science literacy, science literacy, visual literacy, artificial neural network, machine learning, artificial intelligence, mental rotation vi

TABLE OF CONTENTS Acknowledgements.................................................................................................................. iv Abstract ......................................................................................................................................v List of Tables .............................................................................................................................x List of Figures .......................................................................................................................... xi CHAPTER ONE ........................................................................................................................1 Introduction................................................................................................................................1 Science Literacy and COVID-19 ........................................................................................ 1 Educational Neuroscience................................................................................................... 4 Multicomponent Working Memory Framework................................................................. 6 Cognitive Theory of Multimedia Learning......................................................................... 7 Purpose, Research Questions, and Hypothesis ................................................................. 10 Definitions: ....................................................................................................................... 11 CHAPTER TWO ...........................................................................................................................19 Literature Review.....................................................................................................................19 Background, Previous Work, and Justification................................................................. 19 Theoretical Framework ..................................................................................................... 21 Visual Literacy in Science ................................................................................................ 28 Neuroscience of Visual Literacy..............................................................................................36 CHAPTER THREE .......................................................................................................................52 Methods....................................................................................................................................52 Research Design................................................................................................................ 52 Participants........................................................................................................................ 53 Procedure .......................................................................................................................... 53 Measurements ................................................................................................................... 54 vii

Cleaning the data............................................................................................................... 56 Exploratory Factor Analysis ............................................................................................. 57 Mental Rotation for Visual Literacy Assessment.............................................................. 57 Machine Learning with Artificial Neural Networks ......................................................... 59 Developing the Artificial Neural Network in RapidMiner ............................................... 60 Developing the Confirmatory Model in SPSS.................................................................. 65 Rasch Analysis of Visuals ................................................................................................ 66 CHAPTER FOUR..........................................................................................................................69 Results......................................................................................................................................69 Research Question 1 ................................................................................................................69 Exploratory Factor Analysis ............................................................................................. 69 Artificial Neural Network ................................................................................................. 72 Research Question 2 ......................................................................................................... 75 Descriptive Statistics......................................................................................................... 75 Rasch Analysis.................................................................................................................. 76 Convergent Evidence ........................................................................................................ 77 CHAPTER FIVE ...........................................................................................................................78 Discussion ................................................................................................................................78 Research Question 1 ......................................................................................................... 78 Research Question 2 ......................................................................................................... 80 Future Implications ........................................................................................................... 83 Conclusion ........................................................................................................................ 89 References ................................................................................................................................91 Appendix A............................................................................................................................119 Descriptive Statistics....................................................................................................... 119 viii

Appendix B ............................................................................................................................120 SPSS ANN ...................................................................................................................... 120 Appendix C ............................................................................................................................122 Exploratory Factor Analysis in RStudio ......................................................................... 122 Appendix D............................................................................................................................123 Rasch Analysis in RStudio.............................................................................................. 123 Appendix E ............................................................................................................................125 Correlation in RStudio .................................................................................................... 125 Appendix F.............................................................................................................................128 Equations......................................................................................................................... 128 ix

List of Tables Table 1. Model Comparisons........................................................................................................ 63 Table 2. KMO and Bartlett's Test ................................................................................................. 70 Table 3. Task- Factor Breakdown................................................................................................. 72 Table 4. Relative weights of ANN predictors............................................................................... 74 Table 5. Model Fit Statistics ......................................................................................................... 75 Table 6. Mental Rotation Task Successful Completions .............................................................. 75 x

List of Figures Figure 1. .......................................................................................................................................... 2 Figure 2. .......................................................................................................................................... 3 Figure 3. .......................................................................................................................................... 4 Figure 4. Sample Wedge and Dash Mental Rotation Question .................................................... 54 Figure 5. Scree Plot....................................................................................................................... 70 Figure 6. Person-Item Map ........................................................................................................... 76 xi

CHAPTER ONE Introduction The primary goal for science educators is to cultivate students’ scientific habits of mind, such as critical thinking, develop their capability to engage in scientific inquiry, and teach students how to reason in a scientific context to develop knowledge (OECD iLibrary., 2016). To achieve this goal, it is necessary to investigate the specific aspects of the visual (image-based) and textual (text-based) science materials that students find challenging to integrate. Difficulties in integration limit the student’s ability to transfer this material into memory. Examining specific factors that aid in the development of new knowledge is an important consideration for curriculum designers as they develop resources to support students as they acquire skills to interpret and grasp higher-level concepts, extending their awareness and comprehension of natural phenomena. Improving a person’s level of scientific literacy is beneficial for individuals as well as for society since the ability to problem solve and observe the patterns occurring in natural phenomena helps people to make better choices based upon presented evidence (Snow, Dibner, Science, et al., 2016). Science Literacy and COVID-19 The global pandemic of COVID-19 during the winter and spring of 2019-2020 proved to be an ideal example of the importance of science literacy. Misinformation in the United States, circulated and proliferated by social media and polarizing press and encouraged by subcultures intent on embracing pseudoscience and conspiracy theories (Fisher, 2020), confused messages 1

by experts, leading to more cases of COVID-19 per 1 million individuals than any other country. Since individuals who don’t understand written science texts will typically look at the images and develop conclusions, it’s important to see visuals from that time period (Figures 1-3) to see what visualization skills individuals would need to understand to interpret the image. And how, in increasing or varying the complexity of the image, the probability of misinterpretation (L. Mnguni et al., 2016) or transformation to match the individual’s confirmation bias increases (Rajsic et al., 2014). Most individuals would be able to identify the visual elements in a basic visual of science concepts (Mnguni et al., 2016; Figure 1). Facemasks quickly became commonplace in the US in the spring of 2020 and the coronavirus 3D model image seen everywhere from the evening news, social media, and the pharmacy or doctor’s office. However, other than the fact that the individuals are wearing masks to protect themselves from contracting and spreading the virus, little information is conveyed about the pathogen. To do so, would require the encoding of more information in the image’s visual elements requiring a higher level of visual literacy to decode. Figure 1. 2

Although an average skill among university science, technology, engineering and mathematics (STEM) students, the ability to analyze and interpret essential features of a visual to illustrate a process is a difficult visualization skill for average members of society (Mnguni et al., 2016). For example, Figure 2 shows detail as to how COVID-19 infects a host cell, replicates and infects more host cells but it requires science content knowledge in biochemistry and a higher level of visual literacy. Individuals who understand will be encouraged to vaccinate themselves and their children once one is available so that they don’t get sick. Figure 2. Relating concepts using visuals requires an even higher level of visual literacy (Mnguni et al., 2016) and, although Figure 3 may not require the content knowledge in biochemistry as Figure 2, it requires other science foundational literacies such as numeracy to convey a comparison of viruses in relation to mortality and communicability. However, by understanding Figure 3, an individual could understand how serious an infection can be and the extent to which a disease can spread and therefore make informed decisions for social distancing and another mitigation efforts. 3

Figure 3. Individuals who understand these visuals are more likely to understand the dangers of continuing to spread the pathogen to vulnerable populations in the society and with that, the importance personal hygiene and listening to measures supported by trusted medical professionals plays in keeping individuals and the greater population healthy. Those individuals are therefore likely to make better democratic and economic decisions (Snow, Dibner, National Academies of Sciences and Medicine (U.S.). Committee on Science Literacy and Public Perception of Science, et al., 2016). Educational Neuroscience Most learners have a set of skills embedded in their system of perception, such as critical thinking and memory, which develop in their early life (Lamb, Annetta, & Vallet, 2014). These processes evolve into a set of dynamic procedures that are used parallel to processing environmental information (Moreau, 2013). These functional aspects of cognition have been defined as cognitive attributes (Lamb, 2014) which retain specific features of individual cognitive processes. These skills include the ability to understand and generate written and oral 4

statements, and explain the interaction of complex variables like those found in science. Critical thinking, the ability to retrieve from memory, and measure and predict outcomes from simple and complex problems, are other cognitive qualities of interest in science. Studies of neuroimaging demonstrate simultaneous activation and refer to cognitive systems as multiple activations in cognitive attributes (Finn et al., 2014). Psychologists and educators describe critical thinking as the formation of stimulus-based assumptions, inferences or decisions (Neisser, 2014). Science educators, in turn, indicate that critical thinking is a defining attribute used in the process of learning sciences (Erduran & Dagher, 2014). Neuroscientists use functional magnetic resonance imaging (fMRI) in reasoning tasks studies to show activation that occurs in brain areas associated with executive functioning (Barbey, Koenigs & Grafman, 2013). Critical thinking's more generalized attribute appears as behavioral components are seen in interpretation, anticipation, and inference (Ren, Schweizer & Xu 2013). The second area of interest to science education is that of memory, in addition to critical thinking. Memory allows for impact activation and offers a way for students to make use of previous experience and knowledge (Lamb & Annetta, 2012; Lamb, 2014). An individual can retrieve a memory by triggering the specific neurons which serve as a sensory reminder of that particular thought or event. The ability to recall past semantic and episodic memories is termed retrospective memory (Cabeza & Moscovitch, 2013) and memory-related brain areas are located in prefrontal cortex areas and are associated with functional near infrared spectroscopy (fNIRS) Optodes 13 though 16 (Tai & Chau, 2009; Jasińska & Petitto, 2014). In addition to providing basic information on cerebral blood flow, fNIRS allows for direct examination of inaccessible structures and real-time functional analysis as students engage with different instruction modes. 5

fNIRS is a useful tool for connecting structure, function and behaviors. The accuracy of the effectiveness of fNIRS as a tool for examining cognition has been demonstrated such that neurosurgeons make use of this equipment as a functional measure during the treatment of individuals with Parkinson's disease using ventral intermediate (VIM) thalamotomy (Lipsman et al., 2013) and as a means of monitoring patients for other types of surgery during the administration of anesthesia (Zheng et al., 2013). To connect cognitive traits to their associated behaviors, study integrates Baddeley and Hitch’s Multicomponent Working Memory Framework (M-WM; Baddeley, 2011; Hitch, Hu, Allen, & Baddeley, 2018) with Mayer’s (2003) Cognitive Theory of Multimedia Learning. Multicomponent Working Memory Framework This framework expands upon working memory (WM) theory where WM is defined as the limited-capacity system which combines mental operations and temporary information storage to support goal directed cognition and behavior (Hitch et al., 2018) composed of four systems: Central Executive, Phonological Loop, Visuospatial Sketchpad and the Episodic Buffer. More specifically, as we are attempting much more detailed account of WM and developing that model using computer simulation, this study follows a computational model of WM, considered by Baddeley to be a flexible approach which can result in subcomponents resembling aspects of M-WM including, as in this study, the visuospatial sketch pad which is the portion of one's usual mental facility that includes a simulated world for physical simulation, visualization, calculation and visual memory recall (Allen et al., 2017). As posited by Fitch (2014), a successful formal theory of cognition must be compatible with the constraints flowing from such macro-level cognitive concepts, such as working memory and success will require the convergence of three disciplines: neuroscience (with its physical 6

foundations of brain function), the cognitive sciences and cognitive biology. The bridging functions between all three domains should build upon the computational insights of computer science (ie. algorithms). Cognitive Theory of Multimedia Learning According to this theory during the learning process, external pictures first enter the cognitive system through the eyes. Attending to some aspects of the visual model leads to the construction of a mental pictorial image within working memory. Following subsequent construction of mental images, the student arranges the set of images into a coherent mental representation. The process involves the selection, organization and integration of images and is commonly referred to as visuo-spatial thinking (Mayer, 2002; Mnguni, 2014). This theory is related to a constructivist epistemology of learning which equates learning with using experience to create meaning (Ertmer & Newby, 2013). This requires students to be active participants in the learning process which follows three steps; comprehension of visual information, processing of this information in cognitive structures, and externalization of information as visual models. Following the lines of this theory, visualization can be defined as the ability to effectively select and use a set of cognitive skills for perceiving, processing and producing visual models (Mnguni, 2014). However, evaluation of the production and comprehension of visual models is typically subject to the externalization of information in the form of sketches, drawings or verbal descriptions. Since responses are voluntary, they are prone to participant or response bias; a limitation which can be mitigated by using measurements collected from the autonomic or involuntary nervous system (Andreassi, 2007). 7

Cognitive Diagnostics with fNIRS. In place of using self-test assessments, measurement of specific cognitive dynamics in individual students is necessary in order to understand how brain function and cognition relate to the process of learning in science. This has led to the emergence of the field of Educational Neuroscience; bringing together members of multiple communities, all interested in understanding how learning occurs in the science classroom (Lamb et al., 2017). Functional Near-Infrared Spectroscopy (fNIRS) and Functional Magnetic Resonance Imaging (fMRI) are semi-non-invasive and non-invasive imaging methods respectively, which provides basic data about cognitive processing in real-time (Wijeakumar et al., 2017). However, the field of education is far behind that of neurosciences in the adoption of these technologies. Even though fMRI has become the standard for noninvasive neurological imaging in the neurosciences since the invention in the early 1990s, the excessive cost, lack of expertise, as well as a lack of recognition of the usefulness by the field of these tools in clarifying questions and testing hypotheses in the classroom has delayed adoption. In part to address the shortcomings of fMRI, researchers have developed protocols and methods for the use of fNIRS (Lamb et al., 2017). fNIRS is better suited and less expensive than fMRI; fNIRS offers a portable and affordable method to examine cerebral blood flow (hemodynamic responses) and enables direct observation of inaccessible structures and functional evaluation and as students interact with different instructional modes (Lamb et al., 2018). Using Functional Near-Infrared Spectroscopy (fNIRS), a neuroimaging system which shines near-infrared light at two or more different wavelengths through brain tissue to compare oxygenated (HbO) and de-oxygenated hemoglobin (HbR), the ratio of which indicates cognitive dynamics (Wijeakumar et al., 2017), standardized hemodynamic responses were analyzed to measure cognitive demand and processing where the cognitive demand for each of the conditions 8

was defined as the hemodynamic response composite measure gathered over the duration of the experimental condition. The temporal resolution provided by the fNIRS helped to gain useful continuous information on variations and disturbances of cognitive processing in addition to the localization of changes in oxygenated hemoglobin within the brain. As more cognitive processing occurs, neurons in the brain require more oxygen resulting in an area of deoxygenated blood, thereby providing a useful tool for connecting structure, function and behavior through illustrating the cognitive processing of students in the classroom setting while engaging in STEM-based tasks. When combined with computational modeling via machine learning, a pathway is created for educators to understand the complexities involved in the learning of science (Lamb, Firestone, & Ardasheva, 2016; Lamb et al., 2017; Pike et al., 2014). In the past, using technology like this in the classroom was unlikely due to the cost, lack of training for science education researchers and lack of robustness to use in the natural ecology of the classroom, however, now, wearable technology in the form of fitness trackers and smartwatches have become mainstream, creating an environment where the possibility of using technology to track cognitive states in the classroom is close to a reality. The proliferation of biological sensors in commercial wearable technology has resulted in an exponential increase in basic involuntary empirical data related to the systems and mechanisms of learning that we use in formal and informal learning contexts (Llorente & Morant, 2014), furthermore, as the existence and collection of this data has rapidly increased, our translational capacity to discreet education practices and teacher level information has not (Sin & Muthu, 2015). Yet, with the growing capabilities of harnessing big data and with the sheer quantity of information related to basic neurological research available to us growing by the second, neurocognitive processes can be quantitatively examined in real-time for patterns (McKendrick et al., 2015). These patterns, once 9

analyzed using an ANN, as evidenced in this study, can reveal relationships between variables, some never before considered, and can, therefore, illuminate methods for using immersive pedagogy to support those neurocognitive processes. Due to their connection to critical thinking, attention, and memory retrieval respectively, special emphasis was put on signals from Optodes 1 through 4 and Optodes 13 through 16. Purpose, Research Questions, and Hypothesis The purpose of this study is to investigate the visual component of science literacy. Primarily, the goal is to determine a hierarchy of visual components that will be determined through the development and subsequent testing of an Artificial Neural Network (ANN) which relates student science visual literacy and cognition in the science classroom. This cognitive computational model is capable of assessing an individual’s level of scientific visual literacy, a primary component of scientific literacy, using cognitive diagnostics. The second goal of this study is to determine what specific aspects of visual texts students find difficult to interpret. Specifically, this study develops the unidimensional construct known as visual science literacy as it relates to chemical molecular structures. The developed instrument can be used to aide in the assessment of student visual literacy level within the domain of science by evaluating the underlying psychometric properties for the appropriate functioning of the instrument under the formal requirements of the Rasch model. Therefore, this study investigates the following research questions and their associated hypotheses: Research Question 1: How can an artificial neural network (ANN) be developed to assess the visual component of science literacy? 10

Hypothesis 1: By evaluating neurocognitive data collected while individuals are performing science tasks with visuals, it is possible to relate task items with their underlying cognitive attributes to develop a cognitive computational model in the form of an ANN with valid predictive value. Research Question 2: How can the visuals be ranked in order to determine a visual literacy level? Hypothesis 2: The Rasch Analysis can be used to rank visuals from the study in order of their interpretive difficulty. Definitions: The study includes many key terms and definitions that have specific contextual significance for the purposes of this research. The following section includes a list of key definitions as used in the context of this research. Artificial Neural Network (ANN): inspired from the biological neural networks of human brains and a form of machine learning algorithms with the functions of estimation and approximation based on inputs (Kotu & Deshpande, 2014). AAAS: American Science Advancement Association 11

Autonomic Nervous System (ANS): part of the peripheral nervous system that regulates the involuntary physiologic processes such as heart rate, blood pressure, respiration, digestion, and sexual arousal (Waxenbaum & Varacallo, 2019). Augmented Reality (AR): a technology which overlays abstract or virtual objects (augmented components) into the physical world (Akçayır & Akçayır, 2017). Cognitivism: Cognitive theories concentrate on the conceptualization of student learning experiences and discuss the concerns of how information is received, processed, stored and recalled by the mind (Ertmer & Newby, 2013). Connectivism: is an emergent theory of the mind where patterns of input phenomenon create patterns of connections which are distributed in neural networks in the brain (Gerard, 2016). Constructivism: considered to be a branch of cognitivism as both conceive of learning as a mental activity, constructivism is a theory that equates learning with using experience to create meaning (Ertmer & Newby, 2013). Decision Trees: a structure that contains a root node, a branch, and a leaf node. Every internal node denotes an attribute test, each branch denotes the test outcome, and each leaf node carries a class label. The node at the top of the list is the root node (Naik & Samant, 2016). Deep Learning: a set of Machine Learning algorithms which have one or more hidden layers of 12

multiple nonlinear processing units between input and output layers (Ramya, 2017). Episodic Memory: enables the capacity to recollect past events which occurred at a particular place and time (Wixted et al., 2018). Electroencephalography (EEG): also known as the \"brain wave\" and was first described in rabbits and monkeys by Richard Caton in 1875 when a he recorded a change in electrical activity in the occipital area when a visual stimulus (a flashing light) was applied (Andreassi, 2007). Feedback based prediction: predicting an ultimate outcome in an iterative manner, basing each iteration’s operation upon the present outcome (Wu et al., 2016). Feedforward mechanism: the first and simplest type of artificial network devised which does not form a cycle (Schmidhuber, 2015). Functional Near Infrared Spectroscopy (fNIRS): neuroimaging systems which shine near-infrared light at two or more different wavelengths through brain tissue. The two wavelengths of light are absorbed differently by oxygenated (HbO) and de-oxygenated hemoglobin (HbR). The ratio of which indicates cognitive load level (Wijeakumar et al., 2017). Functional Magnetic Resonance Imagery (fMRI): known to be the gold standard for neuroimaging, and while it is expensive, it offers exceptional spatial resolution and has proven useful in a number of clinical and non-clinical applications (Wijeakumar et al., 2017). 13

Generalized Linear Model (GLM): simple mathematical relationships between input and output variables (Kotu & Deshpande, 2014) Gradient Boosted Trees (GBT): a learning procedure consecutively fits new models to provide a more accurate estimate of the response variable (Natekin & Knoll, 2013). Hemodynamic data: the rapid movement of blood to neural tissue to enable different regions of the brain to engage in cognitive processing (Aslin et al., 2015). Hippocampus: location in the brain where episodic memory are retrieved and encoded (Wixted et al., 2018). Immersive Education: see Milgram’s Virtuality Continuum Item Response Theory (IRT): commonly used in psychometrics and educational measurement to provide researchers with information regarding performance of both items on and performance of individuals who take an assessment (Finch & Edwards, 2016). Machine Learning: computer algorithms that improve automatically through experience (Kotu & Deshpande, 2014). Mental rotation: involves the visual inspection and mental simulation of an object’s rotation in 14

space and is predictive of STEM achievement (Stieff et al., 2018). Mental imagery: internal depictive representations as they have intrinsic conceptual features in common with the object displayed (Schnotz et al., 2002). Milgram’s Virtuality Continuum: ranging from face-to-face interaction to digitally improved environments, to digitally enhanced physical worlds (such as digital maps) and finally to virtual worlds (such as virtual reality) (Milgram & Kishino, 1994). Multicomponent Working Memory (M-WM): a more accurate model of primary or short-term memory which splits working memory into multiple components, rather than considering it to be a single construct (Baddeley, 2017). Naïve Bayes: a family of simple probabilistic classifiers which apply Bayes' theorem with a strong independence assumptions between the features (Naik & Samant, 2016). NRC: National Research Council NCES: National Science Education Standards Neurocognitive: “measures of cognitive abilities, such as memory, attention, and executive functions” (Lepage et al., 2014). 15

Neuroimaging: see Functional Near Infrared Spectroscopy (fNIRS) and Functional Magnetic Resonance Imagery (fMRI). NGSS: Next Generation Science Standards Occipitotemporal cortex: the primary cortical region of the brain that receives, integrates, and processes visual information relayed from the retinas. Located in the occipital lobe of the primary cerebral cortex, which is in the most posterior region of the brain, the visual cortex divides into five different areas (V1 to V5) based on function and structure (Huff, 2019) Parietal cortex: a region of the brain associated with attention and episodic memory (Wixted et al., 2018). Person Item Map (Item- Person Map: also called a ‘variable’ map or Wright map (e.g., Wilson, 2005) to honor Ben Wright’s influence on its development and future use show person and item relations in a meaningful pictorial, or ‘map’ form (Bond & Fox, 2015). Proprioception: sometimes considered a sixth sense in that it is a sense of the body's location and movement in the external environment, but around the body (Cameron, 2001). Psychometrics: the measurement of individual differences and psychophysical measurements of similarity (Cripps, 2017). 16

Psychophysiology: is the study of “relations between psychological manipulations and resulting physiological responses, measured in the living organism, to promote understanding of the relation between mental and bodily processes” (Andreassi, 2007, p. 1). Random Forest: generates a set of random trees to create a classification model that predicts the value of a label attribute (Ramya, 2017). Rasch Analysis: an analysis which can be used to 'transform raw human science data into abstract, equal-interval scales,' (Bond & Fox, 2015, p. 7) and is commonly used in psychometrics. Retrospective memory: the ability to recall past semantic and episodic memories (Cabeza & Moscovitch, 2013). Secondary Data: the use of previously recorded data to extract new information (Johnston, 2014). Spatial ability: \"skill in representing, transforming, generating and recalling symbolic, non- linguistic information\" (Linn & Petersen, 1985, p.1482) Spatial computing: see Milgram’s Virtuality Continuum Virtual Reality (VR): allows for a fully immersive sensory experience in almost any space Imaginable often using head mounted displays (Parong & Mayer, 2018). 17

Visuospatial Sketchpad: the portion of one's usual mental facility that includes a simulated world for physical simulation, visualization, calculation and visual memory recall (Allen, Baddeley, & Hitch, 2017). Working Memory: the limited-capacity system which combines mental operations and temporary information storage to support goal directed cognition and behavior (Hitch et al., 2018). 18

CHAPTER TWO Literature Review This section provides an overview of the current literature relating to visual literacy and how visual literacy is a primary component of science literacy. The relationship of visual literacy to the overall comprehension of science concepts ties to biological and cognitive systems associated with spatial ability and episodic memory, associated with the recollection of past events which occurred at a particular place and time (Wixted et al., 2018). These relationships indicate the potential to measure an individual's level of science literacy using neurocognitive measurements. Background, Previous Work, and Justification Although the US is keeping up pace globally (Provasnik et al., 2015), the faster growth in science and technology of other nations may make it more difficult for the United States to stay competitive without a focus on science pedagogy used in classrooms (Fayer, Lacey, & Watson, 2017). Currently, only 28% of adults in the US are considered scientifically literate (Snow, Dibner, Science, et al., 2016). This lack of US achievement in scientific literacy is primarily due to the enormous amount of variation in educational standards across the country as well as a failure by educational researchers to more fully examine the fundamental learning processes in science (García & Weiss, 2017). However, between May 2009 and May 2015, the 10.5% increase in STEM occupations more than doubled the 5.2% net growth in non-STEM fields with the computer market alone projected to yield over 1 million job openings from 2014 to 2024 (Fayer et al., 2017). Without a foundation and understanding of the process of scientific argumentation, the ability to use evidence to logically support a claim (Shivers, Levenson, & 19

Tan, 2017), we will not be able to fill these areas of need. Supporting this foundation requires a focus on what propels innovation and assists the development of modern society, for example, innovation such as incandescent lighting and personal computing which needed exploration into the underlying cognitive tools associated with science literacy (Fayer et al., 2017). However, although we are aware of the importance of the underlying processes of science literacy, limitations within science education often place science educators into the role of only examining the products of science literacy and not the processes which are needed to develop science literacy. Understanding the process of how science literacy is tedious developed through science learning is still not well understood, even though we can identify the underlying cognitive tools and skills, such as using diagrams and drawings to convey meaning (Lee & Spratley, 2010). However, effective measurement tools that can assist and attempt to promote the assessment of an individual’s level of scientific literacy based upon these skills have yet to be developed. This lack of development has occurred despite increased attention to understanding learning processes in education, and importantly integration across disciplines such as cognitive psychology, educational neuroscience, and related fields has not occurred. The rapid development of cognitive sciences has outstripped the abilities of traditional educational research to keep pace with the verification, development, and translation of new ideas from cognitive science to science education. To date, many measures assessing visual science literacy occur using self-test probes (Mnguni, Schönborn, & Anderson, 2016) however, self-test and multiple-choice assessments in content-rich domains, like science, are fundamentally flawed (Opfer, Nehm, & Ha, 2012); since responses are voluntary, they are subject to participant bias. 20

An example of a visual discipline in science is biochemistry due to the requirement that students develop an understanding of numerous representations. However, very little is known about what the students actually understand of the representations used in biochemistry to communicate ideas. The challenge is developing measurements to reduce participant bias through the examination of data not under the direct control of the participant. Data in these approaches arises from direct access to the neurological data (i.e. psychophysiological measurements) from the autonomic nervous system, the part of the peripheral nervous system which regulates involuntary physiological processes like heart rate, blood pressure, respiration, digestion, and sexual arousal (Waxenbaum & Varacallo, 2019). Data from these sources in conjunction with computational modeling via machine learning provides quantitative physiological evidence of associated neurocognitive processes and concept understanding. Computational modeling and simulations of science learning are quite complicated, requiring an understanding of relationships between the cognitive inputs (sensory stimulus), cognitive attributes (neurological processing), and measurable outputs (behaviors) (Lamb, Annetta, Vallett, & Sadler, 2014). One approach in obtaining this information is through the recording and analysis of neurocognitive measurements. This form of measurement results in massive amounts of data from moment to moment streaming electrode outputs, lending itself to effective analysis using Artificial Neural Networks (ANNs). Theoretical Framework Science literacy, commonly associated with content knowledge, understanding scientific practices, and the social practice of how science, is also associated with the foundational literacies of numeracy, visual literacy, and understanding of graphs and charts (Snow et al., 2016). Since the release of the Next Generation Science Standards (NGSS), there has been a 21

demand for educators to understand how students develop the practices of science through the generation of claims and evidence; through the creation of products of learning science (Coburn, Hill, & Spillane, 2016). Research has supported the importance of critical thinking and working memory in learning (Hu, Allen, Baddeley, & Hitch, 2017). However, despite increased attention to understanding learning processes in education, the rapid development of cognitive sciences has outstripped the abilities of traditional educational research to keep pace with the verification, development, and translation of new ideas from cognitive science to education. The key theoretical challenge in educational neuroscience and cognitive science is to bridge levels of analysis, linking brain processes (functions) and behavior in a mutually explanatory manner (Goldwater & Schalk, 2016). The explosive growth of information and psychophysiological measurement technologies in recent years due to the proliferation of biological sensors in the global tech market from wearable technology has resulted in an exponential increase of basic empirical information related to the systems and mechanisms of learning that we use in formal and informal learning contexts. As the existence of this data throughout learning spaces has rapidly increased, our translational capacity to discrete education practices and teacher level information has not. Theories popular among educational practitioners (e.g., schema theory) have attempted to describe the neural basis of information processing through the examination of the products of learning such as mind maps, without linking the products to direct measures of the processes of cognition in learning (Dempsey, Harris, Shumway, Kimball, Herrera, Dsauza, 2015). These examinations and resulting data have not proved to be useful for learning scientists, designers of learning experiences, and educational practitioners such as teachers since there is still tremendous focus on the latent products of learning in education and more importantly in pedagogical practice. Further, with the growing 22

capabilities of harnessing big data and the sheer quantity of information related to basic neurological research available to us, neurocognitive processes can be quantitatively examined for patterns. These patterns, once analyzed with machine learning algorithms reveal relationships between variables not yet considered. An important consideration is how the different disciplinary languages between research and classroom practice interpret and understand each other, requiring transdisciplinary educators to provide links between the fields. This divide arises because behavior is measured in terms of choice and response time, whereas neural activity is measured by spiking activity or blood- oxygen-level-dependent (BOLD) signal in fNIR, for example (Dempsey et al., 2015). One possible bridge between behavior and brain-systems is through the use of cognitive computational models such as the STAC-M (Student Task and Cognition Model; Huys, Maia, & Frank, 2016; Lamb, 2014; Lamb, Annetta, Vallett, & Sadler, 2014). Cognitive computational models are mathematical formalisms that embody psychological principles, often evaluated by their ability to account for behavioral data. The mechanisms in these models can be related to both behavior and to neural measures, thus providing a possible bridge between the abstract cognitive data and practical outcomes the teacher can make use of in the classroom and in the development of supplemental resources. This dissertation details the steps in the development of such a resource; an artificial neural network capable of the assessment of science literacy level. This research integrates the disciplines of integrative learning sciences through a combination of traditional educational research and cognitive science; using tools of artificial intelligence to predict behavior based on psychophysiological measurements. This will address fundamental questions regarding how psychophysiological measurement can help to describe and explain behavioral outcomes (i.e. reading and writing of science texts), the basic processes 23

that underlie unique individual human abilities, and how essential literacies in understanding science are socially acquired. This cognitivist framework aims to link these cognitive processes (eg neurocognitive measurements) to the expressed behavioral outcomes (eg correct or incorrect) in order to evaluate latent constructs (eg visual science literacy). Although minimal, most information on the visual aspect of science literacy has been written in the past decade and referred to as either science visual literacy or visual science literacy and although there are several references found on the visual component of science literacy, the construct has not been fully described as it has usually been overlooked as a part of an individual’s overall scientific literacy but has not fully been detailed as its own literacy that can be developed independently of other science literacies such as numeracy. This lack of theoretical understanding in visual science literacy showed a clear gap in the literature and this study aims to define visual science literacy as a construct. Science Literacy. The term “science literacy” was coined independently by both Hurd (1958) and McCurdy (1958). Science literacy was initially seen as a personal and community issue arising from both the application of science and technology and the development of more knowledge related to both. Each researcher used the phrase to convey the disposition and awareness needed to engage with science. Science literacy directs the development of critical thinking and reasoning skills, which are readily malleable and applicable to lifelong scientific learning (Hand, Lawrence, & Yore, 1999). Science literacy encompasses numerical and visual literacies (Snow et al., 2016); both necessary in understanding the graphs and charts used to represent scientific data. And, since the release of the Common Core Standards (CCSS) and the Next Generation Science Standards (NGSS), there has been demand for educators to understand how students develop the 24

practices of science with the generation of claims and evidence, through the creation of products of learning science (Coburn, Hill, & Spillane, 2016) whereby the incorporation of visual literacy skills in interpreting visual texts and works of art and have been found to stimulate perception and idea production (Shivers et al., 2017), both vital for scientific literacy. Although there is no universally accepted definition of science literacy (Roberts, 2007), scientific literacy is commonly associated with the comprehension of scientific methods, content knowledge and the social practice of science, which is also associated with the foundational literacies of numeracy, visual literacy, and understanding of graphs and charts (Snow et al., 2016). In the late 1950s and through the 1960s, science literacy was seen as less demanding than ‘science' (Hurd, 1958), which was considered desirable for those who did not attend higher education. Since the 1970s, science literacy has been seen as desirable for all students, regardless of background, ability, and interest (DeBoer, 1991). There have been nine elements of science literacy defined; the three basic science literacy elements are (1) understanding the nature of science, (2) understanding basic science concepts and (3) understanding ethics guiding scientists’ work (Shen, 1975). Of intermediate ability, being scientifically literate also requires the understanding of connections, (4) how science and society are interrelated, (5) the interrelationship between science and humanities, and (6) in understanding the relationships and differences between science and technology (Shen, 1975). A more scientifically literate individual will also have the last three types of science literacy which includes (7) practical scientific knowledge which can be used in solving practical problems, (8) civic science literacy which can enable the citizen to become more aware of science-related issues to better participate in the democratic processes, and (9) the knowledge and appreciation of science as a major human achievement and cultural heritage (Shen, 1975). The last three types of science literacy suggest 25

that different people may require different forms of science literacy in different situations. The American Science Advancement Association (AAAS) introduced the broadest notion of science literacy. It defines science literacy as accepting both mathematics and engineering, and natural and social sciences (Lodge, 1989). According to AAAS, a scientifically literate person knows that science, mathematics, and technology are interdependent human enterprises with strengths and limitations, understands the key concepts and principles of science, knows the natural world and recognizes its diversity and unity, and uses scientific knowledge and scientific thinking as an individual (Lodge, 1989). The National Research Council (National Research Council, 1996) offers a more recent and less positive notion of science literacy, which considers that science literacy requires an understanding of unifying science principles and processes, science as research, physical science, life sciences, earth and space sciences, science and technology, personal and social sciences, and science history and nature. In another recent attempt to describe science literacy was made by another National Research Council Committee on Science Education, (National Research Council, 2007), the committee used a different term for the construct, that is, scientific proficiency, its aim to become the target for school science education is the same as that used in literature. According to the committee, scientific proficiency consists of four strands: (1) knowing, using and interpreting scientific explanations of the natural world; (2) generating and evaluating scientific evidence and explanations; (3) understanding the nature and development of scientific knowledge; and (4) participate productively in scientific practices and discourse. The above four strands share several commonalities with scientific literacy established in the National Science Education Standards (National Research Council, 1996). One noticeable difference may be in the last section, where more emphasis is placed on scientific practices and discussion in a mini- 26

society–the classroom (Liu, 2009), and since the introduction of the NGSS, educators have been asked to consider how students build science practices by generating claims and evidence; by developing products of learning science (Coburn, Hill, & Spillane, 2016). Visual Literacy. We live in a world of visuals. Visual literacy refers to the ability to use that visual information to make sense and the combined effects of the stimuli provided by visuals have changed the way students make sense. The “reader” of these images needs the skill or ability to interpret, evaluate and visually represent the meaning (Rowsell, McLean, & Hamilton, 2012). Visual literacy is the ability to read, interpret, understand and communicate that information which is presented in pictorial or graphic images (Elkins, 2009), though diagrams, charts and graphs and is an essential skill in modern society (Miller et al., 2016). Students’ everyday lives reflect the dominance of images on a screen that are colorful, that have animation, texture, and dimensionality; The meaning of images may be at least partly expressed in language and other abstract, symbolic notes with other sensory modalities contributing to these visual models (Ramadas, 2009). Visuals are the foundation of daily experience for people. Human beings have created images that have meaning for thousands of years, but in the last century, the concept of educating visual learning was developed in parallel with new technologies in the communications industry (Felton, 2008). In the late 1960s, John Debes from Eastman Kodak invented the term 'visual literacy' and held the first national conference with a diverse group of academics which soon became the International Visual Literacy Association (IVLA; www.ivla.org). The IVLA hosts an annual conference and sponsors a Web portal (www.ivla.org/portal/intro) that provides links to relevant research, teaching materials, publications, collections, and other resources (Felten, 27

2008). Visual images provide unique information that makes them readily available in a different way from verbal texts. At a glance, the viewer may engage in a dialog with the image, use it as a stimulus to explore personal meanings, and continue to connect them to broader issues. There are, however, different perspectives on how this is achieved cognitively. Purves and Lotto's (2003) study suggests that the visual system interprets a statistical version of visual history rather than a true representation of the physical world, while Stafford (2007) takes a different view of the science of vision, asserting that human perception involves the synthesis of science and art (Felten, 2008). Visual Literacy in Science We recognize that visual literacy is found everywhere but for the purposes of this study we are referring just to the visual literacy required for science. Visual literacy and spatial reasoning are integral to learning in science since complex and real-world phenomena are often simplified into visual representations (Ramadas, 2009); which can be symbolic, including text, drawings, or diagrams or concrete, such as authentic physically manipulatable resources (Stieff, Scopelitis, Lira, & Desutter, 2016) and subject to transformations of image manipulations and for illustrating “dynamic processes” (Ramadas, 2009, p. 312), not static states. Visual literacy, the ability to learn from and communicate through diagrams, charts, and graphs is an essential skill in modern society (Miller, Cromley, & Newcombe, 2016) and the primary barrier to developing this visual literacy and spatial thinking in STEM fields is a lack of representational competence, which includes a distinct set of skills used to generate, transform, analyze and explain representations (Stieff et al., 2018, 2016). Although there is a strong, known relationship between the foundational visual literacy and the domain-specific science literacy (Snow et al., 28

2016), how science literacy is attained through the process of science learning is still not well understood. Moreover, the need for a more reliable measure is necessary to design resources that enhance the fundamental visuospatial cognitive processes behind scientific literacy. This has all led to the question; is visual scientific literacy quantitatively measurable? Researchers have indicated that, in the coming years, jobs requiring at least a basic ability to be scientifically literate (i.e. ability to discern arguments and think critically) is projected to grow twice as fast as jobs requiring other forms of literacy (Galama & Hosek, 2008). Without a foundation and understanding of the process of development of scientific argumentation, we will not fill those jobs – or keep those jobs on our shores – without a concerted effort during critical developmental periods (Burggren & Mueller, 2015). However, although we know the skills required, a measurement tool that can assess an individual’s level of scientific literacy has yet to be developed. To improve the United States’ global competitive advantage in science in technology, necessitates a focus on innovation, needing the development of the underlying cognitive tools of science literacy (Fayer, Lacey, & Watson, 2017). Spatial Ability. Spatial ability refers to \"skill in representing, transforming, generating and recalling symbolic, non-linguistic information\" (Linn & Petersen, 1985, p.1482) which is included in most of the multiple-aptitude batteries; however, there are contradictions in spatial domain literature that make the topic difficult to understand (Yilmaz, 2009). Literature quite consistently shows a broadly defined spatial variable as independent of verbal and quantitative factors (Cooper & Mumaw, 1985; Burke, Greenbowe, & Hand, 2006). Although researchers agree that spatial ability is an important component of intellectual ability, there is no consensus on the nature of the phenomenon. Spatial ability is not a unitary construct, 29

but it combines sub-skills such as using maps, solving geometry issues, and recognizing two- dimensional representation of three-dimensional objects (Linn & Petersen, 1985). Therefore, different types of spatial abilities have been proposed based on factor analytic studies. Spatial ability factor structure has been a research field since the mid-1940s; however, these studies did not provide a clear picture of the subject's underlying factors. Two main factors of spatial ability have been consistently identified and defined: spatial visualization (Vz) and spatial orientation (SO; Lovett & Forbus, 2013). Vz is the more general factor; however, it's hard to identify because the tests that describe it typically require above average mental ability or general intelligence because an important characteristic of the Vz tests is their complexity; some require rotation, reflection, or folding complex figures, others need to combine different figures, while others require multiple transformations. Vz is the capacity to imagine objects moving, rotating, turning, or inverting without self-reference. Complex tests such as Paper Folding assess this ability, while SO is one's ability to visualize an image from different perspectives, such as in mental rotation (Lovett & Forbus, 2013). History of Visual Literacy in Science. The history of visual literacy and the history of science are inextricably entwined. As arguably humanity’s most important invention, writing was originally simple pictures and symbols, such as tally marks keeping count of some quantity of value at the time; these ancient writings, some found from about 30,000 BC, were representations to aid the maker’s memory but contained no context as to their meaning. Then, in 3500 BC, the ancient Sumerians pressed the earliest found symbolic representations or pictograms into clay tablets in Mesopotamia. These pictures were able to convey much more information than the preceding dots and lines used as simplistic observations 30

or reminders. These ancient inscriptions kept track of other primitive man-made technologies, such as in agriculture; Sumerian pictograms originally kept track of cattle and sacks of grain yet evolved, using about 600 signs, into more complex concepts (Calvo, 2001). This was just the beginning of a long entanglement between visual representations and our ability to remember, to record observations, conceive connections and communicate new knowledge. These visual representations facilitated early man’s ability to learn from mistakes and discover patterns since pictorial representations can break time and language barriers with physical records lasting generations and be understood once the language is long gone. Although phonetics spurred our ability to express more abstract concepts (Calvo, 2001) pictures have remained a staple tool in our quest to record our existence. Symbols can quantify and visualize data as is the case with graphs and charts, they can also explain concepts as in flow charts and diagrams; anything that we can see or visualize can be represented by symbols and pictures which is why these visual representations are ubiquitous to science knowledge and learning. Visual Science Literacy. This explains why visual literacy is an integral part of learning science, as complex and real-world concepts are often expressed in visual, symbolic and concrete forms (Ramadas, 2009). The primary barrier to developing this spatial thinking and visual literacy in STEM fields is representational competence, which includes the ability to generate, transform, analyze and explain representations (Stieff et al., 2018). This representational competence along with an individual’s spatial ability are the main components of this visual science literacy that must be transferred to interpret new visual representations as the science content knowledge is embedded in the visual. Since achievement on mental rotation tasks requires both representational 31

competence and spatial ability, mental rotation can be considered an optimal task for assessing an individual’s level of visual science literacy. Graphics can dramatically enhance science learning by offering important information, but they can also add complexity to the comprehension process, particularly for students who are still learning to decode words, therefore this improvement is more consistent with adult learners (McTigue & Flowers, 2011). Hannus and Hyönä (1999) say that the challenge of reading the illustrated scientific text stems from decision-making levels: 1) comprehension of concepts from text and images, 2) deciding on the order in which images and text are to be studied, 3) judging the relevant and superfluous information contained in texts and images, 4) determining the text and graphics related information and 5) then incorporating the related information. This disconnect is unfortunately not solved through the use of visual textbooks since graphics for young students in scientific trade books do not reflect the types used in science textbooks and journals (McTigue & Flowers, 2011). The study of the physical processes involved in visual perception has both facilitated and strengthened the advocacy for visual literacy. Research shows that seeing is not merely a system of passive processing of stimuli, but also an active construction of meaning, because if the physical act of seeing requires active construction, then the intellectual act of understanding and interpreting what is seen must include a critical observer (Felten, 2008). Research into science visuals is typically focused on ways to teach or ways to assess; in this situation, ways to teach the interpretation of the visual or ways to assess what is understood of the visual. James E. Zull's The Art of Changing the Brain draws on \"Biology of Learning\" to argue that the faculty should make \"extensive use of images to help people learn,\" both by teaching visuals and by requiring students to use different visual forms to represent what they 32

know (Zull, 2004; Felten, 2008), in that way, studies involving visual science literacy have both added the necessary dimension to the construct in addition to relating that instruction to the cognitive processes behind attaining visual science literacy. Visual Science Literacy Instruction. A visual literacy approach in the classroom connects with students’ everyday lives as consumers and producers of texts. It helps them to analyze and explore the webs of meaning within which the images exist as artifacts and visual images can often help students relate distant events (Rowsell et al., 2012). Historically and for technical reasons, visuals have not been as common a resource in the classroom as written texts, but we now have the means to explore their potential much more fully (Rowsell et al., 2012). Currently, instructional pedagogy revolves around explicit direct instruction of the conventions of diagrams (Miller et al., 2016) and the assessment of visuals is done by either the ranking of science visuals (McTigue & Flowers, 2011) or using self-test probes to determine whether the individual being assessed understand an aspect of the visual (Mnguni et al., 2016). Visual literacy has sporadically appeared on the fringes of the national liberal education discourse; For example, in the Association of American Colleges & Universities (AAC&U's ) Greater Expectations report (2002), a key feature of an \"empowered learner\" (p. xi) is the ability to communicate effectively orally and visually, as well as, in writing, and in a second language. However, references to visuals disappeared in the AAC&U follow up report Liberal Education Outcomes (2005), while two new literacies, quantitative and information, now are considered essential intellectual and applied skills that complement written and oral communication (Felten, 2008). Visual literacy instruction in the classroom offers the opportunity to create, analyze, and 33

criticize texts and to see how meaning is transferred through visuals (Rowsell et al., 2012). However, as with any literacy skills, it is important to know the current knowledge and skill level of your students before preparing instruction. Nevertheless, visual literacy is missing in most literacy comprehension tests. To date, one of the few reading assessments that directly inquire about students ' interpretation of science graphics is the publicly available comprehension assessment is the Informational Strategic Cloze Assessment (McTigue & Flowers, 2011). Researchers disagree about the more valid measurement of a person’s visual and spatial ability can be obtained when he or she has to solve a set of varied tasks involving hypothetically similar spatial maneuvers rather than solving similar task types (Ramful, Lowrie, & Logan, 2017), while others contend that a person’s spatial ability can be illustrated through the measurement of multiple dependent variables during a set of similar tasks (Lamb, Antonenko, Etopio, & Seccia, 2018). Assessment of Visual Science Literacy. Existing psychometric test development techniques are largely empirical, arising out of a history of test development dominated by correlational methods. These methods have led to a heavy emphasis on the description of tests by factor analytic techniques or examination of predictive validity. Factor analytic studies have resulted in clearer descriptions of the nature of test content and relationships among items within tests. Predictive validity studies provide an estimate of test value in predicting some external criterion. Neither perspective, however, provides information leading to clearer descriptions of the specific human behaviors upon which successful test performance is based. Traditionally, assessment of science literacy has involved the use of multiple-choice tests, self-report surveys and content probes. Nonetheless, student learning evaluation is a multi- 34

billion-dollar industry with many stakeholders within and outside the education system (Flaitz, 2011). Evaluation is also one of the most contentious issues in the current educational environment, with educators and policymakers often lining up on opposite sides as to the role of assessment in the school system (Messick, 1995). Educators suggest that the assessment is not a meaningful measure of student learning and does not account for key student gains. On the other hand, policymakers are demanding the responsibility of the educator as the primary function of the test. This disconnect between educators and policymakers is a stimulus for business and government to seek more appropriate and authentic student learning measures to drive decision- making processes for curriculum and learning (Demarest, 2010). The call by businesses, governments, educators, and policymakers for a more authentic and realistic assessment has driven much of the assessment innovation in recent years (Hall, 2012). Researchers at the beginning of the test movement have argued that the subject's performance on particular test items (tasks) depends on specific cognitive aspects called attributes. At the same time that cognitive psychology has been expanding its contributions to issues close to those traditionally deemed psychometric, increasing demands have been placed upon the test movement to develop instruments that assess more complex levels of knowledge and performance. The two models commonly used for the development of the cognitive diagnosis are the conjunctive model, in which each attribute is necessary to complete the task and the disjunctive model in which each attribute is assumed independent of each other. Using a conjunctive model to link specific items (tasks) to the required cognitive abilities that make up the assessment, patterns seen as acting as an ideal response pattern (task completion pattern) for a specific knowledge state will be developed. The results of the analysis of the response pattern develop into the likelihood of correctly completing the tasks which can be used to drive 35

pedagogical resources. Through connecting measures and cognitive attributes, cognitive psychology and psychometrics can bridge that gap (Kaufman, 2011). Neuroscience of Visual Literacy. But, perhaps most important to the visual literacy problem is the fact that consciousness apparently creates its own material, that is to say, the universe. The idea of a visual prototype is, of course, an ancient one, yet contemporary neural Platonism reveals that even our subjects of experience are not, paradoxically, located in some external event. Our perception of the world is created by our autonomic nervous system, communicating with the environment at any given moment of consciousness. And the resulting type of representation is more a neural network construction than a deliberate reflection or projection of the universe (Elkins, 2009). In reality, the word \"idea\" comes from the Greek verb \"to see,\" and is often associated with the notion of the \"eidolon,\" the \"seen image,\" which is central to ancient optics and theories on perception (Mitchell, 1987). There are also said to be five senses: vision, hearing, touch, smell and taste, however, others incorporate proprioception, a sixth. The role of the senses is also often said to allow the body to communicate with the outside world, but this is oversimplified. Proprioception is about the body's location and movement — in the external environment, but around the body. Taste is about external stimuli but about them entering the body, as is often the case with smell. Furthermore, the senses are about things that are external to the body, but also about the body in how it interacts with the external world (Cameron, 2001). Attention-memory experiences require synchronized processing through various regions of the brain. In the occipitotemporal cortex, the effect of selective attention on memory has been extensively studied, as attention modulates the sensory information stored in these regions as 36

well as the reactivation of sensory information during memory recovery. The top-down signals are thought to derive from the prefrontal cortex through bias encoding and retrieval processes. Nevertheless, in two other brain regions, more recently, there has been an interest in understanding memory-attention interactions: the hippocampus and the parietal cortex. Within the hippocampus, an area that is fundamentally involved in memory (Wixted et al., 2018), questions concern how attention affects hippocampal processing, and how attention is affected by hippocampal learning. Unlike the hippocampus, the parietal cortex is a region traditionally associated with attention but recent evidence of neuroimaging has suggested the involvement of the parietal cortex in episodic memory. This has sparked debate about whether the parietal cortex contributes to memory through more general mechanisms of attention or whether different mechanisms of memory and attention exist within the parietal cortex. Next, we discuss connections between memory-attention within each of those brain regions (Wixted, 2018). The classic evidence of neuropsychology in the nineteenth and twentieth centuries supported inferences in this direction by researching how function deteriorated when parts of the brain were impaired, and such findings have been reinforced in recent decades by modern imaging techniques for the human brain, but such methods look at relative activity in different brain regions, not from the millions of individual neurons (Arbib, 2016). This relative activity can be extrapolated into brain models which may be either an explicit computational model, in conjunction with computer code and simulations, of how a particular brain system operates; a mathematical model to explore; or a conceptual model, in which case the result of a \"simulation\" may simply be a verbal inference (Arbib, 2016). The visual cortex is the primary cortical region of the brain that receives, integrates, and processes visual information relayed from the retinas. Located in the occipital lobe of the 37

primary cerebral cortex, which is in the most posterior region of the brain, the visual cortex divides into five different areas (V1 to V5) based on function and structure (Huff, 2019) V1 is the primary visual cortex which receives, segments, and integrates visual information from the eyes. Information is then passed to V2 which is involved in object orientation and pattern recognition; it is more complex than the V1 area and important for form and orientation; V2 is involved in the visual association as it has a strong feedforward mechanism from V1 and strong feedbacks from V3-V5. The most complex and specified areas of the visual cortex are V3-V5: V3 has distinct connections with other (non-visual) parts of the brain, V4 is important for attentional modulation and color and V5, also known as the Medial Temporal Visual Area (MT) is essential for motion perception. Even though visual stimuli will feedforward from each section from V1 to V5, visual stimuli are picked up by each of the sections of the visual cortex separately as well. Such automatic processes force human beings to identify and attend to the constancies present in their field of vision, filtering out the noisy context. However, an inquiry into constants leads to the top-down problem of universals, or the \"rules\" of the brain — the invisible and autonomous perceptual grouping and joining of features that always take place below the surface and over which we have no control (Arbib, 2016). The size of these visual areas varies greatly. The first two regions, V1 and V2, are massive and together account for about 70% of the number of neurons in the visual cortex (thus 30 percent of the neurons throughout the cortex). The size of V1 is much bigger than the number of fibers that exit the eye, by a factor of at least 200. Nonetheless, it was calculated that this is more by a factor of several hundred than the amount required to represent the information conveyed by the LGN (Arbib & Bonaiuto, 2016) consistent with the idea that V1's intention is to begin interpreting the picture rather than simply encoding it. The important feature of the visual 38