Computational thinking (CT) has been supported as an important skill every young person should possess for the 21st century, with possible implications for problem-solving, self-expression, and creativity. Numerous initiatives, both within and outside classroom settings, have been developed in response to policy mandates aiming at broadening participation for all K-12 learners to acquire CT skills. Consequently, there has been a proliferation of computational toys and tools specifically designed for young learners, including codable robots introduced into classrooms and educational environments. With the growing prevalence of computational devices in educational settings, educators, curriculum designers, and researchers must cultivate diverse teaching approaches and deepen their understanding of how young learners engage with these devices to acquire CT skills effectively within classroom contexts. In this dissertation, I present findings of how elementary-grade learners develop CT skills when they program Sphero robots in mathematics classroom activities. I specifically focused on the kinds of representations students developed, considering their perspectives (understanding) of the environment, and the practices they engaged in to accomplish given tasks.To understand how young learners acquired CT skills, I observed fourth-grade learners as they interacted with activities on the Sphero.Math curriculum to program the Sphero robot in mathematics classrooms. The Sphero.Math curriculum was developed through a collaborative effort between researchers and DCPS partners. Findings from this work revealed that representations play an important role in supporting young learners to engage in CT practices such as Pattern recognition, algorithm design, problem decomposition, and abstraction (PRADA). Findings showed that representations such as (1) concrete manipulatives, (2) language, (3) graphic, (4) symbolic and (5) embodied representations provide scaffolds for learners to gain (PRADA), CT skills through iterating, testing, debugging, abstracting, modularizing, and reusing code. Additionally, the design features of the Sphero robot and its programming environment support CT knowledge acquisition. Features such as (1) programmable LEDs provided opportunities for learners to break down tasks and create opportunities to organize and structure components to get visual feedback that helped them recognize patterns. (2) Taillight (“aim”) LED provided visual cues, that facilitated the involvement of geocentric orientation and embodied practices that empowered students to establish sensorimotor references. (3) Sphero’s virtual protractor supported students through the CT component of abstraction to address the geocentric aspects of the Sphero robot. (4) block-based environment/language, that involves the use of shapes and colors as effective visual aids and abstraction tools, to support the learners’ construct to algorithms. This research can serve as a resource for researchers, curriculum designers, educators, and designers to answer questions about design, choice of computational tools, and their respective programming environments that can afford meaningful CT experiences. Familiarizing learners with representations within CT robotics learning environments serves as a gentle initiation into emerging topics in education such as AI, ML, and data science, given the pivotal role representations play within these fields.