Skip to main content
eScholarship
Open Access Publications from the University of California

UC Davis

UC Davis Electronic Theses and Dissertations bannerUC Davis

UC Davis Electronic Theses and Dissertations

THREE ESSAYS ON THE LONG-RUN CONSEQUENCES OF TEMPORARY SHOCKS

(2024)

This dissertation explores the long-run consequences of temporary macroeconomic shocks. It relies on the combination of historical data, frontier methods in empirical macroeconomics, and macroeconomic models featuring firm heterogeneity. The first chapter provides evidence of substantial long-run effects from monetary shocks on two sources of endogenous growth; human capital and technological adoption. This contribution is the first to test for the presence of this hysteresis in direct measures of the supply-side potential of economies, rather than on indirect measures, e.g., total factor productivity. To estimate the effects of exogenous monetary policy shocks, I improve identification using the trilemma of international finance by developing a mean-unbiased instrumental variable estimator. Results show substantial hysteresis in both human capital and technological adoption. Importantly, monetary shocks have asymmetric effects, as only contractionary shocks result in long lasting responses in productivity. I evaluate the aggregate importance of monetary hysteresis with a growth accounting exercise. Across the 17 countries in the sample, the accumulated average cost of monetary hysteresis ranges between 1.2 and 9.6% of total factor productivity. The second chapter examines the lasting impact of temporary productivity shocks on technological adoption in a model where firms, decide whether to adopt or abandon a new technology after an unexpected productivity shock. The setting is based on firms that face heterogeneous fixed costs to implement and operate said technology. Due to complementarity in the production of the final good, individual technological choices are interlinked, thereby amplifying and prolonging the effects of large shocks. I find that the estimates align closely with the model's predictions, by merging international cross-country data from three databases and using a non-parametric local projections method. A simplified version of the model, calibrated to match the empirical estimates, reveals that subsidizing the operational costs of technology leads to welfare improvements by protecting the economy's productive potential. Lastly, Chapter 3 (prepared as part of a research project with Michaela Elfsbacka Schmöller from the Bank of Finland) explores the sectoral reallocation effects of monetary policy shocks and their transmission to aggregate productivity. Using data for 65 industries for the period 1948-2014, we show by means of local projections that monetary policy shocks reallocate resources towards the services sector and away from the rest of the economy. Results suggest that, since the services sector had lower productivity growth, monetary-originated reallocation contributed negatively to aggregate productivity growth over the sample period.

Cover page of Characterization of Florpyrauxifen-benzyl Herbicide in California Water-Seeded Rice

Characterization of Florpyrauxifen-benzyl Herbicide in California Water-Seeded Rice

(2024)

Rice (Oryza sativa L.) is a staple crop globally and California is the second-largest rice producer in the United States with more than 200,000 hectares of rice planted in the Sacramento Valley and the Sacramento-San Joaquin River Delta. Weeds have historically been one of the biggest challenges of California rice production systems, where herbicide-resistant weeds have increased the complexity of the weed management. Florpyrauxifen-benzyl (FPB) is a new auxin-mimic herbicide with a novel binding site of action for selective grass, sedge, and broadleaf weed control in rice. The objectives of this research were to 1) characterize the effects of FPB on rice crop safety and weed control when applied alone or in mixture with other partner herbicides; 2) determine optimum FPB application timing to control smallflower umbrella sedge; 3) elucidate the response of late FPB applications on rice flower sterility; 4) determine the effects of FPB on nontarget crops including almond, grape, peach, pistachio, plum, and walnut; and 5) compare the onset of foliar symptoms resulting from simulated FPB drift with residues in almond, pistachio, and walnut leaves at several timepoints after exposure.In the first study, FPB was applied at 1/2X, 1X, and 2X field use rates based on 30 g ai ha–1 alone as well as in mixture combinations with bispyribac-sodium, penoxsulam, and propanil in fields near Arbuckle, Biggs, and Willows, CA. FPB applied alone at 30 g ai ha–1 on 4-5-leaf rice stage controlled more than 80% of watergrasses, ricefield bulrush, and smallflower umbrella sedge (SMF) as well as more than 99% of all broadleaf weeds including ducksalad, redstem, waterhyssop, and arrowhead at 28 days after treatment (DAT). The highest rice yield was observed with FPB plus propanil in Arbuckle and Willows, CA. At Biggs, the highest yield (4,626 kg ha–1) was achieved with FPB applied alone at 60 g ai ha–1. In the second study, FPB at maximum use rate of 40 g ai ha–1 was applied to SMF at 1-leaf, 10-, 15-, 20-, and 25-cm tall growth stages. SMF was controlled by 95%, 86%, 89%, 87%, and 85% when FPB applied on 1-leaf, 10-, 15-, 20-, and 25-cm tall growth stage at 42 DAT, respectively. In the third study, FPB at 40 and 80 g ai ha–1 rates were applied at rice panicle initiation (PI) and 50% flowering (FL) growth stages, respectively. While the weed control was more than 90% at 42 DAT for all applications, the FL application caused 26% and 35% rice sterility at the 40 and 80 g ai ha–1 rates, respectively. In the first off-target drift study, fractional rates were 1/200X, 1/100X, 1/33X, and 1/10X of the FPB use rate of 29.4 g ai ha–1 used in 2020 and 2021 on almond, pistachio, and walnut trees treated early in the growing season. Herbicide treatments were applied directly to one side of the canopy of one- to two-year-old almond, pistachio, and walnut trees. The general symptoms were chlorosis, chlorotic spots, leaf curling, leaf narrowing, leaf distortion, leaf malformation, leaf crinkling, shoot curling, stem coloring, stunting, terminal bud death, and twisting. Most symptoms peaked between 14 through 28 DAT with the 1/10X FPB rate, maximum visible injury was 16%, 49%, and 79% on almond, walnut, and pistachio, respectively. The 1/10X FPB treated pistachio trees did not recover as fully as almond and walnut, and injury symptoms persisted for the remainder of the 2021-2022 growing seasons on pistachio. In the second drift study, FPB was applied to one side of the canopy of one- and two-year-old almond, pistachio, and walnut trees at 1/100X and 1/33X of the field use rate of 29.4 g ai ha–1 in 2020 and 2021. Leaf samples were randomly collected for residue analysis at 7, 14, and 28 DAT. Seven DAT with the 1/33X rate, almond, pistachio, and walnut leaves had FPB at 6.06, 5.95, and 13.12 ng g–1 (fresh weight; FW) leaf, respectively. By 28 DAT, all samples from all crops tested with the 1/33X drift rate had FPB at less than 0.25 ng g–1 FW leaf. This study showed that the ideal time frame to collect leaf tissues from trees should be within 14 days after exposure; chemical analysis after this time may underestimate actual exposure. In the third drift study with grapevine, peach, and plum, the fractional rates were 1/200X, 1/100X, 1/33X, and 1/10X of FPB based on 29.4 g ai ha–1. Herbicides treatments were applied on one- to two-year-old peach and plum trees as well as on established grapevines in 2020 and 2021. The general symptoms were chlorosis, chlorotic spots, leaf curling, leaf distortion, leaf malformation, leaf crinkling, necrosis, necrotic spots, and twisting on leaves. Most symptoms appeared at 1/10X FPB rate and peaked from 14 through 42 DAT depending on the species. At 1/10X rate, visible injury was 5%, 50%, and 71% for plum, peach, and grapevine, respectively, at 14 DAT. Some grapevine clusters showed deformation, asymmetrical growth, and fruit dropping. Foliage of all treated crops gradually recovered throughout the growing season regardless of the application rate. Because of low injury symptoms and rapid recovery from herbicide injury in almond, peach, plum, and walnut trees, the proper herbicide drift management, and application precautions are likely to reduce the risk of crop injury from florpyrauxifen-benzyl drift for these crops; however, extra precaution should be taken if there are nearby grapevine vineyards or pistachio orchards because of their greater sensitivity.

Cover page of The function of anaplastic lymphoma kinase receptor and the activities of its inhibitors in neuroendocrine prostate cancer

The function of anaplastic lymphoma kinase receptor and the activities of its inhibitors in neuroendocrine prostate cancer

(2024)

Neuroendocrine prostate cancer (NEPC) represents an aggressive subtype of prostate cancer, which may either manifest de novo or emerge because of anti-androgen receptor (AR) drug treatment. Unfortunately, there is currently no established effective treatment for NEPC. Alectinib and entrectinib are FDA-approved receptor tyrosine kinase inhibitors designed to target the anaplastic lymphoma kinase receptor (ALK) initially developed for NSCLC patients who developed resistance to the receptor tyrosine kinase (RTK) inhibitor crizotinib. In this study, we sought to assess the potential of alectinib and entrectinib as therapeutic agents for NEPC. We conducted experiments using various NEPC cell lines and castration-resistant prostate cancer (CRPC) cells, subjecting them to treatment with alectinib and entrectinib. Notably, treatment with these drugs, particularly in cell lines 42D, NCI-H660, C42B-entry, and LuCaP, significantly inhibited cell growth and proliferation. Further investigation revealed that alectinib and entrectinib exerted their effects by impacting the ALK signaling pathway. Specifically, downstream targets of ALK signaling, such as STAT3 and AKT phosphorylation, exhibited reductions in treated cells. Additionally, the expression of NEPC cellular markers, including enolase2, BRN2, and ASCL1, decreased upon alectinib treatment. Notably, NEPC tumors treated with alectinib or entrectinib in combination with CM 272, a G9a/DMNT1 inhibitor, exhibited a notable decrease in tumor volume. Both alectinib and entrectinib demonstrated their efficacy in inhibiting various cellular pathways, including the Neuroactive ligand-receptor and cytokine-to-cytokine receptor interaction pathways. The findings of our current study provide compelling evidence that alectinib and entrectinib have the potential to effectively inhibit NEPC cell and tumor growth. Consequently, these drugs hold promise as therapeutics for patients grappling with NEPC, offering a potential avenue for improved treatment outcomes.

A Study Of Small Evolution of Vision Transformers For Low Power Devices

(2024)

The emergence of Transformers revolutionized the landscape of Natural Language processing (NLP) algorithms, supplanting traditional approaches like LSTM and RNN withTransformer-based models such as BERT and GPT. Despite their superior performance, the substantial memory storage demands and computational complexity of Transformer-based algorithms pose challenges for their deployment in embedded devices. For instance, GPT-3 alone requires storing and fetching 175 billion parameters, leading to memory bottlenecks and heightened power consumption due to frequent parameter transfers between computational units and memory. The success of Transformer-based models in NLP has spurred their application to computer vision tasks as well. Despite their origins in language processing, models like BERT and GPT have been adapted for image classification, segmentation, and object detection tasks due to their remarkable performance. However, the associated challenges of memory storage requirements and computational complexity persist, hindering their effective deployment on embedded devices in the vision domain. This study tackles the challenges faced by vision Transformer algorithms by introducing an innovative approach to crafting energy-efficient dynamically prunable Vision transformers tailored for edge applications. Termed Incremental Resolution Enhancing Transformer (IRET), our method revolves around sequentially sampling the input image. Notably, our solution utilizes smaller embedding sizes for input tokens in comparison to previous approaches. These embeddings are employed in the initial layers of the IRET vision transformer until a robust attention matrix is established. Subsequently, this attention matrix guides the sampling of additional information via a learnable 2D lifting scheme, focusing solely on important tokens while dropping those with low attention scores. As the model concentrates more on a subset of tokens, its attention and resolution naturally amplify. This incremental attention-driven input sampling and token-dropping mechanism enables IRET to prune its computation tree significantly as needed. By adjusting the threshold for discarding unattended tokens and augmenting the focus on attended ones, we can train a model that dynamically balances complexity with accuracy. Additionally, we explore the application of dynamic inference techniques to the model, enabling it to predict outcomes early. This feature is particularly advantageous for edge devices, where the trade-off between accuracy and complexity can be dynamically adjusted based on factors like battery life and reliability.

Cover page of Exploring the potential of biogenic manganese oxides as water oxidation catalyst and investigating Mn oxidation mechanism using Electron Paramagnetic Resonance (EPR) spectroscopy

Exploring the potential of biogenic manganese oxides as water oxidation catalyst and investigating Mn oxidation mechanism using Electron Paramagnetic Resonance (EPR) spectroscopy

(2024)

Water splitting is a promising but challenging solution to alleviate the urgent fuel crisis. While the hydrogen-evolution reaction provides the powerful hydrogen gas as a renewable energy source, the high energy barrier of the anodic oxygen-evolution reaction (OER) limits the overall water splitting efficiency. While heavy metal oxides have been found to be the highly efficient OER catalysts, nature employs the oxygen-evolving complex (OEC) in the photosystem II, which consists of a Mn4O5Ca cluster. It generates most of the O2 in the world in a highly efficient and persistent manner. Inspired by the OEC cluster, in this dissertation, we synthesized biogenic manganese oxides (BioMnOx) using a multicopper oxidase Mnx for OER catalysts. Chapter 1 will provide background information about the OER catalysts, the Mnx protein and Electron Paramagnetic Resonance (EPR) spectroscopy. Chapter 2 will explore the potential of the BioMnOx as OER catalysts and the structure- function relationship. Chapter 3 will investigate Co-doping effect of BioMnOx as well as the structural elucidation of Co-doped BioMnOx using X-ray Absorption Spectroscopy (XAS). On the mechanistic side of the story, Chapter 4 investigates the first row transition-metal ion-inhibition effect of Mnx. EPR spectroscopy has been proven to be a powerful tool to selectively probe the active sites and the metal binding sites of Mnx. Understanding the mechanism of the inhibition effect provides fundamental knowledge about the Mnx mecha- nism and provide information for formulating and optimizing BioMnOx synthesis. Chapter 5 extends the usage of EPR spectroscopy to other metalloprotein and inorganic systems.

Cover page of A Finite Element Analysis-Based Study of Mechanical Behavior of Nanoporous Gold Thin Films on Silicone Substrates with Varying Effective Stiffness

A Finite Element Analysis-Based Study of Mechanical Behavior of Nanoporous Gold Thin Films on Silicone Substrates with Varying Effective Stiffness

(2024)

Nanostructured materials offer tremendous opportunities for engineering advanced device components for diagnostic and therapeutic applications. One such material, nanoporous gold (np-Au), has found use in applications ranging from catalysis to biosensing, where pore morphology plays a critical role in performance. Np-Au is typically produced by a process known as dealloying, where immersion in nitric acid selectively removes silver from a gold-silver alloy and gold surface atoms diffuse at the metal-electrolyte interface arranging into a bicontinuous ligament network. While morphology evolution of bulk np-Au has been widely studied, knowledge about its thin film form, which is influenced by the underlying substrate, is limited. This thesis delves into the mechanical behavior of nanoporous gold (np-Au) thin films on substrates of varying mechanical compliance, focusing on the role of substrate stiffness controlled by the thickness of polydimethylsiloxane (PDMS) layers anchored onto a rigid glass slide. Using a finite element analysis (FEA) framework, the study simulates the deformation and strain energy characteristics of np-Au films, revealing a nuanced interplay between substrate compliance and film morphology. Simulation results indicate that the effective elastic modulus of PDMS, modulated by its thickness, critically affects the deformation patterns in np-Au thin films. At the film-substrate interface, simulations show that the np-Au/PDMS system undergoes significantly greater deformation than np-Au/glass, characterized by both in-plane compressive and out-of-plane vertical displacements. The study further presents a detailed analysis of strain energies, with the simulations uncovering that the total strain energy of the film-PDMS system decreases as PDMS thickness increases. Corresponding experiments performed in our group show that the decrease in strain energy is associated with the diminished presence of macroscopic cracks in the np-Au films on thicker PDMS substrates from experiments, as opposed to those on glass. The simulations also highlight that the distribution and intensity of microscopic cracks are contingent on the PDMS thickness, confirming experimental observations of the hierarchical crack formation in the np-Au/PDMS system and offering predictive insights into the mechanical stability of the films. In conclusion, the simulations provide compelling evidence that the mechanical characteristics of np-Au films can be finely tuned by adjusting the thickness of the anchored compliant substrate. This paves the way for engineering advanced materials with tailored morphological properties, optimizing np-Au thin films for applications in flexible electronics and wearable sensors.

A Model-Theoretic Logical Inferentialist Account of Three-Valued Strong Kleene Logics

(2024)

Proponents of model-theoretic logical inferentialism defend the view that the meaning of logical vocabulary, such as the truth functions associated with conjunction or negation, is determined by their inferential use. This position provides a naturalist and empiricist account of the meaning of logical vocabulary, since inferences consist of concrete speech acts that can be observed, such as assertions and denials. As a consequence, model-theoretic inferentialism gives an account of a problem widely discussed in philosophy of language as well as cognitive science for the meaning of logical vocabulary, that is, it provides an account of meaning based on the use of language. In this dissertation, we provide a model-theoretic logical inferentialist account to three-valued Strong Kleene Logics. In particular, we will show various solutions to the categoricity problem regarding these logics in order to show that we can determine the intended models of these logics from their inferences, we will give a model-theoretic logical inferentialist account to Quine [94]'s challenge ``Do proponents of different logics talk past each other, i.e., does the meaning of logical operators change when logics change?", and last we will give an account of Prior [93]'s challenge of tonk to inferentialism.

In Chapter 1, we introduce model-thereotic logical inferentialism, and we introduce the technical machinery to determine the inferential semantics from a logic. In Chapter 2, we show how we can determine the intended semantics of Classical Logic and a class of Strong Kleene logics. In other words, we provide a number of well-motivated solutions to the categoricity problem, also known as Carnap's problem, regarding these logics. In Chapter 3, we provide a model-theoretic logical inferentialist account Quine [94]'s question: Do proponents of different logics talk past each other, i.e., does the meaning of logical operators change when logics change? We argue that model-theoretic inferentialists can provide an alternative account to this problem in comparison to the traditional model-theoretic or proof-theoretic accounts, since model-theoretic inferentialists use the inferential semantics of a logic. Chapter 4, we discuss the problem of tonk from a model-theoretic logical inferentialist point of view. We first introduce different tonk-like logical operators in our metainferential system, and define notions of metainferential existence and metainferential uniqueness. Then, we argue that the solution discussed by Fjellstad [39] is favored by model-theoretic logical inferentialist, given that the non-determinate interpretation of tonk discussed in Teijeiro [126] does not yield a completeness result when tonk-like connectives are introduced to our metainferential systems. Then, we argue that the substructural solution to tonk cannot be generalized to all logical operators.

Modeling Revolution: A Global History of Games as Model Systems

(2024)

This dissertation tells a history of how games have been employed as model systems. By focusing on the influence of games upon the theory and design of symbolic computation from the seventeenth century till the end of the twentieth, it argues that games are the paradigmatic technologies for a still-ongoing revolution in modeling practice. From probability theory to general-purpose computers and from game theory to artificial intelligence and machine learning, the domains and technologies emerging from this modeling revolution define every aspect of industrial and post-industrial life.My first chapter argues that games were central to Gottfried Wilhelm Leibniz’s philosophy as modeling tools by which situations arising in the world could be formalized and subsequently rationalized. By contextualizing Leibniz's project within the earlier Llullian art, the theory of probability derived from games of chance, and other discourses of games in medieval and early modern Europe, I describe how his approach to games was unique in European history and synthesized many different discourses into a general science of games. In chapter two, I describe how Charles Babbage’s earliest work on a machine for general computation was inspired by his mathematical study of games. This research on games, which Babbage explicitly credits to Leibniz’s own study of the “geometry of situation” through games, was also a topic of interest in the early correspondence of Babbage and Ada Lovelace. Games provided crucial tools for notational description of mechanisms, the depiction of computational processes using abstract spatial grids, and temporal sequences of operations that would allow for the development of techniques like the “anticipating carry” mechanism. In my third chapter, I take a broader view of twentieth- and twenty-first century systems thinking—comprising formal mathematics, cybernetics, computer science, AI/ML, cognitive science, operations research, economics, and other related domains—to argue that the game of chess modeled for these domains the twentieth-century concept of a formal-symbolic system. I index the ubiquitous discussion and study of chess in systems thinking in the first half of the twentieth century to changes in the concepts, strategies, and institutions of chess in the decades immediately prior. Subsequent changes in chess, including the creative ‘hypermodernist’ movement and its crosspollination with competitive Go, continued to have impacts on how games were understood, discussed, and studied in the sciences. In chapter four, I focus more specifically on chess in the first half of the twentieth century, arguing that it transformed in the early USSR into a tool for modeling dialectical materialism. Chess was understood by Soviet scientists, politicians, and competitors as a tool for cultural uplift and social rationalization, as well as a neutral site for the empirical demonstration of the superiority of socialism over capitalism, because it allowed both ideologies to be represented simultaneously across the board. Finally, I describe a 1926 state-sponsored psychotechnical study on the game of chess as a (now forgotten) exemplar of a Marxist theory of games as modeling technologies that provides a far more nuanced description of games as evolutionary tools than that of Johan Huizinga a decade later. Chapter five returns to the topic of symbolic computation as a global domain of technoscientific research to argue that differing conceptions of chess were central to the twentieth-century development of symbolic computation, cybernetics, organizational research, and artificial intelligence. By recontextualizing the pivotal ITEP-Stanford match between Soviet and American AI researchers in light of Soviet theories of chess and intelligence, I argue that, far from being a ‘fall from grace’ for AI, the ideological neutrality of chess allowed computer chess research to influence systems thinking far more broadly than AI alone could have done. I conclude by describing the chess theory of Soviet control systems engineer, computer chess researcher, and world chess champion Mikhail Botvinnik. I suggest that Botvinnik’s project of formalizing “the algorithm” of historical chess masters into computer language in order to optimize economic planning and other massive social projects represents a revolutionary theory of modeling practice that can help us to reimagine the possibilities of games and computer technology alike in our moment. The mid-century synthesis of communist and capitalist game research produced all the necessary tools for us to construct entirely new futures, and these tools are everywhere around us. It is time to understand them as such.

Cover page of Stochastic Optimization for Machine Learning: Investigations on Bilevel Optimization and Large Learning Rates

Stochastic Optimization for Machine Learning: Investigations on Bilevel Optimization and Large Learning Rates

(2024)

Stochastic optimization is fundamental to modern machine learning and deep learning problems. It provides various algorithmic frameworks, such as stochastic gradient descent (SGD), adaptive gradient algorithm (ADAGRAD) and adaptive moment estimation (ADAM), to efficiently minimize loss functions constructed from large-scale datasets. In this dissertation, we explore the theoretical properties and empirical performance of bilevel optimization algorithms and the phenomenon of large learning rates in machine learning. First, we introduce a novel algorithm, the Moving-Average Stochastic Bilevel Algorithm (MA-SOBA), designed for solving stochastic bilevel optimization under standard smoothness assumptions. Next, we extend the scope of bilevel optimization algorithms from single-agent training to a multi-agent context, i.e., distributed training, by proposing the Moving-Average Decentralized Stochastic Bilevel Optimization (MA-DSBO) algorithm. This approach improves the per-iteration complexity of previous methods, reducing the quadratic dependency on dimensionality to linear dependency. Lastly, inspired by the Edge of Stability (EoS) phenomenon observed in modern deep learning, we examine the training dynamics of gradient descent in a class of quadratic regression models with large learning rates—a scenario that classical optimization theory struggles to explain.

Crafting Disciplines: Women’s Artistic Labor and the Development of Nineteenth-Century American Science

(2024)

Crafting Disciplines interrogates the material dimensions of nineteenth-century scientific objects to reveal a network of craft processes—fancywork, stenography, and photography—that materially and technically supported the emergence of modern American scientific disciplines. Engaging with the histories of nineteenth-century American science, historical recovery projects of women in nineteenth-century American science, and science and technology studies (STS) scholarship on science as practice, this project demonstrates how craft shaped scientific domains through collegiate instruction, public popularization, and internal technical processes. I argue that women’s technical acuity with image and specimen creation, transcription, and notation of experimental procedures were integral to this key period in the development of modern American science. By tracking the material and sociocultural contexts of scientific craft objects made by three elite white women over the course of the century, I demonstrate how craftwork legitimated the burgeoning professional establishment by connecting it to white Christian ideologies associated with domestic production and, through its associations with the mechanical arts, ensured that descriptive and theoretical scientific methodologies remained connected to the social infrastructure of technical expertise required for technological progress.

Each chapter of this project is a case study examining a particular archival collection of crafted scientific objects made by white women embedded in the emerging professional scientific establishment of the nineteenth-century US. I emphasize the materiality of the archival craft objects to ground a discussion of how gendered craft practices shaped professional scientific disciplines and institutional formations. I begin in the college geology classroom, exploring how large visual aids made by Orra White Hitchcock for her husband’s lectures negotiated the unstable theoretical terrain of new geological concepts like deep time for audiences of Christian students. I then turn to the public lecture hall, where Elizabeth Agassiz sat day after day transcribing and later circulating her husband’s lectures on theories of polygenism for massive audiences across the New England and the South. I end in the astronomical observatory, delving into the world of solar photography at the Vassar College Observatory run by famed female astronomer Maria Mitchell as an experimental program in redefining women’s professional scientific work. A short coda explores the possibilities of a new scholarly craft practice—artists’ bookmaking—in researching the interdisciplinary histories of science and art. Situating artists’ bookmaking alongside the already established practices of historical reconstruction and studio art praxis, I discuss how my own engagement in making artists’ books alongside my research into the gendered histories of American scientific craft is exemplary of how we might approach archive-based craft research in the future.