progress—across different fields and over time—than individual scientific discoveries Many discoveries are almost inevitable once we create a new tool that enables seeing and measuring further. These breakthroughs range from how we discovered capillaries in 1661, cells in 1665 and bacteria in 1674 after inventing the microscope in 1590. To how we found mitochondria in 1898 once we designed an improved achromatic microscope in 1841. These breakthroughs span from how we uncovered Jupiter's moons in 1610, Saturn's rings in 1655, galaxies in 1750 and Uranus in 1781 after constructing the telescope in 1608. To how we revealed that the universe is expanding in 1929 after making the needed telescope in 1917 and how we detected the first planet outside our solar system in 1995 after building a high-resolution spectrograph in 1993. These extraordinary new tools sparked these discoveries that no one had ever imagined—by extending what we can observe. And they sparked the later theoretical explanations to describe them. These discoveries were triggered by what the tool in our hands made possible. And they were highly likely once we invented the needed tool. Before these tools, we did not know these even existed—the same with microscopes later triggering breakthroughs from viruses and chromosomes to nerve cells and tissues. It is difficult to argue that—for example in astronomy—one discovery (such as Jupiter's moons) may be more influential for scientific progress in the field than other discoveries (such as galaxies or the expanding universe). Yet none of these discoveries early classification of elements (1829) mechanical computing machine/computer (difference engine) (1833) computer programming (1843) natural experiments in medicine (1855) mathematical law of octaves (1865) discovery of DNA (1869) Periodic table of elements (1869), and discoveries of elements to fill empty places in the table Einstein's general theory of relativity (1915) discovery of insulin (extraction) (1921) modern statistics (1925/1926) instrumental variables (1928) neutrino hypothesis (1930) Econometrics (1933) Turing machine (1936) Digital electronic computer (1940s/1950s) chromatography, paper partition (1944) randomised controlled trial (RCT) method (1948) x-ray crystallography (1913) Structure of DNA molecule (1953) Detection of neutrino (1953) liquid scintillation detector/nuclear reactor (1953) mathematical gauge theory/Yang-Mills model (1954) Structure of insulin (1955) regression discontinuity method (1960) Lab experimentation in economics/experimental economics (1962) theory of Higgs particle (1964) electrophoresis, PAGE (1964) DNA sequencing (1977) Methods for analysing causal relationships in economics - natural experiments (1990) genome mapping, improved (2001) Experimental approach to alleviating poverty (2003) Human genome sequence (2004) large hadron collider (particle accelerator) (2008) Higgs particle/boson (2012) LIGO detector, improved/laser interferometer (2015) Gravitational waves (2015) 1825 1850 1900 1925 1950 1975 2000 Legend Predicted discoveries (in bold) Methods and tools - together with at times existing research - used to make the discovery (not in bold) Figure 1.6 Predicting some discoveries: examples of breakthroughs anticipated before they happened The timeline shows when the central methods and discoveries were made—and the moment when the breakthrough could be predicted. These predicted discoveries (shown in bold) received a Nobel prize, except for the digital electronic computer, the Turing machine and the periodic table of elements. S would have been even conceivable without first inventing the telescope—created by a Dutch lensmaker, Hans Lippershey, in his eyeglass shop in 1608. He designed the first telescope to help magnify distant objects by mounting lenses within a tube. A tool he thought could 'likely to be of utility to the state' especially for military or naval use. And he had no connection to science. From a new highly sensitive spectrograph that enabled discovering many exoplanets since the 1990s, to a new radio telescope that sparked the discovery of pulsars and quasars in the 1960s, this pattern repeats itself. It highlights the fundamental role of a single instrument: one that sparks multiple unintended discoveries and makes it often the more foundational discovery itself. What makes them so powerful is that a single tool triggers a cascade of breakthroughs. Many researchers who would have gotten their hands on the new microscope, optical telescope, spectrograph or radio telescope could have made these discoveries that were not searched for—guided more by empirical observations than theory. They were also made without a large research team. Discovery often comes down to which researchers get access to the new tool first. This often pushes our tool innovations into the foreground, and discoverers and other factors into the background of discovery. We trigger many major scientific advances not through testing a clear hypothesis, but through exploratory research using a new tool innovation—a topic we turn to in the next chapter on serendipity. Exploring science's major discoveries, we uncover the striking rise and importance of the most prominent tools in science: hot streaks of discoveries cluster after we develop powerful new tools (as mapped out in Figure 1.7). The discoveries we unlock bundle together not by chance but by extending our toolbox—from x-ray methods and spectrometers to statistical methods. For each of these tools, the shared link across the diverse discoveries it enabled is not the same theory, teams or funding—it is the tool itself that made them possible. Each one unlocked findings across different disciplines. What commonly makes new tools the crucial driver of progress is not just that each of these tools has made multiple discoveries possible, but that many do so across different fields and in ways their inventors never imagined—and can also do so in the future (Figure 1.1 and Appendix Figure 1.3).
We have seen how discoveries follow the invention of powerful new tools. Now let us dig deeper: how does this cause-and-effect relationship work? There are five pieces of evidence that show how the two link together. First, science's top-ten most influential methods and tools were not designed with different discoveries in mind—but they still made them possible. Charles Townes did not invent the maser (the precursor of the laser) to unlock the structure of molecules. Ernst Ruska did not design the electron microscope to trigger major advances on viruses and nanotechnology. Max von Laue did not create x-ray diffraction to enable breakthroughs on the atomic structure of complex proteins or in materials science. And yet, each of these tools triggered more than ten such major discoveries. In fact, these ten general-purpose tools are responsible for 21% of all nobel-prize discoveries—over 100 of the 533 breakthroughs they could not have predicted. And 52 T HE ENGINE OF SCIENTIFIC DISCOVERY animal experiment techniques (1890) bioassay technique (1894) blotting technique (1975) cathode ray tube (1896) centrifuge (1924) chromatography (1931) computer (1950) electrophoresis (1930) formal economic modelling (1870) game theory analysis (1928) geiger counter (1928) microscope - modern optical (1873) microscope - electron (1933) PCR method (1985) laser/maser (1954) modern statistics (1925) particle accelerator (1929) particle detector (1911) spectrometer (1859) syringe (1853) x-ray method (1895) 1800 1850 1900 1950 2000 Year that discoveries were made using method/tool 12 10 8 6 4 2 # of discoveries Figure 1.7 Timeline of developing and using central methods and tools—each enabling clusters of major discoveries, since 1800 The data reflect science's 308 major discoveries made since 1800—including all nobel-prize discoveries—using central methods and tools that each triggered four or more discoveries. The year shown marks when each method or tool was first developed—but all have since been vastly expanded. For modern statistics, 1925 is used as when the method was created—as it was the year Ronald Fisher, the father of the field, published 'Statistical Methods for Research Workers' . It marked the first full-length book on statistical methods and played a critical role in establishing and spreading modern statistics. For animal experiment techniques, these became standardised and widely adopted around 1890, including producing animals for experimental use. remarkably, about 55 methods and tools have led to two or more nobel-prize discoveries and make up about 80% of all those breakthroughs (426 of 533). This leads to a key insight: each powerful new tool unlocks multiple questions and findings we cannot predict. And because each of the ten most powerful discovery tools uncovered multiple breakthroughs across different fields, their impact often outweighs the varying role of serendipity, funding or specific theories . Second, all top-ten tools, and most of the 55 most used tools in science, were invented in one field but went on to trigger discoveries in other fields. Lasers and electron microscopes came out of physics, but ended up transforming biology, chemistry and medicine. The tools' inventors did not anticipate the different breakthroughs across disciplines (Figure 1.1 and Appendix Figure 1.3). This is one of the strongest signs of a causal relationship: when scientists working in completely different fields pick up a tool and spark major discoveries using it—with supporting factors like collaboration and funding varying by context. It is not necessarily about who holds the tool, but what the tool can do across fields. Third, not just new tools, but extensions S to science's best x-ray devices, lasers and centrifuges, improvements to our best statistical, microscopic and electrophoretic methods, have consistently brought about new breakthrough advances. Fourth, we have not found a major breakthrough that did not rely on a novel method or tool (Figure 1.2). If we want to uncover something new, we need a new lens, detector or technique to reveal it. There is no way around it. Fifth, many discoveries have not been guided by a hypothesis or theory, but by designing a new tool and doing exploratory research with it. In fact, serendipitous discoveries commonly depend on using a new tool that enables the surprising breakthrough observation. We explore this in depth in the next chapter. We can think of a causal relationship between new methods and new discoveries using a kind of experimental logic: a quasi-experiment. How does this causal mechanism work? Before we develop the new enabling tool ( the treatment), we cannot trigger the breakthroughs (the outcome ); but after the tool is created, the breakthroughs follow—often soon. The odds of making many discoveries are zero beforehand, but jump when we develop the tools. The timing matters. The invention year is the baseline, and the discovery year is theendline. The breakthroughs—like uncovering DNA's structure with x-ray crystallography—were not possible without the enabling tools (the counterfactual). The breakthroughs are not bound to occur specifically when they do—but they cannot happen without the right method in hand. Inventing a powerful new method is like getting the clock ticking towards uncovering a discovery that was previously out of reach. The biochemists Archer Martin and Richard Synge invented chromatographic methods that enabled breakthroughs they could not have imagined: from the structure of insulin and sugar nucleotides to carbon dioxide assimilation in plants. The physicist Max von Laue developed x-ray diffraction that led to x-ray crystallography and discoveries he could never have predicted: from the structure of vitamin B12 and penicillin to the DNA's double-helix, and so on. Some major tools began with no specific purpose, like the first maser. Or with no scientific use in mind at all, like the first microscope and telescope. Or are serendipitously found, like x-rays. The physicist Charles Townes famously described his new maser—invented while at Columbia University—as 'a solution looking for a problem' since it had no clear relevance at first. The first microscope and telescope were both constructed not by scientists but eyeglass makers as an extension of eyeglasses and simple magnifying glasses. At the time, no one imagined they would become foundational for modern biology and astronomy—with the telescope for example first thought to be useful for navigating (Chapter 7). Röntgen discovered x-rays by unexpectedly seeing a glowing photographic plate near a discharge tube. This gives us a powerful way to understand cause and effect in science: if a tool is not designed to make specific discoveries, and yet still made them possible—often in completely unrelated fields—then the tool itself is the key cause. This is clear looking at science's most used tools . Using quasi-experimental reasoning, this lack of an initial relationship to the outcome (not originally intended as tools to make different discoveries) is important to identify the causal effect of the new tools in catalysing the discoveries they commonly were not even designed for. The microscope was not invented to discover bacteria, and a sensitive spectrograph was not created to reveal 54 T HE ENGINE OF SCIENTIFIC DISCOVERY planets around distant stars—yet both tools enabled such transformative discoveries. This lack of an initial intended relationship helps isolate the new tool as the causal spark (the independent variable) in driving the new breakthrough (the outcome) that is otherwise not possible—see Figure 1.8. With the tools already developed, it also reduces other explanations—like just funding or research teams as the key triggers. Our tools directly unlock a major advance by providing a completely new perspective. Without them, supporting factors like money and collaborations are not enough. And without something to detect or measure, we cannot generally create and test theories (as we lay out in depth in Chapter 6). It is not only discoveries that are surprising but also some tools are unforeseen. While eyeglass makers unexpectedly developed the first microscope (Zacharias Janssen) and the first telescope (Hans Lippershey) as instruments we could later use in science, they caused unexpected discoveries: from cells and bacteria, to the motion of stars, and galaxies (Figure 1.8). Why are these unintended discoveries? Because they were not just unintentional for the inventors—who had no idea the tools would be used in science—they were often also unintentional for the scientists who made the discoveries using them. Their unexpected observations were not guided by theory. The theories only came after, built to explain what the tools had revealed. Again, before these tools existed, the chance of making such discoveries was zero: there was no way to see exoplanets, or pulsars, or bacteria. And then, suddenly, there was. That is as causal as science can get. New tools do not justhelp us discover—they cause discovery. This causal link holds across time, across fields—from medicine to chemistry—and across expected and unexpected discoveries. And we turn our lens to these unexpected breakthroughs in the next chapter. With this consistent pattern across discoveries, can the powerful role of methods and instruments in science seem self-evident? Yet, what is clear is that there is no consensus among researchers on what powers scientific progress. That is striking, since all scientists share a common goal: discovery. But the current explanations of discovery do not place powerful new methods at the centre of major breakthroughs—as the key spark. Rather, they focus on broad supporting factors like funding, teamwork, chance and serendipity. In practice, science is generally done with existing methods and tools. That is the default—conventional research using current microscopes, statistical models and imaging techniques. But the engine of discovery is observing or measuring something we could not before. And that requires a new tool. That is the key novel insight here: discovering that science's major breakthroughs come from applying a new method or tool (not existing ones) to a problem for the first time. These tools expand the edges of what we can see and understand. They fill in blind spots to studying the world in ways not possible before. This is the missing piece of the puzzle in understanding how we cross a boundary we could not cross before—and it has been overlooked until now. With new methods needed (and not just conventional ones), how do we actually develop them? Scanning science's over 750 major discoveries, we uncover four common ways. First, we create entirely new methods not yet conceived before—like electrophoresis that separates molecules. That technique made discovering DNA sequencing possible. Second, we extend methods in novel ways that also represents a new method not yet used—like the ultracentrifuge that vastly expanded sugar nucleotides (1949) carbon dioxide assimilation in plants (1952) structure of insulin (1955) doppler-free spectroscopy (1958) crystal structure (1913) structure of penicillin (1946) structure of DNA molecule (1953) structure of vitamin B12 (1956) Jupiter's moons (1610) distance to the sun (1672) motion of stars (1719) galaxies (1750) microscope 1590 maser (1954) capillaries (1661) cells (1665) bacteria (1674) 1– 0– x-ray diffraction/ crystallography 1912telescope (1608) paper chromatography (1944)x-rays (1895) characteristic Röntgen radiation (1906) diffraction of x-rays by crystals (1912)
Endline (year discovery made) Baseline (year tool made) Likelihood of making discovery Figure 1.8 The causal power of new methods and tools—unlocking discoveries they are not even designed for Each of the tools—and the discoveries they enabled—earned a Nobel prize, except for those made with the first microscope and telescope. Six examples of tools are provided. 56 T HE ENGINE OF SCIENTIFIC DISCOVERY early centrifuges. That tool led to uncovering blood plasma. Third, we combine methods in novel ways that also reflects a new method not yet leveraged—like x-ray crystallography that merges x-ray methods with crystal analysis. That technique sparked discoveries from the structure of penicillin to DNA's double-helix. Fourth, we adopt new methods developed in other fields—like the electron microscope created in physics by transforming existing light microscopes and used for the first time in fields like biology and medicine. That instrument unlocked breakthroughs from the cell structure to antibiotics against tuberculosis. What all these cases have in common is that we never applied the new method to the problem before . And when applied, discovery followed. New methods are defined as created through these pathways—and science's major discoveries are defined as all nobel-prize and major non-nobel discoveries. This leads us to the key finding—and title—of the book: The engine of scientific discovery: how new methods and tools spark major breakthroughs . So can we better predict the next discoveries? Indeed, this new tools-driven discovery principle not only explains our past discoveries, it expands our current limits of prediction. We cannot know exactly who will achieve what path-breaking discovery. But when we invent, upgrade, combine or adopt a new tool, that is when breakthroughs happen. That is when we can predict where the next breakthroughs can come from, and when. By tracking the pace and direction of method innovations, we can spot powerful signals that discovery is soon to follow. The speed of tool-building today is generally the best indicator we have to predict future discoveries we can spur. Whether a newly invented imaging tool, an upgraded spectrometer, an advanced telescope combined with AI tools, or a new computational method adopted in another field. Look at the tools being developed, and we see where science is heading next.
New methods and tools unlock science's major discoveries by enabling us to see, measure and understand the world in ways that are impossible without them. This key pattern holds across time, fields and all nobel-prize and major non-nobel discoveries. An awareness across science of this new method-to-discovery principle holds the potential for a method revolution in science—a shift in how we think about discovery and focus our time and resources on developing new tools. Imagine that scientists in the future could even look back and divide science into two eras. The era before the method revolution was marked by tool-building that was ad-hoc and improvised. Powerful instruments were developed in scattered ways, often with long delays between invention and discovery (Figure 1.5). We were largely not yet aware that the catalyst of progress is creating the very methods we use to answer our questions. There was no general roadmap, no guiding theory of how discovery happens, no systematic research programmes for tool development. Scientists had to experiment and tinker on their own—researcher by researcher—hoping the right tool would come along. The era of the method revolution can instead be described as tool-building that would be structured, planned and targeted. We would all have become aware of this S powerful principle of new methods propelling science. Future scientists may look back at the advent of this revolution as the point when the scientific community as a whole began to strategically focus on refining, combining, restructuring and inventing tools that accelerate the pace of new breakthroughs. No longer would tool innovation be the side project of a few tool-inclined researchers who deviated from their initial academic training in established fields. Method invention would be at the centre of how science is done. It is a shift from asking what to study to asking how we can study better. It is about pushing science's current boundaries—from gene editing and brain mapping to exoplanet detection and climate modelling—with our hard-won tools. And we lay out how in Chapter 6. As the quantum revolution transformed physics in the 1920s and the cognitive revolution reshaped cognitive science in the 1960s, the method revolution could accelerate science at an unprecedented rate. This principle of method innovation reflects a general rule we can apply across science. The powerful principle has proven very effective in explaining and uncovering new advances—and can better predict them. The pace of tool innovation is our best predictive signal of discoveries. And this principle can also help us lay the foundation for the science of science. We will continue to advance our understanding of the world mainly at the pace of diversifying our toolbox. What is key here is that we need to shift part of our attention and research to refining the very methods and tools we use as a central part of the discovery process. This would open yet unexplored terrain, opportunities and research areas. Scientific progress would then no longer largely remain an uncoordinated and unorganised result of few method-curious scientists. It would be like moving from breakthroughs made by trial and error up to now, to sparking discoveries using systemic controls based on this methods-powered principle. Ultimately, tomorrow's biggest discoveries will be made by those who create and apply the best new methods and tools today. The future of science—our next leaps forward—largely lies in tackling our tools' bottlenecks: the limits of how we measure, observe and test the world.