Book summary “The golem”: summary

Introduction

You likely view science as either an infallible machine delivering absolute truth or a dangerous force destroying nature. This binary view leaves you unprepared when scientific controversies erupt and experts disagree publicly.

H. M. Collins resolves this tension in The Golem: What You Should Know About Science by introducing the golem metaphor. Science is not an infallible machine of absolute truth, but a powerful golem built by human craft that relies on social consensus and human skill to settle controversies.

The book examines famous scientific episodes to reveal how knowledge is actually made. You will explore experiments on memory transfer, tests of relativity, cold fusion, spontaneous generation, gravitational waves, and solar neutrinos.

• How do scientists decide when an experiment succeeds or fails?
• What role does social consensus play in settling technical debates?
• Why do replication attempts often yield conflicting results?

1. Edible Knowledge and Memory Transfer

Experiments on planarian worms and mammals claimed to transfer learned behaviors chemically, but collapsed due to replication failures and the experimenter's regress.

Planarian flatworms

Early experiments by James McConnell on planarian flatworms suggested that training and conditioned responses could be chemically transferred to naive worms through cannibalism or RNA extraction. McConnell trained flatworms to associate a bright light with an electric shock until they learned to scrunch in response to light alone. Because planarian worms incorporate large ingested components into their bodies without fully breaking them down, cannibals eating trained worms exhibited higher response rates.

James McConnell used a plastic trough filled with pond water, brass electrodes, and electric light bulbs to train flatworms to scrunch up their bodies in response to light.

• Imagine a trough gouged out of plastic, 12 inches in length, semicircular in cross-section, and filled with pond water.
• Attach brass electrodes at either end to a power source and place two electric light bulbs above the trough.
• Turn on the lights for 3 seconds when the worm is gliding smoothly in a straight line.
• Add one second of electric shock after the light has been on for two seconds to cause the worm to contract.
• Record the behaviour during the two-second period after the light comes on but before the shock starts.

150 pairings resulted in a 45% scrunch response rate in the 1950s and a 90% response rate in the 1960s. As scientific controversies develop, numerous confounding variables emerge that make exact replication difficult and provide ammunition for both proponents and critics. Disagreements over experimental outcomes often trace back to hidden variables, where proponents attribute failures to lack of experimenter skill and critics dismiss positive results as artefacts of poor controls.

In worm experiments, factors such as slime trails left by previous worms, water temperature, lighting, and even barometric pressure were cited as variables affecting the outcome. Researchers eventually cited 70 variables at one time or another to account for discrepancies in planarian experimental results.

Mammalian memory transfer

Memory transfer research expanded from primitive worms to mammals when independent researchers claimed that behaviors such as drug tolerance, habituation, and fear of the dark could be transferred via brain extracts. Around 1964 and 1965, independent groups including Fjerdingstad, Jacobson, Reinis, and Georges Ungar reported successful transfer of trained behaviors in rats and mice through brain injections, shifting the scientific focus to mammals.

Georges Ungar extracted brain tissue from rats trained to avoid a dark box due to electric shocks and injected it into mice to test whether fear of the dark could be transferred.

Ungar's analysis of transfer experiments in mammals between 1965 and 1975 recorded 105 positive and 23 negative replications. He ground up 50 tolerant rats to extract and transfer morphine tolerance in early 1965. Ungar also required 4000 trained rats to obtain a measurable amount of scotophobin, and 17000 trained goldfish were sacrificed to produce 750 grams of brains for chromodiopsin research.

Replication attempts by major institutions often failed because of subtle procedural differences and disagreements over valid behavioral measurements, leading to intense academic disputes. When a Stanford group led by Avram Goldstein attempted to replicate Ungar's dark-avoidance transfer work, they obtained negative results, leading to a public dispute over testing apparatus, mouse strains, and whether to measure total time in the dark box or latency.

Avram Goldstein spent three days at Ungar's laboratory to learn procedures, yet subsequent blind tests and collaborative efforts yielded completely negative results. The Stanford team used 125 donor rats and 383 recipient mice in their initial eighteen unsuccessful replication experiments.

The field of chemical memory transfer ultimately faded away not because of a definitive technical disproof, but due to shifting scientific priorities, high resource demands, and the loss of credibility of key researchers. Ungar's biochemically ambitious strategy raised the financial and technical stakes too high for other laboratories to compete, and when he died in 1977 at age 71 with no successor, the field dissolved as brain-peptide chemistry moved toward unrelated discoveries.

Three influential publications—including Melvin Calvin's negative worm study, Byrne's multi-laboratory failure to replicate, and Nature's publication of Scotophobin alongside a scathing fifteen-page referee report—undermined the field's credibility.

2. Relativity and Consensus

Famous tests of relativity, such as the Michelson-Morley experiment and Eddington's eclipse observations, achieved acceptance through retrospective reconstruction and an agreement to agree.

The aether sea

Physicists in the late nineteenth century believed light traveled through a universal medium called the aether, creating an aether wind as the earth moved. Albert Michelson and Edward Morley designed an interferometer to detect this wind by comparing the speed of light in different directions, but found no difference whatsoever.

In 1887, Michelson and Morley built an apparatus consisting of a sandstone block floating on a trough of mercury in a basement room in Cleveland, achieving a path length of over 10 metres, yet found no effect resembling the expected aether wind. They tested an estimated speed of the earth in its orbit around the sun of 18.5 miles per second against the known speed of light of 185 000 miles per second. Michelson was disappointed by his null results because he was simply trying to build an earth speedometer and knew nothing of relativity.

Only after Einstein published his papers did the scientific community reinterpret Michelson and Morley's work as a famous proof of the constancy of the speed of light, separating these early observations by a period of approximately 25 years. Subsequent attempts to refine the interferometer experiment led to conflicting positive results that were dismissed as mere anomalies by the mainstream scientific community. In the 1920s, Dayton Miller conducted experiments on Mount Wilson at a height of 6000 feet, without heavy shielding, and claimed to find a persistent positive aether drift.

In 1925, Miller was awarded the American Association for the Advancement of Science prize for his work on aether drift, concluding an observed motion of the earth of 10 kilometres per second. Because the physics community had already embraced relativity, Miller's positive findings were treated as a nuisance anomaly rather than a serious problem. His findings were later countered by heavily shielded, low-altitude null experiments conducted by Michelson and others.

Eclipse expeditions

Sir Arthur Eddington's 1919 solar eclipse observations, which allegedly proved Einstein's theory of general relativity regarding starlight deflection, actually relied on selective data filtering. Measuring the minute bending of starlight passing near the sun required comparing daytime eclipse photographs with nighttime reference plates using fragile equipment in remote locations. The raw data contained massive inconsistencies and conflicting results between different instruments.

During the 1919 eclipse, Eddington's expedition to Principe and Crommelin and Davidson's expedition to Sobral produced messy data, including poor plates from the Sobral astrographic telescope that pointed toward Newtonian predictions rather than Einstein's. Eddington tested Einstein's predicted starlight deflection of 1.7 seconds of arc against Newton's predicted deflection of 0.8 second of arc using exposure times of 5–30 seconds for the eclipse photographs. Acceptance of scientific breakthroughs often proceeds through an agreement to agree where researchers discard inconvenient data points that do not match the desired theoretical outcome.

Eddington justified ignoring the problematic Sobral astrographic plates by citing systematic error, even though he lacked concrete proof. By selecting which measurements to keep and which to treat as noise, scientists aligned their observations with Einstein's predictions in a circular loop of mutual confirmation. In announcing the results on November 6, 1919, the Royal Society leadership endorsed Einstein's theory by relying exclusively on the 4-inch telescope plates from Sobral and Eddington's two usable Principe plates while discarding a total of 18 astrographic plates from Sobral.

The establishment of new scientific facts can happen with sudden speed once a foundational confirmation gives researchers the collective will to reinterpret ambiguous past evidence. Before 1919, solar spectroscopists saw only confusion regarding the gravitational red-shift of light. However, once the eclipse observations were claimed to confirm relativity, researchers suddenly found the courage to throw out contradictory evidence and report finding the Einstein effect everywhere.

3. Cold Fusion and Spontaneous Generation

Cold fusion and spontaneous generation demonstrate how theoretical impossibilities, measurement artifacts, and biased scientific commissions shape the fate of controversial claims.

On March 23, 1989, Martin Fleischmann and Stanley Pons announced at a press conference in Utah that they had discovered fusion in a test tube, sparking a global scientific gold rush and an influx of media attention. You can recreate their setup with a simple apparatus:

• use a beaker of heavy water where hydrogen atoms are replaced by deuterium;
• set up a palladium electrode as the cathode and a platinum electrode as the anode;
• add a small amount of lithiumdeuteroxide as a conductor;
• put a low voltage across the cell for a period of up to several hundred hours.

Passing a low voltage across this electrolytic cell caused deuterium atoms to fuse into helium, releasing excess heat and nuclear byproducts like neutrons and tritium. Before the announcement, palladium cost $145.60 an ounce on March 23, 1989, but reached a peak price of $170.00 an ounce in May 1989 at the height of the cold fusion frenzy. Following the collapse of the controversy, the price dropped to $95.00 an ounce by October 1992. The Utah State Legislature voted $5 million towards the cold fusion effort, and Pons and Fleischmann approached the US Congress for an additional funding amount of $25 million, having initially used $100,000 of their own money to fund experiments.

Theoretical calculations by physicists such as Steve Koonin and Mike Nauenberg showed that the increased pressure inside palladium was insufficient to overcome the Coulomb barrier, making the claimed energy output theoretically impossible. Physicist Steve Koonin illustrated the extreme improbability by stating that a mass of cold deuterium the size of the Sun would undergo only one fusion per year, and comparing cold fusion theories to pigs with wings.

Critics such as Nathan Lewis from Cal Tech and Richard Petrasso from MIT demonstrated that reported neutron peaks and excess heat measurements were either instrumental artifacts or the result of uncalibrated devices and lack of proper controls. The MIT group famously used a video of a television news program showing the inside of Pons and Fleischmann's laboratory to analyze their gamma-ray spectrum and conclude the neutron peak was an artifact.

Pasteur and spontaneous generation

Spontaneous generation is the nineteenth-century doctrine that life can form from dead matter under the right circumstances, a view ultimately defeated by Louis Pasteur through targeted experiments and institutional backing. Henry Bastian remained an unyielding believer in spontaneous generation until his death in 1910, demonstrating that opposing scientific views survived long after official commissions declared them defeated.

Felix Pouchet's experiments under mercury suggested that life grew when purified air met sterilized hay infusions, but Pasteur dismissed these positive results as being confounded by contaminated mercury. You can follow Pouchet's experimental design with these steps:

• perform work with all vessels immersed in a mercury trough so that ordinary air cannot enter;
• introduce specially prepared air into the experimental flask by bubbling it through the mercury trough.

Pasteur admitted to getting positive results for spontaneous generation in 10 percent of his own early mercury-trough experiments, but he chose to suppress them because he believed a hidden flaw existed rather than accepting the hypothesis. Pouchet wrote to Pasteur reporting positive growth from sterilized hay infusions prepared under mercury and exposed to pure air, prompting Pasteur to argue that common air or contaminated mercury dust was unwittingly introduced.

Scientific disputes in nineteenth-century France were settled by official commissions of the Académie des Sciences whose pre-existing biases heavily favored the established viewpoint. When the second commission was established in 1864 after Pouchet's Pyrenean experiments, Pouchet and his collaborators withdrew after failing to alter the narrow terms of reference set by the biased members.

While boiling yeast infusions destroyed all life, it failed to sterilize hay infusions because hay supports resilient spores that survive 100°C temperatures, meaning Pouchet's positive results were driven by unrecognized biological factors. Modern commentators noted that if Pouchet had stayed the course in the commission competition, the facts of hay infusions would have spoken for spontaneous generation, albeit for the wrong reasons.

4. Gravitational Waves and Whiptail Lizards

Weber's gravitational wave claims and Crews's lizard research reveal how scientists rely on non-technical judgments and struggle against artifact disputes.

Weber-bar antennas

Professor Joseph Weber of the University of Maryland pioneered a standard technique for detecting gravitational radiation. He looked for length changes in a massive aluminium alloy bar designed to ring like a bell at the frequency of passing waves.

Weber used a heavy bar suspended in a metal vacuum chamber on a thin wire with a stack of lead and rubber sheets for seismic insulation. Piezoelectric crystals glued to the bar produced tiny electrical potentials when distorted, which were then amplified and recorded.

In 1969, Weber claimed to have found evidence of large amounts of gravitational radiation coming from space. He detected the equivalent of about seven peaks a day above background noise that could not be accounted for by noise.

When testing a controversial phenomenon with unknown outcomes, you face the experimenter's regress. You cannot judge if an experiment is competent without knowing the correct outcome, and you cannot know the correct outcome without a competent experiment.

If an experiment yields a negative result, you face an agonizing choice. Either the phenomenon does not exist, or the experiment was flawed through poor threshold settings, insensitive amplifiers, or improper bar support.

Scientists attempting to replicate Weber's gravitational wave experiments had to decide whether their failure to detect waves meant Weber was wrong or that their own delicate apparatuses were flawed.

During scientific controversies when technical arguments cannot definitively settle a dispute, scientists rely heavily on non-technical grounds. Experimenters cast around for reasons beyond traditional science to believe or disbelieve results, including personal reputation, nationality, university prestige, style of presentation, and perceived dedication.

In evaluating Weber's work and subsequent negative reports, scientists cited factors such as Weber's background as an electrical engineer, personal dedication, and psychological approach to experimentation.

The controversy surrounding Weber's gravitational wave claims was ultimately brought to a close not by pure logic or theory, but by Richard Garwin. Garwin published an outspoken report with uncompromising claims of conflict with Weber's findings, highlighting a specific programming error that created false coincidences.

Garwin published a letter in a popular physics journal stating that the Weber group had published no credible evidence for their claim. This started an avalanche after which nobody saw gravitational radiation fluxes, creating a critical mass of negative reports and halting the debate.

Parthenogenetic lizards

David Crews studied the parthenogenetic whiptail lizard to understand reproductive evolution. He observed a bizarre pseudo-copulatory behavior where all-female lizards mounted each other.

Cnemidophorus is unusual because it breeds parthenogenetically from unfertilized female eggs. Crews observed that non-sexual lizards engaged in courtship and mounting rituals resembling ordinary sexual species. Courting animals were postovulatory, while courted animals had large preovulatory follicles.

Time magazine introduced Crews's observations under the heading "Leapin' lesbian lizards", bringing widespread media attention to the reptile's sexual habits.

Critics of novel scientific discoveries often dismiss striking behaviors as unnatural artefacts of captivity produced by inexperienced observers. This forces researchers into public debates over their personal competence and observational skills.

Established researchers like Orlando Cuellar and C. J. Cole argued that the whiptail lizard pseudo-copulation was an unnatural product of overcrowded cage conditions and poor observation. In response, Crews detailed his exact cage dimensions and husbandry to turn the tables on his critics.

Cole and Townsend emphasized their years of keeping lizards in their offices and caring for them personally to assert their superior observational skills over Crews.

Crews examined 1000 dead female lizards to find that only 3% had marks on their backs and sides. Cuellar had spent fifteen years sporadically observing the behavior without assigning it significance. After five years of research and debate, the reproduction relevance remained unresolved.

5. Solar Neutrinos and Public Understanding

Ray Davis's missing solar neutrinos unmade established theory, highlighting how society flip-flops between viewing science as absolute certainty or charlatanism.

The cleaning fluid detector

In the 1967 year, Ray Davis attempted the first direct test of stellar evolution theory by building a vast underground neutrino detector. He used a tank of chlorine-rich cleaning fluid buried deep underground in a disused mine shaft to capture neutrinos produced by nuclear fusion in the core of the sun, which triggered reactions turning chlorine into radioactive argon.

Ray Davis placed an Olympic-swimming-pool-sized tank of cleaning fluid deep underground in the Homestake mine shaft and used a First-World-War gun barrel made of low-radioactivity steel to shield his counter. The experiment cost 600000 dollars and relied on 550 gallons of cleaning fluid used in Davis's early 1955 detector.

The solar neutrino experiment produced a baffling outcome where the predicted neutrino fluxes were missing, plunging established theory into uncertainty. When Davis detected far fewer neutrinos than predicted by theoretical models, it created a severe confrontation between theory and experiment, temporarily unmaking what was previously taken to be solid scientific knowledge.

Davis obtained a signal no greater than 6 SNU upper limit of the initial signal obtained in August 1967, and by early 1968 lowered the limit to 3 SNU lower limit set by Davis in early 1968, which was vastly lower than the initially predicted 40 SNU predicted flux high when the experiment was funded in 1964 or 19; SNU predicted flux by the time Davis got his first results in 1967. By May 1968, John Bahcall reported a reduced prediction of 7.5 SNU, and by 1978 researchers had published 400 papers proposing solutions to the solar-neutrino problem.

To maintain credibility when results defy expectations, an experimenter must cultivate a profile of openness, caution, and modesty. Because Davis could not easily be replicated due to the immense cost and scale of his detector, he survived intense scrutiny by diligently following up on every external suggestion, calibration test, and criticism. Davis introduced 500 atoms of argon directly into the tank as a test and successfully recovered them to prove he was not losing argon through trapping mechanisms suggested by critics like Kenneth Jacobs.

Flip-flop thinking

Society tends toward unstable flip-flop thinking about science, treating it either as infallible certainty or total charlatanism. When science is expected to deliver absolute certainty, any failure leads to harsh public disillusionment, scapegoating of human error, or an overzealous anti-science backlash.

The public reacts with intense disruption when technologies fail, such as the space shuttle or Chernobyl disasters, attributing everything entirely to avoidable human error rather than inherent risk. Scientific evidence in legal and public policy settings is inherently contestable and should not be presented as immaculate certainty. Miscarriages of justice occur when forensic tests, such as those for nitroglycerine traces, are treated as infallible proofs without proper adversarial scrutiny of their false positive rates and limitations.

The British legal system was rocked when individuals were wrongfully imprisoned for years because forensic tests for explosives failed to account for alternative chemical sources like playing cards. Standard school science experiments mirror the messy realities of frontier research until social negotiation imposes a tidy methodological myth. When students get widely varying measurements in a classroom experiment, the teacher steps in 10 minutes before the end of the experiment when the teacher gathers results and starts social engineering to explain away deviations as experimental errors, mirroring how the scientific community cleans up chaotic results.

A classroom measuring the boiling point of water gets diverse readings like 102 degrees Celsius, 105 degrees Celsius, 99.5 degrees Celsius, 100.2 degrees Celsius, and 54 degrees Celsius, which the teacher rationalizes to uphold the official 100 degrees Celsius standard boiling point of water expected in classroom experiments.

Scientists should be understood not as infallible gods or charlatans, but as expert craftspersons working within the limits of a natural world. You must learn to evaluate scientific expertise using the same everyday skills you use to handle other types of professional advice, recognizing that science works through social agreement rather than absolute external constraints. Citizens cope with divided political expertise every day, and they must learn to apply that same critical judgment to scientific claims rather than expecting absolute certainty.

10 Key Ideas

1. H. M. Collins introduces the golem metaphor to show that science relies on human skill and social consensus rather than absolute truth.
2. Experiments on planarian worms and mammals attempting to prove chemical memory transfer collapsed due to replication failures and the experimenter's regress.
3. Famous tests of relativity, such as the Michelson-Morley experiment, achieved acceptance through retrospective reconstruction long after the initial null results.
4. Sir Arthur Eddington's 1919 solar eclipse observations proved Einstein's theory only by relying on selective data filtering and discarding contradictory plates.
5. Cold fusion claims failed because physicists demonstrated that reported neutron peaks and excess heat measurements were instrumental artifacts.
6. Louis Pasteur defeated the doctrine of spontaneous generation through targeted experiments and institutional backing while suppressing his own anomalous positive results.
7. Professor Joseph Weber's gravitational wave claims were ultimately halted by Richard Garwin publishing uncompromising claims of conflict and highlighting a specific programming error.
8. David Crews's research on parthenogenetic whiptail lizard pseudo-copulation faced intense criticism from researchers who dismissed the behavior as an unnatural artifact of captivity.
9. Ray Davis's cleaning fluid detector produced a baffling outcome where predicted solar neutrinos were missing, plunging established stellar evolution theory into uncertainty.
10. Society flip-flops between viewing science as infallible certainty and total charlatanism, failing to recognize that scientific knowledge is built through social agreement and craft.