Vaccine hesitancy: Wakefield, the MMR-autism fraud, and the public-health communication challenge

Intuition Beginner

In 1998 a British doctor named Andrew Wakefield published a paper in The Lancet claiming the MMR (measles-mumps-rubella) vaccine caused autism and bowel disease. The paper described only 12 children, had no control group, and Wakefield had been paid by lawyers suing vaccine manufacturers — a conflict he did not disclose. The paper was retracted in 2010 and Wakefield was struck off the UK medical register. But the claim spread. Vaccination rates fell, measles returned to countries that had nearly eliminated it, and children died of a preventable disease. The fraud is among the most damaging in medical history.

The modern anti-vaccine movement grew on this foundation. Jenny McCarthy, Robert F. Kennedy Jr., and a small group nicknamed the "Disinformation Dozen" amplified vaccine-autism claims through television, books, and social media. Echo chambers reinforced parents' fears. The psychology is not mere ignorance: cognitive biases make vaccination feel riskier than it is. Omission bias makes "doing nothing" feel safer than acting. Confirmation bias makes one frightening story outweigh a thousand reassuring statistics. By 2019 the World Health Organization named vaccine hesitancy one of the top ten global health threats.

During COVID-19 (2020-2023) the same patterns intensified. An "infodemic" of misinformation spread faster than the virus itself. The challenge for public health is not only to make safe vaccines but to communicate their safety to people who do not trust the institutions recommending them. Trusted messengers, narrative medicine, and motivational interviewing work better than reciting statistics at people. Vaccines exist; this unit is about why some people decline them — and what public health can do about it.

Visual Beginner

The herd-immunity threshold is the coverage above which a disease cannot sustain transmission, and it is set entirely by the pathogen's reproduction number. For measles the threshold sits near $92$ to $94$ percent, the highest of the common vaccine-preventable diseases, which is why measles is the first to return whenever coverage dips. A small absolute drop in coverage — a few hesitant families per hundred — is enough to cross below the threshold and trigger an outbreak.

The picture explains why a handful of refusals can undo a community's protection from measles while barely perturbing protection from a low-reproduction-number infection. The steepness near the threshold is the arithmetic reason the Wakefield fraud — a few percentage points of coverage loss — produced real outbreaks.

Worked example Beginner

Consider the 2014-2015 Disneyland measles outbreak, reconstructed in the CDC's Morbidity and Mortality Weekly Report (Zipprich 2015).

The index case. In December 2014 a person infected with measles visited Disneyland in Anaheim, California while contagious. Measles is the most contagious of the common vaccine-preventable pathogens: in a fully susceptible population one infected person typically infects $12$ to $18$ others, because the virus spreads by airborne transmission and lingers in a room for hours after the infectious person leaves.

The threshold. For a perfectly effective vaccine, the herd-immunity threshold is approximately $1 - 1/R_0$ where $R_0$ is the reproduction number (the average number of new infections from one case). For measles this gives $1 - 1/12 \approx 0.917$ to $1 - 1/18 \approx 0.944$, that is, about $92$ to $94$ percent coverage needed.

The outbreak. From the single index case at Disneyland, measles spread to $147$ people across $7$ US states plus Mexico and Canada. The large majority were unvaccinated, and many of those by parental choice rather than medical exemption. California had been allowing a "personal belief exemption" that let parents opt out of school-entry vaccination for non-medical reasons.

The policy response. In 2015 California passed SB 277, eliminating the personal-belief exemption and leaving only medical exemptions. Kindergarten MMR coverage rose from about $92.8$ percent in 2015 to over $95$ percent by 2018, and measles cases in California fell.

What this tells us: a preventable disease spreads fast in an under-vaccinated community, and measles — with the highest reproduction number — is the first to return whenever coverage dips below about $95$ percent. The outbreak and the policy fix together show how a few hesitant families per hundred can cost a community its herd immunity.

Check your understanding Beginner

Formal definition Intermediate+

Definition (vaccine hesitancy). Vaccine hesitancy is the delay in acceptance of vaccination, or the refusal of vaccination, despite the availability of vaccination services. The WHO SAGE Working Group on Vaccine Hesitancy (MacDonald 2015 ^{[MacDonald2015]}) defined it as a behaviour, not a belief, and as a spectrum ranging from outright refusal through delay to acceptance with doubts. Hesitancy is vaccine-, context-, and time-specific: an individual may accept one vaccine, delay another, and refuse a third.

Definition (the 3Cs model, SAGE 2015). The SAGE framework decomposes the drivers of hesitancy into three categories. Complacency — perceived risk of vaccine-preventable disease is low and vaccination is not judged a necessary preventive action. Confidence — trust in the safety and efficacy of vaccines, in the health system that delivers them, and in the policymakers who recommend them. Convenience — the physical availability, affordability, geographical accessibility, and comfort of the vaccination encounter. Hesitancy is driven by whichever of the three is limiting in a given context.

Definition (the 5As model, Betsch 2015). An alternative taxonomy organising access-side and acceptance-side barriers: Access, Affordability, Awareness, Acceptance, and Activation. The 5As extend the 3Cs by separating the information and motivation layers and by foregrounding the structural (non-psychological) barriers that dominate in low- and middle-income settings.

Definition (herd immunity / community protection). A population has herd immunity against an infection when the fraction effectively immune exceeds the threshold at which a single infective generates fewer than one secondary case on average. With basic reproduction number $R_0$ and vaccine efficacy $e$ (the fraction of vaccinated individuals rendered fully immune), the critical two-dose-equivalent coverage is $c^* = (1 - 1/R_0)/e$. Below $c^*$, sustained transmission is possible; at and above $c^*$, outbreaks die out.

Definition (cognitive biases contributing to hesitancy). Four biases dominate the psychology of vaccine refusal. Omission bias — the action of vaccinating is weighted as riskier than the inaction of not vaccinating, even when the inaction carries the larger expected harm. Base-rate neglect — a vivid adverse-event narrative dominates the population-level frequency, so one feared complication outweighs millions of uneventful vaccinations. Availability cascade — media repetition of a rare adverse event raises its perceived frequency. Confirmation bias — vaccine-hesitant individuals seek and recall information that confirms prior doubts. Kahan's cultural cognition ^[Kahan2010] generalises these: risk perception is conditioned by group values, so information alone polarises rather than persuades.

Counterexamples to common slips Intermediate+

• "Hesitant people are just uninformed; more facts will fix it." This is the deficit model and it fails empirically. Kahan and colleagues' cultural-cognition studies show that numerate, science-literate individuals are more polarised on vaccine risk than less-literate ones, because they interpret evidence through group values. Information campaigns that ignore values backfire.

• "The Wakefield paper was a legitimate study that turned out to be wrong." No. The paper was fabricated — medical records altered, cases selectively reported — and the author was struck off for serious professional misconduct. It was not overturned by later data; it was retracted for fraud. The two are categorically different.

• "Measles is a mild childhood illness." No. Measles has a case-fatality ratio of about 1 to 2 per 1,000 in industrialised settings, causes pneumonia in roughly 1 in 20 cases, and produces immune amnesia (deletion of acquired immunological memory) for two to three years after infection. In the 2019 Samoa measles epidemic, 83 people died in a population of about 200,000.

• "Herd immunity protects the unvaccinated for free, so individual refusal is harmless." Only up to the threshold. Each refusal raises the susceptible fraction and lowers $c$; once $c < c^*$, the unvaccinated cluster sustains transmission and the refusers are the first infected.

• "Vaccines cause the diseases they prevent." For live-attenuated vaccines (oral polio, MMR) rare vaccine-associated cases occur at rates orders of magnitude below the wild-type disease burden; for inactivated/subunit vaccines there is no productive infection at all. The comparison is against the wild-type harm averted, not against zero risk.

Key result: the Wakefield fraud and the erosion of vaccine confidence Intermediate+

The epidemiological content of the Wakefield fraud is that a small absolute drop in vaccine coverage can flip a high-reproduction-number disease from controlled to outbreak, because herd-immunity thresholds for pathogens like measles sit in the narrow band where a few hesitant families per hundred are enough. The formal content is the threshold arithmetic itself.

Theorem (herd-immunity threshold, $R_0$–efficacy form). Let $R_0 > 1$ be the basic reproduction number of a vaccine-preventable infection spreading homogeneously in a large well-mixed population, and let $e \in (0,1]$ be the vaccine efficacy, defined so that a fraction $e$ of vaccinated individuals are fully immune and the remainder are fully susceptible. If a fraction $c$ of the population is vaccinated, the effective reproduction number is

$$R_{\text{eff}} = R_0\,(1 - c\,e),$$

and the critical coverage achieving herd immunity ($R_{\text{eff}} = 1$) is

$$c^* = \frac{1 - 1/R_0}{e}.$$

Proof. In a well-mixed population the probability that a given contact of an infective is susceptible equals the susceptible fraction. Under the vaccination scheme, the effectively immune fraction is $c\,e$ and the remaining susceptible fraction is $1 - c\,e$. By definition, one infective in a fully susceptible population generates $R_0$ secondary infections; restricting to the susceptible fraction $1 - c\,e$ scales the expected count linearly, giving $R_{\text{eff}} = R_0(1 - c\,e)$. A self-sustaining epidemic requires $R_{\text{eff}} > 1$, so the boundary of sustained transmission is $R_0(1 - c\,e) = 1$. Solving: $1 - c\,e = 1/R_0$, so $c\,e = 1 - 1/R_0$, and $c^* = (1 - 1/R_0)/e$. $\square$

Corollary 1 (measles sensitivity). Measles has $R_0 \in [12, 18]$, the highest of the common vaccine-preventable pathogens, and two-dose MMR efficacy is $e \approx 0.97$. This gives $c^* \in [0.943, 0.973]$ — roughly $94$ to $97$ percent two-dose coverage. The lower edge of this band exceeds the coverage that many jurisdictions achieve without active effort, which is why measles is the canonical sentinel of vaccine-hesitancy-driven coverage loss.

Corollary 2 (the fraud's epidemiological cost). UK MMR coverage fell from approximately $92$ percent in 1995-1996 to approximately $80$ percent in 2003-2004 following the Wakefield publication and media amplification, crossing below $c^*$ nationally and well below it in some London boroughs. Confirmed measles cases in England and Wales rose from roughly $100$ in 1996 to nearly $2{,}000$ in 2012, and the first measles death in the UK for 14 years occurred in 2006. The arithmetic is direct: the marginal sensitivity of the effective reproduction number to coverage is $\partial R_{\text{eff}}/\partial c = -R_0\,e$, which for measles is $|{-12 \cdot 0.97}| \approx 11.6$; a $1$-percentage-point drop in coverage raises $R_{\text{eff}}$ by about $0.116$, enough to push a population at the threshold from $R_{\text{eff}} = 1$ to $R_{\text{eff}} \approx 1.12$ and into the outbreak regime.

Bridge. This result builds toward the immunological-memory analysis of 18.10.04, where the $R_0$ threshold is the population-level shadow of the within-host clonal-selection dynamics that vaccines induce, and appears again in 29.01.02, where the replication-crisis and Open-Science critique of low-power, researcher-degrees-of-freedom science is exactly the methodological framework that identifies the 12-child Wakefield case series as a structural null. The foundational reason herd-immunity failure is so sensitive to hesitancy is that $R_{\text{eff}} = R_0(1 - c\,e)$ is linear in $c$ but crosses the critical value $R_{\text{eff}} = 1$ at a coverage set by the pathogen's $R_0$ alone, and this is exactly the structure that identifies a high-$R_0$ pathogen like measles as the case where a few percent of hesitancy is enough to lose control. The bridge is between the population threshold (a coverage fraction) and the individual decision (to vaccinate or not), generalised by the recognition that public-health communication is itself an intervention on $c$.

Exercises Intermediate+

Advanced results Master

Result 1 (Wakefield 1998 — the fraudulent case series). Wakefield and 12 co-authors' 1998 paper in The Lancet ^{[Wakefield1998]} reported 12 children described as presenting with a regressive developmental disorder and chronic enterocolitis, and speculated on an association with MMR vaccination. The paper was a case series with no control group — not an epidemiological study — and its speculative conclusion was amplified at a press conference where Wakefield advocated single-antigen vaccines over the combined MMR. The paper was retracted by The Lancet on 2 February 2010.

Result 2 (Deer 2004-2011 — the investigation). Brian Deer's investigation for The Sunday Times, published from 2004 and extended in a 2011 three-part series in BMJ ^[Deer2011], established that the 12 published case histories had been selectively reported, that the children's medical records did not match the published descriptions, that onset dates had been altered to fit the vaccine-onset temporal claim, and that Wakefield had received over 400,000 pounds from lawyers building litigation against vaccine manufacturers — a conflict he had not disclosed to The Lancet, his co-authors, or the ethics committee. Deer's work is the canonical instance of investigative journalism exposing research fraud.

Result 3 (the 2010 General Medical Council determination). Following a 217-day Fitness to Practise Panel hearing, the UK General Medical Council ^[GMC2010] found Wakefield guilty of serious professional misconduct over the undisclosed conflicts, the ethically unauthorised invasive procedures on children (including lumbar punctures performed without clinical indication), and the falsified case histories, and struck him off the medical register in May 2010. The determination precipitated The Lancet's retraction.

Result 4 (Taylor 2014 — the meta-analytic refutation). Taylor, Swerdfeger, and Eslick's 2014 meta-analysis in Vaccine ^[Taylor2014] pooled five cohort studies (1,256,407 children) and five case-control studies (9,920 children) and found no association between MMR vaccination and autism, no association between vaccine components (thimerosal, mercury) and autism, and no association between autism and any specific developmental window after vaccination. The meta-analysis is the quantitative refutation of the hypothesis the Wakefield paper had launched.

Result 5 (Omer-Salmon 2000s+ — hesitancy as a phenotype). Omer, Salmon, and colleagues' programme, synthesised in a 2009 Lancet Infectious Diseases review ^[Omer2009], established vaccine hesitancy and refusal as a measurable population phenotype with geographic clustering, non-medical-exemption tracking, and measurable downstream consequences for outbreak risk. Permitting philosophical or personal-belief exemptions for school-entry vaccination, Omer showed, is associated with elevated individual risk of measles and pertussis; clusters of exemptors act as ignition points for outbreaks that then reach under-immunised medical-refusers and the small fraction for whom vaccination failed.

Result 6 (Kahan 2010 — cultural cognition of risk). Kahan and colleagues' cultural-cognition programme ^[Kahan2010] reframed the deeper puzzle: vaccine-hesitant individuals are not less informed but differently values-organised. In controlled studies the most numerate, science-literate individuals were the most polarised on vaccine risk, because they used their numeracy to reach the conclusion favoured by their cultural group. The empirical consequence is that the deficit model of communication (more facts will fix it) fails, and value-congruent trusted messengers outperform neutral experts.

Result 7 (SAGE 3Cs 2015 and Betsch 5As — the operational frameworks). The WHO SAGE Working Group on Vaccine Hesitancy (MacDonald 2015 ^{[MacDonald2015]}) produced the 3Cs framework (Complacency, Confidence, Convenience) as the diagnostic taxonomy of hesitancy drivers; Betsch and colleagues' 5As model (Access, Affordability, Awareness, Acceptance, Activation ^[Betsch2015]) extended it to separate structural from psychological barriers. Both frameworks make hesitancy actionable: diagnose the limiting driver, then target the intervention at that driver rather than at a generic information-deficit assumption.

Result 8 (the COVID-19 infodemic, 2020-2023). The COVID-19 pandemic produced the largest infodemic on record, declared by the WHO in February 2020. The same cognitive machinery that operates on MMR — omission bias, base-rate neglect, the availability cascade, cultural cognition — operated on a global scale, compounded by novel-mRNA-platform novelty, rapidly evolving evidence, and social-media amplification. The response — trusted-messenger deployment through community health workers, narrative medicine, and values-congruent communication — empirically outperformed broadcast-facts campaigns, vindicating the framework Kahan had built for the MMR case.

Synthesis. The foundational reason vaccine hesitancy is the canonical public-health-communication problem of the early twenty-first century is that the herd-immunity threshold turns a private decision — to vaccinate one's child — into a public good with a quantified free-rider boundary at $c^* = (1 - 1/R_0)/e$. The central insight is that this threshold is the structural fact that identifies individual refusal with population-level outbreak risk, and the bridge is between the cognitive psychology of risk perception (Kahan's cultural cognition, omission bias, the availability cascade) and the epidemiological arithmetic of $R_{\text{eff}} = R_0(1 - c\,e)$.

This is exactly the dual structure — a fast-spreading pathogen coupled to a fast-spreading belief system about the pathogen — that the COVID-19 infodemic made globally visible. The pattern appears again in 18.10.04 as the immunological substrate (clonal selection and memory) that vaccines exploit and that immune amnesia from wild-type measles erases, and it generalises to every high-$R_0$ vaccine-preventable disease for which $c^*$ sits above the coverage a hesitant population will sustain. Putting these together, the public-health response — trusted messengers, narrative medicine, the 3Cs and 5As frameworks — is itself an intervention on the coverage $c$, and builds toward the broader science-of-science-communication programme in 29.01.02, where the replication crisis and Open-Science reform target the institutional trust that hesitancy erodes.

Full proof set Master

Proposition 1 (herd-immunity threshold, restated). Under homogeneous mixing with reproduction number $R_0 > 1$ and vaccine efficacy $e \in (0, 1]$, the critical coverage is $c^* = (1 - 1/R_0)/e$.

Proof. One infective in a fully susceptible population generates $R_0$ secondary infections. With fraction $c$ vaccinated at efficacy $e$, the effectively immune fraction is $c\,e$ and the susceptible fraction is $1 - c\,e$. Under proportional (frequency-dependent) mixing, the expected secondary count is $R_0(1 - c\,e) = R_{\text{eff}}$. The boundary of sustained transmission satisfies $R_{\text{eff}} = 1$, giving $R_0(1 - c\,e) = 1$. Rearranging: $1 - c\,e = 1/R_0$, so $c\,e = 1 - 1/R_0$, and $c^* = (1 - 1/R_0)/e$. $\square$

Proposition 2 (marginal sensitivity to coverage). The marginal sensitivity of the effective reproduction number to coverage is $\partial R_{\text{eff}}/\partial c = -R_0\,e$. High-$R_0$ pathogens are therefore the most sensitive to coverage loss in absolute terms: for measles the marginal sensitivity is about $-11.6$ per unit coverage, versus about $-1.5$ for seasonal influenza.

Proof. Differentiate $R_{\text{eff}}(c) = R_0(1 - c\,e) = R_0 - R_0\,e\,c$ with respect to $c$. Since $R_0$ and $e$ are parameters, $\partial R_{\text{eff}}/\partial c = -R_0\,e$. The negative sign confirms that raising coverage lowers the effective reproduction number. The magnitude $R_0\,e$ grows with both the pathogen's transmissibility and the vaccine's efficacy, so the same absolute coverage loss produces a much larger swing in $R_{\text{eff}}$ for measles ($R_0\,e \approx 14.6$ for $R_0 = 15, e = 0.97$) than for seasonal influenza ($R_0\,e \approx 1.5$ for $R_0 = 1.5, e = 1$). This is why measles outbreaks are the leading sentinel of vaccine-hesitancy-driven coverage loss: no other common vaccine-preventable disease responds as sharply. $\square$

Proposition 3 (the efficacy ceiling for herd immunity). If vaccine efficacy satisfies $e < 1 - 1/R_0$, then $c^* > 1$ and herd immunity cannot be achieved by vaccination alone, regardless of coverage.

Proof. From Proposition 1, $c^* = (1 - 1/R_0)/e$. If $e < 1 - 1/R_0$, then the numerator $1 - 1/R_0$ exceeds the denominator $e$, so $c^* > 1$. Since coverage is a population fraction bounded by $1$, no achievable $c$ satisfies $c \ge c^*$, and $R_{\text{eff}} > 1$ for every attainable coverage. The disease therefore cannot be eliminated by this vaccine alone, even at universal coverage, because the fraction $1 - e$ of vaccinated-but-unprotected individuals together with the unvaccinated remains a susceptible pool large enough to sustain transmission. This is the structural reason acellular-pertussis vaccines (efficacy around $0.75$ against transmission, with $R_0 \approx 15$ requiring $1 - 1/15 \approx 0.93$) cannot eliminate pertussis, whereas the smallpox vaccine (efficacy above $0.95$ against a pathogen with $1 - 1/R_0 \approx 0.80$) could. The clinical implication is that sub-$c^*$-capable vaccines remain worth giving for individual protection but cannot deliver population-level elimination. $\square$

Connections Master

• Public-health survey: epidemiology and health systems 35.06.01. This unit supplies the public-health-communication depth slice for the survey's chapter on outbreak response and health-system behaviour. The survey's general account of epidemiological study designs, the Bradford Hill criteria for causation, and causal-inference methods over observational data is exactly the toolkit used to establish the MMR-autism refutation (Taylor 2014's meta-analysis of 1.2 million children) and to quantify the Omer-Salmon finding that philosophical-exemption jurisdictions carry elevated measles and pertussis risk. The herd-immunity threshold derived here operationalises the survey's coverage-adequacy criterion as a single closed-form expression in $R_0$ and efficacy.

• Vaccines, immunological memory, and herd immunity 18.10.04. This unit is the population-level complement to the immunological-memory analysis of 18.10.04, where the $R_0$ threshold is derived as the between-host shadow of the within-host clonal-selection dynamics that vaccines induce. The immunology unit supplies the mechanistic substrate (antigen presentation, affinity maturation, the memory B-cell and T-cell compartments, immune amnesia from wild-type measles) that explains why vaccines work in the individual and why measles infection is so destructive; this unit supplies the population arithmetic that determines when the individual protection aggregates into community protection. The two units together close the loop from molecular recognition to outbreak control.

• The opioid epidemic as a parallel public-health-communication challenge 35.06.04. The opioid and vaccine-hesitancy cases are structurally parallel public-health communication failures driven by institutional trust erosion, amplified by interested actors (Purdue's marketing in one case, the Disinformation Dozen and social-media echo chambers in the other), and worsened by policy responses that mistook the dominant driver. In both cases the supply-side intuition (restrict prescriptions; ban exemptions) was necessary but insufficient because it did not address demand (addiction; hesitancy), and the effective response required multi-pronged intervention on coupled drivers. The syndemic concept introduced for opioids reappears here as the dual-epidemic structure of a fast-spreading pathogen coupled to a fast-spreading belief system about the pathogen.

• The replication crisis and Open Science 29.01.02. The Wakefield fraud is a case study in the methodological pathologies that the replication-crisis and Open-Science programme diagnoses and reforms. The 12-child case series with no controls, selective reporting, undisclosed conflicts of interest, and researcher degrees of freedom is the canonical instance of the low-power, low-pre-study-odds, high-bias configuration that Ioannidis 2005 predicted would publish mostly false findings, and the failure of independent groups to replicate the MMR-autism association over the following decade is the empirical confirmation. The institutional trust that vaccine hesitancy erodes is the same trust that preregistration, registered reports, and open data are designed to rebuild.

Historical & philosophical context Master

Wakefield and 12 co-authors' 1998 paper in The Lancet ^{[Wakefield1998]} reported a case series of 12 children described as presenting with a regressive developmental disorder and chronic enterocolitis, and speculated — without an epidemiological study — on an association with MMR vaccination. At a press conference coinciding with publication Wakefield advocated single-antigen vaccines over the combined MMR, a recommendation that had no support in his own data and that triggered a decade of declining UK MMR coverage. Brian Deer's investigation for The Sunday Times, begun in 2004 and extended in a 2011 three-paper BMJ series ^[Deer2011], established that the case histories had been selectively reported, that medical records had been altered to fit the temporal claim, and that Wakefield had received over 400,000 pounds from lawyers building litigation against vaccine manufacturers — a conflict undisclosed to his co-authors, The Lancet, and the ethics committee.

The Lancet retracted the paper on 2 February 2010, and in May 2010 the General Medical Council's Fitness to Practise Panel ^[GMC2010], after a 217-day hearing, found Wakefield guilty of serious professional misconduct over the undisclosed conflicts, the ethically unauthorised invasive procedures on children, and the falsified case histories, and struck him off the medical register. Paul Offit's 2008 Autism's False Prophets ^[Offit2008] synthesised the fraud, the failed replication programme, and the human cost in resurgent measles; Taylor and colleagues' 2014 Vaccine meta-analysis of 1,256,407 children ^[Taylor2014] closed the scientific question with a quantitative null. The institutional and behavioural sequelae — the Omer-Salmon hesitancy phenotype ^[Omer2009], Kahan's cultural-cognition reframing of risk perception ^[Kahan2010], and the SAGE 3Cs and Betsch 5As operational frameworks — reframed the residual problem: vaccines are safe and effective, and the binding constraint on their public-health impact is now the communication of that fact to populations whose trust has been eroded.

Bibliography Master

@article{Wakefield1998,
  author  = {Wakefield, A. J. and Murch, S. H. and Anthony, A. and others},
  title   = {Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and
             pervasive developmental disorder in children},
  journal = {The Lancet},
  volume  = {351},
  number  = {9103},
  pages   = {637--641},
  year    = {1998},
  note    = {RETRACTED 2 February 2010},
  doi     = {10.1016/S0140-6736(97)11096-0},
}

@article{Deer2011,
  author  = {Deer, B.},
  title   = {Secrets of the {MMR} scare: How the case against the {MMR} vaccine was fixed},
  journal = {BMJ},
  volume  = {342},
  pages   = {c5258, c5347, c7452},
  year    = {2011},
  doi     = {10.1136/bmj.c5258},
}

@misc{GMC2010,
  author = {{General Medical Council}},
  title  = {Determination on Dr {Andrew Jeremy Wakefield}},
  howpublished = {Fitness to Practise Panel Hearing, 24 May 2010},
  year   = {2010},
}

@book{Offit2008,
  author    = {Offit, P. A.},
  title     = {Autism's False Prophets: Bad Science, Risky Medicine,
               and the Search for a Cure},
  publisher = {Columbia University Press},
  year      = {2008},
}

@article{Taylor2014,
  author  = {Taylor, L. E. and Swerdfeger, A. L. and Eslick, G. D.},
  title   = {Vaccines are not associated with autism: an evidence-based
             meta-analysis of case-control and cohort studies},
  journal = {Vaccine},
  volume  = {32},
  number  = {29},
  pages   = {3623--3634},
  year    = {2014},
  doi     = {10.1016/j.vaccine.2014.04.085},
}

@article{Kahan2010,
  author  = {Kahan, D. M. and Braman, D. and Cohen, G. L. and Gastil, J. and Slovic, P.},
  title   = {Culture and identity-protective cognition: explaining the
             white-male effect in risk perception},
  journal = {Journal of Empirical Legal Studies},
  volume  = {4},
  number  = {3},
  pages   = {465--505},
  year    = {2007},
}

@article{MacDonald2015,
  author  = {MacDonald, N. E. and the SAGE Working Group on Vaccine Hesitancy},
  title   = {Vaccine hesitancy: Definition, scope, and determinants},
  journal = {Vaccine},
  volume  = {33},
  number  = {34},
  pages   = {4181--4183},
  year    = {2015},
  doi     = {10.1016/j.vaccine.2015.04.036},
}

@article{Betsch2015,
  author  = {Betsch, C. and B{\"o}hm, R. and Korn, L. and Holtmann, P.},
  title   = {On the benefits of explaining herd immunity in vaccine advocacy},
  journal = {Vaccine},
  volume  = {33},
  pages   = {7186},
  year    = {2015},
}

@article{Omer2009,
  author  = {Omer, S. B. and Salmon, D. A. and Orenstein, W. A. and deHart, M. P. and Halsey, N.},
  title   = {Vaccine refusal, mandatory immunization, and the risks of
             vaccine-preventable diseases},
  journal = {New England Journal of Medicine},
  volume  = {360},
  number  = {19},
  pages   = {1981--1988},
  year    = {2009},
}

@article{Zipprich2015,
  author  = {Zipprich, J. and Winter, K. and Hacker, J. and others},
  title   = {Measles outbreak --- {California}, {December} 2014--{February} 2015},
  journal = {Morbidity and Mortality Weekly Report (MMWR)},
  volume  = {64},
  number  = {6},
  pages   = {153--154},
  year    = {2015},
}