How Vaccines Work — and Why Some People Refuse Them

Smallpox, a disease that was fatal in up to 30% of cases, was declared eradicated in 1980, and the tool was a vaccine. Wild polio is now confined to two countries. Yet in the United States, 4,410 measles cases had been confirmed by October 15, 2026, compared with 2,289 in all of 2025. Understanding how vaccines work, and why some people still decline them, means looking at immunology and psychology together, because each explains something the other cannot.

Key takeaways

• Vaccines work by showing the adaptive immune system a harmless piece of a pathogen, or instructions to make one, so that memory cells exist before the real infection arrives.
• A 2024 Lancet modeling study estimated that vaccination averted 154 million deaths between 1974 and 2024, 101 million of them in infants.
• Safety evidence comes in layers: preclinical work, three trial phases, regulatory review, and post-approval surveillance. Trials can miss events rarer than about one in several thousand, which is why surveillance after approval matters.
• Herd immunity protects people who cannot be vaccinated. For measles, the World Health Organization says coverage of 95% or more with two doses is needed.
• Hesitancy is not one thing. Researchers separate confidence, complacency, and convenience, and the fixes differ for each.

What the immune system does, and what a vaccine borrows

Your body fights infection in two phases. The first is innate immunity, a fast, generic response. Neutrophils and macrophages (white blood cells that engulf invaders) and natural killer cells attack within minutes to hours, and the inflammation they trigger produces the redness, heat, swelling, and sometimes fever that you feel as illness. Innate defenses do not remember anything. A pathogen met for the tenth time gets the same response as one met for the first time.

The second phase is adaptive immunity, which is slower and specific. B cells make antibodies, Y-shaped proteins that lock onto particular molecular features (antigens) on a pathogen, neutralizing it or tagging it for destruction. T cells come in several kinds: helper T cells coordinate the response, cytotoxic T cells kill infected cells, and memory B and T cells persist after the fight is over. The antibody structure is what makes this specificity possible, and our broader explainer on how the immune system works covers the full cast.

Here is the key point. Adaptive immunity takes days to weeks to get going the first time, which is why a new infection can run far ahead of it. The US Centers for Disease Control and Prevention notes that immunity can take weeks to develop after vaccination. The second encounter is faster because memory cells are already waiting.

A vaccine is a way of staging that first encounter safely. Think of a fire drill. Nobody has to burn the building down to learn where the exits are, and when the real fire comes, people move without stopping to think. Or think of a wanted poster. Handing the police a picture of the suspect before the crime does not stop the crime, but it makes the response much faster when the suspect turns up.

Edward Jenner's 1796 experiment with cowpox and smallpox is usually cited as the first deliberate vaccination, and our pages on Jenner's smallpox experiment and the broader history of vaccination as an invention cover how the idea spread. The principle has not changed since. What has changed is how precisely modern vaccines choose what to show the immune system.

Five ways to build a vaccine

All vaccines present an antigen. They differ in how they deliver it, and that choice drives how many doses are needed, who can safely receive it, and how it is manufactured.

Platform	How it works	Examples	Trade-offs
Live attenuated	A weakened pathogen that can still replicate, so it mimics a real infection closely	Measles-mumps-rubella (MMR), varicella (chickenpox), yellow fever, oral polio vaccine	Strong, long-lasting immunity, often from one or two doses. Not suitable for people with severely weakened immune systems, and in very rare cases the weakened virus can cause disease
Inactivated	The whole pathogen, killed so it cannot replicate	Injectable flu vaccine, inactivated polio vaccine, hepatitis A, rabies	Cannot cause the infection it prevents. Usually needs multiple doses or boosters
Subunit and conjugate	Only a protein or sugar fragment of the pathogen, often paired with an adjuvant	Hepatitis B, human papillomavirus (HPV), shingles, whooping cough components	Very targeted. Adjuvants (immune stimulants such as aluminum salts) improve the response
mRNA	Messenger RNA instructions for cells to build one viral protein temporarily	Pfizer-BioNTech and Moderna COVID-19 vaccines	Fast to design and update. Needs cold-chain storage. Short-lived mRNA is degraded by the cell within days
Viral vector	A harmless virus, often an adenovirus, carries the gene for an antigen into cells	AstraZeneca and Johnson & Johnson COVID-19 vaccines	Scales well. Pre-existing immunity to the carrier virus can blunt the response

The mRNA platform draws the most questions, and the facts are simple. The mRNA is wrapped in a lipid droplet that fuses with cells and releases the mRNA into the cytoplasm, the cell's working fluid, where ribosomes read it and build the viral protein. The mRNA does not enter the nucleus, where DNA is kept, and it is broken down within days. Our explainer on mRNA vaccines goes deeper on the chemistry. The viral vector approach is related to the biology of adenoviruses, which is useful background on why carrier immunity matters.

For a different treatment of the same immunology, see the dedicated explainer on how vaccines train memory B cells, which follows the cellular mechanism in more detail than this article does. And if the soreness and tiredness after a shot worries you, the CDC says the immune response may cause tiredness and discomfort for a day or two, and there is a separate explainer on why fevers help fight infections.

How a vaccine gets from the lab to your arm

Vaccines are given to very large numbers of healthy people, including infants, so the evidence bar is high. The pipeline has recognizable stages.

Stage	Who is involved	The question it answers
Preclinical	Cell cultures and animal studies	Does it trigger an immune response, and is there an obvious toxicity?
Phase 1	Dozens of volunteers	Is it safe at several doses?
Phase 2	Hundreds	What dose and schedule produce a good immune response? What are common side effects?
Phase 3	Thousands to tens of thousands, randomized against placebo or an existing vaccine	Does it prevent disease, and are there more serious events than in the comparison group?
Regulatory review	Agencies such as the FDA and the European Medicines Agency	Do the data justify approval for a defined population?
Post-approval surveillance	Passive reporting systems and active monitoring of millions of recipients	Are there rare events that trials were too small to detect?

The Pfizer-BioNTech COVID-19 trial shows what a Phase 3 trial looks like in numbers. According to the paper in the New England Journal of Medicine (Polack et al., 2020), 43,548 people aged 16 or older were randomized at 152 sites in six countries, and 43,448 received injections: 21,720 got the vaccine and 21,728 got placebo. Among participants without prior infection, there were 8 cases of COVID-19 in the vaccine group and 162 in the placebo group, which worked out to 95% efficacy. Serious adverse events were similar in the two groups (0.6% against 0.5%), and investigators judged that no deaths were related to either vaccine or placebo. Any adverse event was reported more often by vaccine recipients (27% against 12%), mostly short-lived reactions such as pain at the injection site and fatigue. This kind of randomization is the reason randomized trials are so powerful at separating a vaccine's effects from coincidence.

Why surveillance after approval matters

A trial of about 22,000 vaccinated people can say a lot about common reactions, but there is a mathematical limit to what it can see. A useful rule of thumb in statistics, known as the rule of three, says that if you observe zero events among n people, the upper 95% bound on the true rate is about 3 divided by n. For 21,720 vaccine recipients, that bound is roughly 1 in 7,200. To rule out an event as rare as 1 in 100,000, a trial would need on the order of 300,000 participants. This is not a flaw in the trial. It is arithmetic, and it is why regulators combine trials with monitoring of millions of recipients afterward.

Surveillance has found real signals. After the COVID-19 vaccines were rolled out, health systems detected rare cases of myocarditis (inflammation of the heart muscle) after mRNA vaccination, mostly in younger males. The numbers are best read in context.

Study	Group and window	Exposure	Extra myocarditis events
Patone et al., Nature Medicine 2022 (England, age 16 and older, 28 days)	All eligible recipients	BNT162b2, first dose	1 per million (95% CI 0 to 2)
Same	Same	mRNA-1273, first dose	6 per million (95% CI 2 to 8)
Same	Same	mRNA-1273, second dose	10 per million (95% CI 7 to 11)
Same	People with a positive SARS-CoV-2 test	Infection itself	40 per million (95% CI 38 to 41)
Karlstad et al., JAMA Cardiology 2022 (Nordic countries, 23 million residents)	Males aged 16 to 24, second dose	BNT162b2	5.55 per 100,000 (95% CI 3.70 to 7.39)
Same	Same	mRNA-1273	18.39 per 100,000 (95% CI 9.05 to 27.72)

Two cautions apply when reading this table. First, the groups differ. The England analysis covers everyone aged 16 and older, while the Nordic figures cover the highest-risk subgroup, young males after the second dose, which is why the Nordic numbers look larger. Second, the England authors found that the increased myocarditis risk from the two mRNA vaccines appeared only in people under 40, and that a positive test carried a higher myocarditis risk than the vaccines did for the population as a whole. That comparison does not erase the vaccine risk in the youngest males, and regulators in several countries adjusted recommendations by age and product accordingly. It does show how surveillance is supposed to work: detect a signal, quantify it, and compare it with the harm of the disease. Similar scrutiny followed rare clotting events with adenovirus-vector vaccines. Earlier, the first rotavirus vaccine, RotaShield, was linked to an increased incidence of intussusception (a rare bowel blockage in infants) after its introduction, a reminder that this process has caught problems before.

What vaccines have actually done

Claims about vaccines tend either to dissolve into slogans or to disappear into statistics. A few numbers, all from international health bodies and dated, give a sense of scale.

Measure	Figure	Source
Deaths averted by vaccination, 1974 to 2024	154 million, including 101 million among infants	Shattock et al., The Lancet 2024, with the WHO summary
Share of the decline in global infant mortality attributed to vaccination	40% (more than 50% in the African region)	Same
Measles deaths before the vaccine (introduced in 1963)	About 2.6 million per year	WHO, measles fact sheet (July 2026)
Measles deaths	780,000 in 2000; 95,000 in 2024	Same
Measles deaths averted by vaccination, 2000 to 2024	About 59 million	Same
Wild polio cases	More than 99% lower since 1988, from an estimated 350,000 cases in more than 125 countries; endemic now in Afghanistan and Pakistan	WHO polio fact sheet
Smallpox	Declared eradicated in 1980; last known natural case in Somalia in 1977	WHO
Global coverage with three doses of DTP vaccine	85% in 2025	WHO immunization coverage fact sheet
Children who received no vaccine at all ("zero-dose")	13.5 million in 2025	Same

Sources: WHO news release on the Lancet analysis, WHO measles fact sheet, WHO polio fact sheet, WHO smallpox page, and WHO immunization coverage fact sheet.

Two points deserve emphasis. The Lancet study is a model, not a head count: it compares the world with vaccination to a hypothetical world without it, and the authors report each averted death as gaining an average of 66 years of full health. The authors concluded that the Expanded Programme on Immunization has made the single greatest contribution to improved infant survival over the past 50 years. Second, measles deaths fell by about 88% between 2000 and 2024 (780,000 to 95,000), and measles vaccination accounts for about 60% of the lives saved in the Lancet analysis, which is why measles is the disease that sounds the alarm first when coverage drops.

Herd immunity: the arithmetic of protecting other people

Vaccines protect the person who receives them, and they also protect people nearby. When enough of a community is immune, chains of transmission break, and an outbreak fizzles before it reaches the people who are not protected. That group includes infants too young to be vaccinated (the first MMR dose is recommended at 12 to 15 months, according to the CDC), people receiving chemotherapy, transplant recipients, and people whose immune response to the vaccine was weak.

The basic model is simple. If each infected person would on average infect R0 others in a fully susceptible population, an outbreak cannot sustain itself once the fraction of immune people exceeds 1 minus 1 divided by R0. The WHO says one person with measles can infect up to 18 others, and the virus can remain infectious in the air or on surfaces for up to two hours.

Assumed R0 for measles	Immune fraction needed (1 minus 1/R0)	Vaccine coverage needed if two doses are 97% effective
12	about 92%	about 94.5%
15	about 93%	about 96%
18	about 94%	about 97%

The 97% figure is the CDC's estimate of two-dose MMR effectiveness against measles (93% after one dose). The table shows why public health agencies say "95% or more": it covers the plausible range of R0 and the fact that no vaccine protects everybody. The model assumes people mix at random. Real communities do not, and unvaccinated people tend to cluster in the same schools, congregations, and neighborhoods, so local coverage can fall below the threshold even when the national average looks healthy.

What a threshold breach looks like

The United States declared measles eliminated in 2000, according to the CDC. In 2025 it recorded 2,289 confirmed cases in 48 outbreaks, and as of October 15, 2026, the CDC's count for 2026 had reached 4,410 cases across 42 outbreaks, 96% of them linked to outbreaks. The CDC's most recent breakdown for status and severity, covering 2024, found that 89% of cases were unvaccinated or of unknown status, and 40% (114 of 285) led to hospitalization. These are reporting-based counts that CDC revises, and the 2025 and 2026 breakdowns are not yet available in the same form.

Globally, WHO reported that 33 million children missed a measles vaccine dose in 2022. A serological survey in Samoa during 2018 and 2019, published in Tropical Medicine and Infectious Disease, tested 8,394 valid blood samples from people aged five and older and found measles antibodies in just 45.8% of them, the lowest of the four diseases tested. The authors described pockets of low population immunity. The point is not any single outbreak. It is that herd immunity is a threshold effect, and the cost of crossing below it falls on the people least able to protect themselves.

What the safety evidence says about circulating claims

The claim that the MMR vaccine causes autism has been examined more than almost any other vaccine safety question, and its origin is well documented. It began with a 1998 paper in The Lancet by Andrew Wakefield and colleagues, reporting on 12 children. The Lancet retracted the paper in full on February 2, 2010, after the UK's General Medical Council found that claims about how the children were recruited and about ethical approval were false. Wakefield was struck from the UK medical register on May 24, 2010. In January 2011, the BMJ published journalist Brian Deer's investigation and an editorial calling the paper an "elaborate fraud." Later reporting also showed that the children were recruited through a lawyer preparing litigation against vaccine manufacturers.

Independent research has not found a link. A Danish nationwide cohort study (Hviid et al., Annals of Internal Medicine, 2019) followed 657,461 children born from 1999 through 2010 and found a fully adjusted hazard ratio of 0.93 (95% confidence interval 0.85 to 1.02) for autism in vaccinated compared with unvaccinated children. The authors found no increased risk in subgroups such as siblings of children with autism, and no clustering after vaccination. A 2021 Cochrane review covering 138 studies and 23,480,668 participants concluded that existing evidence supports use of MMR and MMRV vaccines for mass immunization, with a pooled rate ratio of 0.93 (0.85 to 1.01) for autism from two observational studies that included 1,194,764 children.

Why does the belief persist if the evidence is this clear? One reason is timing. Autism is often first noticed at the ages when children get their scheduled vaccines, so many families will see the two events in sequence. Sequence is not cause, a distinction we unpack in our piece on how correlation differs from causation.

Other common claims fall apart on mechanism. mRNA vaccines do not alter DNA, because the mRNA stays in the cytoplasm and is degraded in days. The notion that the childhood schedule "overloads" the immune system is not supported by studies of the schedule, which have found it safe. The difficult-to-communicate truth is that vaccines do have side effects, which range from sore arms to rare serious events, and acknowledging that honestly is part of why trust survives.

Why people say no: a map of hesitancy

The WHO listed "vaccine hesitancy" among its ten threats to global health in 2019, defining it as the reluctance or refusal to vaccinate despite the availability of vaccines. The WHO's SAGE working group framed hesitancy as a delay in accepting or a refusal of vaccination despite available services, and described it as complex and context-specific, shaped by complacency, convenience, and confidence (MacDonald, Vaccine, 2015). Those three words are a useful organizing scheme.

Driver	What it means	Evidence or example	What tends to help
Confidence	Trust in the vaccine, the provider, and the system that delivers it	A 2020 review in the Journal of General Internal Medicine describes research linking the Tuskegee syphilis study to less health-seeking behavior, higher mortality, and greater medical mistrust among Black men who were not enrolled in it	Trusted messengers and honest disclosure of real risks
Complacency	Seeing the disease as unlikely or mild	Vaccine success makes the disease invisible. Measles looks like a historical curiosity until an outbreak arrives	Clear local data on outbreaks and consequences
Convenience	Practical barriers to getting vaccinated	Cost, clinic hours, travel, language. In low-coverage settings, 13.5 million children had no vaccine at all in 2025, which is more often a supply and access failure than a belief problem	Better access, reminders, extended hours
Cognitive biases	Systematic quirks in how people judge risk	A 2023 scoping review in Vaccines found 21 biases related to vaccination across 19 studies, grouped into how information is processed, how decisions are made, and prior beliefs	Framing and messaging that account for them
Social identity	Vaccination as a signal of group membership	Perceived community norms shape decisions	Messengers from within the group

The cognitive-bias row deserves more explanation, since it is the least intuitive. A rare vaccine reaction is vivid, specific, and attributable: a named child, a named day, a named vial. The disease the vaccine prevents is statistical and diffuse. People weigh vivid cases more heavily than statistics. They also tend to treat harm from an action as worse than equal harm from inaction, a tendency called omission bias. If you vaccinate and the child is harmed, you chose that. If you do not vaccinate and the child gets sick, the disease chose it. Our pages on why losses feel stronger than gains and how framing changes risk perception explain the mechanisms behind these asymmetries.

Trust has a history. The Tuskegee study ran for decades without informed consent. The point of the economics research summarized in the 2020 review is that the harm extended beyond participants to a wider community, which is why trust cannot be rebuilt by information alone. A related experiment published in American Economic Journal: Economic Policy in 2024 randomly varied who delivered a vaccine message to Black and white men without a college degree. Laypeople who shared the target group's background were the most effective messengers, especially among those least willing to be vaccinated, and the authors suggest a trade-off between expert credentials and social closeness in low-trust settings.

Misinformation also plays a role, but the picture is not just a lack of facts. Research on motivated reasoning shows that people evaluate evidence through their existing beliefs and identities, and studies of why conspiracy theories spread online describe how those patterns interact with social media. This matters because it means that repeating the facts louder is rarely the whole answer.

What the evidence says helps

Some of the best-supported interventions are unglamorous. A megastudy in PNAS (2021) tested text-message nudges with 47,306 patients ahead of a primary care appointment and found that texts raised flu vaccination by an average of 5% in relative terms. The best messages told patients that a flu shot was reserved or waiting for them. A follow-up in PNAS (2022) with 689,693 Walmart pharmacy patients found that text reminders raised pharmacy flu vaccination by an average of 2.0 percentage points (6.8%) over three months, again with the "waiting for you" framing performing best.

These effects are modest, and they come from flu shots in the United States, so the numbers should not be generalized. What they show is that for a large share of the under-vaccinated, the obstacle is friction and forgetfulness, not conviction. That fits with the WHO's observation that 13.5 million children had no vaccine at all in 2025, a problem fixed by clinics, supply chains, and outreach rather than persuasion.

For people whose hesitancy is rooted in distrust, the evidence points toward messenger and process: who is speaking, whether they acknowledge real risks, and whether the system treats questions as legitimate. This article is not a substitute for individual medical advice, and it does not argue that any one person should or should not receive a particular vaccine. National immunization schedules and the WHO position papers explain the evidence for each vaccine.

Where the science is still unsettled

It would be misleading to say that everything about vaccines is settled. Several questions are active.

• Duration of protection. Protection from some vaccines wanes, and the timing differs by vaccine and by person. Measles vaccination gives long protection for most people, but boosters exist for other diseases for this reason.
• Variable effectiveness. Influenza vaccines must be reformulated each year because circulating strains change. Effectiveness varies by season and by how well the vaccine strains match the circulating ones.
• Rare-event risk by age. The myocarditis data show that risk can differ sharply by age, sex, and product. Evidence is still accumulating on long-term outcomes.
• How hesitancy changes over time. Trust shifts with events, and measures taken in one country do not transfer cleanly to another.

Saying so is not a weakness. A system that publishes its uncertainties, tracks its failures, and revises recommendations is behaving the way a scientific system should, and that transparency is the best defense against the claim that institutions hide things.

Frequently asked questions

Q: Are vaccines ever actually dangerous? A: Every medical intervention carries some risk. Common reactions are mild and short-lived, such as a sore arm or a day or two of tiredness. Rare serious events have been identified, including myocarditis after mRNA COVID-19 vaccines (about 1 to 10 extra cases per million doses in the England data, and roughly 5.5 to 18 per 100,000 in young males after the second dose in the Nordic data). The relevant comparison is always with the risk of the disease the vaccine prevents.

Q: Can you catch the disease from a vaccine? A: Inactivated, subunit, and mRNA vaccines contain no live pathogen capable of causing infection. Live attenuated vaccines contain a weakened pathogen and in rare cases can cause a mild form of the disease, or serious illness in people with severely weakened immune systems. That is why they are handled differently for that group.

Q: Why do flu vaccines work less consistently than measles vaccines? A: Influenza viruses change their surface proteins from year to year, so the vaccine is reformulated in advance and the match between the vaccine and circulating strains varies. The measles virus is more stable, and the CDC estimates two doses of MMR are 97% effective against measles, compared with 93% for one dose.

Q: Is hesitancy the same as refusal? A: No. The WHO's SAGE group defines hesitancy as a delay in acceptance or refusal despite the availability of services, which covers a spectrum from people who accept every vaccine with some worry, to people who delay, to people who refuse. Many people classed as unvaccinated face access problems, not opposition.

Q: Does natural infection give better immunity than a vaccine? A: Sometimes natural infection produces broad immunity, but it comes at the cost of the disease itself. The WHO lists blindness, brain swelling, severe diarrhea, pneumonia, and death among the complications of measles, for example. For several diseases, vaccine-induced protection is comparable to or better than infection-induced protection, which is one reason to compare the two by outcomes and not by immune markers alone.

Q: Where can someone find evidence about a specific vaccine? A: National immunization schedules, WHO position papers, and Cochrane reviews summarize the evidence for each vaccine, including known side effects. For a conversation about a personal medical situation, a clinician who knows the person's history can interpret that evidence in context. Our article on how we know vaccines are safe walks through the process in more detail.

Sources

• WHO, Global immunization efforts have saved at least 154 million lives over the past 50 years (2024)
• WHO, Measles fact sheet (July 2026) and Immunization coverage fact sheet
• WHO, Ten threats to global health in 2019
• CDC, Measles cases and outbreaks (data as of October 15, 2026) and CDC, Measles vaccine
• Polack, F.P., et al. (2020). Safety and efficacy of the BNT162b2 mRNA Covid-19 vaccine. New England Journal of Medicine. Open-access copy
• Patone, M., et al. (2022). Risks of myocarditis, pericarditis, and cardiac arrhythmias associated with COVID-19 vaccination or SARS-CoV-2 infection. Nature Medicine; Karlstad, Ø., et al. (2022). SARS-CoV-2 vaccination and myocarditis in a Nordic cohort study of 23 million residents. JAMA Cardiology.
• Hviid, A., et al. (2019). Measles, mumps, rubella vaccination and autism: a nationwide cohort study. Annals of Internal Medicine, 170(8), 513-520; Cochrane review of vaccines for measles, mumps, rubella, and varicella in children (2021).
• Shattock, A.J., et al. (2024). Contribution of vaccination to improved survival and health: modelling 50 years of the Expanded Programme on Immunization. The Lancet, 403(10441), 2307-2316.
• MacDonald, N.E. (2015). Vaccine hesitancy: definition, scope and determinants. Vaccine, 33(34), 4161-4164.

The takeaway

Vaccines work because they stage a safe rehearsal for an immune system that otherwise learns only by surviving the real thing. The evidence that they save lives at population scale is strong, and so is the evidence that their risks, while real, are rare and monitored. Hesitancy is best understood as several different problems under one name: some people lack access, some lack trust, some are responding to vivid risks that statistics do not match, and some have been given false information. Measles is the early warning, because its high transmissibility means that small gaps in coverage turn into outbreaks quickly, and the cost lands on infants and on people whose immune systems cannot be trained.

Related Reading

• How pandemics spread and what stops them
• How nuclear weapons work and why they have not been used since 1945
• How democracy works and why it is under threat
• How evolution works and the evidence for it
• How black holes form and what happens inside them