We don’t actually know how many people have died of complications from covid-19, the disease caused by the coronavirus that likely arrived in the United States at some point last year. There are a lot of reasons for that, including the uncertainty around when the virus actually arrived.

We don’t know how many people may have died of covid-19 without the disease having been confirmed. We can compare the death toll with prior years and see that the United States recorded hundreds of thousands more deaths in 2020 than in 2019, but at this point we’re talking about a scale at which individual tragedies become blurry. To historians, the difference between 500,500 deaths and 499,500 deaths is a subtle one in the “a million deaths is a statistic” sense. In human terms, those 1,000 deaths are 1,000 people vanished from the Earth sooner than would otherwise have happened.

It’s also unclear how many deaths have occurred because the data we have are necessarily out of date, playing catch-up as deaths occur each day or are tallied from deaths in the past. Virginia, for example, is still logging deaths that occurred during the holidays. As the pandemic emerged, there was no real-time central clearinghouse of data on coronavirus deaths, leading media outlets to generate their own varying counts. Those are usually not in sync.

But even if we knew with certainty that the country’s 500,000th coronavirus death occurred this week, it’s a scale that’s simply beyond our ability to apprehend. We are able to distinguish visually the difference between 5,000, 50,000 and 500,000, but conceptually everything in the multi-thousands tends to blur together. That’s more true at the scale of trillions, a figure nearly incomprehensibly large. But it’s also true to some extent at the scale we’re discussing now.

One way to consider what 500,000 deaths mean is to put it in a scale we’re familiar with: time. As I write, it is about 11 a.m. on Feb. 23, 2021. If each of those deaths were to occur at one-second intervals beginning right now, the 500,000th death wouldn’t occur for more than a week. It takes hours simply to tally the deaths that occurred through the spring of last year. Put in seconds, the toll is agonizingly slow.


The graphic above uses estimated demographic data based on the number of deaths by race and ethnicity at the state level, as well as national gender data. The number of deaths each day in each state is accurate as reported by the COVID Tracking Project.

Even when sped up, it’s grueling. If you show 10 deaths per second, you’re still not going to reach the 500,000th death for more than 13 hours.

This perspective is more useful, I think, than comparisons to other disasters and mass-casualty events like wars. Far more Americans have died of covid-19 than died in World War II, but World War II is barely apprehensible in itself.

There’s another way to look at it, as well: visually. We’re used to seeing abstract representations of the death toll through graphs or, as the New York Times showed on Sunday, as individual dots on a graph. But we can also approximate the toll by substituting each death with an image of a person.

The visualization below will display slightly fewer than 500,000 images, all taken from the artificial-intelligence image generating system at thispersondoesnotexist.com. These people don’t exist, but we’ve blurred the images anyway to reduce the likelihood that they resemble actual people.

It’s enough people that the visualization will likely strain your computer or phone. It will take a while to load, during which time your browser will likely not be responsive. This is not ideal but, in a way, it reinforces the point: The scale is remarkable.

Once the visualization loads, the death toll becomes obvious, even if, scrolling through it, it remains incomprehensible.

This visualization takes a long time to load. You have been warned.