The Hubble Deep Field began as a gamble on a tiny patch of sky that had been chosen because it looked almost empty, and it ended by revealing nearly 3,000 galaxies hiding in what seemed like nothing.

In December 1995, the Hubble Space Telescope spent ten days staring at a small, deliberately unremarkable patch of sky. The resulting image contained around 3,000 galaxies. The decision to do it was contested, and the astronomer who pushed it through did so over the objections of senior colleagues.

What is worth adding is the part the headline compresses. The patch was not chosen because it happened to look empty. It was chosen to be empty, and that distinction is the whole method.

Why “empty” was the point

The field was a small region in the constellation Ursa Major, near the handle of the Big Dipper, about the size of a pinhead held at arm’s length. It was selected against a list of requirements rather than picked at random.

It had to sit at a high galactic latitude, well away from the crowded plane of the Milky Way, so that the telescope’s view would not be blocked by the stars and dust of our own galaxy. It had to be free of bright nearby objects whose glare would drown out anything faint. And it needed to be observable by Hubble without the Earth repeatedly moving into the way. An empty-looking field met all of these conditions. The emptiness was not the gamble. It was the experimental design. The point of pointing at apparent nothing was to clear the foreground so that the faint and distant could register.

The galaxies in the final image were not, in any real sense, hiding. They were simply too faint for a short exposure to detect. Light had been arriving from them all along. What the Deep Field did was hold still and collect that light long enough for it to add up.

The risk that was actually being taken

The person behind it was Robert Williams, then director of the Space Telescope Science Institute. As director, he controlled a block of observing time, Director’s Discretionary Time, that he could assign without going through the usual competitive review. He chose to spend a large share of it on this.

That was the real risk, and it was institutional rather than scientific. Hubble time was scarce and fiercely contested. Worse, the telescope’s reputation was still damaged. The flaw in its primary mirror had made it a public embarrassment after launch in 1990, and a servicing mission had only recently installed the corrective optics that fixed it. Spending ten days of a newly repaired and much-criticised telescope on a blank piece of sky was the kind of decision that would look very bad if it returned nothing.

The opposition was specific. The astronomer John Bahcall, a central figure in Hubble’s creation, went to Williams more than once to urge him to drop the idea, and Bahcall had published a paper around the time of Hubble’s launch arguing that the telescope would not turn up a new population of galaxies. Williams went ahead regardless. He later described himself, without much elaboration, as a risk taker.

What ten days of light added up to

The image was assembled from 342 separate exposures taken over ten consecutive days, with a total exposure time of more than 100 hours, against the few hours that a typical Hubble observation used.

The first count of galaxies in the field came in at roughly 1,500. Later analysis, as NASA notes, raised that to around 3,000. They were not uniform. Alongside familiar spirals and ellipticals were galaxies that looked irregular, disturbed, and oddly shaped, and many of these were the distant ones, seen as they had been billions of years in the past. According to the European Space Agency’s account, the image reached back to an era when many galaxies had not yet had time to form most of their stars.

That mattered for more than its novelty. The disturbed appearance of the distant, younger galaxies fitted the picture in which small galaxies collide and merge over time to build larger ones. The Deep Field did not prove that model on its own, but it gave it direct visual support across a long stretch of the universe’s history.

The method outlived the image

One decision turned the gamble from a single result into something larger. The Deep Field data was released to the whole astronomical community without the usual proprietary period, so that any researcher could work on it immediately. A great deal of the science came from people who had no part in taking the image.

The approach itself then became standard. Staring for a very long time at a carefully chosen blank field was repeated and extended: a southern Deep Field in 1998, the Hubble Ultra Deep Field in 2004, and, more recently, the deep images returned by the James Webb Space Telescope, which reach galaxies from far earlier in cosmic time than Hubble could.

It is worth keeping one limit in view. The Deep Field is a single narrow core sample of the universe, and the often-quoted jumps in the estimated total number of galaxies are extrapolations from that sample and others like it, not a direct count. What the 1995 image settled was narrower and firmer: that a patch of sky chosen for having nothing obvious in it was, in fact, full. The open question it handed forward was how much further back the same method could be pushed, and that question is still being answered.