THE NIGHT WATCH THAT WAS DAY

In the early 1640s, eighteen members of an Amsterdam civic guard company each paid a personal fee towards a single group portrait, commissioned from Rembrandt van Rijn and arranged around their captain, Frans Banninck Cocq. Each paid about a hundred guilders to be recognised, and eighteen men cannot all stand in the best light.

Such a schutterij, or civic guard company, was a part-time militia drawn from a city’s own tradesmen and merchants, and commissioning a large group portrait of its officers and members was already a well-established genre in the Dutch Republic by the 1640s, practised by Frans Hals and several other painters working in Amsterdam and Haarlem. The usual solution to the problem of eighteen paying subjects was a disciplined, near-symmetrical arrangement, faces lined up at similar height and similar brightness, so that no subscriber could reasonably complain of having paid for less prominence than his neighbour.

Rembrandt solved the same problem differently. Movement and selective light replaced the usual rows of equal faces. Figures step forward, turn, gesture, catch a shaft of light or fall into shadow according to where the composition itself needed weight. A hundred guilders or so bought a face and no particular place. Every sitter is present and identifiable somewhere in the canvas, and hardly any two of them stand in quite the same light.

For well over a century the picture carried no fixed name beyond a plain description of its subject, and only later acquired the title by which it is known today, after accumulated layers of darkened varnish had made a bright, daytime militia muster look like a scene lit by moonlight. The name stuck for generations. A cleaning of the surface in the twentieth century confirmed what the composition itself had quietly suggested all along: broad daylight, a company stepping out into an ordinary afternoon, the captain’s own shadow falling crisply on a lieutenant’s coat in a way moonlight could never produce.

The Rijksmuseum has spent years studying and restoring it in public view, with imaging unavailable to any earlier restorer, and each new finding is reported as news. The canvas on the wall has never stopped accumulating a second history on top of the one Rembrandt painted.

Decades after it was painted, workmen relocated the canvas to Amsterdam’s town hall, trimming it on all four sides in 1715, most heavily on the left so that it would fit the space allotted to it between two doors, a piece of practical carpentry that removed a run of figures permanently from view. A smaller painted copy made before the trimming, long attributed to Gerrit Lundens and now held in London, preserves two additional men who once stood at the composition’s edge and survive today only in that copy. Visitors to Amsterdam see a painting still missing part of what its own eighteen subscribers originally agreed to pay for, bordered now by an edge that was decided by a doorway and had no place in the original design.

During the summer of 2021 the museum hung the painting with printed extensions of the missing margins generated by a neural network, which had been taught from the Lundens copy and was made to match Rembrandt’s colours and brushwork and the perspective of the original. The additions stayed up for three months and are not on permanent display, because they are an interpretation and not Rembrandt’s own work. It was a forecast, a gap filled in from a copy, and the museum labelled it as one.

A bond yield curve is a group portrait of its own, painted by lenders who never met. Plotted across maturities from a few months to thirty years, it traces the interest rate the market demands to lend for each length of time. Lenders charge more to be repaid a long way off, so a curve sloping gently upward is the normal shape. When short rates rise above long ones the curve inverts, which is traditionally read as a collective bet that growth or inflation will weaken enough within a year or two to force rates down again. Every American recession since the 1950s had been preceded by such an inversion, which is why forecasters and casual commentators alike treated it as one of the more dependable warnings in economics.

Nobody draws the line. Pension funds, insurers, foreign governments and speculators, mostly strangers, each lend at their own price, and the curve is the sum they leave on the wall. We say that the market thinks, as a guide says that Rembrandt intended. Both are shorthand for a crowd.

Modern distortions reach the curve the way darkened varnish reached the canvas, arriving from outside the original composition entirely. Central banks buying enormous quantities of government bonds as a deliberate tool of monetary policy, first after the 2008 financial crisis and again after 2020, pulled long-term yields down for reasons having nothing to do with what private lenders expected growth or inflation to do. Separately, banking and insurance regulations adopted after 2008 required many institutions to hold large, stable quantities of government debt regardless of their own view on the economy, creating a captive source of demand immune to ordinary changes of opinion. Neither was secret, and both were easy to forget while staring at the finished line.

From July 2022 short rates stood above long ones for longer than at any time on record. The central bank had pushed short rates up sharply against the highest inflation in four decades, while longer rates, held down in part by the institutional holdings described above, rose far more slowly. Nearly every earlier inversion had been followed by recession within six to twenty-four months. This one was not, at least on that schedule, and the argument ran on between those who saw a soft landing and those who thought the old signal had failed, or its timetable lengthened, or that central-bank purchases and new capital rules had bent the curve into a shape that no longer meant what it once had.

By late 2024 the slope had turned upward again, and the explanations arrived on time, each sure that the story had been plain all along.

Which sitter the composition was designed to favour is a question the eighteen might have answered eighteen ways. The picture in Amsterdam is still known by the name of a light the scene never contained, and it still stops where a doorway told it to.