For any frequent rider of the New York City Subway, it’s an archetypical scene: a delayed train, motionless in the depths of a darkened tunnel or marooned atop an aging viaduct. The semi-audible voice of a transit employee crackles through jaundiced fluorescent light, cryptically blaming the culprit: “signal problems.”
Signals have found a role as a convenient scapegoat for the subway’s endless cycle of troubles — perhaps because most subway riders will never take note of the stoplight-like apparatus when they see one.
They’re not a scapegoat without evidence. According to the subway’s own data, consistently around a third of the subway’s “major incidents,” those that delay 50 or more trains, are due to signal problems. As the New York Times has pointed out, some of the subway’s signals date from the 1930s.
If there is a fix, it will be expensive and tortuous. The Metropolitan Transportation Authority (MTA), the agency that operates the subway, estimates that a full upgrade will take something on the order of $20 billion in money and half a century in time.
How did New York’s signals get here? Why have they become a seemingly unsolvable problem that will take ages to solve? The answers lie embedded deep within the subway’s history.
The MTA’s system of train signals might seem like an obscure technical footnote. But leave the comforting station lights behind, wander out onto the tracks or into the tunnels, and the story becomes far more complex — and far more fascinating. The story of New York City’s subway signals is a story of the subway’s own complex and convoluted origins.
More than that, it’s a mirror to over a century’s worth of the city’s troubled history with mass transit.
To understand why the subway’s signals are the way they are, we have to start at the beginning.
Officially, the New York City Subway’s history began on the twenty-seventh of October, 1904, when the Interborough Rapid Transit Company (IRT) opened its first underground lines in Manhattan.
In reality, much like today, the newly opened subway was just another piece of a much larger transit puzzle. It co-existed with streetcars, ferries, and surface rail lines such as the Long Island Rail Road.
The most prominent of its counterparts were the elevated railways. The “elevateds” or “els,” as they were called, were in many ways the precursor to the subway; some would eventually be incorporated into the subway. In later decades, they would come to be associated with noise and grime, but in 1904, they were a backbone of New York City’s transit.
The els’ signals were still in their infancy. “The only thing [the elevated railways] had for signalling were these old-type semaphore-arm signals,” says Joe Cunningham, a subway historian and former subway engineer. Usually operated by electromagnets on the track that would raise a flag when a train passed above them, their main purpose was to prevent collisions around blind curves, where a train operator couldn’t see. The lights that dot darkened tunnels today certainly didn’t exist then.
Enter the subway.
“It’s a totally different ballgame,” says Cunningham. “You’re running fully loaded trains, electrically powered, steel cars, a lot of momentum … You’re running them in very, very close headways. You’re running them through dark tunnels with blind curves.”
Train signals are vital in preventing collisions. Although train signals might look like stoplights, they don’t work quite the same way. There are effectively two types of basic train signalling: block signals and interlocking signals, both of which were built on this early subway.
Block signals are intended to prevent one train from rear-ending another. To do this, the track is divided up into “blocks.” Only one train ought to occupy one block at any given moment; if a train is in a block, the signal behind it isn’t clear. It will, for instance, display a red light, telling the train operator to stop and stay some distance away.
Interlocking signals, meanwhile, are the ones that are indeed like stoplights; they’re situated at junctions, or interlockings, where railway tracks merge or cross. They’re what prevent trains from colliding at angles. In the subway’s early days, interlocking signals were usually manually controlled by a local signal tower; the signalman, or signal operator, would flip the switch if a train was passing through the crossing.
These signals’ effectiveness largely relied on the train operator’s competence, but the early subway also had basic emergency stops. If a train went fast enough to trip a mechanism embedded in the track, it would activate the brakes.
Even with those advances, the original IRT’s signals were limited. Only express tracks had block signals; it was assumed that local trains wouldn’t need them since they stopped often enough.
In 1913, a second company joined the subway: the Brooklyn Rapid Transit Company (BRT).
The BRT already operated its own cobbled-together system of both old elevateds and even older rail surface lines, such as the Sea Beach Line and West End Line in Brooklyn that now carry the N and the D out to Coney Island. Subways, elevateds, and surface rail typically had different signalling equipment.
Partly to corral these systems together, the city government in 1913 issued the so-called Dual Contracts. The IRT and BRT were each contracted not just to operate their existing lines, but also to build more subway — much more.
It was under the Dual Contracts in the 1910s and 1920s that the subway system underwent one of the largest expansions in its history. Much of the network as it exists today sprawled out from Manhattan like the tendrils of an underground organism. Hundreds of miles of trackage were built then.
“Trains were running closer and closer,” says Cunningham. “They were trying to squeeze more and more trains in.”
Trains began to bump into each other on local tracks — or worse. In 1918, a BRT train on what is now the Franklin Avenue Shuttle, then an elevated line that had no signals, jumped the tracks and splintered into the side of a tunnel near Prospect Park. Around a hundred people lost their lives, and the BRT went bankrupt, its assets transferred to the new Brooklyn-Manhattan Transit Corporation (BMT).
You now might expect the regulators to enter, stage left. But American transit policy as a whole was haphazard in the early 20th century. Funding and managing transit was a task left to individual state or city governments.
And the regulation of the Dual Contracts was a maze almost as labyrinthine as the subway under its purview. The New York City government shared in the funding and profits, but the subway itself was run by the Public Service Commission, the state government agency responsible for managing utilities.
“The regulatory agency of these companies went back and forth like a football,” says Cunningham.
It took until 1927 for those regulators to finally mandate that signals be installed on local tracks. The result, even as the Great Depression began to strain the operators’ finances, was a rush to implement signals on the local tracks.
Signals in this era, while still a far cry from their 21st-century counterparts, were growing more sophisticated by the year. Signals could now use electric pulses to flash information in the train equivalent of a car dashboard. And more than merely telling a train operator to stop or go, block signals could set a speed limit that could be lowered depending on track conditions ahead.
In 1932, a third system joined the BMT and IRT: the city-owned Independent Subway System (IND), intended to compete with the private lines and keep costs down. The IND had no previous infrastructure, instead building all-new subway lines from scratch.
And the IND came with quirks. For murky reasons, the IND ensured that every piece of its infrastructure was split between two suppliers. That wouldn’t be unusual, but it extended to anal-retentive levels. The equipment on each IND train, for instance, was split down the middle between two manufacturers.
This same setup extended to the signals. Each of the IND’s lines was split between signals from two different operators. Although the equipment was compatible with each other, and the awkwardness didn’t impede the actual train operations, it made maintenance rather cumbersome. “I’m sure the signal repair crews probably cursed it,” Cunningham says.
In 1940, with the Great Depression taking its toll on the private lines, the three systems — IRT, BMT, IND — were merged into a single system, under the precursor to the modern MTA.
The unified system had multiple different types of signals and infrastructure in place. The IRT, for instance, had a narrower track gauge — and therefore narrower trains — than the BMT and IND. You’ll still notice this today: numbered subway trains, the successors to the IRT, are less spacious than lettered trains.
It also had different types of signalling. And by this time, much of the original signals — those that hadn’t been upgraded or implemented in the 1930s — was starting to age. “A lot of the parts had to be maintained and machined on the property, in the signal shop, because you couldn’t get old parts anymore,” says Cunningham.
At first, the subway was quick to address its own aging infrastructure. A new raft of upgrades began in the mid-1950s. Stations got state-of-the-art fluorescent lighting; station platforms were lengthened to fit longer, higher-capacity trains. And signalling began to drastically change.
Existing signalling followed old railway protocol, meaning each line had multiple signal towers, each responsible for a segment of the line. For instance, the IRT Flushing Line (now the 7) was split between around ten such towers by the 1940s. Those signal towers spoke with each other via telephone, a slow and inefficient method.
But by the post-World War II era, advances in communication technology meant that entire lines could be controlled from a single tower, significantly streamlining the whole process. The subway first trialled this in 1948, on one newly opened section of the IND Fulton Street Line: what is now the A/C between Rockaway Avenue and Euclid Avenue.
Afterward, the 1950s upgrades eliminated most of the original turn-of-the-century signals and replaced them with control towers for each line.
“It was a major engineering feat,” says Cunningham. “It was pulled off on time and under budget.”
As unbelievable as it may be today, this era was a time when the New York City Subway was at the forefront of technology.
It was an age when New York was filled to the brim with symbols of progress. The 1964 World’s Fair lit up Flushing Meadows with visions of the dawning Space Age. The United Nations turned the city into the world’s de facto political capital — for good measure.
In 1968, the city’s transport supremos published a “Program for Action,” an ambitious scheme to let the subway blossom with hundreds of miles of new trackage: a “superexpress” line in Queens; extended lines in Brooklyn; replacements for old elevateds in the Bronx; the Second Avenue Subway in eastern Manhattan; people movers, like the shuttles found in large airports, under major streets in Midtown.
It was that New York City which played host to perhaps the subway’s greatest experiment yet: to run a train automatically. And it was almost fitting that the testbed was one of the subway’s original lines: the 42nd Street Shuttle, the little train linking Times Square with Grand Central.
Trains with varying degrees of automation are common around the world today, but certainly not before the advent of miniaturized computing or network technology.
New York’s system instead relied on very 1960s technology. A room-sized computer in a dispatcher’s office controlled the whole operation. Circuits embedded in the track told the computer where the train was, and the computer responded with commands to the train: when to accelerate, what speed to take, and when to brake. Ultrasound sensors at Grand Central monitored the distance between the train and the end of the platform, ensuring the train was properly stationed.
The mayor appeared for the automated train’s first voyage in January 1962. It’s a trip of less than half a mile, but the whole experiment was a bold step forward. If it worked, it would herald the beginning of gradual automation for the whole system. It was a controversial proposal that would have cost eliminating thousands of jobs.
But the subway was pushing forth with plans to automate other lines: the Franklin Avenue Shuttle (S) in Brooklyn, the IRT Flushing Line (7), the BMT Canarsie Line (L). Those lines are ideal for such trials. They are used by a single service, running in a direct line, with no branches.
It was not to be. In April 1964, one of the automated trains derailed, short-circuiting the signal setup. Days later, a fire ripped through the Grand Central platform, destroying the automated train. Service was restored, but the automation experiment was over.
Of course, back above the surface, trouble was brewing. All those grand plans coincided with an accelerating exodus of the city’s white middle class into the suburbs. And with that migration came perhaps New York City transit’s greatest bane: the automobile. For several decades, Robert Moses — responsible for much of late 2oth-century New York’s infrastructure — had been hard at work disrupting deprived neighborhoods to build expressways right through them.
And even as the Program for Action was in planning, many of the old elevateds were coming down, one by one. Urban renewal proponents saw them not as transit, but as filthy eyesores that blotted out the sun. Elevateds over Third Avenue in the Bronx and on Myrtle Avenue in Brooklyn were demolished. They were never replaced.
The sun seemed to be setting on transit. As the 1970s ticked on, New York City found itself mired in a financial inferno. Thousands of public servants were laid off, President Gerald Ford infamously refused the city a bailout — and the subway was hit particularly hard. One by one, the Program for Action’s grand proposals were scaled back or binned entirely. The automation experiments were consigned to the past and long forgotten.
Meanwhile, transit around the world left New York in the dust. The initial section of the London Underground’s Victoria Line, opened in 1968, brought automatic train operation to a much larger scale. By the end of the 1970s, automation had come to Philadelphia’s PATCO Speedline and the nascent Montréal Métro.
Concurrently, the decaying New York City Subway became an almost legendary symbol of New York City’s decline. As Apollo landed on the Moon, as the personal computer began to proliferate through corporate offices, and as cable television spread its copper tendrils to reach American suburban homes, nobody bothered to spare New York City’s signals the funds for upgrades or even upkeep.
As this story would have it, over the following decades, the state-of-the-art 1950s signals and adjacent track equipment began to corrode.
On a summer night in 1991, a drunk train operator sped a southbound 4 train through Manhattan too fast for the signals’ emergency stops to prevent. Just before Union Square, it jumped the rails and tangled itself around the pillars separating the local and express tracks. Five people died, and hundreds more were injured.
Only then did the subway, barely starting to recover from the depths of the 1970s and 1980s, open its eyes to signals.
In the 1990s, the subway began to implement some of the old IRT lines — 1, 2, 3, 4, 5, 6 — with a technology called automatic train supervision (ATS). ATS is not to be confused with automatic train operation, such as the 42nd Street experiment. It’s more like a backup system that can automatically slow and stop trains as necessary.
ATS is far more sophisticated than prior crude emergency stops. Rather than harshly pulling an emergency brake, ATS can gracefully slow a train down over several signal blocks. To do this, ATS needs to hear from signals en route. If a passing train speeds through a signal above a signal’s adjustable maximum speed, ATS can sound the alarm. If that train continues speeding, ATS can automatically apply the brakes.
But it was hardly pioneering; by the 1990s, forms of ATS had been common for decades. It’s been near-ubiquitous on Japanese railways, for instance, since the 1960s. Even the Long Island Rail Road implemented it in the 1950s.
In any event, implementing ATS meant that the subway had to upgrade some of its other signals. Many of the signal towers consolidated even further, bringing them under the control of a single station on West 53rd Street in Manhattan. On ATS-equipped lines, it was now possible for that control center to know which trains are passing which signals.
Consequently, those upgrades are why you can stand on subway platforms and see a timing board approximating how long your train will take—if, perhaps, with the reliability of your local news meteorologist.
Even after it was planned, ATS was slow to come. In all, it took a decade and a half to fully implement it on those ex-IRT lines. Plans to implement it on the subway’s other lines were scuttled for reasons of high cost and time.
As the 21st century proceeds apace, the MTA seems to be mired in continual funding battles, not helped by the effects of the ongoing pandemic. New York City’s transit expansion plans seem hopelessly quaint in comparison to its counterparts around the world.
Singapore’s MRT boldly claims that, by 2030, eighty percent of the island’s households will be within 10 minutes walk of a train station. The venerable old Paris Métro ambitiously plans four new lines that will almost double its length. Some of Asia’s megacities — Seoul, Shanghai, Delhi, Taipei, the Pearl River Delta megalopolis — are in the midst of decades-long transit expansion sprees that show no sign of stopping.
Even London’s Crossrail, a partly underground line now several billion pounds over budget and plagued by high-profile delays, far eclipses New York’s slow and unsteady construction of the Second Avenue Subway. The days when the Program for Action mapped a bright future now seem further than ever.
But that doesn’t mean changes aren’t coming.
They’re being helped by a recent shift in how the U.S. federal government allocates its transit funds. “Previously … all you could really spend that money on was new extensions,” says Duncan Watry, a planner for Bay Area Rapid Transit (BART).
That federal money can make or break crucial infrastructure projects, but for an older system like New York’s, new extensions often aren’t in the cards.
Fortunately, Watry says the MTA — alongside its counterparts managing similarly antiquated systems in Boston and Philadelphia — helped change that. “The older systems all went to Congress and argued, ‘Wait a minute, we’ve got a lot of needs. We’ve got very crowded cities. We’re not building out into, you know, the farmland. We need money to improve the systems in the urban core.’”
In 2013, they succeeded in lobbying Congress to change that policy, allowing federal funding to be used for upgrades. That money is helping the New York City Subway embrace a new wave of signalling technology. That latest trend — not just in North America, but across the planet — is a move to communications-based train control (CBTC).
Think of CBTC as an evolved form of ATS. CBTC can pinpoint a train’s precise location in real time — and, crucially, pinpoint where other trains can run while keeping a safe distance. In other words, rather than a train moving through signal blocks, the signal blocks move with the train.
The result is that CBTC-equipped lines can run more trains, more efficiently, while staying just as safe — if not safer.
But implementing it is a complex endeavour. “CBTC is a complete infrastructure add-on that needs to be based on ground infrastructure as well as on-board train systems,” says José Soler, a professor at the Technical University of Denmark, who has helped implement CBTC on the Copenhagen S-train and other Danish railways.
CBTC is a thoroughly 21st-century technology. To know a train’s place on the track, that train must be in constant communication with its control systems. That requires the whole line be covered by a network; New York, like most CBTC-equipped systems, uses digital radio links. And if that network goes down, the train needs some sort of backup.
Soler compares CBTC to a home Wi-Fi network. “In principle, the technology is the same, so if your train security depends on it, wouldn’t you like to have … an alternative operational one?”
Additionally, CBTC needs trains that are equipped to handle it. The New York City Subway’s fleet, some of which dates to the 1970s, doesn’t always allow for that. Furthermore, the subway’s convoluted history has left it with a plethora of diverging lines, running independently of each other and slowing the progress it can make.
Compare BART, built to a single plan starting in the 1960s. Any BART train can run on any of its lines. Although BART has expanded in the decades since, the entire system’s technology is compatible with each other. Therefore, Watry says it’s possible to plan one suite of upgrades that would bring CBTC to all of BART’s system in one go.
New York City, meanwhile, has to upgrade its system piecemeal, line by painstaking line. First came a test on the IND Culver Line (carrying the F and G in outer Brooklyn). Then came the L train, in the early 2010s; the 7 followed, having recently finished its upgrade. Slowly, through the early 2020s, others will come: the IND Eighth Avenue Line (carrying the A, B, C, D and E in Manhattan) and the IND Queens Boulevard Line (which carries the E, F and M through, well, Queens).
The subway’s signals might be the bane of commuters’ existences; waiting for upgrades might seem like waiting for a download on a decades-old dial-up internet connection. But that’s only one way of looking at how signals are faring.
The slow pace of upgrades doesn’t change that New York is a pioneer in bringing CBTC technology to the United States. CBTC might be common in Europe and Asia, but amongst heavy rail systems in North America, it’s limited to BART, PATH (also in the New York area), the Panama Metro (whose first line only opened in 2014), a few commuter lines, and a smattering of airport people movers — none of which have the same reach as even a single New York City Subway line.
And it might be easy to know signals by the problems they cause, but it’s just as easy to forget that signals exist to keep trains from colliding, and to keep their riders and operators safe.
That, according to Cunningham, is something they’ve been extraordinarily successful at.
Of the “hundreds of billions” of subway journeys since the system’s inception, only a handful have ended in mechanical disaster, he says. “You’re in more danger from walking out your front door.”
“The fact is, these old systems … they’re old, but they work.”
Interesting review though I have to make a few corrections. The automated train of 1962 did not use computers it had a simple punched film tape that initiated operational sequences. The device was a small work bench size machine originally designed to control industrial machines.There were no derailments recorded. The single three car train had been scheduled as an 18 month test that was extended to 30 at the request of component suppliers. It was due to end ten weeks later on June 30 when the unrelated fire damaged equipment on April 21.
The numbered lines and the lettered lines do not run on tracks of different gauges. In fact, equipment built for the numbered lines is used on work trains throughout the system. The curves on the numbered lines are tighter and the station platform areas narrower than those on the lettered lines, so lettered-line equipment cannot run on numbered lines.
Also, the train destroyed in the 1964 fire was not the automated one, but a conventional train on an adjacent track.
By and large, an excellent treatise on the signal system and the massive challenges facing it.