To Improve Subway Service, Get Rid of the Unnecessary Signals Slowing Trains Down

If timers are needed to safeguard against a clear and present danger, whether that be a curve, switch, or hill - keep them. If not, remove them.

Headways ain't what they used to be.
Headways ain't what they used to be.

Last week, Uday Schultz, a junior at St. Ann’s School, took the top prize at TransitCenter’s annual TransitSlam with a presentation about the MTA’s excessive use of signals that cap subway speeds, which he produced with his classmate, Ivan Specht. We’re pleased to present a blog-ified version of their show below.

For the first time in a generation, New Yorkers are giving up on the subway as severe delays mount and reliability plummets. The trains have simply ceased to be a predictable way to get around the city.

If we are to stop this mayhem, a massive rethink of how our transit system is operated, funded, and managed is needed. But there are also basic steps we can take in the short run to improve service. One of these steps is the rationalization of subway timers.

These signals, which limit train speeds, have been installed unnecessarily, lengthening trip times and hampering subway reliability.

To understand how these devices came to be such a problem, let’s start back in the 1990s. Two fatal train crashes that decade resulted from subways traveling at excessive speeds. In response, the National Transportation Safety Board urged the MTA to address the issue “by converting more automatic signals to grade time signals.”

Instead of simply regulating the distance between trains, grade time signals place strict speed limits on trains. The idea was to prevent crashes by adding these signals in areas where excessive speed could be deadly — near sharp curves, on steep hills, and at busy junctions.

A timer signal on the Williamsburg Bridge.
A timer signal on the Williamsburg Bridge.

Initially, that’s what the MTA did, and there was a logic to the placement of the new, speed-limiting signals. However, the agency did not stop there. The MTA kept adding timers, even to safe areas of track.

The effects are twofold. Most directly, these speed restrictions have lengthened subway trips. Take the 5 train, which serves the busiest subway corridor in the nation. In 2005, a trip from 180th Street to 149th Street took nine minutes. Today, it takes 11 minutes. Similarly, going from Grand Central to Brooklyn Bridge in 2005 took 10 minutes; today it takes 12.

It may not sound like much, but these small increases in travel time compound across the whole subway system, adding up to countless hours lost for riders.

More relevant to the current crisis, timer signals have made the system significantly less reliable. Before their proliferation, a late train could make up time by going just a bit faster. Today, all trains — late or not — have to constantly slow for these speed checkpoints. Trains that fall behind schedule stay behind, preserving service gaps, adding to crowding, and making commutes miserable.

In a city obsessed with time and speed, this unnecessary slowdown is unacceptable. To get subway service back to where it needs to be, something must be done about it.

If timers are needed to safeguard against a clear and present danger, whether that be a curve, switch, or hill — keep them. If not, remove them.

Removing unneeded timers will not solve every problem with the subway system. No one policy fix can do that. But it’s a clear opportunity to make the system perform significantly better. We would be foolish not to take it.

  • joe lee

    Except when they are single tracking thru the Rutgers Street tubes….

  • Anonymity

    That GO hasn’t happened in years. Again, it gets so little use that when CBTC gets installed, it’s slated to be removed.

  • Joe R.

    Note that we’re talking about travel time, not service level capacity. Most of the time the trains aren’t slowed by the throughput of the signal system but rather by the neutering and the delays opening/closing the doors. Ultimately, if a train takes longer to get across the line, you need more trainsets for the same level of service. You also tend to lose ridership with slower trains.

    If you’re just interested in the first 1100′ there is still a significant difference. I’m getting 29.5 seconds with the unneutered trains versus 34 seconds for the neutered ones. The bottom line here is the neutering adding many minutes to longer runs and also made it impossible for trains to make up time if they run late given that they could barely keep up with already padded schedules. It’s also no way to run a system. The T/Os are supposed to regulate train speed. If the MTA really feels they’re incapable of doing that, as the neutering and timers seem to indicate, then perhaps it’s time to either hire different people, or just automate the trains.

  • Kiwiwriter

    The timer relay signals I remember best are on the No. 7 train, from Times Square to Grand Central…basically, the R-33s on that line faced yellow signals that turned green at the very last second. I was once told why that existed, but I don’t remember why any more.

  • AlexB

    The 5 train schedule you linked to shows 8 or 9 minutes from Grand Central to Brooklyn Bridge, not 12 and not 10. The local takes 13 or 14 minutes per the 6 train schedule.

  • AlexB

    How do you have access to real time data of train runs?

  • sbauman

    How do you have access to real time data of train runs?

    The MTA allows anyone to download the feed the count down clocks use. Here’s a link to a page on their website:

    I requested and received a developer’s API key, when the sites that had been archiving the data stopped collecting it.

    The data is updated every 30 seconds but I collect it once per minute. There are 3 parts to the information: an updated schedule for trips currently running or about to run within 30 minutes; the “position” of each train; and any alerts. The “position: reading is the train’s next or current stop; the train’s status (in transit to; approaching; or stopped at) and the time the status was taken.

    That’s enough information to determine that the L trains gain on the published schedule rather quickly in the trip. Unfortunately, the L train feed does not include the final stop. Thus, I cannot document any conga lines as trains approach 8th Ave.

  • AlexB


  • sbauman

    I’m getting 29.5 seconds with the unneutered trains versus 34 seconds for the neutered ones

    What’s the basis for these numbers.

    My calculation is 30.4 seconds, as outlined in my reply to Mr. Wright, above. It’s 12 seconds to accelerate @ 2.5 mph/sec to 30 mph and then traveling for an additional 18.4 seconds at 30 mph.

    the neutering added many minutes to longer runs while also making it impossible for trains to make up time if they run late,

    Schedules should be based on operating trains at the maximum safe speed for each track section. By definition that precludes making up time by increasing speed on these track sections. The place to make up time is during dwell time. That’s padded to allow for load variation at the same station. Three of the four dwell time components don’t involve passenger flow: waiting for the doors to open; doors open after all passengers have crossed the door threshold and waiting to start within the station after the doors have closed.

    the doors should open automatically when a train enters a station once the speed drops below maybe 0.5 mph. I think in France they open the doors when the train is under walking speed.

    It’s been a while, since I’ve been to Paris. There’s was one important difference on the Metro and RER trains. Passengers opened the doors, train operators closed them. This eliminated the obvious safety hazard of the doors opening with the train moving on unsuspecting passengers.

  • Hux

    This article is great, but does not mention that the signal modifications often become miscalibrated… crews don’t trust the posted speeds so they go even slower, and some will go slower than others leading to gaps. Perhaps the source of many “crowding” delays (crowding is rarely an actual cause of delays… this is a subway, it’s supposed to have a lot of people, and normal crowding should be built into dwell times and the schedule.)

    Also, the signal modifications lengthen control lines, leading to less capacity and less resiliency. When there is an incident, there are bound to be more delays, especially due to better data over the years. Everything is slowing down, all the new flagging rules don’t help and most don’t even make things safer. They just slow the trains down so that there’s more time to get hit and less time to do work on the ROW. No wonder there’s less ridership in the system these days (AND more crowding… perhaps because of the crowding, there’s fewer people riding… )

    I’m surprised the NYT hasn’t reported at all on these issues. They just blame things on “crowding” and “not spending enough money”… even though we spend a fortune and waste most of it. If we don’t speed things up soon (or stop slowing down), this city will become unlivable.

  • Anonymity

    I have a feeling that he (?) is comparing rush trips. Take the 8:17 from Nereid — 10 mins between those points. Today, the 8:10 takes 12.

  • Hux

    Don’t forget the flagging rules, many of which are also unnecessary.

    The signal modifications were started due to the crashes in Union Square and on the Williamsburg Bridge. The flagging rules were changed due to fatalities during work on the ROW. These are separate long-term issues. When crews are working on the tracks, they literally bring out flags to tell trains to slow down, and ignore the existing signals. Flagging rules override any signal modifications, which already are slowing things down. The signals are not changed themselves during planned work, they are not connected to anything to tell them to change, unless it is CBTC, in which case the CBTC shuts off in a planned work zone and reverts to manual control, so the operator can follow the flags.

    I think they are starting to understand why delays matter and why speed is important, but I am not sure. Perhaps they only care because of the bad press, which is why they are focusing on preventing major incidents, since those tend to be the ones on the news.

  • Joe R.

    What’s the basis for these numbers?


    0 – 20 mph in 9 seconds (2.22 mph/sec average): 132 feet
    20 to 30 mph in 20 seconds (0.5 mph/sec average): 733 feet
    30 to 32 mph in ~5 seconds (0.4 mph/sec average): 227 feet

    34 seconds to go 1092 feet


    0 – 20 mph in 9 seconds (2.22 mph/sec average): 132 feet
    20 to 30 mph in 7 seconds (1.43 mph/sec average): 257 feet
    30 to 40 mph in 12 seconds (0.83 mph/sec average): 616 feet
    1.5 seconds to travel remaining 87 feet at 40 mph

    29.5 seconds to go 1092 feet

    Obviously a piecewise linear approximation using 1 mph increments would be more accurate but for our purposes this is accurate enough.

    Schedules should be based on operating trains at the maximum safe speed for each track section. By definition that precludes making up time by increasing speed on these track sections. The place to make up time is during dwell time.

    While true the only time there is a maximum safe speed is if the track section has curves or diverging switches. In the absence of either, there is no inherent reason to limit speed. Obviously equipment has design speed limits, but in practice those usually can’t be exceeded. So basically on straight track a T/O should be able to operate at up to design speed to make up time. In practice of course close spacing of stops may preclude being able to reach design speed. Obviously it’s a lot harder for locals to make up time then. Perhaps there should be an emergency acceleration feature which ramps the initial rate to 3 to 4 mph/sec if the train is late so as to allow it to make up time. You could also allow top speeds up to the design speed so expresses can make up time. For example, unneutered NTTs are currently capped at 55 mph. That’s the speed we would base schedules on. However, their design speed is ~70 mph. You might allow running at up to that speed if a train is late and it was safe to do so on a given section of track.

    This article is about artificially limiting speeds with timers on sections of straight track when there is no real reason for doing so. Neutering equipment has the same end result.

  • Joe R.

    They went through Woodhaven at ~55 mph before the trains were neutered. Now it’s closer to 45 mph. Woodhaven is actually the fastest point in the express run.

  • Hux

    Yes, the schedule is important and it is good to adhere to it. It accounts for demand, network (track, signal, terminal) capacity, fleet capacity, load factors, budget, route design (local, express, skip, length, headways), crew work rules, frequencies.

    If we change the schedule, we can increase reliability. Dispatchers and crews can adhere to something attainable. They can minimize the need to intervene, reduce travel time, crowding, and allow us to have better data. The schedule is the benchmark and allows us to identify capital needs.

    Many of our schedules are just wrong and out of date so dispatchers don’t follow them. They can’t make it on time. No wonder trains end up missing merges, short turning or skipping stations. We are not providing the scheduled service as it is, so while a schedule change on paper appears like a service cut (lengthening running times), maybe it can help improve reliability.

  • Hux

    We can also just get our signals and signage correct. Crews always tell us that the signals are not calibrated correctly or not even properly marked, so they go a lot slower than planned and some crews are more cautious than others, so they go even slower and then trains start bunching and causing overcrowding. This is simple stuff that is being ignored and there does not appear to be anyone specifically responsible for checking signals. Line superintendents, train service supervisors, crews… they are all supposed to be reporting these things and I suppose the culture is so bad here that people have given up on trying to improve things.

    Are there any incentives for fixing things or going faster? Crews get OT for going late and going faster can only get them in trouble if they go over a red signal. No one is going to say “go faster”, the unions will get involved, and even the maintenance crews get paid more for working slowly. To stay and get your pension, I think you need to learn to keep quiet and please your boss, like many other companies, so maybe bad news does not even reach top leadership. Or they are focused on messaging and spend their time appeasing the public and politicians, so they have little time to fix things. Or they are too far removed from reality. This can happen in other companies but the difference is that we are government so we don’t need to care about competitors driving us out of business if we don’t fix our problems and keep misstating facts, etc.

    Remember, we were not allowed to say the word “delay” a few years ago. They just wanted to cover things up and pretend delays didn’t matter. But that defeatist attitude, looking for excuses and short-term solutions, seems hard to address. It is easy for them to hear things they want to hear (i.e., blaming ridership) but hard for them to accept bad news. Often we are told not to even share bad news because it will harm our careers and reputation, apparently.

  • Hux

    Adding running time to the schedules appears as a service cut on paper, since we don’t have more cars and we need the ones we have to travel longer journeys. In reality, we already have so many delays and abandonments under our current schedule, that we are short turning a lot of trains and just not running them, so we have a service cut already.
    If we ran to the schedule and were on-time, we could meet merges.

    We need to change our procedures. Stop blaming ridership, start blaming ourselves. Crews are trained to go slower and to not trust the signals, since many of them are not even working properly. Before, if crew went over a red, they were not automatically punished… if it didn’t happen too often, leadership even apparently thought it was a good thing, it meant they were trying to meet the schedule, etc.

  • Hux

    The L is a great line. Even accounting for no merges and that it is relatively short, it still performs better than expected, due to CBTC. PATH also runs 24/7 in NYC, and it has clean stations. What is wrong with our management? Utica Avenue extension makes complete sense. No bedrock there like SAS, would be cheaper. Up-zone it. Increase the supply of housing. Tackle the Mayor’s affordability crisis and extend the subway. So simple. But I think De Blasio truly believes the ferry is going to help fix our problems. The ferry is great, but it is not going to put a dent in our problems. It’s annual ridership is what, the same as 1 subway train?

    I think if we had a leader — city or state — who cared, we’d be OK. We really care a lot about the Governor. If he asks for something, everything stops for him. After him, it’s the press and auditors.

  • Hux

    The system was not always a grade time railroad. Only since the 90s, has it become a problem. At the same time, emphasis on safety means no tolerance for going through signals. Crews are very careful especially because they don’t trust the signals. Management is dishonest, so are the crews. Takes so much effort to report a problem, it is easier to just ignore it until it reaches a crisis. No one wants to risk their career. Signals people will save face and swear that their signals work as designed. Crews swear that their estimates are inaccurate, miscalibrated, and not maintained. And they punish people who speed, rarely people who go slowly.

  • Hux

    Signals are SUPPOSED to make the system FASTER, not slower! The original IRT locals did not have signals… they were added so trains could go faster.

  • Rocketman

    So why have they been so slow to install CBTC?

    Mismanagement, mainly. Their excuse for why the subway has so much trash is that they run 24/7 service… but the PATH is 24/7 and they’re clean. I’m not sure, but I think they may contract out cleaning… at least of the platform. Their station cleaners sometimes just blow the garbage on the tracks for the track cleaners to get (different division), and their train cleaners at the terminals sometimes also throw garbage onto the tracks. Many employees throw their garbage on the tracks too. Their big track fire last summer was actually due to an employee braking incorrectly, causing the train to create sparks. The A train derailment was caused by employees storing maintenance equipment incorrectly on the tracks.

    The signal modifications are especially problematic because they not only lengthen the distance between trains, but often force the trains to go slower by timing themselves between signal circuit blocks. It’s rather complicated, but sometimes they become miscalibrated, and clear at less than the posted speed, so crews don’t trust them and some will go even slower than others, creating variability. Many of these are necessary, but many more aren’t, and they’re all due to a crash on the Williamsburg Bridge in the 1990s, which was caused because they lengthened braking distances on trains in order to protect the wheels – that’s a whole other topic, but basically, the trains themselves are also slower now, they accelerate and brake slower. They continue to add more modifications, more flagging rules, slowing down even more. And since they’re slowing down so much, they need more trains, more crews, and they’re teaching everyone new to go slower.

    Why don’t they just update the schedules? Theyll, hopefully they don’t, until they speed things up… but they also don’t because it would appear as a service cut on paper, since they’d need more trains to meet the lengthened running times, and they don’t have the fleet. There’s serious talk about using the museum trains in regular service.

    I can go on and on… but really, point is, most of these problems are not due “decades of deferred maintenance”, “a century-old system”, “heavy ridership”, and all the other excuses that don’t seem to impact London, Paris, etc… and they probably spend just as much, if not more, on their operations and capital investments. Costs are so high here, there are no good excuses. Bedrock? Deep mining of stations? Union rules? Mismanagement? Corruption? So many excuses and myths, ignorance in the agency.

  • Aaron

    This article does not explain why the MTA continued to place additional timers at locations where there was no danger. The author’s desire to improve service is admirable, but the article states that the MTA continued to place timers at unnecessary locations despite facing repeated funding crises. Why would the MTA place unnecessary timers? It sounds like the MTA’s argument for why those timers should be there is missing.

    To me, it sounds similar to an argument to remove some fire extinguishers from an office building. The argument would go: sure, we don’t need quite so many fire extinguishers, and there are so few fires, we don’t really need them all. What we really need are just a few fire extinguishers strategically placed where there are likely to be fires. While this logic is appealing there is an obvious tradeoff in safety and relies on an assumption that we can correctly identify the areas that are most likely to require a fire extinguisher.

    I am no engineer and I don’t know anything about trains. I am just an average guy trying to make sense of this article that was posted a day after a major derailment. My assumption is that those commenters calling to remove these will likely to be the first to ask why these were removed in the future if there is a crash or derailment. Streetsblog covered the Amtrak derailment in the Bay Area and ended with this quote, “I hope the national media picks up on his story and that it reverberates in the halls of Congress and regulators and the boardrooms of the Class I railroads, all of whom are culpable for dragging their feet on making crash avoidance the cornerstone of U.S. railroad safety culture,” wrote Melzer in his post about his friend Hamre. “Then he and the other victims will not have died in vain.” (

    Though my guess is that the subway probably doesn’t qualify as a Class I railroad, the arguments are essentially that safety has not been a chief concern for trains. It even points to another derailment in NY in 2013 that was “eerily similar” in another article ( The similar derailment talks about a train taking a turn too quickly, which all sounds like the same problem to me and it is hard to understand advocating for the removal of a safety system. Most of the places where timers could be removed according to commenters were argued down by other commenters.

  • gustaajedrez

    I think there’s skeletonized track just south of 96th.

  • BronxSteve

    It often made the full Astoria-Bay Ridge run in an hour flat. When I took it full-length for my draft physical at Fort Hamilton in 1968, that’s how long it took.

  • Anonymity

    The logical conclusion of your line of thought is fire extinguisher wallpaper. And FWIW, extinguishers are placed only in strategic areas — near stairwells, in kitchens, etc. That’s where they’re most useful, and most likely to prevent accidents.

    Look, the Amtrak crash was terrible, but a knee jerk isn’t the correct response. Railroads actually haven’t been dragging their feet on PTC. That’s another blatant media oversimplification of a highly complex issue. If anything, the government has. The FCC is slow to grant the railroads the airwave bands needed for PTC, local authorities are slow to approve PTC installations, and the technology — mandated by Congress to be installed by 2015 in 2008, at which point it was nascent at best — has been developed into a viable technology only in the last couple of years. Timers are the same issue. If you read the comments below, you’ll see that the lack of speed regulation actually didn’t cause the crash cited as the reason for their installation — it was bad braking systems. The MTA jerked.

    And as for the locations, you’ll see 2 of them weren’t argued down, with the 3rd still only questionably necessary. You assume the MTA has a reason for timer installation, but that’s giving them too much credit. Take Bway to Euclid on the A. There are 2 curves there, both timed, but beyond them, the timers continue onto straight track. All the way to Euclid. That’s 3 local stops of useless timing. If we are to have a functional subway, that can’t be the way things work. I’m sure you could furnish something that those timers are preventing, but the fact of the matter is that they serve no truly discernable purpose. It’s those situations where timers must be removed.

  • Joe R.

    One thing to add to your excellent post is the fact even “crowding” is the MTA’s fault. Crowding could be mitigated or eliminated with reliable, frequent service. And one benefit of speeding up trains is you can increase frequency of service with the same number of trainsets.

    The bottom line here is other subway systems operate reliably and safely without also constantly slowing trains to speeds I can ride my bicycle at. Maybe we need to stop the insular attitude the MTA has and emulate what works elsewhere. Better yet, just let the Japanese or Swiss run the system. Both seem to have keeping trains on a schedule down to a science.

  • Joe R.

    Yes, the logical end result of the WB incident would have been to increase the emergency brake rate back up to what it was with asbestos brake shoes. This would have meant increasing brake pressure and perhaps more maintenance, but it was a far better answer than the alternative of adding timers and neutering rolling stock. Much like traffic signals on surface streets, timers have mushroomed to the point most of them are in places where they serve no purpose. It seems we love to slow things down in this city, both above and below ground, for reasons I can’t begin to fathom.

  • Albert

    Holding trains for evenness is not improving WA. It adds time to the scheduled headway. It harms WA — and increases delays, too. Evenness is not the same thing as WA. Delays and WA go together, they are schedule-adherence metrics. Holding trains is important during incidents, to minimize crowding and disruption. Incidents will harm WA and OTP, naturally.

    The MTA needs to stop confusing customer experience indicators with operations. If they don’t like terminal delays, make enroute OTP. Keep to the schedule.

    The priority should be speed. If the priority is literally safety, they should just shut down the system.


Straphangers: Ancient Train Signals a Prime Culprit of Subway Delays

Has your subway been delayed recently? Blame New York City’s aging transit infrastructure, especially its outdated signal system. Then start fighting to make sure Albany fully funds the MTA’s next capital plan. A new report from the Straphangers Campaign shows just how prevalent signals failures are on the subway system. In 2011, the MTA sent […]