Programming Note: I started a Substack! This very same post is there now, and in the future I will be posting “opinion” content like this over there.
All opinions expressed are my own and not representative of my employer.
In July 2026, the MTA published a report titled “Eliminating Bottlenecks: Speeding Up Your Ride”. I want to investigate the claims made by the MTA in this report, their identification of existing bottlenecks, their proposed solutions, and add my own thoughts as to how some of these bottlenecks may be eliminated. I will be quoting extensively from the report, which you should read for yourself here.
The report gives the following list of the types of bottlenecks in the system:
- Interlockings where multiple lines merge
- Sharp curves that force trains to slow down
- Terminal designs that constrain our ability to turn trains around, and
- Legacy signals that require large safety buffers between trains.
The legacy signal system is in the process of being upgraded. Communications-Based Train Control, or CBTC, is an upgraded signal system that allows trains to safely run closer together, which increases speed and capacity. CBTC has been or is being installed on many subway lines around the city, and this work will likely continue until the whole system has been upgraded. However, as the Executive Summary of the report states, “signal modernization is necessary but not sufficient by itself…many bottlenecks require physical changes to track layouts, terminals, and interlockings to unlock additional gains.”
Sharp curves are difficult to eliminate, as there is often not enough space available to easily realign the track. The curves are built into the infrastructure of the subway, and that infrastructure is not easily changed without significant construction projects. The speed limit through tough track geometry may be increased slightly with CBTC signaling, but the two best options for true elimination of these bottlenecks is either significant construction and realignment, or to not run trains through them.
Terminal designs are a major problem in various parts of the system, but like sharp curves, require significant construction to eliminate. On lines with terminal constraints, it is generally best to eliminate as many other bottlenecks as possible, so that the terminal is the only constraint on capacity, and can be used to its fullest extent. At that point, investing in eliminating the terminal bottleneck can be considered as a way to boost capacity on the whole line.
This leaves interlockings, in my opinion the most interesting type of bottleneck to solve, which will be discussed in more depth in the rest of this post. To be clear, since multiple terms are used in the report, merging and branching conflicts at interlockings are the same type of bottleneck. Every interlocking where two lines branch is an interlocking where those same two lines merge in the opposite direction. My high school English teacher told me never to quote long sections of text, but I am going to break that rule here because the author(s) of this report put in a very sober paragraph reflecting on the fundamental root cause of interlocking merge bottlenecks in the system, and I want their words to speak for themselves:
“Another legacy of the way our system was planned which contributes to bottlenecks is heavy reliance on branching: trunk lines that split in multiple directions, allowing trains on one line to reach multiple destinations. Because of branching, most subway routes share tracks with other routes for parts of their journeys. This structure provides choice, but makes the system sensitive to delays and disruptions. This is because any delayed train doesn’t just hold up trains on its own line, but on any other lines with which it’s sharing the track.”
So, of the four types of bottlenecks, the legacy signal system is in the slow but ongoing process of being eliminated, sharp curves are difficult to eliminate without significant construction or avoiding them completely, and terminal bottlenecks are not worth investing in while other bottlenecks on the line remain. So, how should interlocking bottlenecks be eliminated?
Let’s dive into the meat of the report, which highlights two of the most negatively impactful interlocking merge bottlenecks in the system: DeKalb Interlocking, on the Brooklyn side of the Manhattan Bridge, which impacts the (B)(D)(N) and (Q) trains, and Nostrand Junction (which as far as I am aware is usually called Rogers Junction), which impacts the (2)(3)(4) and (5) trains.
DeKalb Interlocking
The problem: “At DeKalb Interlocking in Downtown Brooklyn, the (B)(D) and (N)(Q) trains merge before re-splitting into two branches— one carrying the (D) and (N) and the other carrying the (B) and (Q). For Brooklyn riders, this means that many stations provide direct service to both the 6th Ave line, via the (B) or (D), and the Broadway line, via the (N) or (Q). However, the need for trains to merge at DeKalb Interlocking means that trains can be delayed waiting for another train to join the track ahead.”
The solution: The MTA will upgrade the signaling system to CBTC, which “will allow trains to safely travel faster and more closely spaced through the junction.” Additionally, the MTA will install new switches with a higher speed limit to the interlocking, which together “will help trains flow more smoothly and more reliably through the busy DeKalb Interlocking.”
Nostrand Junction
The problem: “Nostrand Junction is a busy interlocking where (2)(3)(4)(5) trains split into two branches serving different areas of Brooklyn. The service pattern through the junction with (2)(5) trains taking one branch and (3)(4) trains the other—allows riders along those corridors a one-seat ride to either the East or West Side of Manhattan. However, branching at Nostrand Junction causes congestion that results in delays on all four lines and constrains the number of trains that can be operated on these lines.”
The solution: “By adding new switches between the Franklin Av and Crown Hts-Utica Av stations, we will reroute (2)(3)(4)(5) service, streamlining their paths through Nostrand Junction to minimize merge conflicts. The new configuration also allows us to alter the service plan, with (2) and (3) trains taking the Nostrand Av line to Flatbush Av-Brooklyn College and (4) and (5) trains taking the Eastern Parkway line.”
What these two solutions reveal is an inconsistent methodology, at odds with the text of the report that they are found in. At DeKalb Interlocking, the problem is clearly stated as a textbook version of an interlocking where multiple lines merge, but the solution is presented as if the bottleneck is actually the legacy signaling system, and possibly sharp curves on the existing switches. Yet, at Nostrand Junction, where the problem is also stated to be a branching conflict, the solution that is presented is a merging solution: reroute the trains so that the merge conflict is actually eliminated.
An example of a merge conflict that has already been eliminated is mentioned in the report, as a success story. According to this report, the (F)(M) swap, which occurred in December, 2025 and eliminated a merging conflict at Queens Plaza, “has improved service on the (E)(F)(M)(R) lines, speeding up trains in key segments and reducing the frequency of delays.” In this case, like the proposed solution to Nostrand Junction, the routing of the trains was changed to eliminate the merge conflict completely, not merely rely on signal and switch upgrades to improve speeds.
In defense of the MTA, rerouting trains often comes with some political cost. Riders are used to their current trains, and changing them, especially if the change eliminates some one-seat rides, can often engender reflexive local opposition to the change. As a regular rider on the Queens Boulevard Line, I experienced this before the (F)(M) swap was implemented, as people in my neighborhood worried about how their commute might change after the swap. However, this report already clearly laid out the tradeoff: Branching and merging routes “provides choice, but makes the system sensitive to delays and disruptions.” It is therefore frustrating to see a report titled “Eliminating Bottlenecks” not actually aim to eliminate, but merely ameliorate the DeKalb Interlocking bottleneck. What makes the tradeoff of rerouting too steep to consider in the case of DeKalb Interlocking, but not too steep in the case of Nostrand Junction? The report doesn’t say.
The obvious solution to the DeKalb Interlocking bottleneck, which remains unsaid in this report, is to reroute the trains so that the (B)(D) and (N)(Q) trains do not merge with each other at this interlocking. Whether this means sending both (B) and (D) trains to the Brighton Beach Line and both (N) and (Q) trains to the 4th Avenue Line express, or vice versa, I won’t claim to have the answer. But any “solution” which continues to require merging at this location will never truly eliminate the bottleneck.
Secrets of the Subway 