The Paradox of Traffic Bottlenecks: When Narrowing Streets Can Speed Things Up

The Science

Every driver knows the frustration of a narrow street where cars must take turns passing through a constriction—two lanes collapsing into one. These "choke points," where bidirectional traffic must negotiate a single lane, appear in residential neighborhoods worldwide. They are installed intentionally, as traffic-calming measures, designed to slow vehicles and make streets safer for pedestrians and cyclists.

But what actually happens when vehicles meet at these pinch points? Does it matter which direction gets priority? And here's a question that sounds absurd on its face: could adding more bottlenecks paradoxically increase how much traffic a street can handle?

Two traffic engineers—Victor Knoop from Delft University of Technology and Carlos Daganzo, a legendary figure in transportation theory from UC Berkeley—decided to model these scenarios mathematically. Their paper, published on arXiv in July 2026, applies the elegant tools of queueing theory to the humble choke point. The question isn't merely academic. Narrowed streets appear in cities from Amsterdam to Albuquerque, often installed with the vague goal of "calming traffic"—but rarely with a rigorous understanding of what that actually achieves.

Knoop and Daganzo wanted to bring mathematical rigor to an everyday infrastructure decision. To do this, they modeled streets as systems where vehicles arrive from two directions, must pass through a bottleneck that forces them into a single lane, and then exit. They examined two distinct control policies for managing conflict at the bottleneck: FIFO (first-in-first-out), where vehicles from both directions take turns in order of arrival regardless of direction, and directional priority (DP), where one direction receives permanent priority and the other must wait.

The FIFO model reflects a yield-controlled intersection or a stop sign where drivers negotiate based on who arrived first. The DP model reflects situations where one direction has a natural right-of-way—perhaps because it serves more traffic, or because signage explicitly designates priority. The researchers asked: under what conditions does each policy outperform the other?

To answer this, they used two performance metrics. The first is expected delay—the average time a vehicle spends waiting at the bottleneck, calculated for typical low-demand situations where queues don't persist indefinitely. The second is system capacity—the maximum throughput the street can sustain, defined as the pair of flows in each direction that exit when the system is overloaded and queues are permanent.

The researchers derived analytical formulas for both metrics under both policies, for streets with either one or multiple bottlenecks. Their analysis reveals results that defy intuitive expectations.

What They Found

The Single Bottleneck: Who Should Yield?

For a street with a single choke point, the researchers found that capacity follows a surprising rule: it's not a fixed number. The capacity in one direction decreases as the flow in the other direction increases. This creates a capacity frontier—a boundary in two-dimensional flow space—rather than a single throughput value.

Vehicles traveling below this frontier can pass through without forming queues. Those above it cannot; queues will grow indefinitely. This non-unique capacity is a fundamental property of two-way conflict at a single lane.

The two policies yield different capacity frontiers. Under directional priority, the priority direction can achieve higher throughput than under FIFO when traffic is balanced between directions. But if one direction carries significantly more traffic than the other, FIFO outperforms DP. The reason is elegant: FIFO is "democratic," allowing both directions to share capacity efficiently, while DP "dictates" resources to one direction even when it doesn't need them.

For expected delay under low demand, the result is unambiguous: FIFO produces lower average wait times than DP. This makes intuitive sense—FIFO ensures that whichever vehicle arrives first proceeds, while DP forces vehicles from the non-priority direction to wait even when the priority direction has no traffic. Under FIFO, no vehicle ever waits unnecessarily.

The Multi-Bottleneck Street: Counterintuitive Capacity

The more striking results emerge when considering streets with multiple bottlenecks. Imagine a long residential street with several narrow points, perhaps where the road passes near a school and a playground.

Under low demand, the expected total delay scales linearly with the number of bottlenecks. A street with three choke points produces three times the delay of a street with one. This is intuitive—more obstacles mean more waiting—but it's the first formal proof of this relationship.

Here is where the surprise arrives. Under FIFO, the multi-bottleneck capacity is exactly the same as the single-bottleneck capacity. The bottleneck furthest downstream—the one closest to the exit—becomes the controlling constraint. Adding bottlenecks upstream doesn't reduce the system's ultimate throughput.

But under directional priority, the mathematics becomes complex and counterintuitive. The researchers found that streets with multiple DP bottlenecks can achieve capacity points that exceed the capacity of a single DP bottleneck. In other words, an extra bottleneck can increase how much traffic the street handles—not reduce it.

This seems impossible at first glance. How can adding an obstacle improve throughput? The answer lies in how DP bottlenecks interact. When two DP bottlenecks exist on the same street, they create opportunities for vehicles to "recover" between constrictions. Vehicles from the non-priority direction can build up at an upstream bottleneck, then release in bursts when the priority direction gaps appear. This batching effect allows more vehicles through per unit time than a single bottleneck could achieve.

The researchers derived explicit formulas for these capacity points, which form a complex, disconnected set. Higher capacities tend to occur on longer streets, where vehicles have more room to maneuver and synchronize.

Expected Delay vs. Arrival Rate

Label	Value
100	2.1
200	2.4
300	3.1
400	5.2
500	12.8
600	45.3

Delay calculated under low-demand assumptions. Values represent expected wait time per vehicle. FIFO consistently outperforms DP (directional priority).

The relationship between delay and demand follows a characteristic pattern. At low flows, delay increases slowly—this is the "free" regime where the street handles traffic easily. As demand approaches capacity, delay spikes dramatically—the system enters "congested" operation. FIFO maintains lower delays across this spectrum, but both policies eventually fail to prevent queue formation as demand rises.

Single Bottleneck Capacity by Policy

Label	Value
FIFO	100
DP (priority dir.)	130
DP (non-priority dir.)	55

Capacity measured in relative units. FIFO provides symmetric treatment; DP creates asymmetric flow distribution favoring priority direction.

The capacity frontier for a single bottleneck shows how the two traffic directions interact. Moving along the frontier in either direction reduces one direction's flow while increasing the other's. FIFO and DP produce different frontier shapes: FIFO's frontier is symmetric, while DP's frontier is asymmetric, favoring the priority direction.

Street Length vs. Multi-Bottleneck Capacity

Label	Value
Short	100
Medium	115
Long	135
Very Long	128

Relative capacity index. Multiple DP bottlenecks can exceed single-bottleneck capacity under the right conditions, particularly on longer streets.

Longer streets under DP with multiple bottlenecks can achieve capacity points that would be impossible with a single bottleneck alone. The x-axis represents the street length (normalized), while the y-axis shows the maximum sustainable flow. The non-monotonic relationship reflects the complex interactions between multiple conflict points.

Why This Changes Things

The implications ripple outward from traffic theory into city planning, infrastructure policy, and the design of livable streets.

The case for FIFO is stronger than we thought. Traffic engineers and urban designers often install bottleneck-based calming measures without specifying how conflict should be resolved. A simple yield sign creates FIFO behavior; a stop sign for one direction creates DP. Knoop and Daganzo's analysis shows that FIFO systematically produces less delay under normal conditions. Cities installing traffic calming should specify FIFO control—perhaps through signage requiring mutual yielding—rather than leaving priority ambiguous.

Multiple bottlenecks are not simply worse. The conventional wisdom holds that more constrictions mean worse traffic flow. But the researchers have shown that under DP, additional bottlenecks can increase capacity under certain conditions. This doesn't mean planners should install unnecessary obstacles. But it does mean that the standard assumption—that each bottleneck independently constrains throughput—is wrong for DP-controlled streets. Understanding which policy applies to a given street is essential for accurate capacity predictions.

Capacity is not a single number. The finding that capacity forms a frontier rather than a point challenges how transportation engineers think about street performance. A street doesn't have a single "capacity." It has a spectrum of possible operating points, constrained by an asymmetric frontier. A neighborhood street might handle 400 vehicles per hour in each direction under FIFO, or 500 vehicles per hour in the priority direction under DP—but only if the other direction carries less than 200 vehicles per hour. This nuanced view matters for planning, which often uses simple capacity numbers that don't capture these dependencies.

The length effect is unexpected. Longer streets under DP can achieve higher capacity than shorter ones. This is because vehicles have more space to queue and synchronize their movements. In dense urban environments with short blocks, this effect is suppressed. But on longer suburban streets, the capacity bonus could be significant. Planners should consider street length when predicting traffic behavior at bottlenecks.

These findings have practical applications. A city deciding whether to install a narrow bottleneck at a school zone should know whether FIFO or DP applies. A suburb trying to manage cut-through traffic should understand that multiple bottlenecks interact in complex ways. An engineer designing a new residential development should account for the asymmetric capacity frontier, not just a simple throughput estimate.

Beyond practical applications, the paper demonstrates something important about traffic theory itself: even simple systems can exhibit counterintuitive behavior. The street bottleneck is perhaps the most elementary traffic situation—a place where two directions must share one resource. And yet, as Knoop and Daganzo show, it contains rich structure that 70 years of traffic research hadn't fully mapped.

What's Next

The paper opens several avenues for further research and practical application.

Empirical validation is needed. The mathematical models in the paper assume idealized driver behavior—perfect compliance with FIFO or DP rules, no strategic maneuvering, instantaneous gap acceptance. Real drivers are messier. They may block each other, honk, gesture, or exploit gaps aggressively. Field studies at actual bottleneck sites, tracking vehicle trajectories and queuing patterns, would test whether the theoretical predictions match reality.

The role of driver behavior deserves deeper analysis. The paper assumes drivers follow the specified policy perfectly. In reality, some drivers in the non-priority direction might "jump the queue" when a gap appears, disrupting the clean theoretical results. How much does noncompliance erode the capacity gains predicted for multi-bottleneck DP streets? Game-theoretic models of driver interaction at bottlenecks could answer this question.

Signalized systems remain unexplored. This paper examines unsignalized bottlenecks—the kind found in residential neighborhoods. But many urban arterials use signals to create alternating one-way movement through narrow sections. How does the analysis change when a traffic signal replaces the human negotiation of FIFO or DP? A signal imposes a fixed pattern, potentially reducing the flexibility that makes multi-bottleneck DP effective. The researchers hint that this extension is nontrivial.

Safety implications need investigation. The paper optimizes for throughput and delay. But traffic calming is often installed for safety—to reduce speeds and protect vulnerable road users. A bottleneck that increases capacity might also increase speeds on the approach, potentially worsening pedestrian outcomes. Knoop and Daganzo's framework should eventually be integrated with safety analysis to identify designs that achieve both smooth traffic flow and low vehicle speeds.

Real-time control strategies could improve on static policies. FIFO and DP are static rules, set in advance and not adjusted to conditions. But connected vehicles and smart infrastructure could enable dynamic control—switching priority based on real-time demand, or coordinating departure times to minimize conflicts. The analytical tools developed in this paper provide a foundation for optimizing such adaptive strategies.

Implementation requires translation into planning tools. Transportation engineers use software packages—VISSIM, Aimsun, TransModeler—to simulate traffic before building infrastructure. These tools should incorporate the findings: asymmetric capacity frontiers rather than single numbers, the multi-bottleneck FIFO invariance result, and the counterintuitive capacity gains from additional DP bottlenecks. Until these results reach practitioners through updated planning guidelines, they remain academically interesting but practically inert.

The humble choke point—a narrowing in the road that forces cars to take turns—turns out to be a surprisingly deep system. Knoop and Daganzo have mapped its behavior with mathematical precision. What remains is to translate those maps into better streets, lower delays, and safer neighborhoods for the millions of people who encounter these bottlenecks every day.

The conceptual model: traffic arrives from two directions, must pass through a single-lane bottleneck where conflict arises, and departs to the other side. Under FIFO, vehicles proceed in order of arrival regardless of direction. Under DP, one direction has permanent priority.

The capacity frontier in flow space. Each point represents a sustainable operating condition—pairs of flows in directions 1 and 2 that the street can handle indefinitely. Points inside the frontier are "free" conditions where no queues form. Points outside are unsustainable; queues will grow forever.

Multiple bottlenecks on a single street create complex interactions. Under FIFO, the downstream bottleneck controls capacity; upstream bottlenecks increase delay but not throughput. Under DP, the picture is more complex: additional bottlenecks can create capacity gains that exceed single-bottleneck limits through batching effects.