On the evening of December 10, 1868, a revolving gas-lit semaphore arm was erected outside the Houses of Parliament on Bridge Street in London. It was operated manually by a police constable who stood beside it, turning a lever to raise or lower the semaphore arm and rotating a lantern of red and green lights for visibility after dark. Within a month, the gas supply leaked and the lantern exploded, injuring the constable on duty. The signal was removed and not replaced.
For the next forty-five years, traffic at busy intersections was managed by human police officers, hand signals, and the informal negotiated chaos of horse-drawn vehicles, bicycles, and — increasingly — early automobiles sharing the same streets with no coordination mechanism beyond social convention and mutual fear.
The Electric Reinvention
The gap between Knight's exploding semaphore and the electric traffic signal is not a story of slow progress. It's a story of waiting for the right preconditions: the automobile, electric power infrastructure, and the bureaucratic will to standardize.
In 1912, Lester Wire, a police officer in Salt Lake City, Utah, constructed what is generally recognized as the first electric traffic signal. It had two colors — red and green — and was mounted on a pole at an intersection. Wire never patented his design.
Two years later, in 1914, James Hoge filed a patent for a system of four interconnected traffic signals in Cleveland, Ohio. The signals were controlled from a single police booth via electrical switching, which meant one officer could coordinate an entire intersection rather than standing in the street. This was the first interconnected system — a meaningful step from individual signal to network.
The Yellow Light
Both Wire's and Hoge's systems had a critical limitation: binary state. Red or green, stop or go, with no warning transition. Drivers discovered what every later traffic engineer would formalize: the interval between states is where accidents happen.
In 1920, William Potts, a police officer in Detroit, added the amber light. Potts was managing an intersection where the binary stop/go signal was causing collisions as drivers tried to race through on the late green — and got caught by the immediate red on the cross street. The yellow light created the warning phase: the system's acknowledgment that stopping takes time and that a transition state needed to be visible.
The three-phase signal is, at its core, a UX insight. Binary states are cognitively simple but operationally incomplete. The transition state costs time but prevents accidents. Potts arrived at this through traffic management, not engineering theory.
Garrett Morgan and the Patent
In 1923, Garrett Morgan — the same inventor who had developed the safety hood (an early gas mask used in the Triangle Shirtwaist Factory fire rescue and later adopted by fire departments) — patented a different three-position traffic signal. Morgan's design included a "caution" position with an extended arm, distinct from the stop-and-go positions, intended to halt all traffic before the cross-street green.
Morgan's approach to the transition problem was structural rather than chromatic: halt everything, then release. He sold the patent to General Electric. Morgan was Black, and navigating the patent system and commercial negotiation during the Jim Crow era required not only technical skill but deliberate strategy around when to reveal his identity to potential buyers — a dimension of his story that tends to be omitted in the abbreviated versions.
Standardization
By the late 1920s, a proliferation of incompatible signals across American cities was creating its own problems. The 1935 Manual on Uniform Traffic Control Devices (MUTCD) standardized color order (red on top, green on bottom), signal timing ranges, and placement conventions. The placement order is itself a concession to colorblindness — someone who cannot distinguish red from green can still read the signal by position.
Adaptive systems arrived in the 1960s via inductive loop detectors embedded in pavement, measuring vehicle presence and adjusting cycle timing accordingly. Camera-based detection followed in the 1980s and 1990s. Machine-learning optimization of signal timing across networks is a current research area, though its deployment in real infrastructure has been slower than the research suggests it should be.
Three Observations
First: the forty-five-year gap between Knight's gas lantern and Wire's electric signal is instructive. The technology for an electric signal existed well before 1912. What didn't exist was a sufficient density of automobiles to make the problem urgent, electrical infrastructure ubiquitous enough to power signals at scale, and institutional willingness in municipal governments to mandate and maintain them. Technology readiness and institutional readiness are different clocks.
Second: the three-phase signal took fifty years to emerge after the two-phase signal. This isn't because the warning interval was a subtle insight — it's because binary systems feel complete until you accumulate enough accident data to see the gap. The transition state is always the last thing added.
Third: Morgan's patent story runs alongside the mainstream account without quite joining it. General Electric acquired it, the signal type spread, and the inventor's name was not attached. This is a recurring pattern in the history of invention — the gap between who creates a thing and whose name travels with it.
Read more at anethoth.com — or explore indie SaaS projects at builds.anethoth.com.