Previously I covered some of the history of the traffic signals in service in the area, as well as how to recognize them in the field. Due to length I didn’t really have time to detour into more of the general history of signals, so I’m taking a chance to do so now.
The Early Days
There’s not really a distinct “first traffic signal”, it was more an evolution of early mechanical aids to a policeman directing traffic. These involved rotatable stop/go signs or various semaphores and arrows up to elaborate structures resembling railroad control towers. Some of these were naturally illuminated to aid visibility at night. As such there’s various conflicting stories about the first “real” traffic signal, but one claim is a device that was built by policeman Lester Wire in Salt Lake City in 1912. It resembled a birdhouse and used standard 5-1/2″ red and green railroad lenses (it was eventually discarded in the 1960s by the museum that had it). From this, it wasn’t too much of a leap to realize that you could install a mechanical timer and have it run automatically. The first three color traffic signal was installed in Detroit in 1920 by William Potts, but the sequence was originally different: the yellow would flash briefly to warn of an impending change between red and green. Railroad lenses proved suboptimal, being both too small and too focused to work good for city streets. To address the problem, the lenses were increased in size to 8”, and designed to spread light more.

Elaborate manually operated signalling tower, this is one of the later ones. By this time they had dispensed with signs and flags and were using lights exclusively,
Replica of the first traffic light (although this appears to use all red lenses). Utah Department of Transportation, CC License.
From the 1950s on, even the 8” lenses were becoming too small for the higher speed, higher volume, wider roads becoming in use, so the 12” version was introduced. (And of course signals companies were eager to sell bigger, more expensive traffic signals. The Crouse-Hinds “type K”, which had 12” red lenses, was the first. Arrows were hard to see in 8” so they quickly increase in size, the green and yellow balls were slower to be changed out. To this day 8” indications are permitted where the speed limit it 30 mph or less, and some agencies, including New York, take advantage of that.
Why Red, Yellow, and Green?
These are copied from railroad signaling. Red is the obvious choice for “stop”, since being the color of blood it has meant “danger” since time immemorial. The original choice for “caution” was green, and “go” was white. (I’d speculate green was chosen at random because it was noticeably different from red, and a green filter was more efficient than say blue or purple. Probably they just didn’t hire colorblind engineers). The obvious problem is you have a “fail deadly” situation if the red filter doesn’t engage or the lens falls off, which supposedly actually happened around 1914 causing a crash. A second problem was even back then there were a lot of white lights around making it hard to tell a “go” signal from a light in a barn. When the colors were reassigned, they kept green and made it “go”, and added yellow as “caution”. The meaning of those colors entered culture because of signal lights, not the other way around.
Two section vs Three section
The yellow “caution” indication was an option almost from the beginning, but with slower and lighter traffic in those days some agencies decided it was an unnecessary expense, and dispensed with it. Notably New York City primarily installed two section displays up to 1952 (In defiance of a federal ban in 1935) , some of which persisted into the 1980s. Sometimes the signal would go dark to signal a phase change, but later both lights would illuminate, and red and green blended together to form yellow farther back. Perhaps the ultimate in cheap was the two lamp Darley. There was one lamp in each section, with red on top for the main street and green on top for the side street. Each lamp was projected in all four directions, so the top section would show red for the main street and green for the side street simultaneously. A simple three position controller was included in the light, so all the city had to do was hang it and hook up mains power.
The Acme, Wiley, and Mercury Signals.
Los Angeles, San Francisco, and New York each had unique designs specific to that city. Los Angeles had the Acmes, which in addition to the lights had a bell and the stop/go flags which would come out of the housing- you see these in old Hollywood movies and cartoons. Wiley “birdcage” signals were in San Francisco, and had lighted red and green “stop” and “go” panels that rotate around. New York had the Mercury signals, with conventional red and green lights, but made of bronze, very ornate, and with a statue of the goddess Mercury on top. The Mercury signals were all destroyed by New York City, although a few of the statues survived. There are a few Acmes and Wileys around in private collections, and they normally fetch 5 figures when sold.

Acme Traffic Signal. Late at night the large red and green lights would go out, and the small red light in the stop sign would flash indicating a four-way stop. Alan Weeks, Metro Library and Archive, CC License

Wiley “Birdcage” signal. the top was internally lit, and a rotating drum would change each side between stop and go. Metro Library and Archive, CC License
Traffic Signal Lamps
Signals always have and most still do run on mains electricity (although some of the very newest ones run on 48 volts DC). The lamps are specialized incandescent lamps optimized for long life and an even light dispersion rather than efficiency. 8” and 9″ signals would use 69 watt lamps, 12” and 16″ would normally use 116-150 watt versions, although often yellow balls, which pass more light through the lens than other colors, would use 69 watts.

Traffic signal bulb. The filament is supported in a way that throws light down towards the reflector. Not all of them have this silvered band around the middle which helps the lens light more evenly.
Paint Colors
Originally signals tended to be black, green, or bare aluminum. The yellow as used here and other areas comes from a 1950s Manual of Uniform Traffic Control Devices recommendation that was eventually removed. Yellow makes the overall installation more visible, but black provides more contrast making the indications easier to see. Minnesota used the best of both colors by using yellow bodies and black doors. Poles here were also traditionally painted yellow. Mn/DOT and most counties have recently switched from yellow aluminum signals and painted steel poles to black polycarbonate signals and galvanized steel poles that need no paint.
The Future
Even the new LED indications mimic the same basic three “lamp” configuration that’s more than 100 years old even with the different abilities of LEDs. Other nations have been more adventurous- Asia has animated pedestrian “mans” that speed up as the time counts down. For vehicles they horizontal bars that light up red, yellow, or green that as needed, and shrink as the time expires. In Germany pedestrians can even play a game on the push-buttons with the person across the street. One proposal for the US that made it into prototype was the “Unilight” that used a single section to light up a red square, yellow diamond, or green circle. There’s been various proposals that would give drivers feedback as to how much red and green time they have left (presumably for pre-timed installations only). And perhaps with self-driving cars the overhead installations will go away entirely and the controller will talk to the cars directly and tell them they need to stop, the very few people still driving “manual ” cars will need an adapter that lights up an indication inside their car.
This concludes the post, you may have noticed there’s nothing about lenses, left turns and pedestrian signals. These are long stories in themselves and I hope to write about them individually in future articles.
One thing I’d add: I’ve seen several East Coast jurisdictions (most notably in New Jersey) utilizing LEDs, taking what’s physically a 4-lens signal but operating it as a 5-lens (with the green and yellow arrows sharing the same lens).
Yes, I plan to cover bimodal fiber optic and LED indications later in an article on left turns or lenses / LED module.
Not only do I vividly remember Darleys around the mid-Atlantic, I still know where a few can be found in operation!
Also, I’ve seen signals with motorist-oriented countdowns in use at a couple spots in Eastern Europe — particularly in Russia.
Despite their “cheap” origins, Darleys, particularly the one lamp per section (they made a more conventional design later) are highly desired by collectors due to their uniqueness; they usually sell for $1000-$2000 depending on configuration and condition.
Do you have any information on the signals in Minneapolis that were mounted on springs right in the middle of the intersection?
Ihttp://collections.mnhs.org/cms/display.php?irn=10720682
Not this specifically. One thing I’ve tried to get at is that in the early days before mandatory standardization there were all sorts of signalling devices that ranged from practical to impractical, and clever to wacky. Some of this stuff exists only in patents, some only had a few prototypes made, and some was used widely on a local basis. This persisted later for left turns and pedestrian signals due to those coming later in general. And now we’re starting to see it again in areas of the world without rigid standards, now that with LEDs the paradigm of the bulb/reflector/lens is broken.
“And perhaps with self-driving cars the overhead installations will go away entirely and the controller will talk to the cars directly and tell them they need to stop, the very few people still driving ‘manual’ cars will need an adapter that lights up an indication inside their car.”
And cyclists are left to guess? 🙂
One thing I’ve really liked in recent years is the prevalence of video detection. Hennepin County now uses this on all new signals. Minneapolis uses it selectively, and Mn/DOT doesn’t appear to use them at all yet. (At least, haven’t spotted it myself.) This is much more reliable for detecting cars and bicycles alike.
I know similar technology exists for detecting pedestrians (live right now at 56th and 53rd at Lyndale), although it still doesn’t address the basic disadvantage that you can’t happen upon a walk signal or a countdown. If the signal is programmed to do 7 seconds walk signal and 30 seconds pedestrian clearance interval and you arrive with only 36 seconds left on the green, you’ll be ignored and forced to wait until another cycle.
Maybe everyone will have a smartphones or smart watches or bionic implants that will tell when it’s OK to go. Or at the minimum we could go back to a short pole on two kitty-corner sides of the intersection with 8″ lights, instead of the huge expensive overhead masts for non-motorized and “legacy” users. A 8″ LED indication only draws about 10 watts, and could draw even less if fault monitors were redesigned so they wouldn’t trip with less.
As far as video detection vs inductive loops, there’s not one that’s clearly better all the time. Loops go bad due to salt spray and time, need the street torn up to replace, and have trouble picking up bicycles. Video detection requires more sophisticated (=expensive) equipment, and is vulnerable to snow and rain on the lens, and will generate false calls if the wind blows the mast too hard (it sees what it’s focused on move relative to the frame). Video is getting a lot better, which is why you see more of it.
I deliberately excluded pedestrian related trivia here due to space (and I still went way over my 1000 word limit) and to devote a future article to it, but yes, I’m a big advocate of pedestrian sensors. The only downsides are the cost and integrating it into existing systems, there are upsides for both pedestrians and vehicles.
1) Pedestrians no longer have to push a button. I’m not sure people that object to it on a idealistic philosophical ground exist much beyond this and similar blogs, but I do get that sometimes snow removal is less than optimal, and they’re difficult to push when mounted on a bicycle.
2) Solutions to the above that severely degrade vehicle performance by always giving a walk signal, even if it’s just a single car on narrow side street waiting to cross a wide suburban road, or even giving a walk phase periodically with no vehicle calls just in case their might be a pedestrian there, aren’t necessary.
2) With additional sensors mounted on the crosswalk, you could extend the walk and clearance interval for slow elderly individuals. Or terminate it early and let vehicles go if a single young guy runs across.
“Acme Traffic Signal. Late at night the large red and green lights would go out, and the small red light in the stop sign would flash indicating a four-way stop.”
I want this back! I’m so tired of sitting at red lights when there’s no traffic. I’m not the only one.
Madison, WI does this still (and perhaps other Wisconsin cities?). The major streets gets a flashing yellow, and the minor street a flashing red. Makes a world of sense for late night.
Minnesota engineers are more than most oriented towards safety rather than efficiency. Witness also the number of protected only left turns here compared with other areas.
Minnesota also has rather stringent criteria on protected vs protected-permitted lefts. Though believe it or not, they’re not as bad as New Hampshire, where practically every left is either permitted-only or protected-only (I know of only two protected-permitted signals in the whole state).