For more than half a century, the Search for Extraterrestrial Intelligence (SETI) has looked for artificial signals from space using the communication technologies humans themselves rely on. Typical targets include narrowband radio waves and regular laser pulses, which are thought to rarely occur in nature. But there is no guarantee that an extraterrestrial civilization uses the same communication technology or the same way of representing information as humans. If we search only for signals that are easy for humans to understand, we may overlook a civilization that tries to announce its presence in a completely different way.
On September 7, 2026, a research team consisting of Cameron Brooks, Estelle Janin, Gage Siebert, Cole Mathis, Orit Peleg, and Sara Imari Walker brought a new perspective to this problem. They published a paper titled "A firefly-inspired model for detecting the alien" in the journal Scientific Reports. The team focused on how fireflies announce themselves to potential mates with distinctive flashing patterns so that they are not drowned out by the light of those around them. They applied this biological communication strategy as a mathematical model and devised a way to derive artificial signal structures that are easy to distinguish from natural cosmic emissions.
That said, this study has not discovered a signal from an extraterrestrial civilization. What the team carried out was a verification using a simplified model (a toy model) for proof of concept, combining data on known pulsars with mathematical optimization techniques. The authors themselves state clearly that this is a quantitative thought experiment, not a finished method capable of predicting real observational results. It is positioned not as a replacement for existing SETI, but as a new way of evaluating signals in order to narrow down which ones deserve closer examination.
Looking for differences from natural signals, not decoding messages
One of the difficulties in SETI is how to distinguish signals emitted by intelligent life from emissions produced by natural phenomena. Methods that search for mathematical sequences or artificial modulation patterns risk missing a signal if an unknown civilization uses an entirely different communication scheme. On the other hand, an extremely regular signal is not necessarily one emitted by intelligent life. The classic example is the pulsar, a rapidly rotating neutron star.
Pulsars are celestial objects that emit regular radio pulses as they rotate. When the first pulsar was discovered in 1967, its remarkably precise periodicity led researchers to consider the possibility that it was a signal sent by extraterrestrial intelligence. The source was nicknamed "LGM-1" (Little Green Men), but it was later shown to be a natural astronomical phenomenon. This history illustrates that regularity alone cannot separate artificial signals from natural ones.
In this study, the team placed more weight on how different a signal's structure is from surrounding natural signals than on understanding the meaning of the information it carries. Specifically, they compare candidates against the radio pulses of background pulsars and mathematically find the signal whose period and transmission timing overlap least with them. The idea is that even if the content cannot be decoded, a signal with characteristics unlike natural emissions can be a candidate for follow-up investigation.
Behind this idea lies an interest in the communication methods of non-human organisms. On Earth, a wide variety of creatures, including birds that sing, whales that produce songs, and insects that emit light, have developed ways to convey information to their own kind without being buried in ambient noise or the signals of other individuals. If an extraterrestrial civilization wanted to announce its existence to distant observers, its signal might likewise be designed so that it can be easily distinguished from natural emissions. The team chose firefly courtship behavior as the subject for testing this hypothesis.
Applying firefly flashing strategies to a model of radio signals
During the breeding season, fireflies emit light in species-specific flashing patterns to attract mates. When several firefly species live in the same place, lights of various periods blink all around, so each individual's signal needs to be distinguishable from the flashes of others. Simply glowing brightly is not enough; varying the timing and duration of flashes is important for letting others of the same species know where one is.
However, light emission requires metabolic energy and also carries the risk of being spotted by predators. For this reason, firefly flash patterns are thought to involve a trade-off between being easy to distinguish from other species and keeping down the costs and dangers that come with flashing. Building on a model of the evolution of firefly flash patterns reported in 2022, the team applied this relationship to radio signals from space. They did not run any experiment on space communication with real fireflies; the research uses the mathematical mechanism that explains flashing behavior.
In this model, a signal is represented as a pulse train that repeats two states, "on" and "off." The main elements that determine its structure are the period ((P)), which represents the repetition interval, the duration of a single pulse ((\tau)), and the phase, which represents the temporal position of the signal. The fraction of the period during which the signal is on is called the "absolute duty cycle" and is expressed by the following equation:
[ D=\frac{\tau}{P} ]
The absolute duty cycle takes a value between 0 and 1. For example, a value of 0.36 means the signal is being emitted for 36% of each period. The team adopted this duty cycle as a proxy for the energy required for transmission. This is a simplifying assumption that a shorter transmission time means lower energy consumption; it is not a calculation of the actual power consumption or energy requirements of a transmitter.
The cost function used for optimization combines similarity to the background signals with an energy metric based on transmission time. Expressed in the range 0 to 1, the cost is:
[ C=w_sS+(1-w_s)D ]
Here, (S) is the average similarity to the group of background pulsars, (D) is the duty cycle, and (w_s) is a weighting coefficient between 0 and 1 that determines how much importance is placed on similarity. The paper also uses a percentage cost obtained by multiplying this value by 100. Similarity is evaluated by finding the longest time interval over which the on and off states of two signals coincide, taking differences in phase into account. The lower the cost, the more desirable the signal structure is judged to be under the specified conditions.
When (w_s=1.0), energy consumption is ignored and the goal is solely to minimize similarity to the background signals. As (w_s) is reduced, weight is also placed on shortening the time the signal is emitted while maintaining its difference from the background. In other words, the model seeks to balance being easy to distinguish from other signals against keeping down the costs of transmission. However, because the energy evaluation depends on the duty cycle, this optimal solution is not necessarily the most energy-efficient one in a real communication device.
Reproducing natural radio signals with data on 3,724 pulsars
To distinguish artificial signals from natural emissions, one needs to understand the characteristics of the natural signals used for comparison. The team chose pulsars, which have strong periodicity. A pulsar is a rapidly rotating neutron star whose emission beam is observed as a regular pulse each time it sweeps toward Earth. Because these features resemble the signal of a transmitter with a deliberately set period, pulsars were adopted in this study as the comparison target for telling artificial signals apart.
The team used data on 3,724 pulsars recorded in version 2.8.1 of the pulsar catalogue managed by the Australia Telescope National Facility (ATNF). However, the radio emission of pulsars as actually observed has a complex waveform whose intensity varies continuously. In this model, the parts where the radio flux density exceeds a set threshold were converted to "1" and everything else to "0," producing a binary signal. This processing simplifies comparison, but it loses information such as the rise and decay of pulses and fine internal intensity fluctuations.
For the main analysis, the team selected 158 pulsars located within a radius of 5 kiloparsecs (about 16,300 light-years) of Earth and with a mean flux density at 1,400 MHz of more than 1 mJy (millijansky). The 1,400 MHz frequency is close to 1,420.4 MHz, which corresponds to the 21 cm line of neutral hydrogen, a radio band long regarded as important in SETI. The team treated the pulse structures of these 158 pulsars as the naturally originating background signal and generated hypothetical artificial signals that are easy to distinguish from them.
The search space for signals was set as a 100×100 grid, with period and absolute duty cycle each divided into 100 steps. The range of each parameter was determined from the mean and standard deviation of the background pulsar population. As a result, the approach does not search without limit for extreme signals far removed from real pulsars; instead, it searches for signals with low similarity to the background within a certain range anchored on pulsar characteristics.
What was evaluated here, however, was how far the mathematically defined cost function could be reduced. The team did not measure the probability of detecting an artificial signal with a real radio telescope, nor the false-positive rate at which a natural phenomenon is mistaken for an artificial signal. Judging detection performance in actual observations would require separate verification.
Adding an energy constraint cuts pulse transmission time to about 1/92
Over the 100×100 search space, the team varied the weighting coefficient (w_s) among 0.0, 0.2, 0.5, 0.8, and 1.0, and found the signal that minimized cost under each condition. The results showed that the optimal signal structure changes dramatically depending on whether energy consumption is taken into account.
Under the condition (w_s=1.0), which ignores energy consumption entirely and emphasizes only the difference from the background, a signal with a period of just under about 25 milliseconds and an absolute duty cycle of about 0.36 (36%) was selected. This is a pulse train with a relatively long on-time that emits radio waves for more than a third of each period. As far as the goal of minimizing similarity to the background is concerned, this structure was judged optimal.
However, when the weighting coefficient was lowered to (w_s=0.8), giving weight to shorter transmission times as well, the optimal duty cycle dropped sharply to about 0.0039 (0.39%). Compared with the (w_s=1.0) case, the fraction of time the signal is on is about 1/92. Furthermore, the same short pulse structure was selected under the other conditions tested, (w_s=0.5), 0.2, and 0.0. Within the model, the optimal solution thus split sharply between the case where energy consumption is ignored and the case where it is considered even slightly.
データを表で見る
| 最適デューティ比 (%) | |
|---|---|
| ws=1.0（際立ち重視） | 36 |
| ws=0.8以下（省エネ考慮） | 0.39 |
This result suggests that when the conditions require a signal that is easy to distinguish from regular natural radio signals while also keeping transmission time low, a signal that repeats very short pulses may be selected. However, this is a result based on the cost function and search range set in this study, and it does not mean that an extraterrestrial civilization would actually adopt such a communication scheme. Nor does the reduction in duty cycle, which represents the fraction of transmission time, demonstrate that actual energy consumption falls to 1/92.
The team further applied the same cost function to compare the optimized hypothetical signals with the 158 background pulsars. Under the condition (w_s=0.0), the hypothetical signal's cost was lower than that of 84.30% of the background pulsars. For (w_s=1.0), which emphasizes only the difference from the background, the proportion reached 99.69%, according to the description of Figure 3 in the paper. Still, this is a cost comparison within the mathematical model that was set up; it does not mean that artificial signals could be detected with 99.69% accuracy.
Interestingly, under the (w_s=1.0) condition, two natural objects showed a cost equal to or lower than that of the optimized hypothetical signal: "J1748-2446C" and "J1824-2452A." Both are millisecond pulsars with rotation periods of 1 to 10 milliseconds, natural objects that can be explained as having been spun up to high speeds through a special evolutionary process. The team does not claim that these are signals from an extraterrestrial civilization.
The existence of these two objects shows that obtaining an unusual score in the model does not allow one to conclude that the signal was artificially produced. The team also considered a case in which the search range was widened to match millisecond pulsars, and within a period range of 1 to 10 milliseconds they were able to generate hypothetical signals with even lower costs. Because the optimal solution can change depending on the search range and the fineness of the calculation, no particular score can serve as an absolute criterion for judging a signal artificial.
The team also analyzed what happens when the conditions for selecting background pulsars are changed. In the main analysis, 158 pulsars were selected on the basis of mean flux density, but when a condition using an approximate pulse flux was used, the number of targets rose to 803. Even so, the qualitative trend that adding an energy constraint greatly shortens the duty cycle was maintained. However, confirming stability against such changes in conditions is a separate matter from working effectively on real observational data.
The differences between this study and typical conventional SETI methods can be summarized as follows.
| Item | Typical conventional SETI methods | This firefly-inspired model |
|---|---|---|
| Main search targets | Narrowband radio, laser pulses, artificial modulation, etc. | Pulse trains structurally different from background pulsars |
| Features used for judgment | Frequency, bandwidth, periodicity, modulation, etc. | Period, duty cycle, similarity to background signals |
| Handling of natural background | Remove or identify natural emissions and noise | Use the distribution of natural signals as a reference for evaluating signal distinctiveness |
| Decoding of signal meaning | Not required at the detection stage, though decoding content is also a subject of study | Does not presuppose decoding of meaning |
| Output obtained | Candidates among observed signals that meet the conditions | Signal structures that differ greatly from the background under the set conditions |
| Main limitations | May miss unexpected communication schemes | Structural uniqueness alone cannot establish a signal as artificial |
Conventional SETI also includes methods that look for artificial features without decoding a signal's meaning, so this study is not the first to propose a search that does not decode. What is new is that, taking as a clue the communication strategies developed by non-human organisms such as fireflies, it mathematically evaluates the relationship between structural difference from the natural background and transmission cost. The team believes that such a biological perspective could offer a way to search for kinds of signals different from those targeted so far.
A signal that stands out is not evidence that it was artificially made
The biggest problem in applying this model to real SETI is that being very different from natural signals does not necessarily imply an artificial origin. In astronomy, there have been repeated cases in which unusual, previously unknown signals were later explained as new natural phenomena. In addition to the discovery of pulsars in 1967, fast radio bursts (FRBs) and objects emitting radio waves with ultra-long periods are also being studied as signals that are difficult to explain with conventional understanding. Given such phenomena, even if a signal different from background pulsars were detected, the possibility of an unknown astrophysical phenomenon would have to be considered first.
If the approach proves useful, it would be not for directly proving that a signal is artificial but for selecting signals worth examining in detail. One conceivable use is to extract, from the vast amounts of data collected by radio telescopes, candidates whose period and transmission time differ from known natural signals, and then investigate them through follow-up observations and other analytical methods. However, this study did not verify what detection rate or false-positive rate such an operation would achieve.
The model itself also has several important limitations. First, simplifying the waveforms of real pulsar radio emission into binary on/off signals loses things like the rise and fall of pulses and fine internal intensity fluctuations. Features of such continuous waveforms may contain important clues for judging a signal's generation mechanism and whether it carries information. The authors themselves acknowledge that developing analysis methods that retain this information is a future task.
Furthermore, because this optimization is based on just two main parameters, period and duty cycle, solutions tend to be biased toward the edges of the search range. The team confirmed a tendency for extreme values of period and duty cycle to be selected as optimal solutions as a result of avoiding regions where many background pulsars are distributed. This may mean that the model is choosing regions with few comparison objects within the set search range, rather than having found structures fundamentally different from natural signals. If the search range or grid fineness were changed, the minimum cost value and the optimal solution could also change.
The word "evolution" used in the model also calls for caution. The original firefly research used an evolutionary model in which flash patterns change over generations. The SETI model in this study, by contrast, does not reproduce a process in which multiple extraterrestrial civilizations evolve communication technologies through generational turnover and competition. It calculates a cost function within a preset search range and selects the signal structure that minimizes it.
Assuming that a real extraterrestrial civilization adopts such a communication scheme would also require additional conditions. For example, one would have to consider the possibility that the transmitting side knows which pulsars are observable from Earth and deliberately chooses a pattern that is easy to distinguish from that background. Radio waves arriving from a civilization thousands of light-years away were emitted thousands of years ago, so the state of the background celestial objects may have changed between transmission and reception. The team also explains that the current model does not adequately address the effect of pulsar periods changing over time.
Even so, this study has significance in broadening the range of SETI search targets. By not assuming that an extraterrestrial civilization uses the same communication technology as humans, and instead taking as a clue how diverse organisms on Earth announce their presence, it opens the way to considering new methods of evaluating signals. Going forward, it will be necessary to apply the approach to real radio telescope data and to test whether it can detect candidates amid complex background noise and how correctly it can distinguish known natural phenomena.
What this study shows is not the communication method of an extraterrestrial civilization itself, but a new way of thinking about searching for unknown signals. Whether the perspective derived from firefly flashing, a signal that is easy to distinguish from its surroundings while keeping transmission costs low, becomes an effective search metric in real SETI depends on verification with future observational data.