A flattened absorption profile centered at 78 MHz was detected in the sky-averaged radio spectrum, with a full-width at half-maximum of 19 MHz and an amplitude of 0.5 K. This profile is consistent with the 21-centimetre signal expected from early stars, but its amplitude is more than twice the largest predictions. This discrepancy suggests that the primordial gas was colder or the background radiation temperature was hotter than expected. Astrophysical phenomena are unlikely to explain this, but dark matter-baryon interactions could account for the observed amplitude. The profile's edges indicate that stars existed and produced Lyman-α photons by 180 million years after the Big Bang, and the gas was heated to above the radiation temperature less than 100 million years later. The detection was made using the EDGES low-band instruments, which recorded spectra between 50 and 100 MHz. The data were calibrated, filtered, and integrated over hundreds of hours. The absorption profile was detected in data spanning nearly two years and was observed by two identical instruments located 150 meters apart. The profile was detected with various hardware modifications and using two independent processing pipelines. The profile is not explained by known astronomical or atmospheric mechanisms. The 21 cm line has a rest-frame frequency of 1420 MHz, and its observed frequency is redshifted according to v = 1420 / (1 + z) MHz. The observed absorption profile is the continuous superposition of lines from gas across the observed redshift range and cosmological volume, tracing the history of the gas across cosmic time. The profile is centered at z ≈ 17 and spans approximately 20 > z > 15. The observed profile amplitude could be explained if gas and background radiation temperatures decoupled by z ≈ 250, allowing the gas to begin cooling adiabatically earlier. The presence of stars should eventually halt the cooling of the gas and heat it, as stellar radiation deposits energy into the gas. The z=15 edge of the observed profile places this transition around 270 million years after the Big Bang. The ages derived for the events fall within the range expected in many theoretical models. The flattened shape of the observed absorption profile is uncommon in existing models and could indicate that the initial flux of Lyman-α radiation from early stars was sufficiently large to quickly saturate the spin temperature to the gas temperature. High Lyman-α flux models were probed at z<14 using EDGES high-band measurements and a large fraction were found to be inconsistent with the data. The observed profile amplitude could be explained by dark matter-baryon interactions if the dark matter particle mass is below a few GeV and the interaction cross-section is greater than ~10^-21 cm². Other non-standard physics models predict increased gas temperatures and are unlikely to accountA flattened absorption profile centered at 78 MHz was detected in the sky-averaged radio spectrum, with a full-width at half-maximum of 19 MHz and an amplitude of 0.5 K. This profile is consistent with the 21-centimetre signal expected from early stars, but its amplitude is more than twice the largest predictions. This discrepancy suggests that the primordial gas was colder or the background radiation temperature was hotter than expected. Astrophysical phenomena are unlikely to explain this, but dark matter-baryon interactions could account for the observed amplitude. The profile's edges indicate that stars existed and produced Lyman-α photons by 180 million years after the Big Bang, and the gas was heated to above the radiation temperature less than 100 million years later. The detection was made using the EDGES low-band instruments, which recorded spectra between 50 and 100 MHz. The data were calibrated, filtered, and integrated over hundreds of hours. The absorption profile was detected in data spanning nearly two years and was observed by two identical instruments located 150 meters apart. The profile was detected with various hardware modifications and using two independent processing pipelines. The profile is not explained by known astronomical or atmospheric mechanisms. The 21 cm line has a rest-frame frequency of 1420 MHz, and its observed frequency is redshifted according to v = 1420 / (1 + z) MHz. The observed absorption profile is the continuous superposition of lines from gas across the observed redshift range and cosmological volume, tracing the history of the gas across cosmic time. The profile is centered at z ≈ 17 and spans approximately 20 > z > 15. The observed profile amplitude could be explained if gas and background radiation temperatures decoupled by z ≈ 250, allowing the gas to begin cooling adiabatically earlier. The presence of stars should eventually halt the cooling of the gas and heat it, as stellar radiation deposits energy into the gas. The z=15 edge of the observed profile places this transition around 270 million years after the Big Bang. The ages derived for the events fall within the range expected in many theoretical models. The flattened shape of the observed absorption profile is uncommon in existing models and could indicate that the initial flux of Lyman-α radiation from early stars was sufficiently large to quickly saturate the spin temperature to the gas temperature. High Lyman-α flux models were probed at z<14 using EDGES high-band measurements and a large fraction were found to be inconsistent with the data. The observed profile amplitude could be explained by dark matter-baryon interactions if the dark matter particle mass is below a few GeV and the interaction cross-section is greater than ~10^-21 cm². Other non-standard physics models predict increased gas temperatures and are unlikely to account