Music by the Hour: What Listening Data Says About Time of Day

The cheapest sensor in any phone is its clock. It needs no strap, no camera and no setup, and it always knows whether it is 7 a.m. or 11 p.m. On Sunday, November 1, clocks in most of the United States are due to go back an hour at 2 a.m. That makes this a good week to ask what the hour tells us about the music people want.

Two large studies of streaming data give a clearer answer than you might expect. They also show where the clock stops being useful, and that matters for anyone building music that adapts.

765 million plays, one daily rhythm

In 2019, Minsu Park and Michael Macy of Cornell, together with colleagues at Spotify, published a study in Nature Human Behaviour. They analyzed 765 million plays that 1 million people in 51 countries streamed on Spotify in 2016. Cornell's news office explains that tracks were rated from "highly relaxing (acoustic, instrumental, ambient and flat or low tempo) to highly energetic (strong beat, danceable, loud and bouncy)."

The daily pattern was similar across cultures and demographic groups. People played more relaxing music late at night and more energetic music during normal business hours. That held in mid-afternoon too, a time when earlier research on text posts found people's expressed mood at its lowest.

Some baselines differed. Younger listeners played more intense music. Listeners in Latin America played more arousing music than other regions, and listeners in Asia played more relaxing music. People who are habitually awake at night, the so-called night owls, played less intense music overall. Seasons mattered too: relaxing music was preferred in cold seasons and arousing music in warmer ones, and absolute day length tracked intensity more closely than changes in day length did.

The authors are careful. The study is observational, it did not measure mood itself, and it offers no causal explanation. Macy called the data "a remarkable opportunity to advance our understanding, empirically, of people's emotional management based on how they use music."

Five blocks of the day

A second team came at the same question with a different dataset. Ole Adrian Heggli, Jan Stupacher and Peter Vuust of Aarhus University analyzed 2,026,529,205 plays from a public Spotify dataset of listening sessions sampled over eight weeks. In Royal Society Open Science in 2021, they reported that listening divides into five blocks: morning, afternoon, evening, night, and late night or early morning.

The blocks typically began around 6 a.m., noon, 8 p.m., 11 p.m. and 4 a.m. They always ran in the same order, but weekends changed their shape: the evening block grew longer and pushed the night block later. Loudness rose toward the end of the morning and tapered off at night. Tempo and danceability dipped around 2 p.m. and were highest in the evening, while the night block had the lowest loudness and tempo.

Here is the part worth remembering. The differences in averages were tiny. Mean tempo stayed between 122.3 and 122.8 beats per minute in every block, which the authors call non-discernible in musical terms. What separates the blocks is the range of music people draw on, not a different average sound.

The small hours were the most varied of all. Audio features spread out most then, and listeners agreed least with the model's guess for that block. The authors suggest who is awake then: shift workers, night owls heading to bed, early risers, and people up for a one-off reason such as a party or an early flight. Each has a different need.

Playlists carry the hour, single songs mostly do not

The Aarhus team also asked listeners. In one test, 89 people ran 253 of their own playlists through a model trained on the five blocks, and they agreed with its guess for 63 percent of them. In another, 176 people rated individual tracks. They preferred morning and late-night tracks at their intended times, showed no clear preference for afternoon or evening tracks, and leaned toward playing the night tracks at other times.

The authors conclude that "playlists of multiple tracks, but not individual tracks, reflect diurnal musical preference," and they note that "music listening is rarely a one-track event."

An old idea

Matching music to the hour is much older than streaming. The Aarhus paper notes that Western serenades and nocturnes were often meant for the evening, and that in the Hindustani tradition a raga is often intended for particular times of day. It quotes a treatise on that tradition, the Sangita Makaranda: "one who sings knowing the proper time remains happy."

What this means for adaptive music

For music that adapts, the clock is a free input. These studies suggest a few ways to use it well.

• Treat the hour as a nudge, not a rule. Averages barely moved between blocks, so an evening setting should tilt the odds toward calmer choices, not force them.
• Think in sessions. People feel the hour in a run of music, not in one song, so the useful unit is the arc of the next stretch of listening.
• Expect people to differ. Age, region, season and chronotype all shifted what people played, and the Aarhus authors note that Spotify data lean toward Western listeners.
• Keep the listener in charge. People often use music to change how they feel, not to mirror it, so what suits the hour on average may not suit you right now.
• Pair slow signals with fast ones. The hour shifts the mood slowly, while steps, heart rate or a crowd change the detail quickly, as we describe in our piece on responsive installations.

Try it yourself

Macy said listeners could use findings like these to be more strategic about choosing music to reinforce or influence their emotions. So try a small experiment. For a few days, jot down the time whenever you choose music, along with what you picked. See where your own curve follows the average and where it does not. The gaps are the interesting part, and they are the reason the listener should have the last word.

A song can follow the hour too. One composition can be heard in a calmer or livelier production and nudged a little slower or faster, so the same song may suit more than one time of day. On Starchild, every song can be heard in different styles and voices before licensing, with live key and tempo control. For the wider idea of responsive sound, see music that knows the room.

Sources

• Nature Human Behaviour: Global music streaming data reveal diurnal and seasonal patterns of affective preference (Park et al., 2019)
• Cornell Chronicle: Streaming chill vibes? Spotify data says the season is the reason (2019)
• Royal Society Open Science via PubMed Central: Diurnal fluctuations in musical preference (Heggli, Stupacher and Vuust, 2021)
• NIST: Daylight Saving Time