A bass line can make you reach for the skip button before the singer takes a breath. Let the track play a few more bars and that same bass line might be the reason you stay. So how long does it take to judge a song? It depends on whether you're asking for a first impression, a considered opinion, or a preference that other listeners might share.
In a 2018 study on the timing of musical judgments, Amy Belfi and colleagues found that ratings based on half a second of electronica or three quarters of a second of jazz or classical music already gave them a clue as to how longer clips would be rated. Those numbers describe how much audio people heard. They show how little music it can take to form a first impression.
That impression can move. In a 2007 study of piano performance, Sam Thompson, Aaron Williamon, and Guy Valentine asked 33 listeners to rate recordings of Bach and Chopin preludes as they played. Listeners typically made an initial evaluation after 15 to 20 seconds. Ratings shifted between that first decision and the end, with the biggest revisions coming in the first minute. With more of the performance to go on, listeners could change their minds.
Still, a few seconds can tell you quite a bit about whether you'll like the rest of a song. A 2023 study in Music Perception used 3,120 excerpts from 260 songs. Across the songs studied, ratings of short excerpts tracked ratings of the full tracks closely. The researchers tested clips of 5, 10, and 15 seconds and found that clip length and placement made relatively little difference to that overall relationship. Any one excerpt can still miss the moment that wins you over.
Now put those headphones on two groups of people. Will they hear the same thing? In a 2024 study of singing preferences, two groups rated the same voice clips, one online in the US and one in a lab in Germany. When the researchers averaged the ratings for each clip, the two sets of averages were strongly correlated (r = 0.85). Agreement between individual listeners was low. The groups tended to favor similar clips, even though the people in them had plenty of room to disagree.
That's the kind of shared taste worth looking for in blind music voting. On Kallitone, you hear two tracks without seeing the artists' names. Playback starts at a random point within each track. In a standard vote, you need to hear at least 15 seconds of each before you pick one. A single vote captures a preference for what you heard that time. When many listeners make those choices, we can look for preferences that repeat across comparisons.
We chose 15 seconds per track as a practical sweet spot. A few bars give you time to hear what the music does after that first sound, and the minimum is short enough to compare several pairs in one sitting. At 15 seconds per track, ten votes take at least five minutes of audio. At 30 seconds per track, that doubles to ten minutes of audio. The studies give us reason to take short listening seriously. The exact cutoff is our decision about how much listening to ask for in each comparison.
A direct test would give separate groups 5, 15, or 30 seconds per track on the same set of song pairs. How often do their picks hold up after a full listen? Do separate groups favor the same tracks? How many listeners finish each vote? Together, those answers would tell us what the extra listening time buys.
You can try a version with a friend. Pick two songs neither of you knows and play a 15-second excerpt from each. Write down your picks privately, then listen to both songs in full. Did you pick the same song? Did either of you change your mind?