This was actually a question from one of my children.
Songs seem to be around three minutes - but of course not all songs are that long. I know there are some very long songs by both Led Zeppelin and Pink Floyd. Though this seems mostly correct. Three minutes is a very common time length for a song.
I asked a knowledgeable friend and he suggested that the time length of popular songs was based on the phonograph. You know, the original medium for recording sound. At first, these phonographs were cylinder shaped. Later, they came in the form of a 10 inch disk. These disks rotated at 78 rpm and could hold about 3 minutes worth of music. This format was popular until the 1960s when the 45 came out. The 45 was a smaller disk that rotated at 45 rpm (thus the name). However, they could still only hold "about 3 minutes".
How Long is a 45 rpm Record?
Personally, I was confused with the "about 3 minutes". Didn't people know how much music a record could hold? Well, it turns out that there wasn't a clear answer to this question. In order to understand why, you need to understand how a record works. The record is all about grooves. Sound is recorded using grooves. Here is a very basic side view of a record's grooves (I am showing two different grooves in the same image).
This needle moves along the groove. The groove can get both deeper and wider and as it does so making the needle vary in two diagonal directions. The variations of the needle in these two directions correspond to the sound in the left and right stereo channels. I know that's not the best explanation, but honestly I didn't hang around records too much. Check out these awesome animated gifs showing both mono and stereo record grooves - those might help.
As the record spins, this groove spirals towards the center of the record. You could make a tighter spiral (and thus longer song duration). However, you would have to make thinner grooves. Thinner grooves means less dynamic range to the sound (I'm not even sure what "dynamic range" means but it sounds cool). This is why they say "about 3 minutes". Surely some record engineer will come along and point out just how utterly wrong I am. That's fine, my point is not to explain records.
Does the Song Length Change Over the Years?
The idea is that songs are about 3 minutes because the 45 record is 3 minutes. How do you get your song to the radio station or in a jukebox? You need that song on a record. It's just that simple.
But what about other music formats? Eventually, people in the industry started using Compact Discs (CD). These have a 75 - 80 minute recording time (again, depending on the format) and became commercially available in 1982. You could easily put a 20 minute song on there and send it to the radio station. Of course now, everyone just uses some type of digital format for music. Again, these could be any length.
If the recording device is setting limits on the length of popular songs, shouldn't there be a change in this song length as new technologies emerge? Or maybe the human brain only likes 3 minute songs. Or maybe 3 minute songs is the most optimal song length so that you can get enough commercial advertisements in. I just don't know.
How about massive amounts of data? Yes? Yes. Here is a giant database of music. From that, I can get songs from different years and look at the duration. Let's just plot this stuff. Since most years had many different songs, I have calculated the average song length (in seconds) and included error bars that represent the standard deviation of the distribution.
From this it seems like songs are longer now than they were long ago. Still, is this because of the format? How about this. What if I look at songs in 1975 and 1985. I would assume in 75, the songs were all on records and in 85 they were on CDs. I'm just guessing here. Here is a the distribution of song lengths for the two years.
Here you can see these are two song distributions that look different. The 1985 songs seem longer than the 1975 songs. Why? Is it because of the CD? Maybe. Just maybe. Well, if this indeed the case then there shouldn't be a huge difference in song length between 1970 and 1980, right? I am assuming that in 1980 there weren't any radio stations using CDs (could be wrong here). This is a plot of the distribution of songs in both 70 and 80.
This seems to suggest that songs got longer even without new technology. Yes, there was still new technology. There was the extended play 45 record (EP). It still seems like that still would reduce the quality of the recording.
What about now? Since 1990, it seems that the average song length has sort of stabilized around 250 seconds (over 4 minutes). Maybe that's because humans prefer 4 minute songs. Clearly there is no technological limit to song length anymore, right?
So, did new technologies influence song length? I am going to say that it's plausible but not for certain. I still like the graphs though.
Homework
Here are some other questions. Just because....
- The 45 record spins at an angular speed of 45 rpm (thus the name). Estimate the length of the groove in this record. I'm pretty sure it's a 7 inch diameter record, but I don't know how far towards the center the groove goes.
- What if the average song length varies in the same way that fashion does? What if the 3 minute song length was a human preference that will return? See if you can fit an oscillating function to the above data and use it to predict when the songs will be 3 minutes long once again.
- Maybe song lengths just keep on increasing. Fit a linear function to the above data (you can do this right in plotly) and estimate the average length of a song in the year 2258 (the date of Star Trek the movie). At what year would the song length be zero seconds?
