>        From: Rob Naylor
>
>        How did they decide what's "pop" when selecting the half million songs to base their analysis on? Since it was Spanish researchers, did they use Spanish charts, UK charts, US charts of what as selection criteria?
>
>        Did they include, with hindsight, a lot of older songs that didn't figure as "pop" at the time, but as "underground rock" and if so did they balance them up with current non-chart "indie" songs that are relatively unknown in the mainstream?
>
>        In fact, I looked up the "million song dataset"
>        ...
>        (From "The Million Song Dataset Challenge" by Brian McFee et al)

A link would have made it easier to comment on your comments.

>        Having gone through this paper, which deals with the "Taste Profile Subset" I'm still not much wiser as to how they define "pop" music. From what I can see, in this dataset, a song like Pink Floyd's "Wish You Were Here" would make the list on play-count alone, though it would never have been regarded as a "pop" song when released as an album track in 1975 and wasn't even ever released as a single until 1995! the same applies to many songs from the 60s and 70s which were never in any "commercial pop" when originally released but have since become iconic anthems of the period.
>
>        How to compare these "fairly" with what's being produced "underground" these days, and which may well be as varied in its own terms as the "non-pop-pop" of 40-50 years ago was?
>
>        If the review was confined to just singles chart tracks of the relevant period, I suspect the results would show that "top 20 hits" of every period have been structurally limited, similar and "battery farm pop". When I look back to stuff that was in the charts in the days I started listening to "underground music" it was mostly as bland and anodyne (though not quite as loud) as what's in the charts now.

This is the first critical comment I've seen that might have some genuine scientific foundation. One of the surest ways to lie with statistics is in the choice of the dataset, or the criteria applied to it.

Derek Brimstone used to say something like: "What these people say is that the first day of your life is the most dangerous, but the last one's pretty dodgy too!"

In the 80s the police began a campaign against drunken driving, using a statistic like: "X% of accidents involve alcohol!" I can't remember exactly what X was, but my point is, if a drunken pedestrian suddenly lurched out in front of a stone cold sober driver and was hit, that would be included in X, so the choice of that particular statistic was deliberately misleading, and was presumably done to give a bigger value of X than would have a more appropriate statistic such as the percentage of accidents involving drivers found to have alcohol in their blood.

The points you raise about song selection for the database are valid. It would certainly be interesting to know more.

>        From: GUEST,Stim
>
>        As I may have let on, I think the research is lame, in addition to being completely wrong.

Surely it can't be both? It's either lame, in which case it's only partially flawed, or it's completely wrong, in which case it's totally flawed. Given that these things are peer-reviewed, etc, I think the latter is unlikely.

>        Here is a list of The Top 100 Songs from 1962

"100 Greatest Songs From 1962

Criteria: - These records were chosen and ranked based on their initial and lasting popularity, and on their impact on the overall scope of musical history. Records are listed based on the year that they were released."

So that list is flawed as well, because it's partly based on hindsight. If it is true that pop music today is more simplified than that of fifty years ago, those making a selection today such as the above are likely to skew the results by applying modern criteria to produce the selection. You could try this:

1962

>        this statement, "In particular, we obtained numerical indicators that the diversity of transitions between note combinations - chords plus melodies t has consistently lessened in the last 50 years." doesn't even make sense.

Can't find that sentence in the original research page.