Ever wonder what adjectives Richard Thompson likes most?

Richard Thompson’s adjectives

I am a huge fan of Richard Thompson; I think he's an incredible musician and a brilliant songwriter.

In his book about Richard and Linda Thompson’s album Shoot Out the Lights, Hayden Childs makes an intriguing suggestion in Appendix F, “Other Detritus Too Esoteric for Even These Appendices”: “A pie chart showing the most frequently employed adjectives in all Richard Thompson songs.” Well, I like this kind of thing, and I have a computer, so why not?


First, I must apologize for not being Mark Liberman; he would have written this up as a Breakfast Experiment™ on Language Log using more sophisticated tools and with cool graphs that dance around the screen whilst humming RT melodies.

Richard Thompson has conveniently put up lyrics to all his songs on his website. I wrote a script to download all of these (pausing after each song so as not to overload the server); I cleaned up each file to leave only the lyrics, without any of the boilerplate links, formatting, etc. There are various issues: Some songs, especially from the Fairport Convention days, are covers, e.g. of Bob Dylan songs or Child Ballads; some songs appear in different versions. I did not try to correct for these, figuring that even the covers are at least representative of the kind of lyrics that Thompson saw fit to include on albums.

A further issue is that the repetition of a word in a particular song skews the results, or at least may be regarded as skewing the results. For example, 20 of the 21 examples of reckless come from Reckless Kind. I figured, though, that a word repeated frequently in a song is at least important enough to repeat; that is, the repetition itself indicates that it's important.

The entire corpus came to 87,866 words. I ran this though an automated part-of-speech tagger (I used TreeTagger, developed by Helmut Schmid in the TC project at the Institute for Computational Linguistics of the University of Stuttgart), and then counted up the results. Like many computational taggers, TreeTagger uses a more sophisticated analysis of parts of speech than most people are familiar with, so I conformed the results to the familiar English p-o-s descriptions. Finally, I sorted the results.

It would be useful to compare this to baseline frequencies of these adjectives in a standard English corpus, but I didn't have the time.

The results

The full list of course goes on for quite some length. But here are the first 100 (or so; there were ties) most common adjectives in Richard Thompson’s songs:

old 149
little 139
good 129
more 77
bad 71
long 65
poor 64
wrong 62
hard 58
other 55
sweet 55
dead 51
young 50
many 49
dark 48
cold 48
best 46
last 44
new 41
rich 40
next 38
better 37
blue 37
true 36
free 34
black 33
first 32
lonely 31
own 31
same 31
right 31
much 30
left 30
small 30
only 27
real 25
blind 24
high 23
precious 22
whole 22
wide 22
crazy 22
big 21
reckless 21
mine 20
loving 20
strange 20
empty 20
fine 19
restless 19
easy 18
white 18
great 18
sure 18
grey 18
cruel 17
happy 17
perfect 17
sad 17
wild 16
red 16
bright 15
dumb 15
live 15
fair 14
slow 14
strong 14
few 14
twisted 14
late 14
alive 14
full 14
such 13
shady 13
human 13
weary 13
different 12
dirty 12
wise 12
lucky 12
low 12
green 12
sorry 12
rare 12
sick 12
hungry 12
light 11
tired 11
deep 11
bonny 11
afraid 11
plain 11
mad 11
worth 11
simple 11
famous 10
wonderful 10
two-faced 10
back 10
enough 10
ready 10

The Analysis

Um, no. I’m not a critic, I’m just a lexicographer and sometime programmer. I’ll leave the analysis for people smarter than I am. I will, however, mention that though this list does include a lot of words people would not be surprised to see in RT songs (poor, wrong, dead, dark, cold, lonely, crazy, empty, restless), it is nonetheless the case that good is much more common than bad.