Benchmarking on the JVM is tricky, beyond pure algorithmic efficiency (assuming all memory accesses are equal), there's cache effects, JVM heap and GC effects and the usual warmup/inlining concerns of benchmarking. On page 551 of the excellent scala staircase book (2nd edition[1]) there's a collection efficiency table w.r.t. head/tail/indexing/insert, but that's a jumping off point (sorry, I can't find an online version of the table).