Last night Jon Stewart cracked a Cheney-heart-condition-joke with “defibulators” as the punch line. Now I’m not one to judge, but I had to point it out. Despite the 29K ghits that “defibulator” gets, and the 12K ghits that “defibulators” gets, Google still asks you if you mean “defibrillator(s)”. Oxford lists defibrillator under defibrillation.
There’s a parallel here with vascillations like nuclear/nucular and parap[a]legic, discussed a while back by Arnold Zwicky on Language Log and Eric Bakovic on phonoloblog. Both refer to a piece by Geoff Nunberg that links nucular to a morphological analogy from words like molecular and particular. Eric extends the discussion to parap[a]legic, suggesting a rhythmic analogy is better suited in this case. I added two cents there in a comment about the possible role of splitting the obstruent-liquid cluster.
Defibulator seems to fit both the morphological analogy and the cluster splitting accounts. Morphological analogy would attribute the appearance of defibulator to the extistence of other words that have the sequence –bulation, –bulate, or –bulator, and the absence of other words that have the sequence –brillation, –brillate, or –brillator. There are words like tabulate and discombobulate, but the only other word with –brillat– is apparently fibrillate. The cluster splitting analysis would attribute defibulator to an avoidance of the [br] sequence in defibrillator.
Seems to me that these analyses aren’t competing – they both presume a lack of rigid specification on the part of the speaker as to what the word is. So if the speaker is not quite sure what the word is, then a constraint against [br] clusters (usually violable in English) can have an observed effect. Emergence of the Unmarked in infrequent words?