Many thanks to all who have commented so incisively on ‘Sometimes behave so strangely’ – the feedback was great. Including those who wrote to me privately, I received over 40 responses – most with multiple comments and questions. I’ll reply to all individually over the next couple of days, but thought I should make a few general points.
A number of people have pointed out that the timing characteristics of the pattern could be important. I agree – the pattern has an emphatic rhythm, and the repetition of this rhythm may help to induce listening in ‘music mode’. Also, in a study in progress we’re looking at the effect of the durations of the pauses between repetitions, and it appears that lengthening these pauses reduces the effect - or at least slows down its development– the formal experiment has yet to be completed. And it’s true that I intentionally configured the pauses in the published example so that the entire sequence of repetitions should be metrically coherent.
About the intonation contours, what I had intended in saying that they were ‘flattened in the subject’s rendition’ is that, in the subject’s rendition, the contour of each syllable spans a smaller range in terms of f0 than it does in the original – considerably smaller for some syllables. The contours of the syllables in the original are not flat – some of them, such as for ‘some’ and ‘so’ span a large range – but they become flat when listeners who hear the effect as music repeat what they hear.
The fact that the contours in my original are not flat has led to some small disagreement as to what the ‘notes’ actually are. Although the large majority hear the phrase as I notate it, some people argue that the first syllable ‘some’ should be notated a semitone lower, and others argue that the last syllable ‘ly’ should be notated a semitone lower (see for example, Linda Seltzer’s posting). There is in fact no ‘correct’ answer, because the pattern of f0s is consistent with a number of interpretations. (This is true even taking amplitudes into account.) It may well be that most of us ‘opt’ for the phrase as I notate it because it enables an interpretation in terms of a well-formed tonal melody.
I’m flattered that Eliot Handelman writes that ‘Diana actually IS singing’, but can assure you that before I noticed this effect in my commentary on my CD ‘Musical illusions and paradoxes’ neither I nor anyone else believed I was singing. And in one experiment described in the ASA/ASJ meetings, we had subjects listen to the phrase ten times, and each time they notated what they heard on a five point scale from ‘exactly like speech’ to ‘exactly like singing’ . The subjects overwhelmingly chose ‘speech’ on the first iteration, but by the tenth they had swung over to ‘singing’.
Many thanks again,