Professor Rollman's response

1
382 Pain, 9 (1980) 382 © ~lsevier/North-Holland Biomedical Press Ottawa, Ont., 1 August 1980 re: Professor Rollman's response Dear Editor: I think Professor Rollman and I are in substantial agr~ment. Somepain researchers have made quite erroneous claims about the value of SDT tech- niques in pain research. On the other hand, I do not think that methodologi- cal points were only a trifling part of Professor Rollman's original paper [2]. The point he now raises about maximum-likelihood estimates of ROC again seems likely to cloud the issue and suggest to pain researchers that they have little of practical value to gain from SDT techniques. The point I wished to make [1 ] is that simple methods are available based upon the use of propor- tio~ of area under the ROC, P(A), which do not require any curve-fitting at all. Unlike Professor Rollman I don't think this is merely a quibble. It is of course often the case the assumption of equal-variance Gaussian distributions of signal + noise and noise is often unjustified (in fact one might say that except for some auditory detection experiment, these assumptions are always difficult to justify). However, I must make the same point as I made in my ar*Acle" d' is not the only statistic available. Easily computable mea- sures like P(A) may be used when distribution assumptions are hard to jus- tify. It may of course be the case that my speculative analysis of Rollman's data is incorrect. Sensible application of SDT techniques may be the only way to settle the mater. I should not conclude on a note of discord. I continue to think, despite ~ome differences, that Professor Rollman has effectively exposed a logical flaw at the heart of mucl~ psychophysical research on the problem of pain. BILL JONES Department of Psychology, Carleton University, Ottawa, Ont. KIS 5B6, (Canada) REFERENCES 1 Jones, B., Signal detection theory and pain research, Pain, 7 (1979) 305--312. 2 Rollman. G.B.~ ~ignal detection theorv mo~s~remen~ of pain: a review and critique, Pain, 3 (1977) 187--211. 3 Rollman, G.B., Signal detection theory pain measures: empirical validation studies and adaptation-level effects, Pain, 6 (1979) 9--21.

Transcript of Professor Rollman's response

Page 1: Professor Rollman's response

382 Pain, 9 (1980) 382 © ~lsevier/North-Holland Biomedical Press

Ottawa, Ont., 1 August 1980

re: Professor Rollman's response

Dear Editor:

I think Professor Rollman and I are in substantial agr~ment . Somepa in researchers have made quite erroneous claims about the value of SDT tech- niques in pain research. On the other hand, I do not think that methodologi- cal points were only a trifling part of Professor Rollman's original paper [2]. The point he now raises about maximum-likelihood estimates of ROC again seems likely to cloud the issue and suggest to pain researchers that they have little of practical value to gain from SDT techniques. The point I wished to make [1 ] is that simple methods are available based upon the use of propor- tio~ of area under the ROC, P(A), which do not require any curve-fitting at all. Unlike Professor Rollman I don ' t think this is merely a quibble. It is of course often the case the assumption of equal-variance Gaussian distributions of signal + noise and noise is often unjustified (in fact one might say that except for some auditory detection experiment, these assumptions are always difficult to justify). However, I must make the same point as I made in my ar*Acle" d' is not the only statistic available. Easily computable mea- sures like P(A) may be used when distribution assumptions are hard to jus- tify.

It may of course be the case that my speculative analysis of Rollman's data is incorrect. Sensible application of SDT techniques may be the only way to settle the m a t e r .

I should not conclude on a note of discord. I continue to think, despite ~ome differences, that Professor Rollman has effectively exposed a logical flaw at the heart of mucl~ psychophysical research on the problem of pain.

BILL JONES Department of Psychology,

Carleton University, Ottawa, Ont. KIS 5B6,

(Canada) REFERENCES

1 Jones, B., Signal detection theory and pain research, Pain, 7 (1979) 305--312. 2 Rollman. G.B.~ ~ignal detection theorv mo~s~remen~ of pain: a review and critique,

Pain, 3 (1977) 187--211. 3 Rollman, G.B., Signal detection theory pain measures: empirical validation studies and

adaptation-level effects, Pain, 6 (1979) 9--21.