Dear Piotr
Thank you for the comment and the references (I was however, already
familiar with them).
For me, the problem on the statistical anaylsis is simple: what
null hypotheses are being rejected by the result's statistical
significance.
To me, the two null hypotheses as formulated in the analysis are: 1)
the DCN has no effect on sound localization; and 2) the DCN's effect
on sound localization is equal in both azimuth and elevation. Both
of these null hypotheses can be rejected (no surprise here).
The difficulty here is that the null hypotheses do not match to the
actual desired hypothesis; the desired hypothesis is not that the
DCN has some effect on localization, but rather that the DCN is in
the critical path for elevation localization. To match that actual
desired hypothesis, the null hypothesis would then be formulated as
something like: there is SOME elevation localization if the DCN
output is disabled.
OK, determining the statistical level of "SOME elevation
localization" may be problematic, except in this case for the (-30,
0, +30) elevations. The (for me, the correctly formulated) null
hypothesis cannot reasonably be rejected -- there is still a lot of
elevation localization occurring. [Even for the other cat, this
still holds.]
Is this an incorrect analysis?
If my analysis is correct, then the experiment still has an
extremely important result; it's conclusion is just the opposite of
what has been interpreted: Localization based on spectral cues is
processed someplace else other than the DCN. For higher elevations
the DCN may play a role.
Best regards,
Mark
On 10/7/2011 11:50 AM, Piotr Majdak wrote:
Dear Mark,
When I look only at a single spatial position in your document, I
agree, the statistical evidence is also not obvious to me. But
when you average over all the tested positions, the statistical
variance would decrease and the decrease in the elevation slopes
(response vs. target) after the lesion would be more prominent.
Looking at the statistical results in Table I of May (2000) - it
seems like the average effects (column EL error) are significant.
BTW, the data in your document show the results for the cat having
less effect (p < 0.05). The other cat showed even more
statistical significance (p < 0.01).
I think that the common assumption is that DCN is an important
stage in the localization process but also other stages are also
involved. The DCN type IV cells have been shown to be tuned to
spectral notches also by others [1, 2]. The type IV cells project
to the inferior colliculus (ICC), where further neural basis for
sound localization has been found [3]. The ICC projects further
(but not only) to the superior colliculus (SC), where a systematic
map of auditory space could be confirmed in birds (in their SC's
equivalent, optic tectum) but the situation seems to be more
complicated for mammals [4]. If you'd like to read more, I'd like
to suggest you this excellent review: Grothe, B., Pecka, M., and
McAlpine, D. (2010). "Mechanisms of sound localization in
mammals," Physiol Rev 90, 983-1012.
[1] Imig, T. J., Bibikov, N. G., Poirier, P., and Samson, F. K.
(2000). "Directionality derived from pinna-cue spectral notches in
cat dorsal cochlear nucleus," J Neurophysiol 83, 907-925.
[2] Hancock, K. E., and Voigt, H. F. (1999). "Wideband inhibition
of dorsal cochlear nucleus type IV units in cat: a computational
model," Ann Biomed Eng 27, 73-87.
[3] Davis, K. A., Ramachandran, R., and May, B. J. (2003).
"Auditory processing of spectral cues for sound localization in
the inferior colliculus," J Assoc Res Otolaryngol 4, 148-163.
[4] King, A. J. (2004). "The superior colliculus," Curr Biol 14,
R335-8.
Best regards,
Piotr Majdak
|