[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Cochlea Amplifier models : a new list

I think reasonable treatment of data would be to tell us the confidence level of your statements. If you think the data supports your hypothesis very strongly, you should be able to back this up with some statistics. That might convince people, whereas saying "thus it is not possible to attribute the findings of these authors to stochastic fluctuation" won't convince anyone as it is a self-contradictory statement at worst and unconvincing at best.

Also, I didn't claim to be able to explain the fiber data in all its details, I merely objected to your interpretation.

I'm perfectly happy to leave it at this. I think several people have described how they think about these things adequately. Everyone can now form their own opinion. Its fine to have disagreement, and I respect that you have a different point of view. We will all hopefully learn more when new data becomes available, and then we can take it from there.

Erik

Martin Braun wrote:
Caution is good, but there is also something like reasonable treatment of data. Both you and Erik have not mentioned the fact that the peak split at 100 dB showed up in two separate experiments in the same animal (Figs. 7A and 7B). Further both of you have disregarded the finding that in both experiments (Fig. 7A and 7B) there are two (Fig. 7A) and three (Figs. 7B) lower rate level points between the two peaks, not one. Thus it is simply not possible to attribute the findings of these authors to stochastic fluctuation.

Further, the fiber responses in Figs. 7A and 7B show no sign of saturation at all. This is seen in Fig. 7D only (different animal).

Most importantly, none of your or Erik's concerns question the finding that the first fiber neural data do NOT mirror the half-octave shift seen in basilar membrane (BM) behavior.

Martin



----- Original Message ----- From: "Gestur Björn Christianson" <g.christianson@xxxxxxxxx>
To: <AUDITORY@xxxxxxxxxxxxxxx>
Sent: Friday, October 12, 2007 9:32 AM
Subject: Re: Cochlea Amplifier models : a new list


On 11 Oct 2007, at 9:55, Martin Braun wrote:
A "stochastic model of the fiber response" is not needed. The rate count figures are so high that their significance need not be tested by further mathematics.

This is a claim that requires justification, I think. As a rule, variability of response increases proportionally to the mean rate of discharge. While the proportionality is normally less than that seen in cortical areas, it is still significant. In this case, as Erik pointed out, it is very important to demonstrate that the response at 5.5 kHz is significantly lower that the response to the surrounding frequencies, as in the absence of that single point all that is evident is that the discharge rate saturates at 100 dB.

In the context of this discussion, there is also a significant methodological confound in this study. Geisler et al. did not use pseudo-random stimulus delivery. Instead, they presented a single frequency repeatedly, stepping up the intensity from the minimum to the maximum. This raises concerns about habituation. Note that at both 5.5 and 5.75 kHz, you can see an increase in discharge rate with increasing intensity until the peak discharge rate of slightly more than 250 spikes/sec is achieved, at which case response begins to drop off. Because of their methodology, this is a progression in time, and consistent with known spike-rate adaptation effects; thus, even if the response to 5.5 kHz at 100 dB is significantly lower than the response to surrounding frequencies, it potentially reflects spiking mechanics and not contributions from the cochlea (or otherwise). I'd suggest that caution should be taken when trying to interpret the highest stimulus intensity plots in their data.

Bjorn


--
Erik Larsen
PhD candidate Speech and Hearing Bioscience and Technology
http://web.mit.edu/shbt

You will stop at nothing to reach your objective, but only because your
brakes are defective.