Subject: Re: Cochlea Amplifier models : a new list From: "Richard F. Lyon" <DickLyon@xxxxxxxx> Date: Thu, 11 Oct 2007 07:57:02 -0700 List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>Martin, That's the same Ruggero paper that I meant, so yes I'm familiar with it. The mechanical data of figure 1A are sliced at a threshold level of 164 microns/second and compared with neural threshold data, in figure 3. They're also sliced and compared at 1.5 nm displacement. The result is that the neural response is somewhere between the velocity and the displacement, but is otherwise about the same. But you are right that unfortunately, this single slice does not get at the issue you are are trying to address. I do see now that the CF would shift to lower if it was defined via a sufficiently high threshold level; such high mechanical thresholds, reached at around 80 dB SPL, would be hard to compare with anything neural, since it will be hard to find a neural rate threshold that will exceeded at 80 dB for some frequency but not at lower levels for any frequency (because rate is very well saturated by that point). So I don't see how there can be corresponding data of the type that you suggest the authors have suppressed, or if it exists it's won't show much interesting in the region you're referring to due to saturation. Phase data, on the other hand, do exist and should be robust, and would be good to look at for differences. Their Fig. 5, phase data at 75 dB SPL, does not really address the level dependence, and there's an apparent slope disagreement that probably comes from assuming a 1 ms delay instead of fitting a delay that makes the slopes match; Fig. 6 looks at level dependence, and shows an interesting discrepancy above 80 dB SPL, in the case a stimulus frequency far below CF; this is the interesting one to try to explain, though it says nothing about what happens within an octave of CF. I don't see anything about level dependence of neural phase near CF, which is too bad. I don't have access to figure 7B of the second paper; I'm sure it would be fair use for you to email me a scan of it. I'm not able to follow your logic in saying that the non-existence of some figure that you imagine is evidence of the data relationship that you claim exists in many publications. Here's a suggestion: if the data show the effect you are claiming, interpolate and replot from some data that do exist, and show us a figure of the sort you mean, constructed at least approximately from real data. Then maybe we'll be able to see what it is that you're seeing in the data, and why you disagree with Ruggero et al on the interpretation of their measurements. Ruggero et al explain that the lack of understanding of the relationship between neural and mechanical is due to a dearth of mechanical data. Now that they've got the mechanical data, you can't appeal to missing neural data to support a discrepancy; so show us some data that illustrates your point. Dick At 3:18 PM +0200 10/11/07, Martin Braun wrote: >Richard F. Lyon asked: > >>>>Where should I look to study the data on this idea that underlies your >>>>modeling approach? >>> >>>You can take any data on level dependence of neural responses. None of >>>them mirrors the half-octave shift of BM tuning. >> >>I've looked at lots of data, and you must be interpreting it differently >>from how I am. So if you have something specific that we can look at and >>discuss, we can try to resolve that difference. Lacking that, I'll stick >>with Ruggero's interpretation that says mechanical and neural are >>essentially the same, not different. > >OK, here it goes. > >1) Mario Ruggero only compared mechanical and neural behavior at threshold. >Had he compared mechanical and neural behavior at high sound levels, he >would have seen the striking dissociation between the two. > >2) The mechanical data (basilar membrane BM): >The literature is full of data showing the half octave shift of BM behavior >between low/medium sound levels and high sound levels. A freely available, >and recent, example is this one: > >Fig. 1A of http://www.pnas.org/cgi/content/full/97/22/11744 > >Here we see that, at one single recording location, the BM reacted most >strongly to sound levels of 0, 20, and 40 dB, when the probe tones had a >frequency 9.5 kHz. At a sound level of 100 dB, however, the BM reacted most >strongly, when the probe tones had a frequency 6 kHz. > >Interpretation: The low level data reveal the tuning of the outer hair cells >(OHC), whose motility primarily excites the adjacent inner hair cells (IHC) >and secondarily, as a side effect, cause local BM excursions. The high level >data reveal the tuning of the BM proper, the passive BM. The protective >effect of this passive BM tuning: high level sound of 9.5 kHz vibrates the >BM *basalward* of the place of the OHCs tuned to 9.5 kHz. This takes out >energy at the most critical frequency (9.5 kHz) that otherwise might damage >the OHCs. > >3) The neural data (auditory nerve fiber): >If neural data would correspond to BM data, the figure above would have a >corresponding figure, where the fiber fired most strongly at sound levels of >0, 20, and 40 dB, when the probe tones had a frequency 9.5 kHz, but most >strongly at sound levels of 100 dB, when the probe tones had a frequency 6 >kHz. > >Now - and this is the crucial point - such figures do not exist. Did they >exist, you can be certain that Mario Ruggero would have published one, >together with the figure above. The experiments were done, and the figures >were plotted. But they do NOT show a parallel to BM behavior. Actually, I >know that Mario is aware of these figures! > >For an example, take Fig. 7B in this paper: > >http://www.ncbi.nlm.nih.gov/sites/entrez?Db=pubmed&Cmd=ShowDetailView&TermToSearch=4215872&ordinalpos=2&itool=EntrezSystem2.PEntrez.Pubmed.Pubmed_ResultsPanel.Pubmed_RVDocSum > >Here we see that at one single auditory nerve fiber recording site the >firing rate was strongest at the lowest sound levels of 40 and 50 dB, when >the probe tones had a frequency 6 kHz. At a sound level of 100 dB the firing >rate was *again* strongest, when the probe tones had a frequency 6 kHz. Most >interestingly, the data for the 100 dB probe tones show a second peak at 5.2 >kHz. So, the neural data mainly reflect the OHC tuning, and secondarily also >reflect the passive BM tuning. At following stages of neural processing >these secondary peaks are then filtered out by lateral inhibition. > >4) Conclusion: >The dissociation of BM tuning and auditory nerve fiber tuning is perfect and >obvious, and it has a physiological explanation. It is one of the many >pieces of evidence that demonstrate the *real* function of BM mechanics. > >I hope these data can be of use. I would >appreciate if Richard F. Lyon, or somebody else >from the list, could check the referenced data >and confirm for the list that they are real. > >Martin > >--------------------------------------------------------------------- >Martin Braun >Neuroscience of Music >S-671 95 Klässbol >Sweden >web site: http://w1.570.telia.com/~u57011259/index.htm > > > > >----- Original Message ----- From: "Richard F. Lyon" <DickLyon@xxxxxxxx> >To: <AUDITORY@xxxxxxxx> >Sent: Wednesday, October 10, 2007 9:55 PM >Subject: Re: Cochlea Amplifier models : a new list > >>At 5:39 PM +0200 10/10/07, Martin Braun wrote: >>>>I haven't been able to find the evidence for the assertion that neural >>>>tuning remains unchanged in some way that differs from the sense in which >>>>BM tuning remains unchanged; >>>[.......] >>>>Where should I look to study the data on this idea that underlies your >>>>modeling approach? >>> >>>You can take any data on level dependence of neural responses. None of >>>them mirrors the half-octave shift of BM tuning. >> >>I've looked at lots of data, and you must be interpreting it differently >>from how I am. So if you have something specific that we can look at and >>discuss, we can try to resolve that difference. Lacking that, I'll stick >>with Ruggero's interpretation that says mechanical and neural are >>essentially the same, not different. >> >>Dick